text
stringlengths 254
1.16M
|
|---|
---
title: 'Tele-pharmacy perception, knowledge and associated factors among pharmacy
students in northwest Ethiopia: an input for implementers'
authors:
- Masresha Derese Tegegne
- Sisay Maru Wubante
- Mequannent Sharew Melaku
- Nebyu Demeke Mengiste
- Ashenafi Fentahun
- Wondwossen Zemene
- Tirualem Zeleke
- Agmasie Damtew Walle
- Getnet Tadesse Lakew
- Yonas Tsegaw Tareke
- Mubarek Suleman Abdi
- Hawariyat Mamuye Alemayehu
- Eskedar Menkir Girma
- Gizaw Getye Tilahun
- Addisalem Workie Demsash
- Hiwote Simane Dessie
journal: BMC Medical Education
year: 2023
pmcid: PMC9969706
doi: 10.1186/s12909-023-04111-9
license: CC BY 4.0
---
# Tele-pharmacy perception, knowledge and associated factors among pharmacy students in northwest Ethiopia: an input for implementers
## Abstracts
### Background
Tele-pharmacy is a subset of telemedicine in which pharmacies use telecommunication technology to provide patient care. Tele-pharmacy can improve pharmaceutical care service delivery by reducing medication errors, improving access to health professionals and facilities in remote and rural areas, and minimizing adverse drug events. However, there is limited evidence regarding future pharmacists' knowledge and perceptions of the Tele-pharmacy system in Ethiopia. As a result, this study aimed to assess tele-Pharmacy perception, knowledge and associated factors among pharmacy students in Northwest Ethiopia.
### Methods
An institutional-based cross-sectional study was conducted among 376 pharmacy students in Northwest Ethiopia between July 15 and August 27, 2022. A pre-tested self-administered questionnaire was used to collect data. The data were entered using Epi info version 7.0 and analyzed using SPSS version 25. Descriptive statistics, bivariable and multivariable logistic regression analysis were used to describe pharmacy students' knowledge and perceptions of Tele-pharmacy and identify associated factors. An adjusted odds ratio (OR) and a p-value with a $95\%$ confidence interval (CI) were calculated to declare statistical significance.
### Results
From a total of 352 participants, about $32.4\%$ with [$95\%$ CI ($27\%$-$37\%$)] and $48.6\%$ with [$95\%$ CI ($43\%$—$54\%$)] had good knowledge and a positive perception toward Tele-pharmacy, respectively. Being age group of 26–30 (AOR = 0.35, $95\%$ CI: 0.17–0.68), being male (AOR = 2.38, $95\%$ CI: 1.26–4.49), *Having a* CPGA of > 3.5 (AOR = 2.28, $95\%$ CI: 1.24–4.19), Taking basic computer training (AOR = 2.00, $95\%$ CI: 1.17–3.39), Management support (AOR = 1.84, $95\%$ CI: 1.06–3.19) were found to be significantly associated with pharmacy students' knowledge of Tele-pharmacy. Similarly, having access to electronic devices (AOR = 3.80, $95\%$ CI: 1.81–7.97), training related to pharmacy information systems (AOR = 6.66, $95\%$ CI: 3.34–13.29), availability of guidelines (AOR = 2.99, $95\%$ CI: 1.62–5.50) were found to be significantly associated with pharmacy students' perceptions of Tele-pharmacy.
### Conclusion
This study found that pharmacy students have limited knowledge and perceptions of the Tele-pharmacy system. A continuing Tele-pharmacy training package, incorporating pharmacy information system guidelines as part of their education, and providing managerial support could be recommended to improve pharmacy students' knowledge and perception of Tele-pharmacy.
## Introduction
Telemedicine refers to delivering health care and public education in rural and remote areas [1]. Telemedicine has grown steadily over the last decade as telecommunication technology has advanced and costs have decreased. Tele-pharmacy is a subset of telemedicine in which pharmacies use telecommunication technology to provide patient care [2]. Tele-pharmacy can potentially improve pharmaceutical care service delivery by reducing medication errors and adverse drug events [3]. Furthermore, Tele-pharmacy has the potential to benefit remote and rural areas with limited access to health professionals and facilities [4]. The practical and efficient use of health information technology will be of infinite importance; it will increase pharmacist accessibility, improve patient quality of life and satisfaction with healthcare services, minimize resources, and improve patient clinical outcomes [5].
Pharmacists today want to broaden their profession to provide more services to the rural community while also improving patient outcomes. As a result, Tele-pharmacy services such as medication orders, medication history reviews, dispensing drugs, remote patient consultation, therapeutic drug monitoring, and medication therapy management are becoming more common [6]. These services can be provided using eHealth tools like mobile consultation, software applications, and automated dispensing machines [7].
Pharmacists and student pharmacists should understand the application of telecommunication technology in the pharmacy field to provide the best services. Worldwide there is evidence showing that the proportion of pharmacy student knowledge of tele-pharmacy was $60.3\%$ in the United States [8], $42\%$ in Riyadh city of Saudi Arabia [9], and $67\%$ in Malaysia [4]. Whereas the proportion of perception towards tele-pharmacy was $87\%$ in the University of Tennessee [8], $70.6\%$ in Jordan [10], $61\%$ in Malaysia [4], and $40\%$ in Pakistan [11]. Furthermore, evidences revealed that technological variables and socio-demographic characteristics were discovered to be the determinant factors associated to students' knowledge of and perceptions of the tele-pharmacy system [9–12].
Tele-pharmacy is one of the best options for providing community-based medication-related healthcare services, allowing pharmacists to address healthcare needs in developing countries like Ethiopia, where health professionals and healthcare facilities are scarce [13]. Furthermore, Tele-pharmacy was the best option for patients living in rural areas to reduce travel distance, save time, and access health care services, particularly for those over the age of 65 and those with disabilities [13, 14].
Ethiopian health sector transformation plan mentioned using health-related information technologies such as Tele-pharmacy as an essential transformation device to improve the quality of health care services [15]. Even though it is widely accepted that tele-pharmacy can help improve access and quality of healthcare delivery when distance is an issue, there is little evidence in Ethiopia about future pharmacists' knowledge and perceptions of tele-pharmacy.
The knowledge and perception of pharmacists students about Tele-pharmacy is a determining factor in the successful implementation of Tele-pharmacy services [16]. Since they play a vital role in the health care system and the functioning of Tele-pharmacy, investigating their knowledge and perception of tele-pharmacy is mandatory [17]. As a result, this study aimed to examine pharmacy students' knowledge and perception of tele-pharmacy.
The result of this study will provide baseline data for policymakers in solving the problem regarding the limited utilization of Tele-pharmacy and improving the knowledge and perception of pharmacy students. It will help them in planning an intervention based on the evidence generated by this study.
## Study settings and design
The study was conducted among pharmacy students at the University of Gondar College of Medicine and Health Science, located in the historic town of Gondar, 726 kms northwest of Addis Abeba. The University of Gondar, formerly known as the Gondar College of Medical Sciences until 2003, is Ethiopia's oldest medical school, founded in 1954 as a Public Health College. It is located in northwest Ethiopia, 726 kms from Addis Ababa, Ethiopia's capital. According to data from the University of Gondar's college of medicine and health science school of pharmacy, 376 students were enrolled in their course. An institutional-based cross-sectional study was used to assess pharmacy students’ knowledge and perception of Tele-pharmacy and the associated factors.
## Study population and eligibility criteria
The study was carried out among pharmacy students at the University of Gondar College of Medicine and Health Science. The study included all pharmacy students enrolled in their courses at the University of Gondar College of Medicine and Health Science who were available during the data collection period. However, students who were ill and unable to complete the questionnaire were excluded from this study.
## Sample size determination and sampling procedure
There are currently 376 students enrolled in the pharmacy department at the University of Gondar College of Medicine and Health Sciences. The first-year students were not enrolled in the department at the time of data collection. This is because Ethiopia’s current ministry of health schedule requires all natural science students to attend a one-year common course before enrolling in their specific department. In the remaining years, students attended classes in the pharmacy department. There are 51 second-year students, 54 third-year students, 88 fourth-year students, 82 fifth-year students, and 101 post-basic and postgraduate students. Finally, all active pharmacy students in the pharmacy department at the University of Gondar College of Medicine and Health Science ($$n = 376$$) were invited to participate in this study.
## Study variables
The primary outcome variable of the study was perception and knowledge of Tele-pharmacy. The tools for this study were adapted from a review of related literature [4, 11, 18]. Some independent variables include socio-demographic and technological variables related to Tele-pharmacy perception and knowledge.
## Operational definitions
Knowledge of Tele-pharmacy: Ten items with "yes" or "no" responses were used to assess knowledge of Tele-pharmacy. For a total possible score of ten, each correct answer was worth one point, while each incorrect answer was worth zero points. A median of ten questions about Tele-pharmacy Knowledge was calculated. Those who scored higher than the median value were thought to have "Good knowledge about Tele-pharmacy," while those who scored a median value and lower were supposed to have "Poor knowledge about Tele-pharmacy" [4, 18].
Perception towards Tele-pharmacy was assessed using a 5-point Likert scale ranging from "strongly disagree" (score 1) to "strongly agree" (score 5). A median of 14 questions about Perception toward Tele-pharmacy was calculated. Those who scored higher than the median value were thought to have a "Good perception of Tele-pharmacy," while those who scored a median value and lower were supposed to have a "Poor perception of Tele-pharmacy” [4, 11, 18].
## Data collection procedure and quality control
A structured, pre-tested, and self-administered questionnaire was used to collect data. Six health information technology professionals collected the required data, And Two health informatics professionals with master's degrees and research experience oversaw the data collection. The principal investigators provided training for data collectors and supervisors two days before the start of data collection. Supervisors strictly supervised the data collection process and provided regular onsite advice and feedback to data collectors. The principal investigators and supervisor exchanged information face-to-face daily.
Before data collection, a pre-test was conducted on $10\%$ of the sample size among pharmacy students at Bahirdar University. The questionnaire was checked for clarity, simplicity, understandability, completeness, consistency, and coherency during the pre-testing. Appropriate corrections were taken on time for completeness and accuracy before the beginning of data collection. The pre-test results were also used to assess the internal consistency of the questionnaire. Cronbach's alpha was used to determine the internal validity of the data collection instrument, and the scores on knowledge and perception of Tele-pharmacy were 0.86 and 0.98, respectively.
## Data processing and analysis
The collected data were entered into Epi info version 7.0 and transferred into SPSS version 25.0 software for further analysis. A table, graph, and text were used to present descriptive statistics. A bivariable logistic regression analysis was performed to determine each study variable's effect on the outcome variable. Variables with a p-value of 0.2 in the bivariate analysis would be entered into a multivariable logistic regression analysis to check for confounding effects on the bivariate analysis's association. The strength of the association would be determined using a $95\%$ confidence interval odds ratio, and a p-value less than 0.05 would be considered a significant variable. The model was fitted with $$p \leq 0.34$$ according to homer's goodness of fit test, multi-collinearity was checked between independent variables, and all variance inflation factors were less than 3.
## Socio-demographic characteristics of participants
A total of 352 study subjects participated, with a response rate of $93.62\%$. The mean age of the study participants was 24.26 with an SD ± 3.419 years with ranges from 20 to 37 years. About 259 ($73.6\%$) of the respondents were in the age category of 20–25. The majority, 274($77.8\%$) of the participants, were male, and 295 ($83.8\%$) were single. Regarding religion, around 286 ($81.3\%$) of the study participants were Orthodox, and 93 ($26.4\%$) were post-basic and postgraduate students Table 1.Table 1Socio-demographic characteristics of pharmacy studentsVariablesCategoryFrequencyPercentAge in years20–2525973.626–307521.3> = 31185.1SexMale27477.8Female7822.2ReligionOrthodox28681.3Muslim4211.9Protestant246.8Marital statusSingle29583.8Married4011.4Divorced174.8Year of Study2nd year4713.43rd year4813.64th year8423.95th year8022.7Post basic and Postgraduate9326.4Cumulative GPA< 3.0011332.13.0–3.510830.7> 3.513137.2ResidencyUrban27377.6Rural7922.4
## Technological and organizational characteristics
More than two-thirds of pharmacy students, 286 ($81.3\%$), have access to one of the electronic devices, with the majority of respondents, 203 ($57.7\%$), having access to smartphones. The findings show that approximately 207 ($58.8\%$) of study participants did not receive basic computer training, and only half, 178 ($50.6\%$) of the study participants, had sufficient skill to use computer systems, enabling them to use the Tele-pharmacy system. The majority of respondents, 298 ($84.7\%$), said they had internet access in their learning environment, and roughly three-fourths, 210 ($59.7\%$), said they used the internet to access health-related information.
Most of the 260 ($73.9\%$) pharmacy students did not receive training in pharmacy information systems. Furthermore, more than two-thirds of 237 ($67.3\%$) reported no pharmacy information system implementation guideline, and approximately 262 ($74.4\%$) reported a lack of management support from their department to implement a Tele-pharmacy system Table 2.Table 2Technological and organizational characteristics of pharmacy studentsTechnological and organizational variablesResponseFrequencyPercentageAccess to electronic devicesYes28681.3No6618.8Which devices, did you have access to?Smartphone20357.7Laptop13337.8Desktop123.4Tablet computer82.3Have you ever taken basic computer training before?Yes14541.2No20758.8Self-reported basic computer skillsSufficient17850.6Not sufficient17449.4Access to the internetYes29884.7No5415.3Mainly for what purpose do you use the internetTo get health information21059.7To communicate with my friends10329.3To get daily news6217.6To manage patients’ health data185.1For reporting purpose123.4Others277.7Have you ever taken any training related to pharmacy information systems?Yes9226.1No26073.9Availability of pharmacy information system implementation guidelineYes11532.7No23767.3Is there a management support to implement pharmacy information system from your department?Yes9025.6No26274.4
## Pharmacy student’s knowledge regarding Tele-pharmacy
Of the study participants, 114 ($32.4\%$ CI = $27\%$-$37\%$) had adequate knowledge of the Tele-pharmacy system (Fig. 1. More than three-quarters, 304($86.4\%$) of the participants stated that they are unaware of Tele-pharmacy systems in Ethiopia. Approximately 277 ($78.7\%$) of participants agreed that pharmacists should know about information and communication technology to practice Tele-pharmacy. Fig. 1Pharmacy students’ knowledge and perception of Tele-pharmacy Approximately $80\%$ of participants believed that tele-pharmacy played a significant role in outbreaks worldwide. The majority of 283 ($80.4\%$) of the study participants agreed that Tele-pharmacy provides better counseling in terms of privacy and length of the session, Table 3.Table 3Knowledge of Tele-pharmacy among pharmacy studentsKnowledge regarding Tele-pharmacy statementsYesNoTele-pharmacy is available in Ethiopia48(13.6)304(86.4)Information Communication Technology (ICT) knowledge is important for pharmacists on how to conduct Tele-pharmacy277(78.7)75(21.3)Tele-pharmacy played a big role during outbreak around the world280(79.5)72(20.5)Tele-pharmacy does require a strong internet connection or high-performance technology245(69.6)107(30.4)Tele-pharmacy provides better counseling in terms of privacy and length of the session283(80.4)69(19.6)Tele-pharmacy solves the waiting time problem in most general hospitals283(80.4)69(19.6)Tele-pharmacy is also involved in ADR monitoring and reporting238(67.6)114(32.4)*In* general hospitals, Tele-pharmacy is conducted by drug information service during office hours and by emergency departments after office hours233(66.2)119(33.8)Patients from rural areas can have more medication access and information via Tele-pharmacy168(47.7)184(52.3)Tele-pharmacy services can extend hospital pharmacy services outside office hours that do not offer round-the-clock pharmacy services227(64.5)125(35.5)
## Pharmacy student’s perception of Tele-pharmacy
Of the total study participants, 171 ($48.6\%$ CI = $43\%$—$54\%$) had a positive perception of Tele-pharmacy Fig. 1. Table 4 illustrates that about 195($55.4\%$) agreed that Tele-pharmacy can help patients save their money and travel time to reach healthcare facilities. The majority of 245($69.6\%$) of the study participants agreed that Tele-pharmacy could minimize the cost of establishing a pharmaceutical business compared to a regular pharmacy. Moreover, about 216(61.4) of the study participants agreed that pharmacy schools should provide education programs on computers, IT, and Tele-pharmacy to assist in the future utilization of Tele-pharmacy Table 4.Table 4Perception towards Tele-pharmacy among pharmacy studentsPerception towards Tele-pharmacy QuestionsStrongly DisagreeDisagreeNeutralAgreeStrongly agreeDo you think Tele-pharmacy will improve the patient’s adherence to the medication?12(3.4)24(6.8)17(4.8)224(63.6)75(21.3)Do you agree Tele-pharmacy will have a higher error rate for medication dispensing and filling, as compared to traditional pharmacy?39(11.1)53(15.1)47(13.4)177(50.3)36(10.2)Do you feel Tele-pharmacy will enhance the patient’s access to the medication, especially those who are in rural areas?12(3.4)61(17.3)21(6.0)189(53.7)69(19.6)Do you think Tele-pharmacy will provide the complete privacy setting during the consultation period?9(2.6)27(7.7)51(14.5)193(54.8)72(20.5)Based on your opinion, do you agree Tele-pharmacy will increase the pharmacist’s workload and commitment?30(8.5)76(21.6)30(8.5)162(46.0)54(15.3)Do you think Tele-pharmacy is able to help patients save their money and travel time to reach the healthcare facilities?3(0.9)17(4.8)9(2.6)195(55.4)128(36.4)Are you willing to share your personal information on the online database when using Tele-pharmacy services?11(3.1)21(6.0)54(15.3)191(54.3)75(21.3)Do you think Tele-pharmacy can minimize the cost to establish pharmaceutical business in comparison to regular pharmacy?-29(8.2)24(6.8)245(69.6)54(15.3)Do you think patient consultation via Tele-pharmacy will be effective?6(1.7)26(7.4)33(9.4)239(67.9)48(13.6)Do you think pharmacy schools should provide education programs on computers, IT, and Tele-pharmacy to assist in future utilization of Tele-pharmacy?8(2.3)9(2.6)18(5.1)216(61.4)101(28.7)Do you think therapeutic drug monitoring via Tele-pharmacy in rural areas will be easily monitored?30(8.5)72(20.5)43(12.2)141(40.1)66(18.8)Do you agree that security is a greater concern in a remote site Tele-pharmacy than in a traditional community pharmacy?9(2.6)56(15.9)45(12.8)173(49.1)69(19.6)Scarcity of pharmacists has caused a situation where medications are supplied without the involvement of pharmacists18(5.1)30(8.5)54(15.3)163(46.3)87(24.7)Do you agree that Tele-pharmacy is able to help to minimize this scarcity?15(4.3)6(1.7)39(11.1)211(59.9)81(23.0)
## Factors associated with pharmacy students’ knowledge about Tele-pharmacy
Bivariate and multivariable analyses were used to investigate the factors associated with students' knowledge of Tele-pharmacy. In a bivariate analysis, the candidate variables for the multivariable logistic regression analysis were Sex, Age, Student’s grade (CPGA), Training related to pharmacy information systems, Device access, Basic computer training, computer skill, internet access, availability of pharmacy information system implementation guidelines, and management support to implement a pharmacy information system.
According to the findings of multivariable logistic regression analysis, being age group of 26–30 (AOR = 0.35, $95\%$ CI: 0.17–0.68), being male (AOR = 2.38, $95\%$ CI: 1.26–4.49), *Having a* CPGA of > 3.5 (AOR = 2.28, $95\%$ CI: 1.24–4.19), Taking basic computer training (AOR = 2.00, $95\%$ CI: 1.17–3.39), Management support to implement pharmacy information system (AOR = 1.84, $95\%$ CI: 1.06–3.19) were found to be significantly associated with knowledge towards tele-pharamcy among pharmacy students Table 5.Table 5Factors associated with pharmacy students’ knowledge about Tele-pharmacy ($$n = 352$$)CharacteristicsKnowledgeCOR (CI $95\%$)AOR (CI $95\%$)Good (%)Poor (%)Age20–2593(26.4)166(47.2)1126–3015(4.3)60(17.0)0.44(0.24–0.83)0.35(0.17–0.68) *> 316(1.7)12(3.4)0.89(0.32–2.45)0.97(0.36–3.15)SexMale96(27.3)178(50.6)1.79(1.00–3.21)2.38(1.26–4.49) *Female18(5.1)60(17.0)11CGPA< 3.0027(7.7)86(24.4)113.0–3.536(10.2)72(20.5)1.59(0.88–2.87)1.70(0.90–3.21)> 3.551(14.5)80(22.7)2.03(1.16–3.54)2.28(1.24–4.19) *Access to electronic devicesYes97(27.6)189(53.7)1.47(0.80–2.70)1.36(0.70–2.63)No17(4.8)49(13.9)11Computer trainingYes60(17.0)85(24.1)2.00(1.27–3.14)2.00(1.17–3.39) *No54(15.3)153(43.5)11Computer skillYes65(18.1)113(32.1)1.47(0.93–2.30)1.27(0.74–2.17)No49(13.9)125(35.5)11Internet accessYes100(28.4)198(56.3)1.44(0.75–2.77)1.23(0.60–2.51)No14(4.0)40(11.4)11Training related to pharmacy information systemsYes36(10.2)56(15.9)1.50(0.91–2.46)1.01(0.56–1.82)No78(22.2)182(51.7)11Availability of pharmacy information system implementation guidelineYes44(12.5)71(20.2)1.47(0.92–2.36)1.16(0.66–2.03)No70(19.9)167(47.4)11Management support to implement pharmacy information systemYes39(11.1)51(14.5)1.90(1.16–3.13)1.84(1.06–3.19) *No75(21.3)187(53.1)11N:B *P-value < 0.05 for the multivariable analysis
## Factors associated with pharmacy students’ perception of Tele-pharmacy
Based on the multivariable logistic regression analysis in Table 5, Having access to electronic devices (AOR = 3.80, $95\%$ CI: 1.81–7.97), training related to pharmacy information systems (AOR = 6.66, $95\%$ CI: 3.34–13.29), availability of pharmacy information system implementation guideline (AOR = 2.99, $95\%$ CI: 1.62–5.50) were found to be significantly associated with perception towards Tele-pharmacy among pharmacy students Table 6.Table 6Factors associated with pharmacy students’ perception of Tele-pharmacy ($$n = 352$$)CharacteristicsPerceptionCOR (CI $95\%$)AOR (CI $95\%$)Good (%)Poor (%)SexMale130(36.9)144(40.9)0.81(0.49–1.34)0.64(0.35–1.19)Female41(11.6)37(10.5)11Access to electronic devicesYes153(43.5)133(37.8)3.06(1.70–5.53)3.80(1.81–7.97) *No18(5.1)48(13.6)11Computer trainingYes95(27.0)50(14.2)3.27(2.10–5.10)1.58(0.92–2.72)No76(21.6)131(37.2)11Computer skillYes106(30.1)72(20.5)2.46(1.60–3.79)1.23(0.72–2.08)No65(18.5)109(31.0)11Internet accessYes152(43.2)146(41.5)1.91(1.04–3.50)1.87(0.92–3.83)No19(5.4)35(9.9)11Training related to pharmacy information systemsYes79(22.4)13(3.7)11.09(5.85–21.0)6.66(3.34–13.29) *No92(26.1)168(47.7)11Availability of pharmacy information system implementation guidelineYes81(23.0)34(9.7)3.89(2.41–6.28)2.99(1.62–5.50) *No90(25.6)147(41.8)11Management support to implement pharmacy information systemYes59(16.8)31(8.8)2.54(1.54–4.19)1.62(0.86–3.01)No112(31.8)150(42.6)11KnowledgeGood62(17.6)52(14.8)1.41(0.90–2.20)1.09(0.63–1.88)Poor109(31.0)129(36.6)11N:B *P-value < 0.05 for the multivariable analysis
## Discussion
The finding of this study revealed that $32.4\%$ ($95\%$ CI = $27\%$-$37\%$) of pharmacy students had adequate knowledge regarding telepharmacy. The result of this study is lower as compared with studies conducted in Malaysia $67\%$ [4], Saudi Arabia $42\%$ [19], and the United States of America $60\%$ [8]. The significant disparity could be attributed to the fact that developing countries use fewer eHealth applications than middle-income and developed countries [20, 21]. Because of the lower use of Tele-pharmacy in developing countries, such as Ethiopia, there is a significant knowledge gap between developed and developing countries. Another reason for the disparity could be country-specific differences in information and communication technology infrastructure and socioeconomic status [22].
In this study, $48.6\%$ ($95\%$ CI = $43\%$—$54\%$) of pharmacy students had a favorable perception of the Tele-pharmacy system. This finding is lower as compared with a study conducted in Malaysia with $61\%$ [4], Saudi Arabia with $87\%$ [11], Jordan $70.6\%$ [10], and the United States $87\%$ [8]. This could be due to technological advancement and ICT infrastructure differences between countries. Furthermore, the difference could be due to differences in educational curricula between countries; for example, in Ethiopia, pharmacy students only take fundamentals of health informatics courses. As a result, delivering an eHealth application course that includes Tele-pharmacy as part of their educational curriculum is highly recommended.
This research also found factors associated with pharmacy students' knowledge and perception of the Tele-pharmacy system.
Among the factors associated with knowledge, students in older age groups were $65\%$ less likely to have sufficient knowledge of tele-pharmacy compared to students in younger age groups. This outcome is consistent with past research that showed younger students had a higher comprehension of health information technologies [23–25]. This was explained by the fact that most university students in the Ethiopian were under the age of 25, and younger students were more active in using and accessing information and communication technology [26].
This study found that male respondents were 2.38 times more likely to have adequate knowledge about Tele-pharmacy than females. Other research has also found that men understand eHealth applications better than women [27–29]. The digital divide could explain the difference and gender inequality in access to technology continue to be challenging in low-income countries like Ethiopia. This implied that female students would receive more attention to improve their understanding of eHealth applications. Moreover, students *Having a* CPGA of > 3.5 were 2.28 times more likely to have adequate knowledge of the Tele-pharmacy system. This was explained by the fact that students with higher CPGA were more likely to understand the Tele-pharmacy system than students with lower CPGA.
Pharmacy students who took training on a basic computer were 2.00 times more likely to have adequate knowledge of Tele-pharmacy than students who did not receive basic computer training. This study's findings are consistent with previous research indicating that computer training can improve understanding of the Telehealth applications [30–32]. This is because basic computer training is the most critical factor in improving students' knowledge of Tele-pharmacy [33]. This implies that computer training is an integral part of successful Tele-pharmacy adoption. Furthermore, those with management support were 1.84 times more likely to have good knowledge of Tele-pharmacy than those who did not have management support. This implies that health administrators should pay attention and provide ongoing support to future pharmacists to understand better and implement a Tele-pharmacy system.
Among the factors significantly associated with pharmacy students' perceptions of Tele-pharmacy, having access to electronic devices was 3.80 times more likely to have a positive perception than not having access to electronic devices. Previous research backs up this evidence that having access to electronic devices is the most critical factor in having a positive perception of the Telemedicine system [34, 35]. This implied that access to electronic devices was required to implement Ethiopia's Tele-pharmacy system successfully.
Those who received pharmacy information system training were 6.66 times more likely to have a positive perception of the Tele-pharmacy system than those who did not receive pharmacy technology training. This is because students who have received training in pharmacy technology may have a favorable opinion of the Tele-pharmacy system [10, 36]. Furthermore, the availability of pharmacy information system implementation guidelines was linked to a positive perception of the Tele-pharmacy system. Those with guidelines on pharmacy information system implementation were 2.99 times more likely to have a positive perception of Tele-pharmacy than their counterparts. This implies that health managers should prepare a user manual when planning to implement Telehealth applications in the Ethiopian healthcare system.
## Conclusion and recomendation
This study found that pharmacy students have limited knowledge and perceptions of the Tele-pharmacy system. A continuing Tele-pharmacy training package, incorporating pharmacy information system guidelines as part of their education, and providing managerial support could be recommended to improve pharmacy students' knowledge and perception of Tele-pharmacy. Based on the findings, policymakers and other stakeholders can develop a plan to implement Tele-pharmacy in the health care system.
## Strength and limitations of the study
This study is unique in Ethiopia because it discovered future pharmacists' knowledge and perceptions of Tele-pharmacy. However, the study has some limitations. The study's findings could be influenced by response bias. However, we attempted to reduce this bias by improving survey designs through pre-testing feedback.
## References
1. Wootton R. *Telemedicine BMJ* (2001.0) **323** 557-560. PMID: 11546704
2. Angaran DM. **Telemedicine and telepharmacy: Current status and future implications**. *Am J Health Syst Pharm* (1999.0) **56** 1405-1426. DOI: 10.1093/ajhp/56.14.1405
3. 3.Petropoulou S, Bekakos M, Gravvanis G, editors. E-prescribing-Telepharmacy. 7th Hellenic European Conference on Computers; 2005: Citeseer.
4. Elnaem MH, Akkawi ME, Al-Shami AK, Elkalmi R. **Telepharmacy knowledge, perceptions, and readiness among future Malaysian pharmacists amid the COVID-19 pandemic**. *Ind J Pharm Educ Res* (2022.0) **56** 09-16. DOI: 10.5530/ijper.56.1.2
5. Belay TS. *Utilization of Telemedicine in Tikur Anbessa Specialized Hospital, Addis Ababa* (2013.0)
6. Baldoni S, Amenta F, Ricci G. **Telepharmacy services: present status and future perspectives: a review**. *Medicina* (2019.0) **55** 327. DOI: 10.3390/medicina55070327
7. Keeys C, Kalejaiye B, Skinner M, Eimen M, Neufer J, Sidbury G. **Pharmacist-managed inpatient discharge medication reconciliation: a combined onsite and telepharmacy model**. *Am J Health Syst Pharm* (2014.0) **71** 2159-2166. DOI: 10.2146/ajhp130650
8. 8.Patel K. Assessment of Knowledge, Attitude, Perception of Pharmacy Students Towards Telepharmacy. Applied Research Projects. 2021;75. 10.21007/chp.hiim.0072, https://dc.uthsc.edu/hiimappliedresearch/75.
9. Dat TV, Tran TD, My NT, Nguyen TTH, Quang NNA, Tra Vo Nguyen M. **Pharmacists’ Perspectives on the Use of Telepharmacy in Response to COVID-19 Pandemic in Ho Chi Minh City, Vietnam**. *J Pharm Tech* (2022.0) **38** 106-14. DOI: 10.1177/87551225221076327
10. Muflih SM, Al-Azzam S, Abuhammad S, Jaradat SK, Karasneh R, Shawaqfeh MS. **Pharmacists’ experience, competence and perception of telepharmacy technology in response to COVID-19**. *Int J Clin Pract* (2021.0) **75** e14209. DOI: 10.1111/ijcp.14209
11. Muhammad K, Baraka MA, Shah SS, Butt MH, Wali H, Saqlain M. **Exploring the perception and readiness of Pharmacists towards telepharmacy implementation; a cross sectional analysis**. *PeerJ* (2022.0) **10** e13296. DOI: 10.7717/peerj.13296
12. Umayam KAD, Rosadia ANN, Tan RNR, Salazar DJR, Masakayan RLL, Santiago GMB. **Knowledge, Attitudes and Perceptions on the Use of Telemedicine Among Adults Aged 18–34 in Manila, Philippines During the COVID-19 Pandemic**. *J Med Univ Santo Tomas* (2022.0) **6** 858-867. DOI: 10.35460/2546-1621.2021-0144
13. Win AZ. **Telepharmacy: Time to pick up the line**. *Res Social Adm Pharm: RSAP* (2017.0) **13** 882-883. DOI: 10.1016/j.sapharm.2015.06.002
14. Poudel A, Nissen LM. **Telepharmacy: a pharmacist’s perspective on the clinical benefits and challenges**. *Integr Pharm Res Pract* (2016.0) **5** 75. DOI: 10.2147/IPRP.S101685
15. Seboka BT, Yilma TM, Birhanu AY. **Factors influencing healthcare providers’ attitude and willingness to use information technology in diabetes management**. *BMC Med Inform Decis Mak* (2021.0) **21** 1-10. DOI: 10.1186/s12911-021-01398-w
16. Ameri A, Salmanizadeh F, Keshvardoost S, Bahaadinbeigy K. **Investigating pharmacists’ views on telepharmacy: prioritizing key relationships, barriers, and benefits**. *J Pharm Technol* (2020.0) **36** 171-178. DOI: 10.1177/8755122520931442
17. Ayatollahi H, Sarabi FZP, Langarizadeh M. **Clinicians’ knowledge and perception of telemedicine technology**. *Perspect Health Inf Manag* (2015.0) **12** 1c
18. Tjiptoatmadja NN, Alfian SD. **Knowledge, Perception, and Willingness to Use Telepharmacy Among the General Population in Indonesia**. *Front Public health* (2022.0) **10** 825554. DOI: 10.3389/fpubh.2022.825554
19. Alanazi A, Albarrak A, Muawad R. *5PSQ-184 Knowledge and attitude assessment of pharmacists toward telepharmacy in Riyadh City* (2021.0)
20. Kirigia JM, Seddoh A, Gatwiri D, Muthuri LH, Seddoh J. **E-health: determinants, opportunities, challenges and the way forward for countries in the WHO African Region**. *BMC Public Health* (2005.0) **5** 1-11. DOI: 10.1186/1471-2458-5-137
21. Quaglio G, Dario C, Karapiperis T, Delponte L, Mccormack S, Tomson G. **Information and communications technologies in low and middle-income countries: Survey results on economic development and health**. *Health Policy Technol* (2016.0) **5** 318-329. DOI: 10.1016/j.hlpt.2016.07.003
22. 22.Mekuria F, Nigussie EE, Dargie W, Edward M, Tegegne T. Information and Communication Technology for Development for Africa: First International Conference, ICT4DA 2017, Bahir Dar, Ethiopia, September 25–27, 2017, Proceedings: Springer; 2018.
23. Alwan K, Ayele TA, Tilahun B. **Knowledge and utilization of computers among health professionals in a developing country: a cross-sectional study**. *JMIR Hum Factors* (2015.0) **2** e4184. DOI: 10.2196/humanfactors.4184
24. 24.Tegegne MD, Endehabtu BF, Klein J, Gullslett MK, Yilma TM. Use of social media for COVID-19-related information and associated factors among health professionals in Northwest Ethiopia: A cross-sectional study. Digital Health. 2022;8:20552076221113390.
25. Shiferaw KB, Mehari EA. **Internet use and eHealth literacy among health-care professionals in a resource limited setting: a cross-sectional survey**. *Adv Med Educ Pract* (2019.0) **10** 563. DOI: 10.2147/AMEP.S205414
26. Engel A, Salvador CC, Membrive A, Badenas JO. **Information and communication technologies and students’ out-of-school learning experiences**. *Digital Education Review* (2018.0) **33** 130-149. DOI: 10.1344/der.2018.33.130-149
27. Jain R, Dupare R, Bhanushali N, Kumar V. **Knowledge and utilization of computer among health-care professionals in Mumbai**. *J Indian Assoc Public Health Dentist* (2020.0) **18** 97
28. Kay R. **Addressing gender differences in computer ability, attitudes and use: The laptop effect**. *J Educ Computing Res* (2006.0) **34** 187-211. DOI: 10.2190/9BLQ-883Y-XQMA-FCAH
29. Shiferaw KB, Tilahun BC, Endehabtu BF. **Healthcare providers’ digital competency: a cross-sectional survey in a low-income country setting**. *BMC Health Serv Res* (2020.0) **20** 1-7. DOI: 10.1186/s12913-020-05848-5
30. 30.Weldegebrial TT, Berhie G. Telehealth in Ethiopia–The Barriers Vs. The Success Factors. Marshall University. 2017;2016.
31. Tegegne MD, Wubante SM. **Identifying barriers to the adoption of information communication technology in ethiopian healthcare systems. a systematic review**. *Adv Med Educ Pract* (2022.0) **13** 821. DOI: 10.2147/AMEP.S374207
32. Wubante SM, Tegegne MD. **Health professionals knowledge of telemedicine and its associated factors working at private hospitals in resource-limited settings**. *Front Digit Health* (2022.0) **4** 976566. DOI: 10.3389/fdgth.2022.976566
33. Aruru M, Truong H-A, Clark S. **Pharmacy Emergency Preparedness and Response (PEPR): a proposed framework for expanding pharmacy professionals’ roles and contributions to emergency preparedness and response during the COVID-19 pandemic and beyond**. *Res Social Adm Pharm* (2021.0) **17** 1967-1977. DOI: 10.1016/j.sapharm.2020.04.002
34. Sagaro GG, Battineni G, Amenta F. **Barriers to sustainable telemedicine implementation in Ethiopia: A systematic review**. *Telemed Rep* (2020.0) **1** 8-15. PMID: 35722252
35. Biruk K, Abetu E. **Knowledge and attitude of health professionals toward telemedicine in resource-limited settings: a cross-sectional study in North West Ethiopia**. *J Healthc Eng* (2018.0) **2018** 2389268. DOI: 10.1155/2018/2389268
36. 36.Francis SG. Pharmacists’ Perceptions About the Effect of Work Environment Factors on Patient Safety in Large-Chain Retail Pharmacies. J Pharm Technol. 2022;38(6):376–8. 10.1177/87551225221116000.
|
---
title: Enlarged waist combined with elevated triglycerides (hypertriglyceridemic waist
phenotype) and HDL-cholesterol in patients with heart failure
authors:
- Camila Weschenfelder
- Aline Marcadenti
- Airton Tetelbom Stein
- Catarina Bertaso Andreatta Gottschall
journal: São Paulo Medical Journal
year: 2017
pmcid: PMC9969720
doi: 10.1590/1516-3180.2016.004519102016
license: CC BY 4.0
---
# Enlarged waist combined with elevated triglycerides (hypertriglyceridemic waist phenotype) and HDL-cholesterol in patients with heart failure
## ABSTRACT
### CONTEXT AND OBJECTIVE:
The association of serum triglycerides plus waist circumference seems to be a good marker of cardiovascular risk and has been named the “hypertriglyceridemic waist” phenotype. The aim of our study was to investigate the association between the hypertriglyceridemic waist phenotype and HDL-cholesterol among patients with heart failure.
### DESIGN AND SETTING:
Cross-sectional study in a tertiary-level hospital in southern Brazil.
### METHODS:
We included patients with heart failure aged > 40 years. Anthropometric assessment (weight, height, waist and hip circumferences) was performed; body mass index (BMI) and waist-hip ratio were calculated and lipid measurements (serum total cholesterol, LDL-cholesterol, HDL-cholesterol and triglycerides) were collected. In men and women, respectively, waist circumference ≥ 94 cm and ≥ 80 cm, and triglycerides ≥ 150 mg/dl were considered abnormal and were used to identify the hypertriglyceridemic waist phenotype. Analyses of covariance were used to evaluate possible associations between levels of HDL-cholesterol and the hypertriglyceridemic waist phenotype, according to sex.
### RESULTS:
112 participants were included, of whom $62.5\%$ were men. The mean age was 61.8 ± 12.3 years and the mean ejection fraction was 40.1 ± $14.7\%$. Men and woman presented mean HDL-cholesterol of 40.5 ± 14.6 and 40.9 ± 12.7 mg/dl, respectively. The prevalence of the hypertriglyceridemic waist phenotype was $25\%$. There was a significant difference in mean HDL-cholesterol between men with and without the hypertriglyceridemic waist phenotype (32.8 ± 14.2 versus 42.1 ± 13.7 mg/dl respectively; $$P \leq 0.04$$), even after adjustment for age, body mass index, type 2 diabetes mellitus, use of statins and heart failure etiology.
### CONCLUSIONS:
The hypertriglyceridemic waist phenotype is significantly associated with lower HDL-cholesterol levels in men with heart failure.
## CONTEXTO E OBJETIVO:
A associação de triglicerídeos séricos e circunferência da cintura parece ser um bom marcador de risco cardiovascular e é denominada fenótipo da cintura hipertrigliceridêmica. O objetivo do estudo foi avaliar a associação entre o fenótipo da cintura hipertrigliceridêmica e o HDL-colesterol em pacientes portadores de insuficiência cardíaca.
## TIPO DE ESTUDO E LOCAL:
Estudo transversal em um hospital terciário no sul do Brasil.
## MÉTODOS:
Foram incluídos indivíduos com insuficiência cardíaca com idade > 40 anos. Foram realizadas as medidas antropométricas (peso, estatura, circunferência da cintura e do quadril) e calculados índice de massa corporal e relação cintura quadril, e foi avaliado o perfil lipídico (colesterol total, LDL-colesterol, HDL-colesterol e triglicerídeos séricos). Em homens e mulheres, respectivamente, circunferência da cintura ≥ 94 cm e ≥ 80 cm e triglicerídeos ≥ 150 mg/dl foram considerados anormais e usados para identificação do fenótipo da cintura hipertrigliceridêmica. Análises de covariância foram usadas para avaliar possíveis associações entre níveis de HDL-colesterol e o fenótipo da cintura hipertrigliceridêmica de acordo com o sexo.
## RESULTADOS:
Foram incluídos 112 participantes e 62,$5\%$ eram homens. A média de idade foi de 61,8 ± 12,3 anos e a fração de ejeção média foi 40,1 ± 14,$7\%$. Homens e mulheres apresentaram médias de HDL-colesterol 40,5 ± 14,6 e 40,9 ± 12,7 mg/dl, respectivamente. A prevalência do fenótipo da cintura hipertrigliceridêmica na amostra foi de $25\%$. Observou-se diferença significativa entre as médias de HDL-colesterol entre homens com e sem o fenótipo da cintura hipertrigliceridêmica (32,8 ±14,2 versus 42,1 ± 13,7 mg/dl, $$P \leq 0$$,04), mesmo após ajuste para idade, índice de massa corporal, diabetes mellitus tipo 2, uso de estatinas e etiologia da insuficiência cardíaca.
## CONCLUSÕES:
O fenótipo da cintura hipertrigliceridêmica está associado significativamente com menores níveis de HDL-colesterol em homens com insuficiência cardíaca.
## INTRODUCTION
Heart failure is a complex systemic clinical syndrome1 and coronary artery disease is the main cause of heart failure of ischemic origin.2 An obesity paradox is commonly reported among patients with heart failure, in which patients with high adiposity have a better prognosis than do individuals who are normal or underweight.3 The prognostic value of indexes that detect excess abdominal body fat, such as waist circumference (the traditional tool) and the visceral adiposity index (an alternative and emerging tool) have been evaluated among individuals with ischemic heart failure,4 since abdominal obesity is also associated with coronary heart disease.
In addition to abdominal obesity, there has been increasing interest in the role of the atherogenic lipid triad, i.e. hyperinsulinemia, elevated apolipoprotein B and small, dense low density lipoprotein (LDL) particles, in the genesis of coronary artery disease.5,6 However, difficulties in obtaining these parameters in routine practice hinder their use in screening for individuals at high cardiovascular risk. The hypertriglyceridemic waist phenotype (enlarged waist and elevated triglycerides, EWET), defined as simultaneous presence of increased waist circumference and elevated triglycerides, seems to more accurately identify individuals who are at risk, compared with isolated measurements of waist circumference or serum triglycerides,7 and can be applied in clinical practice. In addition to the strong association of the hypertriglyceridemic waist phenotype with the atherogenic triad,8,9 it is related to increased visceral adipose tissue,10 worse cardiometabolic profile (both in the general population11,12,13 and in individuals who are at risk14,15 or who present cardiovascular disease16), higher incidence of coronary artery disease and cardiovascular mortality.17 Low high-density lipoprotein-cholesterol (HDL-c) levels are negatively associated with cardiovascular events in individuals with cardiovascular diseases.18,19 Individuals with the hypertriglyceridemic waist phenotype have been found to present decreased HDL-c levels11,12 and smaller HDL particles.20 Gomez-Huelgas et al.12 showed that subjects without cardiovascular disease but with the hypertriglyceridemic waist phenotype had lower HDL-c levels independently of sex and age. However, the prevalence of the hypertriglyceridemic waist phenotype was higher in men and it was positively associated with age. In a multiethnic population also without cardiovascular disease,11 men with the hypertriglyceridemic waist phenotype showed lower HDL-c levels than women, while HDL-c levels were significantly lower in women with hypertriglyceridemic waist than in those without this phenotype.
Lower levels of HDL-c and higher levels of serum triglycerides may lead to a worse prognosis for ischemic heart disease patients.21 Moreover, adipokines secreted by visceral adipocytes may negatively contribute towards decreased HDL-c levels in individuals with heart failure.22 Although the hypertriglyceridemic waist phenotype has been investigated in populations in which the obesity paradox is common,23 it has not yet been evaluated in heart failure patients.
## OBJECTIVE
To evaluate a possible association between HDL-cholesterol and hypertriglyceridemic waist in men and women with heart failure.
## METHODS
We performed a cross-sectional analysis among patients who had previously been diagnosed with heart failure and who were enrolled at the baseline of a cohort study conducted in a public tertiary hospital. Between 2011 and 2012, these patients were consecutively enrolled if they met the following inclusion criteria: history of New York Heart Association class I-IV heart failure defined by cardiologists in accordance with the American College of Cardiology Foundation/American Heart Association (ACCF/AHA) criteria;24 age between 40 and 90 years; no history or clinical evidence of severe heart failure comorbidities (coronary artery disease, cerebrovascular disease or severe kidney disease) over the last six months; and residency in the Porto Alegre metropolitan area (southern Brazil). The following were excluded: patients with lower limb amputation, sequelae of stroke, acute coronary syndrome in the last 90 days or valvular heart disease; pregnant women; candidates for myocardial revascularization; patients in the postoperative period of cardiac surgery (myocardial revascularization or heart valve surgery performed less than one year earlier); and individuals with a history of cancer within the last two years.
Dietitians, medical students and nutrition students administered a questionnaire that asked for clinical data (use of medications, history of diseases, hospitalizations, etc.) and sociodemographic data (age, sex, educational attainment and self-reported skin color). A field coordinator (local cardiologist) was responsible for quality control in relation to the interviews. Patients were also asked about alcohol consumption (alcohol abuse was defined as ethanol consumption per day of 30 g or more among men and 15 g or more among women) and smoking habits, in which they were classified as current smokers, ex-smokers or never smokers.
An anthropometric assessment was performed at the first clinical evaluation. Weight and height were measured with the patient wearing lightweight clothing and standing barefoot on a flat surface, in accordance with the method proposed by Lohman.25 Weight was measured to the nearest 100 g using a calibrated scale with a capacity of 150 kg (Cauduro, Brazil). Height was measured to the nearest 0.1 cm using a stadiometer with a measuring rod of 205 cm (Sanny, Brazil). Body mass index (BMI) was calculated in accordance with the World Health Organization criteria, using a cutoff point of 30 kg/m2 for the diagnosis of obesity.
Waist and hip circumferences were measured in cm, using an inelastic measuring tape. Waist circumference was measured at the midpoint between the lowest rib and the upper border of iliac crest,26 and hip circumference was measured at the maximum protuberance of the buttocks. The waist-hip ratio was calculated by dividing the waist circumference by the hip circumference, and an elevated waist-hip ratio was defined as > 0.90 for men and > 0.85 for women.27 The ejection fraction (%) was determined during a transthoracic echocardiogram, using color Doppler and tissue Doppler imaging (GE VIVID 3, General Electric, Norway).2 *These data* were obtained from patients’ medical records. Heart failure etiology was diagnosed by the cardiology staff and was registered in the medical records: ischemic etiology was defined if the individual had a previous diagnosis of ischemic heart disease.
For lipid measurements (serum total cholesterol, LDL-cholesterol, HDL-c and triglycerides), 10 ml of venous blood was collected from each participant. Lipid concentrations were determined using a standard colorimetric enzymatic method. HDL-c levels (dependent variable) were treated as continuous values for statistical analysis. The lipid profile was considered to be altered if the HDL-c level was below 40 mg/dl in men and 50 mg/dl in women, and if serum triglycerides were above 150 mg/dl in men and women,28 in addition to the medical diagnosis.
Patients were deemed to present hypertriglyceridemic waist (main independent variable) if they had waist circumference ≥ 94 cm (men) or ≥ 80 cm (women) + serum triglycerides ≥ 150 mg/dl.28,29 Thus, these patients were considered were considered to present the hypertriglyceridemic waist phenotype. Blood pressure was determined using standard techniques, and patients were considered hypertensive if they had previously been diagnosed with hypertension (collected from the medical records), if they had systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg, or if they were taking antihypertensive drugs.23 Fasting blood glucose ≥ 126 mg/dl or glycated hemoglobin ≥ $6.5\%$ or a previous medical diagnosis were used to detect patients with type 2 diabetes mellitus.30 Sample size was calculated using the WinPepi software, version 11.18. The total sample size required for the study was calculated as 76 individuals, by making the assumptions that the prevalence of hypertriglyceridemic waist phenotype would be at least $20\%$ in the sample, with a difference of at least 7 mg/dl in HDL-c levels between patients with and without the hypertriglyceridemic waist phenotype (standard deviations of 12.3 and 9.4 mg/dl, respectively),13 a power of $80\%$ and a significance level of $5\%$.
Analyses were performed using the Statistical Package for the Social Sciences (SPSS) software, version 17.0 (SPSS, IL, USA). Continuous variables were expressed as means and standard deviations and categorical variables as absolute values and percentages. Student’s t test (continuous variables) and Pearson’s chi-square or Fisher’s exact test (categorical variables) were used for comparisons. Analyses of covariance (ANCOVA) were used to evaluate possible associations between mean HDL-c and hypertriglyceridemic waist after adjustment for potential confounding factors (age, BMI, diagnoses of type 2 diabetes mellitus, statin use and heart failure etiology), separately according to gender. For each analysis, an α-level = 0.05 was considered significant, and $95\%$ confidence intervals (CI) were shown.
The study was approved by the local Research Ethics Committee (CEP-GHC number 10-118), and all patients signed an informed consent statement. There was no external funding for the study.
## RESULTS
Between July 2011 and January 2012, 112 patients were included, of whom 70 ($62.5\%$) were men. Eighty-five patients (approximately $76\%$) were classified as New York Heart Association grade III-IV. The patients had a mean age of 61.8 ± 12.3 years, and a mean of 5 ± 3.3 years of educational attainment. Thirteen patients ($12\%$) were smokers, 55 ($49\%$) ex-smokers, and 44 ($39\%$) never smoked; 10 patients ($9\%$) were identified as alcohol abusers. Thirty-seven patients ($33\%$) were diagnosed with type 2 diabetes mellitus, 86 ($77\%$) had hypertension and 38 ($34\%$) had dyslipidemia. The mean ejection fraction was 40.1 ± $14.7\%$, and 19 patients ($17\%$) were diagnosed with ischemic heart failure. The prevalence of hypertriglyceridemic waist phenotype was $25\%$ ($95\%$ CI: 16.8-35.6).
The mean BMI was 28.4 ± 6.5 kg/m2, and 36 patients ($32\%$) were considered obese (BMI ≥ 30 kg/m2). BMI was higher among women (29.7 ± 7.6 kg/m2) than among men (27.6 ± 5.8 kg/m2), but with no statistical difference. Elevated waist-hip ratio was identified in 91 patients ($81\%$), and the waist-hip ratio values were higher among men (0.99 ± 0.11) than among women (0.93 ± 0.07), but with no statistical difference. Regarding the prevalence of enlarged waist circumference according to different cutoff points for detecting higher cardiovascular risk, for ≥ 102 cm among men and ≥ 88 cm among women, there were 26 cases ($23.2\%$) and 32 cases ($28.6\%$), respectively; for ≥ 94 among men and ≥ 80 among women, there were 37 cases ($33\%$) and 46 cases ($41.1\%$). Triglyceride levels ≥ 150 mg/dl were detected in 32 individuals ($28.6\%$).
No differences between men and women were observed regarding HDL-c levels (40.5 ± 14.6 mg/dl in men and 40.9 ± 12.7 mg/dl in women), systolic arterial pressure (120.1 ± 17.6 mmHg in men and 124.5 ± 18.7 mmHg in women) or diastolic arterial pressure (74.1 ± 11.8 mmHg in men and 75.1 ± 10.9 mmHg in women).
Regarding patients diagnosed with ischemic heart failure, 17 were using statins, of whom three were classified as New York Heart Association grades I and II, and 14 as New York Heart Association grades III and IV, with no statistical difference ($$P \leq 0.3$$) between them. Among the patients with nonischemic heart failure, 36 were using these medications, of whom nine were classified as New York Heart Association grades I and II, and 27 as New York Heart Association grades III and IV, also with no statistical difference ($$P \leq 0.9$$).
Table 1 shows the characteristics of the study group according to presence or absence of the hypertriglyceridemic waist phenotype. Patients with the hypertriglyceridemic waist phenotype had higher prevalence of type 2 diabetes mellitus, dyslipidemia and statin use, higher BMI and ejection fraction and lower HDL-c levels, compared with patients without the hypertriglyceridemic waist phenotype. No statistical difference was observed regarding age, self-reported skin color, educational attainment, smoking, hypertension, New York Heart Association functional classification of heart failure or waist-hip ratio. The prevalence of the hypertriglyceridemic waist phenotype was significantly higher among women than among men ($$P \leq 0.01$$). No patient classified as an alcohol abuser had the hypertriglyceridemic waist phenotype.
Table 1:Participants’ characteristics according to presence or absence of hypertriglyceridemic waist (enlarged waist and elevated triglycerides, EWET) [mean ± standard deviation, SD, or n (%)]*Student’s t test; †Pearson’s chi-square test; ‡Fisher’s exact test. NYHA = New York Heart Association.
Mean HDL-c levels in men and women according to presence or absence of the hypertriglyceridemic waist phenotype are shown in Table 2. In univariate analysis, men with the hypertriglyceridemic waist phenotype had significantly lower ($$P \leq 0.001$$) HDL-c levels than men without the hypertriglyceridemic waist phenotype, but this was not observed among women ($$P \leq 0.2$$). The significant association between the hypertriglyceridemic waist phenotype and HDL-c ($$P \leq 0.04$$) among men was observed even after adjusting for age, BMI, diagnosis of type 2 diabetes mellitus, statin use and heart failure etiology (ischemic/nonischemic) in the multivariate analysis.
Table 2:Mean high density lipoprotein-cholesterol (HDL-c) levels in men and women according to presence or absence of hypertriglyceridemic waist (enlarged waist and elevated triglycerides, EWET) [mean ± standard deviation, ($95\%$ confidence interval)]*Univariate analysis, Student’s t test; †Multivariate analysis, using analysis of covariance (ANCOVA) model: mean adjusted for age, body mass index, medical diagnosis of type 2 diabetes mellitus, statin use (yes/no) and heart failure etiology (ischemic/nonischemic).
## DISCUSSION
To our knowledge, this is the first study to evaluate the presence of the hypertriglyceridemic waist phenotype among individuals with heart failure, and also the association of this phenotype with HDL-c levels. We observed high prevalence of the hypertriglyceridemic waist phenotype in the study group (higher among women than among men), which was associated with HDL-c levels in men after adjusting for age, BMI, diagnosis of type 2 diabetes mellitus, statin use and heart failure etiology. Few studies have investigated the hypertriglyceridemic waist phenotype in Brazil; prevalence of $4.5\%$ was reported among young adults31 and $33\%$ among Brazilian women with hypertension.14 The prevalence of the hypertriglyceridemic waist phenotype varies according to the population studied. Gasevic et al.11 compared the prevalence of the hypertriglyceridemic waist phenotype between Aboriginals, Chinese, Europeans and South Asians, and higher prevalence was found among Chinese people, in both genders. The hypertriglyceridemic waist phenotype was reported in $14.5\%$ of the participants in a study conducted in Spain,12 and in $41.3\%$ of the individuals with diabetes mellitus in a Serbian population.14 The notable differences in prevalence of the hypertriglyceridemic waist phenotype in previous studies may be due to different cutoff points for defining elevated waist circumference in different ethnic groups, and different serum triglyceride values for calculating the hypertriglyceridemic waist phenotype. In the present study, we used the waist circumference and serum triglyceride values proposed in Brazilian guidelines.
Body fat distribution differs between men and women in the general population,32 but in our study the frequency of individuals with elevated waist-hip ratio was higher than that of obesity (defined according to BMI), in both genders. Measurement of abdominal adiposity is useful for assessing the risks associated with obesity and excess visceral fat. Visceral adipose tissue, in turn, is metabolically active and associated with insulin resistance, hypertriglyceridemia, small LDL particles and low HDL-c levels.33 However, an increased waist-hip ratio may also result from loss of muscle and fat mass from the lower limbs, which is usually associated with the aging process and the pathophysiology of heart failure, particularly the more severe forms. In a study by Fülster et al.34 on heart failure patients with a mean age of 66 years, muscle wasting was more pronounced in these individuals than what would be expected for subjects of the same age group. These authors suggested that cachexia relating to chronic heart failure prevails over aging-related loss of lean mass. Therefore, an elevated waist-hip ratio may reflect not only excess abdominal fat accumulation, but also a risk of loss of muscle mass or subcutaneous fat. It is worth mentioning that cardiac cachexia is strongly associated with an inflammatory process.35 Hypertrophied visceral adipocytes increase the release of fatty acids via lipolysis and may also contribute towards activation of adipokines involved in inflammation.36 As previously mentioned, visceral adiposity plays a role in the pathophysiology of type 2 diabetes mellitus and dyslipidemia. The hypertriglyceridemic waist phenotype can be considered to be an indicator of visceral adiposity that includes anthropometric and biochemical components that are highly associated with a worse cardiometabolic profile and higher prevalence of diabetes, dyslipidemia and statin use. In addition, the higher ejection fraction values observed in patients with the hypertriglyceridemic waist phenotype could be a reflection of the obesity paradox in cases of heart failure, i.e. higher adiposity levels would be associated with lower mortality and hospitalization rates.3 In our study, no patients who were identified as alcohol abusers had the hypertriglyceridemic waist phenotype. HDL-c plays a key role in reverse cholesterol transport and attenuates the levels of serum triglycerides. Additionally, ethanol seems to increase HDL apolipoprotein A-I and A-II transport rates by increasing hepatic production.37 Therefore, increased HDL-c levels may have contributed towards maintenance of serum triglyceride levels within the normal range (< 150 mg/dl) in the alcohol abusers of our study group. However, we did not evaluate potential associations between other cardiometabolic factors and alcohol consumption.
We found no significant differences in statin use, heart failure functional class and heart failure etiology between patients with and without the hypertriglyceridemic waist phenotype. According to the American Heart Association,2 statins should not be used as adjunct therapy in cases of heart failure alone, when no other criteria for their use are met (presence of metabolic syndrome and coronary artery disease). Statin therapy in heart failure patients is controversial, because despite its pleiotropic anti-inflammatory effect, the most effective lipoprotein within the context of cardiovascular risk and protection has not yet been identified.38 Higher levels of serum LDL-cholesterol, HDL-c, ApoA-I, ApoB and triglycerides seem to be associated with a better prognosis.39 A significant association between the hypertriglyceridemic waist phenotype and HDL-c levels was found among men but not among women, even after adjusting for some confounding variables. This finding may be explained by several factors: first, the markedly higher visceral fat accumulation in men in comparison with women, which is accompanied by elevated serum triglycerides and reduced HDL-c40 (although not statistically different, the mean BMI among the women in this study was higher than that of the men, thus suggesting greater subcutaneous fat deposition);41 second, the effect of abdominal obesity on proinflammatory states and their atherogenic consequences, including reduction in HDL-c levels;33 and finally, changes in HDL-c levels that are commonly observed in heart failure patients, especially those with ischemic heart failure.22 The inflammatory process involved in the pathophysiology of heart failure per se leads to reduction of HDL-c, which plays a significant anti-inflammatory role in the etiology of the disease. HDL-c inhibits expression of cell adhesion molecules that promote monocyte infiltration through the endothelium, and decreases the inflammatory process that precedes development of heart failure.42 Some of the limitations of our study include the facts that this was an exploratory analysis and that the cross-sectional design of the study might point to reverse causality; the small sample size, which may have conferred higher variability and may have lacked power to detect some associations, especially among women; and the fact that the study was carried out in a public tertiary-level hospital that deals with patients with higher prevalence of more severe forms of heart failure, which may limit the generalization of these results.
## CONCLUSION
The prevalence of the hypertriglyceridemic waist phenotype among our patients with heart failure was high. Reduced HDL-c levels were observed in men with the hypertriglyceridemic waist phenotype, even after adjusting for age, general adiposity, statin use and diagnosis of type 2 diabetes mellitus. Further studies are still needed to identify better anthropometric indicators for altered metabolic profiles and better predictors of the risk of cardiovascular events in heart failure patients. Also, further studies on other populations would enable discussion and comparison of our findings.
## References
1. McMurray JJ, Pfeffer MA. **Heart failure**. *Lancet* (2005) **365** 1877-1889. PMID: 15924986
2. Yancy CW, Jessup M, Bozkurt B. **2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines**. *J Am Coll Cardiol* (2013) **62** e147-e239. PMID: 23747642
3. Sharma A, Lavie CJ, Borer JS. **Meta-analysis of the relation of body mass index to all-cause and cardiovascular mortality and hospitalization in patients with chronic heart failure**. *Am J Cardiol* (2015) **115** 1428-1434. PMID: 25772740
4. Vogel P, Stein A, Marcadenti A. **Visceral adiposity index and prognosis among patients with ischemic heart failure**. *Sao Paulo Med J* (2016) **134** 211-218. PMID: 27191246
5. Lamarche B, Tchernof A, Mauriège P. **Fasting insulin and apolipoprotein B levels and low-density lipoprotein particle size as risk factors for ischemic heart disease**. *JAMA* (1998) **279** 1955-1961. PMID: 9643858
6. Blackburn P, Lemieux I, Lamarche B. **Type 2 diabetes without the atherogenic metabolic triad does not predict angiographically assessed coronary artery disease in women**. *Diabetes Care* (2008) **31** 170-172. PMID: 17934152
7. Kahn HS, Valdez R. **Metabolic risks identified by the combination of enlarged waist and elevated triacylglycerol concentration**. *Am J Clin Nutr* (2003) **78** 928-934. PMID: 14594778
8. Lemiux I, Pascot A, Couillard C. **Hypertriglyceridemic waist: A marker of the atherogenic metabolic triad (hyperinsulinemia; hyperapolipoprotein B; small, dense LDL) in men?**. *Circulation* (2000) **102** 179-184. PMID: 10889128
9. Hobkirk JP, King RF, Gately P. **The predictive ability of triglycerides and waist (hypertriglyceridemic waist) in assessing metabolic triad change in obese children and adolescents**. *Metab Syndr Relat Disord* (2013) **11** 336-342. PMID: 23758076
10. Sam S, Haffner S, Davidson MH. **Hypertriglyceridemic waist phenotype predicts increased visceral fat in subjects with type 2 diabetes**. *Diabetes Care* (2009) **32** 1916-1920. PMID: 19592623
11. Gasevic D, Carlsson AC, Lesser IA, Mancini GJ, Lear SA. **The association between “hypertriglyceridemic waist” and sub-clinical atherosclerosis in a multiethnic population: a cross-sectional study**. *Lipids Health Dis* (2014) **13** 38-38. PMID: 24558974
12. Gomez-Huelgas R, Bernal-López MR, Villalobos A. **Hypertriglyceridemic waist: an alternative to the metabolic syndrome? Results of the IMAP Study (multidisciplinary intervention in primary care).**. *Int J Obes (Lond)* (2011) **35** 292-299. PMID: 20548300
13. Solati M, Ghanbarian A, Rahmani M. **Cardiovascular risk factors in males with hypertriglycemic waist (Tehran Lipid and Glucose Study)**. *Int J Obes Relat Metab Disord* (2004) **28** 706-709. PMID: 14770189
14. Cabral NAL, Ribeiro VS, França AKTC. **Cintura hipertrigliceridêmica e risco cardiometabólico em mulheres hipertensas [Hypertriglyceridemic waist and cardiometabolic risk in hypertensive women]**. *Rev Assoc Med Bras (1992)* (2012) **58** 568-573. PMID: 23090228
15. Radenković SP, Kocić RD, Pešić MM. **The hypertriglyceridemic waist phenotype and metabolic syndrome by differing criteria in type 2 diabetic patients and their relation to lipids and blood glucose control**. *Endokrynol Pol* (2011) **62** 316-323. PMID: 21879471
16. Yang RF, Lin Z, Liu XY, Zhang G. **A clinical study of patients with coronary heart disease complicated with hypertriglyceridemic waist phenotype**. *Cell Biochem Biophys* (2014) **70** 289-293. PMID: 24740454
17. Tankó LB, Bagger YZ, Qin G. **Enlarged waist combined with elevated triglycerides is a strong predictor of accelerated atherogenesis and related cardiovascular mortality in postmenopausal women**. *Circulation* (2005) **111** 1883-1890. PMID: 15837940
18. Arsenault BJ, Barter P, DeMicco DA. **Prediction of cardiovascular events in statin-treated stable coronary patients by lipid and nonlipid biomarkers**. *J Am Coll Cardiol* (2011) **57** 63-69. PMID: 21185503
19. Yunke Z, Guoping L, Zhenyue C. **Triglyceride-to-HDL cholesterol ratio. Predictive value for CHD severity and new-onset heart failure**. *Herz* (2014) **39** 105-110. PMID: 23588603
20. Blackburn P, Lemieux I, Lamarche B. **Angiographically-assessed coronary artery disease associates with HDL particle size in women**. *Atherosclerosis* (2012) **223** 359-364. PMID: 22695528
21. Sakatani T, Shirayama T, Suzaki Y. **The association between cholesterol and mortality in heart failure. Comparison between patients with and without coronary artery disease**. *Int Heart J* (2005) **46** 619-629. PMID: 16157953
22. Karadag MK, Akbulut M. **Low HDL levels as the most common metabolic syndrome risk factor in heart failure**. *Int Heart J* (2009) **50** 571-580. PMID: 19809206
23. Zhe X, Bai Y, Cheng Y. **Hypertriglyceridemic waist is associated with increased carotid atherosclerosis in chronic kidney disease patients**. *Nephron Clin Pract* (2012) **122** 146-152. PMID: 23736857
24. Yancy CW, Jessup M, Bozkurt B. **2013 ACCF/AHA guideline for the management of heart failure: executive summary: a report of the American College of Cardiology Foundation/American Heart Association Task Force on practice guidelines**. *Circulation* (2013) **128** 1810-1852. PMID: 23741057
25. 25
Vigilância alimentar e nutricional
Sisvan: orientações básicas para a coleta, processamento, análise de dados e informação em serviços de saúde
Brasília
Ministério da Saúde
2004. *Sisvan: orientações básicas para a coleta, processamento, análise de dados e informação em serviços de saúde* (2004)
26. 26
World Health Organization. Division of Noncommunicable Diseases. Programme of Nutrition Family and Reproductive Health
Obesity: preventing and managing the global epidemic: report of a WHO consultation on obesity
Geneva
World Health Organization
1998. *Obesity: preventing and managing the global epidemic: report of a WHO consultation on obesity* (1998)
27. Brandão AA, Rodrigues CIS, Consolim-Colombo F. **VI Diretrizes Brasileiras de Hipertensão**. *Arq Bras Cardiol* (2010) **95** 1-51
28. Sposito AC, Caramelli B, Fonseca FAH. **IV Diretriz Brasileira sobre Dislipidemias e Prevenção da Aterosclerose: Departamento de Aterosclerose da Sociedade Brasileira de Cardiologia**. *Arq Bras Cardiol* (2007) **88** 2-19. PMID: 17515982
29. 29
Associação Brasileira para o Estudo da Obesidade e da Síndrome Metabólica
Diretrizes brasileiras de obesidade 2009/2010
3a
Itapevi
AC Farmacêutica
2009
Available from: http://www.abeso.org.br/pdf/diretrizes_brasileiras_obesidade_2009_2010_1.pdf
Accessed in 2016 (Nov 17). *Diretrizes brasileiras de obesidade 2009/2010* (2009)
30. **Diagnosis and classification of diabetes mellitus**. *Diabetes Care* (2014) **37** 81-90. PMID: 23959568
31. Haack RL, Horta BL, Gicante DP. **Cintura hipertrigliceridêmica em adultos jovens no Sul do Brasil [The hypertriglyceridemic waist phenotype in young adults from the Southern Region of Brazil]**. *Cad Saúde Pública* (2013) **29** 999-1007. PMID: 23703005
32. Lemieux S, Prud’homme D, Bouchard C, Tremblay A, Després JP. **Sex differences in the relation of visceral adipose tissue accumulation to total body fatness**. *Am J Clin Nutr* (1993) **58** 463-467. PMID: 8379501
33. Després JP, Lemieux I. **Abdominal obesity and metabolic syndrome**. *Nature* (2006) **444** 881-887. PMID: 17167477
34. Fülster S, Tacke M, Sandek A. **Muscle wasting in patients with chronic heart failure: results from the studies investigating co-morbidities aggravating heart failure (SICA-HF)**. *Eur Heart J* (2013) **34** 512-519. PMID: 23178647
35. Okoshi MP, Romeiro FG, Paiva SAR, Okoshi K. **Caquexia associada à insuficiência cardíaca [Heart failure-induced cachexia]**. *Arq Bras Cardiol* (2013) **100** 476-482. PMID: 23568095
36. Eguchi K, Manabe I. **Toll-like receptor, lipotoxicity and chronic inflammation: the pathological link between obesity and cardiometabolic disease**. *J Atheroscler Thromb* (2014) **21** 629-639. PMID: 24695021
37. De Oliveira E, Silva ER, Foster D, McGee Harper M. **Alcohol consumption raises HDL cholesterol levels by increasing the transport rate of apolipoproteins A-I and A-II**. *Circulation* (2000) **102** 2347-2352. PMID: 11067787
38. Miura S, Saku K. **Effects of statin and lipoprotein metabolism in heart failure**. *J Cardiol* (2010) **55** 287-290. PMID: 20350519
39. Wedel H, McMurray JJ, Lindberg M. **Predictors of fatal and non-fatal outcomes in the Controlled Rosuvastatin Multinational Trial in Heart Failure (CORONA): incremental value of apolipoprotein A-I, high-sensitivity C-reactive peptide and N-terminal pro B-type natriuretic peptide**. *Eur Heart Fail* (2009) **11** 281-291
40. Després JP. **Cardiovascular disease under the influence of excess visceral fat**. *Crit Pathw Cardiol* (2007) **6** 51-59. PMID: 17667865
41. Blaak E. **Gender differences in fat metabolism**. *Curr Opin Clin Nutr Metab Care* (2001) **4** 499-502. PMID: 11706283
42. Barter PJ, Baker PW, Rye KA. **Effect of high-density lipoproteins on the expression of adhesion molecules in endothelial cells**. *Curr Opin Lipidol* (2002) **13** 285-288. PMID: 12045398
|
---
title: Association between asthma and female sex hormones
authors:
- Raquel Prudente de Carvalho Baldaçara
- Ivaldo Silva
journal: São Paulo Medical Journal
year: 2017
pmcid: PMC9969728
doi: 10.1590/1516-3180.2016.011827016
license: CC BY 4.0
---
# Association between asthma and female sex hormones
## ABSTRACT
### CONTEXT AND OBJECTIVE:
The relationship between sex hormones and asthma has been evaluated in several studies. The aim of this review article was to investigate the association between asthma and female sex hormones, under different conditions (premenstrual asthma, use of oral contraceptives, menopause, hormone replacement therapy and pregnancy).
### DESIGN AND SETTING:
Narrative review of the medical literature, Universidade Federal do Tocantins (UFT) and Universidade Federal de São Paulo (Unifesp).
### METHODS:
We searched the CAPES journal portal, a Brazilian platform that provides access to articles in the MEDLINE, PubMed, SciELO, and LILACS databases. The following keywords were used based on Medical Subject Headings: asthma, sex hormones, women and use of oral contraceptives.
### RESULTS:
The associations between sex hormones and asthma remain obscure. In adults, asthma is more common in women than in men. In addition, mortality due to asthma is significantly higher among females. The immune system is influenced by sex hormones: either because progesterone stimulates progesterone-induced blocking factor and Th2 cytokines or because contraceptives derived from progesterone and estrogen stimulate the transcription factor GATA-3.
### CONCLUSIONS:
The associations between asthma and female sex hormones remain obscure. We speculate that estrogen fluctuations are responsible for asthma exacerbations that occur in women. Because of the anti-inflammatory action of estrogen, it decreases TNF-α production, interferon-γ expression and NK cell activity. We suggest that further studies that highlight the underlying physiopathological mechanisms contributing towards these interactions should be conducted.
## CONTEXTO E OBJETIVO:
A relação entre os hormônios sexuais e a asma tem sido investigada em diversos estudos. Esta revisão tem como objetivo descrever a relação entre hormônios sexuais (endógenos e exógenos) e a inflamação nas vias aéreas, especialmente na asma, em eventos diferentes (na asma pré-menstrual, durante o uso de anticoncepcionais, na menopausa, no uso de terapia hormonal e na gestação).
## TIPO DE ESTUDO E LOCAL:
Revisão narrativa da literatura médica, Universidade Federal do Tocantins (UFT) e Universidade Federal de São Paulo (Unifesp).
## MÉTODO:
Pesquisamos o Portal de Periódicos Capes, uma plataforma brasileira que fornece acesso a artigos nas bases de dados MEDLINE, PubMed, SciELO e LILACS. Os descritores utilizados foram asma, hormônios sexuais, mulheres e uso de anticoncepcionais, com base no "Medical Subject Headings".
## RESULTADOS:
As associações entre hormônios sexuais e asma ainda permanecem obscuras. Em adultos, a asma é mais frequente em mulheres do que em homens. Além disso, a mortalidade por asma é significativamente maior no sexo feminino, destacando-se que o sistema imunológico sofre influência de hormônios sexuais, seja porque a progesterona estimula o fator bloqueador induzido pela progesterona e citocinas Th2 ou porque contraceptivos derivados de progesterona e estrógeno estimulam o fator de transcrição GATA-3.
## CONCLUSÕES:
A associação entre asma e hormônios sexuais femininos permanece obscura. Nós especulamos que as flutuações do estrogênio são responsáveis pelas exacerbações da asma que ocorrem nas mulheres. Devido à ação anti-inflamatória do estrogênio há redução da produção de TNF-α, da expressão do interferon-γ e da atividade das células NK. Sugerimos que sejam realizados novos estudos para esclarecer os mecanismos fisiopatológicos dessas interações.
## INTRODUCTION
Asthma is a heterogeneous process that displays considerable phenotypic variability and affects 300 million people globally.1,2 *It is* characterized by the presence of inflammation, hyperresponsiveness and reversible airway obstruction. It is considered to be a public health problem that affects $21\%$ of the Brazilian population.3,4 In Brazil, the mortality rate due to asthma among women is 0.241 per 100,000 inhabitants, whereas among men, it is 0.193 per 100,000 inhabitants.5 Among adults, epidemiological studies have demonstrated that the prevalence of asthma is higher among females than among males.6,7,8 The relationship between sex hormones and asthma has been evaluated in several studies.9,10 Sex-related differences in the risk, incidence and pathogenesis of a variety of lung diseases exist in humans.11 Among children, the prevalence is higher in boys than in girls.12 Interestingly, after puberty, the frequency and severity of asthma increase among girls, such that it becomes more common among women by the age of 20 years.13,14 After the menopause, the difference in asthma prevalence between men and women decreases.14 Thus, in the United States, $65\%$ of all deaths due to asthma occur among women.11 The current paradigm for the pathogenesis of asthma is directly related to gene-environment interaction. Production of Th2 cells (T helper 2) involves the 5q32 region, which is located on the long arm of chromosome 5, in a cluster of genes encoding IL-4 (interleukin 4), IL-5 (interleukin 5), IL-13 (interleukin 13) and IgE (immunoglobulin E) levels.15 The transcription factors that relate to increased Th2 cytokine levels include STAT-5 (signal transducer and activator transducing-5) and GATA-3 (a transcription factor that promotes differentiation of Th2 cells from naïve T lymphocytes). GATA-3 stimulates growth of Th2 cells and inhibits differentiation to Th1 (T helper 1).16,17 T lymphocytes are important effector cells in relation to asthma, and activation of Th2 cells is considered to be important, especially in cases of asthma relating to atopy. However, immune responses to Th1 lymphocyte activation may be responsible for epithelial changes and activation of airway smooth muscle. In addition, as the disease becomes chronic, it may cause activation of Th1 lymphocytes with increased TNF-α expression (tumor necrosis factor) and IFN-γ (interferon gamma). In non-atopic asthma, a neutrophil inflammatory process may occur.18 Tregs (regulatory T cells) reduce proliferation and decrease Th2 levels and hence the inflammatory process in asthma cases.19 Tregs are essential for induction and maintenance of tolerance against antigens.20 In asthmatic patients, Tregs become reduced in number and function.20 Recently, other T helper cells were discovered (Th9 and Th17), and these cells are related to the physiopathological process and worsening asthma.21 The role of IL-17 in asthma is often investigated in patients with non-IgE-mediated non-atopic asthma with a predominance of neutrophils, because Th17 cell levels correlate with disease severity.22
Sex hormones play an important role in respiratory health, and hormone fluctuations may be responsible for exacerbations of asthma in women. Hormone fluctuations occur cyclically in reproductive-age women. For four days after menstruation, follicle-stimulating hormone (FSH), luteinizing hormne (LH) and 17-β-estradiol levels are low. During the follicular phase of the menstrual cycle (days 12-16), progesterone levels remain low, while FSH, LH and 17-β-estradiol levels reach a peak. Finally, during the luteal phase (days 24-28 of the cycle), FSH and LH levels are low, whereas progesterone and 17-β-estradiol levels are moderately high.23 If pregnancy occurs, luteolysis is prevented and the progesterone and 17-β-estradiol levels remain high. After many years, as follicles are depleted and women reaches menopausal status, their sex hormone concentrations decrease to very low levels. In women using oral contraceptives, the progestin component suppresses secretion of LH, and the estrogenic component suppresses secretion of FSH, thus preventing ovulation.12 Asthmatic women need to be monitored for hormonal changes.24 *In a* study conducted by Scichilone that included eight healthy women, the progesterone levels during the menstrual cycle influenced the concentration of nitric oxide in exhaled air (FeNO) and alveolar exhaled nitric oxide (CANO).25 *There is* evidence suggesting that both endogenous and exogenous sex steroids modulate inflammatory processes in the lungs and in smooth muscle tissue during different phases of the hormonal cycle in women.26,27 A relationship between sex hormones and inflammatory responses in the lower airways, especially with regard to asthma, has been observed in several studies.9,10,11,12,13,14However, the mechanism for this interaction remains obscure. Thus, it is very important to review the main findings regarding interactions between sex hormones and to understand the pathophysiological mechanisms of this association.
## OBJECTIVE
To investigate the association between asthma and female sex hormones, at different conditions (premenstrual asthma, use of oral contraceptives, menopause, hormone replacement therapy and pregnancy).
## METHODS
For this narrative review, we searched for articles that addressed association between female sex hormones and asthma regardless of clinical situation, which could encompasse premenstrual period, pregnancy, post-menopause period, use of hormone replacement therapy or oral contraceptives. To do this, we searched the journals in the portal of the Coordination Office for Improvement of High-Education Personnel (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, CAPES). This is a Brazilian platform that provides access to bibliographic sources from various locations around the world, including the following: MEDLINE, PubMed, SciELO, and LILACS. The following keywords were used (based on Medical Subject Headings: https://www.nlm.nih.gov/mesh/): asthma and sex hormones (for the initial search); and women and oral contraceptives (included to refine the analysis). The inclusion criteria were the following: complete articles, published over the last 20 years and written in English or French. The exclusion criteria were the following: items for which the full content was not available, letters to the editor, editorials and articles published in non-scientific journals.
The search was performed in four steps: Keywords search. Preliminary search to include and exclude articles by using their abstracts. Complete articles were read and additional exclusions were made. Synthesis.
## Results from search
In the initial search, we identified 447 references. However, through the preliminary analysis, only 68 references were selected. Only 16 were original articles. The process of study selection is presented in a flow diagram (Figure 1).
Figure 1:Flow diagram showing study selection for review of studies on association between asthma and female sex hormones.
## Menstrual cycle and asthma
There is little data about airway physiology and hormonal fluctuations.28 Exacerbation of asthma in the form of premenstrual asthma (PMA) affects $30\%$ to $40\%$ of women with asthma.29,30 PMA was described for the first time by Frank in 1931, who reported on a woman who experienced severe attacks of asthma that occurred before her menstrual period.31 Some studies have reported a decrease in pulmonary function during the premenstrual portion of the cycle, with a decreased peak expiratory flow rate.24 *There is* also evidence for increased airway inflammation in patients with PMA, as demonstrated by increased levels of eosinophils in sputum and increased levels of fractionated exhaled nitric oxide.32 Tan et al. reported on abnormal regulation of beta2-adrenoreceptors, which was proposed as a possible mechanism for PMA during the period of the cycle when progesterone levels are high.33 The peak incidence of PMA complaints is two to three days before the onset of menstruation, but this phenomenon can also occur during both the menstrual and premenstrual intervals.31 *In a* prospective study on 182 female patients with asthma, $46\%$ of all admissions to emergency departments due to acute asthma occurred during the perimenstrual period.29,34 Murphy reported that use of oral contraceptives was not protective, and further investigation was required to determine the mechanisms involved in PMA.35 A few studies have described treatments for PMA, with conflicting results. Several small series have described use of leukotriene receptor antagonists, exogenous intramuscular progesterone, xanthines,14,24 increased doses of inhaled corticosteroids, addition of long-acting beta2 agonists during the second half of the menstrual cycle, oral contraceptives, a single dose of estradiol (2 mg) during the luteal phase and gonadotropin-releasing-hormone (GnRH) analogues.29 However, more studies are needed in order to determine the appropriate treatment for PMA.
## Use of hormone contraceptives and asthma
Contraceptives have frequently been used over the last 50 years for indications including hirsutism, irregular menstruation, dysmenorrhea, polycystic ovarian disease and contraception. Recently, clinical evidence has suggested that use of contraceptives is associated with impaired lung function.7,36 Some studies have suggested that use of contraceptives is a risk factor for development or exacerbation of asthma crises.7,36 The association between asthma and use of combined contraceptives (estrogen and progesterone) is unclear. The findings in the literature are divergent, given that some studies have reported that estrogen and progesterone improve total lung capacity and reduce the exacerbation of asthma symptoms, such as coughing, wheezing and dyspnea.37,38,39,40 *In a* study by Carlson et al., use of oral contraceptives (combined contraceptives) and unopposed forms of estrogens reduced hormone fluctuations and decreased premenstrual asthma.41 *In a* study by Lange, no relationship was found between use of oral contraceptives and asthma.42 Erkoçoğlu et al.45 found in a survey on 487 women by means of a questionnaire that 196 ($40.2\%$) reported using oral contraceptives. This use was associated with higher risk of current wheezing among adolescents and young adults, but only among those who had taken the oral contraceptives recently during the previous year. In a study by Macsali et al. ,7 women taking oral contraceptives had more asthma and allergies, but this association was not present in lean women, and there was an additional association with body mass index (BMI).
The association between asthma, obesity and sex hormones has been discussed in the medical literature. Obesity has been correlated with higher estrogen levels and with the enzyme aromatase, which in adipose tissue can convert androgens into estrogens.43,44 The Tucson Children's Respiratory Study showed a significant positive association between obesity and wheezing among women who reached puberty when they were under 11 years of age, while obesity was not associated with wheezing among women in whom puberty occurred after they were 11 years old. In the study by Erkoçoğlu et al., there was no evidence of a relationship between BMI and current wheezing.45 *In a* study by Nwaru and Sheikh, hormonal contraceptives reduced exacerbation of asthma and the number of episodes requiring care. That study also showed that overweight and obese women who do not use contraceptives may be at higher risk of asthma.38 *In a* study by Dratva et al., oral contraceptives also appeared to have a protective effect, through decreasing bronchial hyperreactivity.39 The cohort study by Jenkins et al. was the first to report an association between parity, use of oral contraceptives and the onset of asthma among women. In this study, women without asthma or wheezing by the age of seven years showed a lower risk of developing asthma, and the risk decreased by $7\%$ per year of oral contraceptive pill use, independent of parity history. In this group (women without previous asthma or wheezing), the risk of current asthma increased for each birth (odds ratio, OR: 1.50; CI: 1.03-2.23; $$P \leq 0.04$$). Moreover, in the same group, the risk of current asthma was greater among women who were parous, according to the number of births. Women with one birth were at lower risk than nulliparous women. Among women who did have asthma or wheezy breathing by the age of seven years, neither reproductive history nor oral contraceptive pill use predicted current asthma.46 Some authors have suggested mechanisms to explain the complex interaction between hormonal contraceptives and asthma. Velez-Ortega reported on the impact of oral contraceptives on generation of induced regulatory T cells (iTregs).37 Dysregulation of iTregs plays a major role in the pathophysiology of asthma. In this study, patients taking oral contraceptives showed reduced serum sex hormone levels, and this was associated with higher rates of iTreg induction, better asthma control test scores and a tendency towards lower exhaled nitric oxide (eNO) levels.37 On the other hand, Guthikonda et al.47 reported that oral contraceptives and early menarche (via exogenous or endogenous hormones) were associated with the DNA methylation level of the Th2 transcription factor gene and GATA-3 and that they increased the risk of asthma among girls, possibly through interaction with genetic variants. This factor may explain how endogenous and exogenous hormones can, in women, increase the prevalence of asthma after puberty.47 Another mechanism was reported by Tan et al., who proposed that exogenous progesterone but not estradiol induces paradoxical downregulation and desensitization of β2-adrenoceptors in asthmatic women, compared with non-asthmatic subjects.48,49 Moreover, in another study on eleven women with stable mild to moderate asthma, Tan et al. reported that oral contraceptives did not alter β2-adrenoceptor regulation and function in stable female asthmatic patients.33
Finally, Salam et al.26 linked oral contraceptive use and asthma, both of which are common in young women. The outcomes from their study demonstrated that among women without asthma, oral contraceptive use was associated with higher risk of current wheezing. In contrast, in the same study, oral contraceptive use was associated with reduced prevalence of current wheezing among women with asthma. This paradox between hormonal contraceptives and immunologically unclear characteristics of sex hormones emphasizes the need for further research and the importance of knowing a patient's medical history, including the gynecological and hormonal characteristics of asthmatic women.26 In Table 1, we have summarized the differences between the results from different studies on asthma and hormone contraceptives. In Table 2, we have reported the main outcomes from animal model studies on sex hormones and asthma.
Table 1:Animal models for sex hormones and airway inflammationAuthorsMethodResults and conclusionsHellings et al.63 BALB/c male mice of 6 weeks of age were sensitized to ovalbumin (Ova) using intraperitoneal injections. Medroxyprogesterone or placebo was instilled daily into the esophagus before and during the inhalatory Ova challenge phase. Progesterone worsened allergic airway inflammation in Ova-challenged mice. Progesterone increased IL-5 levels and elevated airway eosinophilia. Progesterone did not influence allergen-specific IgE production. Progesterone aggravates the phenotype of eosinophilic airway inflammation in mice by enhancing systemic IL-5 production. Degano et al.64 Ovariectomized seven-week-old female Wistar rats received either placebo or 17β-estradiol (E2) (10 to 100 mcg/kg/day) for 21 days. They were administered an aerosol of saline and increasing concentrations of acetylcholine (Ach) until lung resistance was observed. Rats treated with low-dose E2 were less responsive to Ach than rats given either placebo or high-dose E2 were. Treatment with E2 had a differential, dose-dependent effect on airway responsiveness to Ach.de Oliveira et al.65 The authors evaluated the roles of estradiol and progesterone in allergic lung inflammation. Female Wistar rats were ovariectomized (Ovx) and then sensitized with ovalbumin (OA). They received estradiol and progesterone. In Ovx-allergic rats, treatment with estradiol decreased the amount of IL-10 and increased the amount of IL-4 produced by bone marrow (BM) cells. Estradiol increased IL1β and TNFα levels in BAL (bronchoalveolar lavage) cells. Progesterone increased the release of IL-10, IL-1β and TNFα by BAL cells and increased the production of IL-4 by BM cells. The existence of such dual hormonal effects suggests that hormone therapy in asthmatic postmenopausal women and women who suffer from premenstrual asthma should take into account the possibility that these treatments may worsen pulmonary conditions. Mitchell et al.66 Adult female BALB/c mice were ovariectomized and implanted with time-release progesterone pallets. They were housed in filtered air or ETS (environmental tobacco smoke) for 6 weeks and exposed to HDMA (house dust mite allergen) by inhalation. Progesterone alone did not increase mucous cell mass or abundance of eosinophils, but ETS coupled with progesterone exposure resulted in a significant increase in mucous cell metaplasia and increased accumulation of eosinophils in the asthma model. Progesterone, in the absence of estrogen, exacerbated the airway inflammation and airway remodeling that was induced by the toxicant ETS.Matsubara et al.67 The authors compared sex differences in the development of airway hyperresponsiveness (AHR) following allergen exposure exclusively via the airways. Ovalbumin was administered via nebulization on 10 consecutive days in 8 to 10-week male and female BALB/c mice. After methacholine challenge, significant AHR developed in male mice but not in female mice. Ovariectomized female mice showed significant AHR after 10 days of Ova inhalation. ICI182,780, an estrogen antagonist, similarly enhanced airway responsiveness even when administered 1 hour before the assay. The results showed that 17 beta-estradiol dose-dependently suppressed AHR in male mice. In all cases, airway responsiveness was inhibited by administration of a neurokinin 1 receptor antagonist. The neurokinin 1 receptor antagonist attenuated the effect that the estrogen receptor antagonist had in enhancing AHR in female mice in vivo. Endogenous estrogen may regulate the neurokinin 1-dependent prejunctional activation of airway smooth muscles in allergen-exposed mice.
Table 2:Hormone contraceptives and asthmaAuthors and type of studyMethodResults and conclusionsMacsali et al.7 Cross-sectional surveyPostal questionnaires were sent to subjects in Denmark, Estonia, Iceland, Norway and Sweden from 1999 to 2001 (response rate in women, $77\%$). The analyses included 5791 women who were 25 to 44 years old, of whom 961 ($17\%$) used oral contraceptives. Oral contraceptive pills were associated with an increased risk of asthma, asthma with hay fever, wheezing and shortness of breath, hay fever and ≥ 3 asthma symptoms. Associations were present. Women using oral contraceptive pills had more asthma. This was found only in the normal weight and overweight women and not in lean women, thus indicating an interplay between sex hormones and metabolic status in their effects on airways. Guthikonda et al.47 CohortBlood samples were collected from 245 female participants aged 18 years old. Subjects with genotype AG showed an increase in the average risk ratio (RR) from 0.31 ($95\%$ CI: 0.10 to 0.8) to 11.65 ($95\%$ CI: 1.71 to 79.5) when the methylation level increased from 0.02 to 0.12 relative to the risk in genotype AA. A two-stage model that takes into account genetic variants of the GATA-3 gene, oral contraceptive use, age at menarche and DNA-methylation may explain how sex hormones can increase the prevalence of asthma after puberty. Erkoçoğlu et al.45 Cross-sectionalThe ISAAC questionnaire was provided to 487 women between 11.3 and 25.6 years of age. Questions on oral contraceptives were also asked. In this study, $$n = 487$$ (ages ranged from 11.3 to 25.6 years old), 196 ($40\%$) reported using an oral contraceptive, $7.4\%$ had a diagnosis of asthma from a physician and $10.3\%$ of them were active smokers. Young women taking oral contraceptives had a higher rate of current wheezing, thus suggesting that sex steroids may be important for respiratory health. Dratva et al.39 SPALDIA 2 Cohort571 women aged 28 to 58 years who had menstrual periods without hormone treatment were subjected to methacholine challenge. In a second step, 130 women taking oral contraceptives were subjected to methacholine challenge. An effect of modification according to asthma status and oral contraceptive use was found, with a lower odds ratio (OR) among subjects without asthma. An OR < 1 was found among woman taking oral contraceptives. Oral contraceptives appeared to have a protective effect through which they decreased bronchial hyperreactivity. Vélez-Ortega et al.37 CohortThirteen patients were included in this pilot study. During three distinct phases of their menstrual cycles, the authors measured exhaled nitric oxide (eNO) levels, forced expiratory volume at 1 second (FEV1), asthma control test (ACT) scores, sex steroid hormone levels in serum, natural Tregs levels in peripheral blood, and the ability of CD4+ T cells to generate iTregs ex vivo. Patients taking oral contraceptives showed reduced serum sex hormone levels in association with higher levels of iTreg induction, better ACT scores and a tendency to have lower eNO levels. The impact of sex hormones on the capacity of T cells to polarize towards a regulatory phenotype suggests that regulation of peripheral T cell lineage plasticity is a potential mechanism that may underlie the beneficial effects of oral contraceptives among women with asthma. Tan et al.33 Cohort with intragroup analysisThe study population comprised 11 women aged 19 to 40 years with stable and moderate asthma. The patients were evaluated while on (day 20 to 21) and off (day 5 to 7) oral contraceptives during a 28-day calendar period. Baseline FEV1 did not differ between patients who were on and off oral contraceptives. These did not alter beta2-adrenoreceptor regulation or function in stable female asthmatic patients. Tan et al.48 TrialSeven nonsmoking females aged 26 years with mild asthma completed the study. They were evaluated through two successive menstrual cycles during the follicular phase (days 1 to 6). They were randomized to receive single oral doses of either ethinyl estradiol or medroxyprogesterone. The results showed that exogenous progesterone, but not estrogen, when given during the follicular phase, decreased beta2- adrenoreceptor density and cyclic-adenosine monophosphate (AMP) responses in female asthmatics. The beta2-adrenoreceptor was abnormally regulated in female asthmatics, and this might be a potential mechanism through which premenstrual asthma could be triggered when progesterone levels are high. Salam et al.26 Cohort905 women who had undergone menarche were included. The subjects ranged in age from 13 to 28 years and had participated in the Children's Health Study. In women without asthma, oral contraceptive use was associated with higher risk of current wheezing. In contrast, oral contraceptive use was associated with reduced prevalence of current wheezing in women with asthma. These associations showed significant trends with duration of oral contraceptive use. Age at menarche was associated with new-onset asthma after puberty. Compared with women who had their menarche after they were 12 years old, women who reached their menarche before they were 12 years old were at higher risk of asthma after puberty. Because women have a higher risk of asthma after puberty, and because oral contraceptive use is common among young women, clinicians should inform women with asthma about the potential effects of oral contraceptives on asthma-related respiratory symptoms. Jenkins et al.46 Cohort681 women aged 29-32 years were randomly sampled from participants who were first surveyed at the age of 7 years in the 1968 Tasmanian Asthma Survey, which was a study of all children born in 1961 who attended school. Current asthma was defined as reporting asthma or wheezy breathing during the past 12 months. The risk of current asthma in individuals who were parous increased with the number of births, while women with one birth were at lower risk than nulliparous women. Independent of parity, the risk decreased by $7\%$ per year of oral contraceptive pill use. In women who had asthma or wheezy breathing by the age of 7 years old, neither reproductive history nor oral contraceptive pill use predicted current asthma. Parity and decreased oral contraceptive use predicted asthma in women, and these results are consistent with the hypothesis that the asthma that develops after childhood is in part a response to endogenous and exogenous female hormones. Nwaru and Sheikh38 Cross-sectional surveyA population-based analysis using serial data from the *Scottish* general population. A total of 3257 non- pregnant, 16-45-year-old women were included. The use of any hormonal contraceptive was associated with a reduced risk of current physician-diagnosed asthma. The use of a hormonal contraceptive may reduce asthma exacerbations. Overweight and obese non-contraceptive-using women may be at increasing risk of asthma. Lange et al.42 Cross-sectionalData from a study on women who were selected from the general population were used to correlate the effect of treatment with oral contraceptives and hormonal replacement therapy (HRT) with asthma indications. 377 women were on oral contraceptives ($24.5\%$ of the premenopausal women) and 458 were on HRT ($15.2\%$ of the postmenopausal women). The age span of the premenopausal women was 21-49 years and of the postmenopausal women, 27-90 years. A weak association was observed between HRT and self-reported asthma. No relationship was found between the use of oral contraceptives and asthma, although an association was observed between asthma and HRT.
## Postmenopausal hormone replacement therapy (HRT) and asthma
Among women over 50 years of age, the menopause can either coincide with the onset of asthma or be associated with deterioration of a pre-existing asthma condition.50 The definition of menopause is the cessation of menstruation for 12 months.51 The overall incidence of asthma decreases after the menopause,14 although in the Nurses' Health Study, use of hormone replacement therapy (HRT) approximately doubled the risk of asthma, compared with postmenopausal women without HRT. In that study, a $35\%$ decrease in the incidence of asthma was observed among postmenopausal women without HRT.10 *In a* cohort study, Romieu et al. reported that the increase in the risk of asthma onset at the time of the menopause was only significant among women who reported using estrogen alone, especially among those who had never been smokers and those who had had an allergic disease before the onset of asthma. A small increase in the risk of asthma among women who used estrogen/progestogen was found in these subgroups.52 *In a* systematic review and meta-analysis, Zemp et al. found that there was no significant association between menopause with asthma prevalence or incidence except for women who reported using HRT.53 *In a* study by Carlson et al., HRT was associated with better lung function and an increase in forced expiratory volume at one second (FEV1).41 The mechanisms that link asthma and the menopause are unclear. After the menopause, FSH and LH levels are elevated, and estrogen levels decrease to the levels observed in patients with surgical oophorectomy, who also show extremely low progesterone levels. The incidence of asthma may be associated with decreased estrogen levels and a protective effect against the relative androgen excess that occurs during the menopausal transition.53,54 Clinical studies have indicated that the menopause is associated with exacerbation of pre-existing asthma. Thus, the onset of asthma is characterized by absence of atopy, absence of a family history and associations with urticaria and/or recurrent sinusitis of high severity.23 Balzano et al.55 showed that eosinophil levels were higher in the induced sputum of menopausal asthmatics, but Foschino Barbaro et al. reported that there were high sputum levels of neutrophils and exhaled interleukin (IL)-6 in women with menopausal asthma.50 Few studies have explored the link between the menopause and asthma. Hormonal processes and other factors, including genetics and inflammatory and metabolic characteristics, need to be taken into consideration. Studies have indicated that obesity has an effect on the severity of asthma and that this relationship is modified by gender. Estrogen and leptin levels (which have been correlated with increased airway inflammation in animal models)56 are higher in obese women than in non-obese women.54 Moreover, obesity has been shown to increase the risk of developing asthma. Interestingly, Gómez Real reported that lean women presented a higher risk of postmenopausal asthma than did obese women using HRT.57 This phenomenon can be explained by the notion that in lean women without insulin resistance, the pro-inflammatory effect of estrogens may predominate; while in obese women, the pro-inflammatory effects of estrogens are decreased through insulin resistance.53
## Pregnancy and asthma
Asthma affects $3.7\%$ to $8.4\%$ of all pregnant women in the United States. Maternal asthma is associated with an increased risk of both maternal and fetal adverse perinatal outcomes,58 such that $20\%$-$30\%$ of women with asthma experience exacerbations that require medical intervention during pregnancy.43 *There is* also evidence of an increased risk of maternal mortality among some asthmatic women.59 A number of the physiological changes that occur during pregnancy can affect asthma status, including mechanical, immunological and hormonal alterations. Estradiol and progesterone levels are highest during pregnancy.60 Moreover, one third of women experience improved asthma, while another third of women retain the same asthma status and the remaining third experience worse asthma. Pregnancy is also marked by a state of Th2 dominance, and asthma is generally characterized by Th2 inflammation. Progesterone receptors are present in large quantities on the surface of lymphocytes, and binding of progesterone to its receptor induces stimulation and release of progesterone-induced blocking factor (PIBF) in a Th2 cytokine expression pattern (IL-4, IL-5, IL-6, IL-9, IL-10 and IL-13). The effects of these proteins are reduced in natural killer (NK) cells, in which expression of IFN-γ is decreased. NK cells are mainly observed in the endometrium of pregnant women.12,61 During the first trimester of pregnancy, the numbers of circulating and decidual regulatory T cells (Tregs) increase to promote tolerance at the maternal-fetal interface.62 Interestingly, fetal sex may influence asthma. Kwon et al. examined pregnant asthmatic women and found that carrying a female fetus was associated with worse maternal asthma than carrying a male fetus was.60 The mechanism that contributes towards this result is unclear, but there is evidence showing that testosterone potentiates the β-adrenergic-mediated relaxation of bronchial tissues and inhibits responses to histamine. Female sex is associated with higher maternal circulation of monocytes and upregulation of maternal inflammatory pathways.58 The mechanisms through which sex hormones influence asthma and the immunological characteristics of pregnancy at the maternal-fetal interface remain obscure, and new studies are needed in order to increase our understanding of and ability to manage asthmatic women.
## DISCUSSION
Studies examining the role of hormonal factors in asthma among women have been conducted on human subjects and animal models, and the results have been described in reviews. In an attempt to understand the influence of sex hormones on pulmonary inflammatory responses, we discuss the main immunological aspects of sex hormones here.
Studies using animal models have demonstrated that both progesterone and estrogen can directly affect the lungs.63,64,65,66,67 Sex steroid hormones influence the immune system by acting on the structure and function of the thymus, thereby modulating the activity of B and T cells, mast cells and natural killer cells (NK cells), and affecting phagocytic cells and cytokine production. These hormones act via a variety of receptors (including the estrogen receptors ERα and ERβ; and the progesterone receptors PR-A and PR-B), and these steroid receptors have been described as nuclear receptors that act as transcription factors to regulate gene expression.23 However, it has been shown that some steroid receptors are located at the plasma membrane (e.g. membrane-bound G-protein-coupled receptors).68,69 These receptors are also expressed in the human lungs, such that sex hormones play a role in development of the lungs and androgen receptors are expressed in the mesenchymal and epithelial cells of the lungs.
Gender differences have been observed in relation to development of the lungs. For example, production of surfactants appears earlier in female than in male neonatal lungs, and male preterm infants are at higher risk of experiencing developmental distress syndrome. In addition, before puberty, the prevalence of asthma is higher among boys.43 Both male and female fetuses express androgen receptors (AR-A, AR-B) in non-reproductive tissues, with significantly higher numbers of AR-B than AR-A receptors expressed in the lungs. However, few studies have examined expression of androgens in inflammatory airways, and testosterone has been shown to cause relaxation of airway smooth muscles.70 Testosterone may increase apoptosis in T cells, thus resulting in a lower percentage of T lymphocytes in the total pool of lymphocytes in males than in females.12 In allergic asthma, airway inflammation is mainly characterized by Th2-mediated processes, including secretion of the cytokines IL-4, IL-5, IL-6, IL-9 and IL-13, secretion of chemokines, regulation of the activation of normal T cells (RANTES), and production of granulocyte macrophage colony-stimulating factor (GM-CSF). In patients with asthma and in allergic animal models (e.g. allergen-challenged mice), bronchoalveolar lavage contains large numbers of eosinophils, M2-polarized macrophages and activated mast cells. In several cases, the numbers of neutrophils in the bronchoalveolar lavage have been found to be higher as a result of Th17-mediated responses and production of IL-8.68,69 The airway epithelium in asthmatic patients recruits innate and adaptive cells via cytokines, including IL-25 and IL-33, and chemokines such as CCL2, CCL17 and CCL20, and it secretes transforming growth factor beta (TGFβ), which is responsible for airway remodelling.69 The transition of monocytes along the monocyte-macrophage axis is accompanied by upregulation of the 46 kDa ERα.35 Activated monocytes and macrophages show increased tumor necrosis factor-alpha (TNFα) secretion. TNFα is a cytokine produced by Th1 cells and is an important mediator in pro-inflammatory responses. Female reproductive phases also influence the production of TNFα by monocytes. In the luteal phase, higher plasma levels of TNFα have been observed.12 However, 17β estradiol may decrease TNFα levels via an anti-inflammatory effect caused by estrogen.71 Few studies have examined the effects of sex hormones on the bronchial epithelium. The human bronchial epithelium expresses both ERα and ERβ. In patients with asthma, estrogens facilitate dissociation of endothelial nitric oxide synthetase, which results in activation of the NO pathway, vasodilatation and increased inflammation.72 In another study, treatment of bronchial epithelial cells with 10 nM estrogen induced expression of NOS and production of nitric oxide, thus resulting in bronchodilation.69,73 *In a* study by Mandhane et al., among women who were not using oral contraceptives, an increase in progesterone level was associated with an increase in exhaled nitric oxide levels, thus indicating that an inflammatory process was associated with progesterone.74
Stimulation of Th2-mediated inflammatory responses and asthma by progesterone has been considered by many studies to represent a typical Th2 disorder.69,75 *In a* study by Loza et al., increased accumulation of IL-13+T cells (Th2) was observed in female but not in male asthmatics, and this association was maintained when the analysis was restricted to atopic subjects.75 In an animal model, ovariectomized or estradiol antagonist-treated mice developed reduced IL-5 dependent eosinophilia during allergic inflammation.76 However, depending on the concentration of estrogen, it may play dual pro and anti-inflammatory roles.64,77
## CONCLUSIONS
We have attempted to discuss the characteristics that are affected by sexual hormones during pulmonary inflammatory responses. However, the associations between these factors remain obscure. We speculate that estrogen fluctuations are responsible for asthma exacerbations that occur in women. Because of the anti-inflammatory action of estrogen, as this hormone decreases TNF-α production, it reduces IFN-γ expression, and NK cell activity. We suggest that further studies that highlight the underlying physiopathological mechanisms contributing towards these interactions should be conducted.
## References
1. 1
Global Strategy for Asthma Management and Prevention (GINA)
Available from: http://ginasthma.org/wp-content/uploads/2016/04/GINA-2016-main-report_tracked.pdf
Accessed in 2016 (Aug 30). *Global Strategy for Asthma Management and Prevention (GINA)*
2. 2
National Institute of Allergy and Infectious Diseases
Asthma
Available from: http://www.niaid.nih.gov/topics/asthma/Pages/default.aspx
Accessed in 2016 (Aug 30). *Asthma*
3. **IV Diretrizes Brasileiras para o Manejo da Asma [IV Brazilian Guidelines for the management of asthma]**. *J Bras Pneumol* (2006.0) **32** s447-s474. PMID: 17420905
4. Solé D, Yamada E, Vana AT. **International Study of Asthma and Allergies in Childhood (ISAAC): prevalence of asthma and asthma-related symptoms among Brazilian schoolchildren**. *J Investig Allergol Clin Immunol* (2001.0) **11** 123-128
5. Lotufo PA, Bensenor IM. **Temporal trends of asthma mortality rates in Brazil from 1980 to 2010**. *J Asthma* (2012.0) **49** 779-784. PMID: 22953750
6. Siroux V, Oryszczyn MP, Paty E. **Relationships of allergic sensitization, total immunoglobulin E and blood eosinophils to asthma severity in children of the EGEA Study**. *Clin Exp Allergy* (2003.0) **33** 746-751. PMID: 12801307
7. Macsali F, Real FG, Omenaas ER. **Oral contraception, body mass index, and asthma a cross-sectional Nordic-Baltic population survey**. *J Allergy Clin Immunol* (2009.0) **123** 391-397. PMID: 19121863
8. Melgert BN, Ray A, Hylkema MN, Timens W, Postma DS. **Are there reasons why adult asthma is more common in females?**. *Curr Allergy Asthma Rep* (2007.0) **7** 143-150. PMID: 17437685
9. Barr RG, Wentowski CC, Grodstein F. **Prospective study of postmenopausal hormone use and newly diagnosed asthma and chronic obstructive pulmonary disease**. *Arch Intern Med* (2004.0) **164** 379-386. PMID: 14980988
10. Troisi RJ, Speizer FE, Willett WC, Trichopoulos D, Rosner B. **Menopause, postmenopausal estrogen preparations, and the risk of adult-onset asthma. A prospective cohort study**. *Am J Respir Crit Care Med* (1995.0) **152** 1183-1188. PMID: 7551368
11. Matteis M, Polverino F, Spaziano G. **Effects of sex hormones on bronchial reactivity during the menstrual cycle**. *BMC Pulm Med* (2014.0) **14** 108-108. PMID: 24984749
12. Bouman A, Heineman MJ, Faas MM. **Sex hormones and the immune response in humans**. *Hum Reprod Update* (2005.0) **11** 411-423. PMID: 15817524
13. Varraso R, Siroux V, Maccario J. **Asthma severity is associated with body mass index and early menarche in women**. *Am J Respir Crit Care Med* (2005.0) **171** 334-339. PMID: 15557134
14. Kynyk JA, Mastronarde JG, McCallister JW. **Asthma, the sex difference**. *Curr Opin Pulm Med* (2011.0) **17** 6-11. PMID: 21045697
15. Wills-Karp M, Ewart SL. **Time to draw breath asthma-susceptibility genes are identified**. *Nat Rev Genet* (2004.0) **5** 376-387. PMID: 15143320
16. Zhu J, Yamane H, Cote-Sierra J, Guo L, Paul WE. **GATA-3 promotes Th2 responses through three different mechanisms: induction of Th2 cytokine production, selective growth of Th2 cells and inhibition of Th1 cell-specific factors**. *Cell Res* (2006.0) **16** 3-10. PMID: 16467870
17. Höfer T, Nathansen H, Löhning M, Radbruch A, Heinrich R. **GATA-3 transcriptional imprinting in Th2 lymphocytes: a mathematical model**. *Proc Natl Acad Sci U S A* (2002.0) **99** 9364-9368. PMID: 12087127
18. Magnan AO, Mély LG, Camilla CA. **Assessment of the Th1/Th2 paradigm in whole blood in atopy and asthma. Increased IFN-gamma-producing CD8(+) T cells in asthma**. *Am J Respir Crit Care Med* (2000.0) **161** 1790-1796. PMID: 10852746
19. Lowder TW, Kunz HE. **Regulatory T Cells in Asthma and Airway Hyperresponsiveness**. *Journal of Allergy & Therapy* (2011.0) 002-002
20. Langier S, Sade K, Kivity S. **Regulatory T cells in allergic asthma**. *Isr Med Assoc J* (2012.0) **14** 180-183. PMID: 22675860
21. Vock C, Hauber HP, Wegmann M. **The other T helper cells in asthma pathogenesis**. *J Allergy (Cairo)* (2010.0) **2010** 519298-519298. PMID: 20976014
22. Cosmi L, Liotta F, Maggi E, Romagnani S, Annunziato F. **Th17 cells: new players in asthma pathogenesis**. *Allergy* (2011.0) **66** 989-998. PMID: 21375540
23. Balzano G, Fuschillo S, Melillo G, Bonini S. **Asthma and sex hormones**. *Allergy* (2001.0) **56** 13-20. PMID: 11167347
24. Karpel JP, Wait JL. **Asthma in women. Part 3: Perimenstrual asthma, effects of hormone therapy**. *Journal of Critical Illness* (2000.0) **15** 265-272
25. Scichilone N, Battaglia S, Braido F. **Exhaled nitric oxide is associated with cyclic changes in sexual hormones**. *Pulm Pharmacol Ther* (2013.0) **26** 644-648. PMID: 23665049
26. Salam MT, Wenten M, Gilliland FD. **Endogenous and exogenous sex steroid hormones and asthma and wheeze in young women**. *J Allergy Clin Immunol* (2006.0) **117** 1001-1007. PMID: 16675325
27. Haggerty CL, Ness RB, Kelsey S, Waterer GW. **The impact of estrogen and progesterone on asthma**. *Ann Allergy Asthma Immunol* (2003.0) **90** 284-291. PMID: 12669890
28. Macsali F, Svanes C, Sothern RB. **Menstrual cycle and respiratory symptoms in a general Nordic-Baltic population**. *Am J Respir Crit Care Med* (2013.0) **187** 366-373. PMID: 23204251
29. Tan KS. **Premenstrual asthma epidemiology, pathogenesis and treatment**. *Drugs* (2001.0) **61** 2079-2086. PMID: 11735634
30. Pereira-Vega A, Sánchez Ramos JL, Vázquez Oliva R. **Premenstrual asthma and female sex hormones**. *J Investig Allergol Clin Immunol* (2012.0) **22** 437-439
31. Vrieze A, Postma DS, Kerstjens HA. **Perimenstrual asthma a syndrome without known cause or cure**. *J Allergy Clin Immunol* (2003.0) **112** 271-282. PMID: 12897732
32. Redmond AM, James AW, Nolan SH, Self TH. **Premenstrual asthma: emphasis on drug therapy options**. *J Asthma* (2004.0) **41** 687-693. PMID: 15584626
33. Tan KS, McFarlane LC, Lipworth BJ. **Beta2-adrenoceptor regulation and function in female asthmatic patients receiving the oral combined contraceptive pill**. *Chest* (1998.0) **113** 278-282. PMID: 9498939
34. Skobeloff EM, Spivey WH, Silverman R. **The effect of the menstrual cycle on asthma presentations in the emergency department**. *Arch Intern Med* (1996.0) **156** 1837-1840. PMID: 8790078
35. Murphy VE, Gibson PG. **Premenstrual asthma: prevalence, cycle-to-cycle variability and relationship to oral contraceptive use and menstrual symptoms**. *J Asthma* (2008.0) **45** 696-704. PMID: 18951263
36. Real FG, Svanes C, Macsali F, Omenaas ER. **Hormonal factors and respiratory health in women--a review**. *Clin Respir J* (2008.0) **2** 111-119. PMID: 20298359
37. Vélez-Ortega AC, Temprano J, Reneer MC. **Enhanced generation of suppressor T cells in patients with asthma taking oral contraceptives**. *J Asthma* (2013.0) **50** 223-230. PMID: 23259774
38. Nwaru BI, Sheikh A. **Hormonal contraceptives and asthma in women of reproductive age: analysis of data from serial national Scottish Health Surveys**. *J R Soc Med* (2015.0) **108** 358-371. PMID: 26152676
39. Dratva J, Schindler C, Curjuric I. **Perimenstrual increase in bronchial hyperreactivity in premenopausal women: results from the population-based SAPALDIA 2 cohort**. *J Allergy Clin Immunol* (2010.0) **125** 823-829. PMID: 20227756
40. Caracta CF. **Gender differences in pulmonary disease**. *Mt Sinai J Med* (2003.0) **70** 215-224. PMID: 12968194
41. Carlson CL, Cushman M, Enright PL. **Hormone replacement therapy is associated with higher FEV1 in elderly women**. *Am J Respir Crit Care Med* (2001.0) **163** 423-428. PMID: 11179117
42. Lange P, Parner J, Prescott E, Ulrik CS, Vestbo J. **Exogenous female sex steroid hormones and risk of asthma and asthma-like symptoms: a cross sectional study of the general population**. *Thorax* (2001.0) **56** 613-616. PMID: 11462063
43. Carey MA, Card JW, Voltz JW. **It's all about sex gender, lung development and lung disease**. *Trends Endocrinol Metab* (2007.0) **18** 308-313. PMID: 17764971
44. Weiss ST. **Obesity: insight into the origins of asthma**. *Nat Immunol* (2005.0) **6** 537-539. PMID: 15908930
45. Erkoçoglu M, Kaya A, Azkur D. **The effect of oral contraceptives on current wheezing in young women**. *Allergol Immunopathol (Madr)* (2013.0) **41** 169-175. PMID: 22959299
46. Jenkins MA, Dharmage SC, Flander LB. **Parity and decreased use of oral contraceptives as predictors of asthma in young women**. *Clin Exp Allergy* (2006.0) **36** 609-613. PMID: 16650045
47. Guthikonda K, Zhang H, Nolan VG. **Oral contraceptives modify the effect of GATA3 polymorphisms on the risk of asthma at the age of 18 years via DNA methylation**. *Clin Epigenetics* (2014.0) **6** 17-17. PMID: 25250096
48. Tan KS, McFarlane LC, Lipworth BJ. **Paradoxical down-regulation and desensitization of beta2-adrenoceptors by exogenous progesterone in female asthmatics**. *Chest* (1997.0) **111** 847-851. PMID: 9106558
49. Tan KS, McFarlane LC, Lipworth BJ. **Loss of normal cyclical beta 2 adrenoceptor regulation and increased premenstrual responsiveness to adenosine monophosphate in stable female asthmatic patients**. *Thorax* (1997.0) **52** 608-611. PMID: 9246131
50. Foschino Barbaro MP, Costa VR, Resta O. **Menopausal asthma: a new biological phenotype?**. *Allergy* (2010.0) **65** 1306-1312. PMID: 20557302
51. Macsali F, Svanes C, Bjørge L, Omenaas ER, Gómez Real F. **Respiratory health in women: from menarche to menopause**. *Expert Rev Respir Med* (2012.0) **6** 187-200. PMID: 22455491
52. Romieu I, Fabre A, Fournier A. **Postmenopausal hormone therapy and asthma onset in the E3N cohort**. *Thorax* (2010.0) **65** 292-297. PMID: 20142267
53. Zemp E, Schikowski T, Dratva J, Schindler C, Probst-Hensch N. **Asthma and the menopause a systematic review and meta-analysis**. *Maturitas* (2012.0) **73** 212-217. PMID: 22964072
54. Zein JG, Erzurum SC. **Asthma is Different in Women**. *Curr Allergy Asthma Rep* (2015.0) **15** 28-28. PMID: 26141573
55. Balzano G, Fuschillo S, De Angelis E. **Persistent airway inflammation and high exacerbation rate in asthma that starts at menopause**. *Monaldi Arch Chest Dis* (2007.0) **67** 135-141. PMID: 18018752
56. Shore SA, Fredberg JJ. **Obesity, smooth muscle, and airway hyperresponsiveness**. *J Allergy Clin Immunol* (2005.0) **115** 925-927. PMID: 15867846
57. Gómez Real F, Svanes C, Björnsson EH. **Hormone replacement therapy, body mass index and asthma in perimenopausal women a cross sectional survey**. *Thorax* (2006.0) **61** 34-40. PMID: 16244093
58. Bakhireva LN, Schatz M, Jones KL. **Fetal sex and maternal asthma control in pregnancy**. *J Asthma* (2008.0) **45** 403-407. PMID: 18569234
59. Tan KS, Thomson NC. **Asthma in pregnancy**. *Am J Med* (2000.0) **109** 727-733. PMID: 11137489
60. Kwon HL, Belanger K, Holford TR, Bracken MB. **Effect of fetal sex on airway lability in pregnant women with asthma**. *Am J Epidemiol* (2006.0) **163** 217-221. PMID: 16357112
61. Broide DH. **Molecular and cellular mechanisms of allergic disease**. *J Allergy Clin Immunol* (2001.0) **108** S65-S71. PMID: 11498675
62. Oertelt-Prigione S. **Immunology and the menstrual cycle**. *Autoimmun Rev* (2012.0) **11** A486-A492. PMID: 22155200
63. Hellings PW, Vandekerckhove P, Claeys R. **Progesterone increases airway eosinophilia and hyper-responsiveness in a murine model of allergic asthma**. *Clin Exp Allergy* (2003.0) **33** 1457-1463. PMID: 14519155
64. Degano B, Mourlanette P, Valmary S. **Differential effects of low and high-dose estradiol on airway reactivity in ovariectomized rats**. *Respir Physiol Neurobiol* (2003.0) **138** 265-274. PMID: 14609515
65. de Oliveira AP, Domingos HV, Cavriani G. **Cellular recruitment and cytokine generation in a rat model of allergic lung inflammation are differentially modulated by progesterone and estradiol**. *Am J Physiol Cell Physiol* (2007.0) **293** C1120-C1128. PMID: 17634417
66. Mitchell VL, Van Winkle LS, Gershwin LJ. **Environmental tobacco smoke and progesterone alter lung inflammation and mucous metaplasia in a mouse model of allergic airway disease**. *Clin Rev Allergy Immunol* (2012.0) **43** 57-68. PMID: 21837398
67. Matsubara S, Swasey CH, Loader JE. **Estrogen determines sex differences in airway responsiveness after allergen exposure**. *Am J Respir Cell Mol Biol* (2008.0) **38** 501-508. PMID: 18063836
68. Zierau O, Zenclussen AC, Jensen F. **Role of female sex hormones, estradiol and progesterone, in mast cell behavior**. *Front Immunol* (2012.0) **3** 169-169. PMID: 22723800
69. Keselman A, Heller N. **Estrogen Signaling Modulates Allergic Inflammation and Contributes to Sex Differences in Asthma**. *Front Immunol* (2015.0) **6** 568-568. PMID: 26635789
70. Chang HY, Mitzner W. **Sex differences in mouse models of asthma**. *Can J Physiol Pharmacol* (2007.0) **85** 1226-1235. PMID: 18066124
71. Ito A, Bebo BF, Matejuk A. **Estrogen treatment down-regulates TNF-alpha production and reduces the severity of experimental autoimmune encephalomyelitis in cytokine knockout mice**. *J Immunol* (2001.0) **167** 542-552. PMID: 11418693
72. Sathish V, Martin YN, Prakash YS. **Sex steroid signaling: implications for lung diseases**. *Pharmacol Ther* (2015.0) **150** 94-108. PMID: 25595323
73. Townsend EA, Meuchel LW, Thompson MA, Pabelick CM, Prakash YS. **Estrogen increases nitric-oxide production in human bronchial epithelium**. *J Pharmacol Exp Ther* (2011.0) **339** 815-824. PMID: 21940647
74. Mandhane PJ, Hanna SE, Inman MD. **Changes in exhaled nitric oxide related to estrogen and progesterone during the menstrual cycle**. *Chest* (2009.0) **136** 1301-1307. PMID: 19617403
75. Loza MJ, Foster S, Bleecker ER, Peters SP, Penn RB. **Asthma and gender impact accumulation of T cell subtypes**. *Respir Res* (2010.0) **11** 103-103. PMID: 20667106
76. Riffo-Vasquez Y, Ligeiro de Oliveira AP, Page CP, Spina D, Tavares-de-Lima W. **Role of sex hormones in allergic inflammation in mice**. *Clin Exp Allergy* (2007.0) **37** 459-470. PMID: 17359396
77. Straub RH. **The complex role of estrogens in inflammation**. *Endocr Rev* (2007.0) **28** 521-574. PMID: 17640948
|
---
title: 'Fatty Acid-Binding Protein 5 Gene Deletion Enhances Nicotine-Conditioned Place
Preference: Illuminating the Putative Gateway Mechanisms'
authors:
- Nicole Roeder
- Brittany Richardson
- Abrianna Mihalkovic
- Samantha Penman
- Olivia White
- John Hamilton
- Ashim Gupta
- Kenneth Blum
- Mark S. Gold
- Panayotis K. Thanos
journal: Future pharmacology
year: 2023
pmcid: PMC9969817
doi: 10.3390/futurepharmacol3010007
license: CC BY 4.0
---
# Fatty Acid-Binding Protein 5 Gene Deletion Enhances Nicotine-Conditioned Place Preference: Illuminating the Putative Gateway Mechanisms
## Abstract
Emerging evidence indicates that the endogenous cannabinoid system modulates the behavioral and physiological effects of nicotine. Fatty acid-binding proteins (FABPs) are among the primary intracellular trafficking mechanisms of endogenous cannabinoids, such as anandamide. To this end, changes in FABP expression may similarly impact the behavioral manifestations associated with nicotine, particularly its addictive properties. FABP5+/+ and FABP5−/− mice were tested for nicotine-conditioned place preference (CPP) at two different doses (0.1 or 0.5 mg/kg). The nicotine-paired chamber was assigned as their least preferred chamber during preconditioning. Following 8 days of conditioning, the mice were injected with either nicotine or saline. The mice were allowed to access to all the chambers on the test day, and their times spent in the drug chamber on the preconditioning versus the test days were used to examine the drug preference score. The CPP results showed that the FABP5−/− mice displayed a higher place preference for 0.1 mg/kg nicotine than the FABP5+/+ mice, while no CPP difference was observed for 0.5 mg/kg nicotine between the genotypes. In conclusion, FABP5 plays an important role in regulating nicotine place preference. Further research is warranted to identify the precise mechanisms. The results suggest that dysregulated cannabinoid signaling may impact nicotine-seeking behavior.
## Introduction
Cigarette smoking remains a leading preventable cause of death in the United States [1]. Although most of the toxicity of cigarette smoking is related to other components of combustible cigarettes, nicotine is the primary addictive component [2]. When inhaled, nicotine readily infuses into the brain and binds to the nicotinic acetylcholine receptors, releasing neurotransmitters, including dopamine (DA), into the mesolimbic regions of the brain [2]. Specifically, nicotine stimulates the activation of DAergic neurons within the ventral tegmental area (VTA) and the nucleus accumbens (NAc) shell, which are critical regions for the rewarding properties of nicotine [2,3]. Additionally, DAergic signaling is an important player in the transmission of reward-related information [4] and is believed to be the modulator of behaviors associated with nicotine use [5].
The DAergic signaling within the mesolimbic reward pathway is modulated by the endocannabinoid (eCB) system [6,7]. This system is comprised of cannabinoid type-1 and type-2 G-protein-coupled receptors (CB1R and CB2R, respectively) and two major endogenous ligands, anandamide (AEA) and 2-arachidonoyl-glycerol (2-AG), as well as the enzymes involved in their synthesis and metabolism, fatty acid amide hydrolase (FAAH) and monoacylglycerol lipase (MAGL) [8]. The dysfunction of eCB signaling has been implicated in the pathophysiology of psychiatric disorders, such as schizophrenia, substance abuse, depression, and anxiety disorders [9–12]. Cannabinoid receptor signaling is tied to the reinforcing properties associated with nicotine. Rimonabant, a CB1R inverse agonist, has been shown to decrease nicotine self-administration and conditioned place preference in the translational literature [13,14]. This CB1R inverse agonist has also been shown to prolong abstinence rates among smokers who are motivated to quit smoking [15]. The blocking of CB1R activation has also been shown to attenuate nicotine-induced DA increases in the NAc [16–18]. Furthermore, reduced nicotine-seeking behavior in animal models of relapse has been observed with the application of rimonabant [19]. These findings indicate that CB1Rs and eCBs are important for nicotine and the motivation to seek the drug. Kandel and Kandel have published several papers on the potential gateway theories of certain drugs, such as nicotine, acting as a gateway to other drugs of abuse [20].
In addition, other research has examined eCB signaling, such as FAAH, which is involved in the synthesis and metabolism of eCBs. Specifically, FAAH is known to rapidly metabolize AEA, which has been shown to have a high affinity with CB1Rs on the presynaptic neurons that activate the mesolimbic DA system [21]. AEA is synthesized on demand, and upon release, it is quickly degraded by FAAH into arachidonic acid and ethanolamine [21–23]. FAAH−/− mice, which have a 10- to 15-fold increase in the brain AEA levels, show enhanced CPP in response to a low dose of nicotine (0.1 mg/kg), whereas their FAAH+/+ counterparts had no nicotine acquisition [24]. These effects were reversed following the blockade of CB1Rs, indicating that these actions are CB1R-mediated. At higher doses, such as 0.5 and 1 mg/kg nicotine, both the FAAH+/+ and FAAH−/− mice displayed no difference in nicotine CPP. A more recent microdialysis investigation of a dose of 0.1 mg/kg nicotine observed elevated DA concentrations within the NAc of FAAH−/− mice [25], thus suggesting that elevated eCB levels enhance the rewarding properties associated with nicotine.
More recently, fatty acid-binding proteins (FABPs) have been described as intracellular chaperone proteins that facilitate the uptake and transport of AEA to FAAH for degradation [26]. Genetic deletion of the FABP genes elevates the whole-brain AEA levels [27]. FABP5−/− mice have also been shown to have heightened levels of AEA and 2-AG in the midbrain when compared to FABP5+/+ mice [28]. Previous research has indicated that 2-AG plays a role in glutamate signaling on DA neurons in the VTA, which may play a role in tobacco addiction [29,30]. Given the potential impact of eCB signaling on nicotine reward, the current study sought to determine the specific behavioral effects of the FABP5 gene in regulating the rewarding effects of nicotine. We hypothesized that because FABP5 serves as an intracellular carrier of AEA, the deletion of this gene would enhance the rewarding properties of nicotine in FABP5−/− mice.
## Animals
Male and female FABP5+/+ and FABP5−/− mice on a C57B6 background, as described previously [31], were kindly provided by Dr. Hotamisligil at Harvard University. The mice were bred in-house, as previously described. All the mice were habituated to the holding room for at least one week and tested between 10 and 14 weeks of age. Before testing, the mice were habituated to handling and subcutaneous (sc) injections. All the mice were drug-naive at the start of the testing and single-housed in a temperature-controlled room on a reverse 12 h light/dark cycle (lights off from 0900–1800). The animals were provided with ad libitum access to food and water throughout the experiments. All the experiments and procedures conformed to the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. The protocol was approved by the Institutional Animal Care and Use Committee at the University of Buffalo, NY, USA.
## Drugs
Nicotine hydrogen tartrate salt was purchased from Sigma-Aldrich (St. Louis, MO, USA). On the first day of each experiment, the animal body weights were recorded, and nicotine solutions were prepared by dissolving 0.01 mg of nicotine per 1 mL of saline. Each administered nicotine dose was based on the individual animal body weight, and all the doses were injected via sc injections at 10 mL/kg. On the drug days of the nicotine-conditioned place preference paradigm, the mice were injected (sc) with either 0.1 or 0.5 mg/kg nicotine immediately before being placed in the nicotine-paired conditioning chamber. These doses were based on previous studies of nicotine CPP [24].
## Statistical Analysis
All statistical analyses were conducted using GraphPad Prism v. 9.3.0 (GraphPad Software Inc., San Diego, CA, USA). The nicotine-conditioned place preference data analysis was conducted using an unpaired two-tailed t-test (comparing differences between the time spent (delta) in the preconditioning versus the test phases, as well as the average locomotor activity per drug dose). The sexes were compared first (males versus females) to determine the potential sex differences, followed by the comparison of the genotypes (FABP+/+ versus FABP5−/−).
## Nicotine-Conditioned Place Preference (CPP)
The FABP5+/+ and FABP5−/− mice were tested for nicotine CPP using commercially available equipment (Coulbourn Instruments, Allentown, PA, USA). The mice were tested in three phases: preconditioning, conditioning, and the test day, as previously described [32–36]. Briefly, the place-conditioning boxes contained three compartments with distinct flooring and wall patterns for tactile and visual differentiation (black/white polka dots with plexiglass flooring or black/white stripes with metal flooring), which were separated by a neutral clear plexiglass compartment. During the preconditioning phase, movement between the distinct polka or striped compartments was possible through the use of two manual guillotine doors, which provided access to either chamber when opened. The entrances to both chambers were closed during the vehicle and drug conditioning days. During the intertrial intervals, the chambers were wiped clean. The testing took place for a total of 10 days and was conducted between 1200 and 1700 each day during the animal’s dark cycle. The mice were subjected to one of two experiments. Experiment one (Exp 1) tested the mice for nicotine CPP at the 0.1 mg/kg nicotine dose. Experiment two (Exp 2) tested the mice for nicotine CPP at the 0.5 mg/kg nicotine dose. Below is a brief summary of the procedure.
## Day 1: Preconditioning phase.
The mice were placed in the neutral center chamber and allowed to access the distinct compartments, with the guillotine doors open, for a total of 15 min. The time spent on each side was recorded in seconds and compiled as a percentage of the time spent in each chamber to determine each subject’s baseline place preference. Animals who exhibited an equal preference between chambers were randomized for the following conditioning phase.
## Days 2 to 9: Conditioning phase.
Both the FABP5+/+ and FABP5−/− mice received sc injections of either saline or nicotine on alternating days and were immediately placed in their respective chambers for the next 8 days of testing. The conditioning took place for 20 min, in which the mice were free to roam the chamber corresponding to the injection received. The drug-paired sides were pre-determined as the opposite of the mice’s initial baseline preference, and the initially preferred chambers were pre-determined as the vehicle treatment (i.e., the biased paradigm). The injections were counterbalanced with respect to the side of the chamber on which the animal was placed. For example, if the animal’s least preferred chamber had the stripped walls and plexiglass flooring during the preconditioning phase, they would be placed in this chamber on the days when they received nicotine injections to measure the difference in their drug-induced place preference later on the test day (day 10). On the saline days, they would be placed in the polka dot chamber with metal flooring, and vice versa. The total number of conditioning days for nicotine and saline was equally divided: four days of nicotine conditioning exposure and four days of saline conditioning exposure.
## Day 10: Test phase.
On the final day, all the animals were placed in the center neutral compartment without exposure to either saline or nicotine. The guillotine doors were opened, and the subjects were given free-roam access to either chamber for 15 min. The time spent on each side was automatically recorded in seconds, and the mice’s preference for the drug-paired chamber was expressed as a percentage of the time spent on the drug-paired side on the test day (day 10) minus their baseline percentage of time spent in their assigned drug chamber on the preconditioning day (day 1). A positive number indicated a preference for the drug-paired chamber, whereas a negative number indicated aversion. A value of zero indicated no preference for either side.
## Nicotine CPP
In each experiment, the treatment group was tested for their change (delta) in preference to the nicotine-paired chamber by measuring the total time spent (seconds) in the nicotine-paired chamber on the preconditioning day versus the test day. Outlier testing (ROUT, $Q = 1$%) was completed for both the experiments, and none were observed.
## Exp 1 (0.1 mg/kg Nicotine CPP):
Potential sex differences in the delta preference scores between the FABP5+/+ and FABP5−/− groups were assessed as previously described, and none were observed ($p \leq 0.05$). The sexes were then collapsed to compare the genotypes within the groups, and the results showed that at the 0.1 mg/kg nicotine dose, the FABP5−/− mice had a significantly higher preference score for the nicotine-paired chamber, t[35] = 2.18, * $$p \leq 0.036$$, compared with their FABP5+/+ counterparts (see Figure 1).
## Exp 2 (0.5 mg/kg Nicotine CPP):
Potential sex differences in the delta preference scores between the FABP5+/+ and FABP5−/− groups were assessed as previously described, and none were observed ($p \leq 0.05$). The sexes were then collapsed to compare the genotypes within the groups, and the results showed no significant difference ($p \leq 0.05$) between the FABP5−/− mice and their FABP5+/+ counterparts (see Figure 2).
## Nicotine CPP Locomotor Activity
Each group’s locomotor activity was assessed for each dose, as measured by the average photobeam breaks on the conditioning days. For both Exp 1 (0.1 mg/kg nicotine CPP) and Exp 2 (0.5 mg/kg nicotine CPP), there was no significant difference ($p \leq 0.05$) in the locomotor activity when comparing saline injections to the nicotine injections in the case of either genotype.
## Discussion
The present study examined the role of the gene encoding of the endocannabinoid-trafficking protein, FABP5, on nicotine CPP. For the first time, we demonstrated the novel role of this protein in regulating the rewarding properties associated with nicotine. *Mice* genetically deficient in FABP5 showed greater acquisition of a nicotine place preference at a low nicotine dose of 0.1 mg/kg (Figure 1). Unlike the FABP5+/+ mice, the FABP5−/− did not show a dose-dependent effect of nicotine CPP. This enhanced CPP acquisition for 0.1 mg/kg nicotine supports our hypothesis regarding what is known about the role of cannabinoids and nicotine reward.
The observed phenotype in the FABP5−/− mice, associated with a subthreshold nicotine dose of 0.1 mg/kg, is supported by the findings of Merritt and colleagues, who tested FAAH−/− mice and demonstrated the enhanced acquisition of nicotine place preference with the same low dose of nicotine (0.1 mg/kg) but not at the high doses of nicotine (0.5 or 1 mg/kg) [24]. However, the effect on the FAAH−/− mice was greater in magnitude compared with the FABP5−/− mice in the current study, which is plausible, since FAAH−/− mice show a ~15-fold increase in the AEA levels in the brain, while FABP5−/− mice only show a −1.5-fold increase in AEA [37]. Microdialysis studies of FAAH−/− mice indicated that 0.1 mg/kg of nicotine significantly increases the DA concentrations in the NAc [25], which may underlie the enhanced place preference for nicotine in CPP paradigms. A similar mechanism of action may underlie the behavior observed in FABP5−/− mice, though this has yet to be evaluated.
Additionally, it is also possible that the enhanced nicotine CPP acquisition observed in FABP5−/− mice could be due not only to the AEA levels but also to heightened 2-AG levels. Heightened 2-AG levels have been reported in the midbrain of FABP5−/− mice when compared to their FABP5+/+ counterparts [28], including areas such as the VTA. Previous research has determined that nicotine increases the VTA dialysate 2-AG levels under conditions of acute and chronic administration [30]. It was found that 2-AG plays a key role in the plasticity of glutamate signaling to DA neurons in the VTA [38], which may be a critical component of the mechanisms of tobacco addiction [39]. When there are higher levels of 2-AG in the VTA, a cue-evoked increase in DA within the NAc is potentiated, which has been associated with reward-seeking behavior [40]. In addition, VTA AEA signaling may enhance DA cell activity through CB1R-mediated decreases in GABA release [41,42]. Indeed, the CB1R-mediated suppression of VTA glutamate release has also been reported [41,43], which may contribute to the CB1R-mediated attenuation of the nicotine-induced excitation of DA cells in the VTA following FAAH inhibition [44]. The global deletion of the FABP5 gene significantly decreased tonic 2-AG and AEA signaling in the GABA synapses of medium spiny neurons. Phasic 2-AG-mediated short-term plasticity was also blunted, but this did not impact CB1R function or expression, indicating that the FABP5 gene plays a role in central excitatory and inhibitory synapse signaling [28]. While not much is known regarding how the eCB levels influence the metabolism of nicotine, it is clear that nicotine influences the eCB levels which, in turn, enhances the reinforcing effects of the drug.
Our previous work showed a significant decrease in ethanol consumption among mice treated with an inhibitor of FABPs (SBFI26). Specifically, male and female mice treated with SBFI26 consumed $24\%$ and $42\%$ less compared with their FABP5+/+ counterparts receiving the vehicle, respectively. This supports the interrelationship between nicotine, cannabis, and ethanol [45]. While this seems paradoxical, it suggests that the reduction in FABPs can result in a blunted response in the pre-neuronal release of DA, thereby reducing the ethanol-induced euphoria followed by the attenuation of ethanol-seeking behavior.
To date, however, few studies have examined the impact of the co-exposure of cannabinoids and nicotine on locomotor activity. The previously referenced work by Merritt and colleagues reported no difference in the locomotor activity of C57BL/6J mice following nicotine injections of 0.1 mg/kg compared to saline injections [24]. These data support our current findings, as we did not observe a significant effect on the locomotor activity after the sc injections of 0.1 mg/kg nicotine in the case of either genotype. Based on this, it is likely that the increased nicotine preference, which is believed to be CB1R-dependent, is not influenced by locomotor activity.
Future studies should aim to explore the effects of FABP5−/− on nicotine self-administration and withdrawal. Blocking CB1R activation via antagonists or inverse agonists has been shown to decrease nicotine-seeking behavior and self-administration and lessen nicotine withdrawal symptoms [14,24,46]. Therefore, it is likely that FABP5−/− mice would display opposite effects due to their heightened AEA levels and CB1R activation in the mesolimbic reward pathway. While this enhanced nicotine preference appears to be CB1R-dependent, researchers should examine potential treatment methods for nicotine and other substances of abuse. For example, inhibiting the activity of CB1R has been explored for Δ9-tetrahydrocannabinol (THC), but the long-term blocking of CB1R would disinhibit GABA signaling. As a result, the neuronal release of DA would be reduced, which may, in turn, lead to enhanced substance use and abuse [47]. It is crucial to understand how both nicotine and THC interact with the eCB system, as tobacco use commonly follows or coincides with cannabis use [48]. Additionally, further studies should be conducted to determine the long-term effects of heightened AEA levels on nicotine metabolism in FABP5−/− mice. The observed increase in nicotine-seeking behavior may be due to differences in metabolism, but this has not yet been confirmed. While our study focuses on the influence of the eCB system on nicotine-seeking behavior, it is important to examine other mechanisms involved in nicotine use in order to determine the best potential treatment methods. Previous studies found that tobacco smoke exposure leads to nicotine dependence in rats, which resulted in increased alpha-7 nicotinic acetylcholine receptors (nAChR) in the hippocampus and was correlated with increased somatic symptoms of withdrawal [49]. Additionally, corticotropin-releasing-factor (CRF)-like peptides have been linked to prolonged symptoms of withdrawal from cannabis, alcohol, and tobacco. While no studies have observed a direct effect of nicotine withdrawal on CRF production, the chronic administration of nicotine may alter the sensitivity of CRF-like peptides to their receptors [50]. When examining both nicotine self-administration and withdrawal symptoms in future studies, it is important to consider not only the eCBs but also changes in the nAChRs and CRF-like peptides.
While our study does point to the potential importance of the FABP5 gene for nicotine-seeking behavior, it is not without limitations. Specifically, we do not know the exact effect of the eCB levels on nicotine metabolism. Higher eCB levels may slow down the metabolism of nicotine which may, therefore, potentiate its effects, but this is unknown. Future studies should aim to examine the pharmacological metabolism of nicotine in FABP5−/− mice or other genetic models which display enhanced eCB levels.
Our findings support the conclusion that the eCB levels have an important influence on nicotine preference. We showed, for the first time, that the global deletion of FABP5 potentiates the reinforcing aspects of low doses of nicotine, as measured by CPP. Future research should aim to directly examine the eCB levels in genetic models of FABP5−/− in response to nicotine and the influence of the eCB levels on nicotine metabolism in order to confirm these notions.
## Funding:
This research was funded by the NY Research Foundation (RIAQ0940) and NIH (DA045640).
## Data Availability Statement:
Not applicable.
## References
1. Lariscy JT, Hummer R, Rogers R. **Cigarette Smoking and All-Cause and Cause-Specific Adult Mortality in the United States**. *Demography* (2018) **55** 1855-1885. PMID: 30232778
2. Benowitz NL. **Pharmacology of Nicotine: Addiction, Smoking-Induced Disease, and Therapeutics**. *Annu. Rev. Pharmacol. Toxicol* (2009) **49** 57-71. PMID: 18834313
3. Zhang T, Zhang L, Liang Y, Siapas AG, Zhou F-M, Dani JA. **Dopamine Signaling Differences in the Nucleus Accumbens and Dorsal Striatum Exploited by Nicotine**. *J. Neurosci* (2009) **29** 4035. PMID: 19339599
4. Schultz W. **Dopamine signals for reward value and risk: Basic and recent data**. *Behav. Brain Funct* (2010) **6** 24. PMID: 20416052
5. Febo M, Blum K, Badgaiyan RD, Baron D, Thanos PK, Colon-Perez LM, Demotrovics Z, Gold MS. **Dopamine homeostasis: Brain functional connectivity in reward deficiency syndrome**. *Front. Biosci* (2017) **22** 669-691
6. Melis M, Muntoni A, Pistis M. **Endocannabinoids and the Processing of Value-Related Signals**. *Front. Pharmacol* (2012) **3** 7. PMID: 22347186
7. Parsons LH, Hurd Y. **Endocannabinoid signalling in reward and addiction**. *Nat. Rev. Neurosci* (2015) **16** 579-594. PMID: 26373473
8. Murillo-Rodríguez E, Vázquez E, Millán-Aldaco D, Palomero-Rivero M, Drucker-Colin R. **Effects of the fatty acid amide hydrolase inhibitor URB597 on the sleep-wake cycle, c-Fos expression and dopamine levels of the rat**. *Eur. J. Pharmacol* (2007) **562** 82-91. PMID: 17336288
9. Zhang L-Y, Zhou Y-Q, Yu Z-P, Zhang X-Q, Shi J, Shen H-W. **Restoring glutamate homeostasis in the nucleus accumbens via endocannabinoid-mimetic drug prevents relapse to cocaine seeking behavior in rats**. *Neuropsychopharmacology* (2021) **46** 970-981. PMID: 33514875
10. Gunduz-Cinar O, MacPherson KP, Cinar R, Gamble-George J, Sugden K, Williams B, Godlewski G, Ramikie TS, Gorka AX, Alapafuja SO. **Convergent translational evidence of a role for anandamide in amygdala-mediated fear extinction, threat processing and stress-reactivity**. *Mol. Psychiatry* (2013) **18** 813-823. PMID: 22688188
11. Garani R, Watts J, Mizrahi R. **Endocannabinoid system in psychotic and mood disorders, a review of human studies**. *Prog. Neuro-Psychopharmacol. Biol. Psychiatry* (2021) **106** 110096
12. Blum K, Khalsa J, Cadet JL, Baron D, Bowirrat A, Boyett B, Lott L, Brewer R, Gondré-Lewis M, Bunt G. **Cannabis-Induced Hypodopaminergic Anhedonia and Cognitive Decline in Humans: Embracing Putative Induction of Dopamine Homeostasis**. *Front. Psychiatry* (2021) **12** 623403. PMID: 33868044
13. Le Foll B, Goldberg S. **Rimonabant, a CB1 antagonist, blocks nicotine-conditioned place preferences**. *Neuroreport* (2004) **15** 2139-2143. PMID: 15486497
14. Cohen C, Perrault G, Griebel G, Soubrié P. **Nicotine-associated cues maintain nicotine-seeking behavior in rats several weeks after nicotine withdrawal: Reversal by the cannabinoid (CB1) receptor antagonist, rimonabant (SR141716)**. *Neuropsychopharmacology* (2005) **30** 145-155. PMID: 15292905
15. Le Foll B, Perrault G, Griebel G, Soubrié P. **Blocking cannabinoid CB1 receptors for the treatment of nicotine dependence: Insights from pre-clinical and clinical studies**. *Addict. Biol* (2008) **13** 239-252. PMID: 18482433
16. Cheer JF, Wassum KM, Sombers LA, Heien MLAV, Ariansen JL, Aragona BJ, Phillips PEM, Wightman RM. **Phasic dopamine release evoked by abused substances requires cannabinoid receptor activation**. *J. Neurosci* (2007) **27** 791-795. PMID: 17251418
17. Cohen C, Perrault G, Voltz C, Steinberg R, Soubrié P. **SR141716, a central cannabinoid (CB(1)) receptor antagonist, blocks the motivational and dopamine-releasing effects of nicotine in rats**. *Behav. Pharmacol* (2002) **13** 451-463. PMID: 12394421
18. Grieder TE, George O, Tan H, George SR, Le Foll B, Laviolette SR, van der Kooy D. **Phasic D1 and tonic D2 dopamine receptor signaling double dissociate the motivational effects of acute nicotine and chronic nicotine withdrawal**. *Proc. Natl. Acad. Sci. USA* (2012) **109** 3101-3106. PMID: 22308372
19. Forget B, Coen K, Le Foll B. **Inhibition of fatty acid amide hydrolase reduces reinstatement of nicotine seeking but not break point for nicotine self-administration—Comparison with CB(1) receptor blockade**. *Psychopharmacology* (2009) **205** 613-624. PMID: 19484221
20. Kandel ER, Kandel DB. **A molecular basis for nicotine as a gateway drug**. *N. Engl. J. Med* (2014) **371** 932-943. PMID: 25184865
21. Rodríguez de Fonseca F, del Arco I, Bermudez-Silva FJ, Bilbao A, Cippitelli A, Navarro M. **The endocannabinoid system: Physiology and pharmacology**. *Alcohol Alcohol* (2005) **40** 2-14. PMID: 15550444
22. Devane WA, Hanuš L, Breuer A, Pertwee RG, Stevenson LA, Griffin G, Gibson D, Mandelbaum A, Etinger A, Mechoulam R. **Isolation and structure of a brain constituent that binds to the cannabinoid receptor**. *Science* (1992) **258** 1946-1949. PMID: 1470919
23. Cravatt BF, Giang DK, Mayfield SP, Boger DL, Lerner RA, Gilula NB. **Molecular characterization of an enzyme that degrades neuromodulatory fatty-acid amides**. *Nature* (1996) **384** 83-87. PMID: 8900284
24. Merritt LL, Martin BR, Walters C, Lichtman AH, Damaj MI. **The endogenous cannabinoid system modulates nicotine reward and dependence**. *J. Pharmacol. Exp. Ther* (2008) **326** 483-492. PMID: 18451315
25. Pavon FJ, Serrano A, Sidhpura N, Polis I, Stouffer D, de Fonseca FR, Cravatt BF, Martin-Fardon R, Parsons LH. **Fatty acid amide hydrolase (FAAH) inactivation confers enhanced sensitivity to nicotine-induced dopamine release in the mouse nucleus accumbens**. *Addict. Biol* (2018) **23** 723-734. PMID: 28660730
26. Kaczocha M, Glaser S, Deutsch D. **Identification of intracellular carriers for the endocannabinoid anandamide**. *Proc. Natl. Acad. Sci. USA* (2009) **106** 6375-6380. PMID: 19307565
27. Kaczocha M, Glaser ST, Maher T, Clavin B, Hamilton J, Joseph O, Rebecchi M, Puopolo M, Owada Y, Thanos PK. **Fatty acid binding protein deletion suppresses inflammatory pain through endocannabinoid/N-acylethanolamine-dependent mechanisms**. *Mol. Pain* (2015) **11** 52. PMID: 26311517
28. Fauzan M, Oubraim S, Yu M, Glaser ST, Kaczocha M, Haj-Dahmane S. **Fatty Acid-Binding Protein 5 Modulates Brain Endocannabinoid Tone and Retrograde Signaling in the Striatum**. *Front. Cell. Neurosci* (2022) **16** 936939. PMID: 35875351
29. Yu S, Levi L, Casadesus G, Kunos G, Noy N. **Fatty acid-binding protein 5 (FABP5) regulates cognitive function both by decreasing anandamide levels and by activating the nuclear receptor peroxisome proliferator-activated receptor β/δ (PPARβ/δ) in the brain**. *J. Biol. Chem* (2014) **289** 12748-12758. PMID: 24644281
30. Buczynski MW, Polis I, Parsons L. **The volitional nature of nicotine exposure alters anandamide and oleoylethanolamide levels in the ventral tegmental area**. *Neuropsychopharmacology* (2013) **38** 574-584. PMID: 23169348
31. Maeda K, Uysal KT, Makowski L, Görgün CZ, Atsumi G, Parker RA, Brüning J, Hertzel AV, Bernlohr DA, Hotamisligil GS. **Role of the fatty acid binding protein mal1 in obesity and insulin resistance**. *Diabetes* (2003) **52** 300-307. PMID: 12540600
32. Thanos K, Bermeo C, Wang G-J, Volkowa ND. **D-cycloserine accelerates the extinction of cocaine-induced conditioned place preference in C57bL/c mice**. *Behav. Brain Res* (2009) **199** 345-349. PMID: 19152811
33. Thanos K, Malave L, Delis F, Mangine P, Kane K, Grunseich A, Vitale M, Greengard P, Volkow ND. **Knockout of p11 attenuates the acquisition and reinstatement of cocaine conditioned place preference in male but not in female mice**. *Synapse* (2016) **70** 293-301. PMID: 26990537
34. Thanos PK, Bermeo C, Rubinstein M, Suchland KL, Wang GJ, Grandy DK, Volkow ND. **Conditioned place preference and locomotor activity in response to methylphenidate, amphetamine and cocaine in mice lacking dopamine D4 receptors**. *J. Psychopharmacol* (2010) **24** 897-904. PMID: 19282420
35. Hamilton J, Marion M, Figueiredo A, Clavin BH, Deutsch D, Kaczocha M, Haj-Dahmane S, Thanos PK. **Fatty acid binding protein deletion prevents stress-induced preference for cocaine and dampens stress-induced corticosterone levels**. *Synapse* (2018) **72** e22031. PMID: 29457656
36. Ananth M, Hetelekides EM, Hamilton J, Thanos PK. **Dopamine D4 receptor gene expression plays important role in extinction and reinstatement of cocaine-seeking behavior in mice**. *Behav. Brain Res* (2019) **365** 1-6. PMID: 30797855
37. Cravatt BF, Demarest K, Patricelli MP, Bracey MH, Giang DK, Martin BR, Aron H. **Lichtman Supersensitivity to anandamide and enhanced endogenous cannabinoid signaling in mice lacking fatty acid amide hydrolase**. *Proc. Natl. Acad. Sci. USA* (2001) **98** 9371. PMID: 11470906
38. Haj-Dahmane S, Shen R. **Regulation of plasticity of glutamate synapses by endocannabinoids and the cyclic-AMP/protein kinase A pathway in midbrain dopamine neurons**. *J. Physiol* (2010) **588** 2589-2604. PMID: 20498231
39. Mansvelder HD, Mertz M, Role L. **Nicotinic modulation of synaptic transmission and plasticity in cortico-limbic circuits**. *Semin. Cell. Dev. Biol* (2009) **20** 432-440. PMID: 19560048
40. Oleson EB, Beckert MV, Morra JT, Lansink CS, Cachope R, Abdullah RA, Loriaux AL, Schetters D, Pattij T, Roitman MF. **Endocannabinoids shape accumbal encoding of cue-motivated behavior via CB1 receptor activation in the ventral tegmentum**. *Neuron* (2012) **73** 360-373. PMID: 22284189
41. Lupica CR, Riegel AC. **Endocannabinoid release from midbrain dopamine neurons: A potential substrate for cannabinoid receptor antagonist treatment of addiction**. *Neuropharmacology* (2005) **48** 1105-1116. PMID: 15878779
42. Szabo B, Siemes S, Wallmichrath I. **Inhibition of GABAergic neurotransmission in the ventral tegmental area by cannabinoids**. *Eur. J. Neurosci* (2002) **15** 2057-2061. PMID: 12099913
43. Melis M, Pistis M, Perra S, Muntoni AL, Pillolla G, Gessa GL. **Endocannabinoids mediate presynaptic inhibition of glutamatergic transmission in rat ventral tegmental area dopamine neurons through activation of CB1 receptors**. *J. Neurosci* (2004) **24** 53-62. PMID: 14715937
44. Melis M, Pillolla G, Luchicchi A, Muntoni AL, Yasar S, Goldberg SR, Pistis M. **Endogenous fatty acid ethanolamides suppress nicotine-induced activation of mesolimbic dopamine neurons through nuclear receptors**. *J. Neurosci* (2008) **28** 13985-13994. PMID: 19091987
45. Figueiredo A, Hamilton J, Marion M, Blum K, Kaczocha M, Haj-Dahmane S, Deutsch D, Thanos PK. **Pharmacological Inhibition of Brain Fatty Acid Binding Protein Reduces Ethanol Consumption in Mice**. *J. Reward Defic. Syndr. Addict. Sci* (2017) **3** 21-27. PMID: 29367955
46. Cohen C, Kodas E, Griebel G. **CB1 receptor antagonists for the treatment of nicotine addiction**. *Pharmacol. Biochem. Behav* (2005) **81** 387-395. PMID: 15935455
47. Blum K, Oscar-Berman M, Braverman ER, Febo M, Li M, Gold MS. **Enhancing Brain Pregnenolone May Protect Cannabis Intoxication but Should Not Be Considered as an Anti-addiction Therapeutic: Hypothesizing Dopaminergic Blockade and Promoting Anti-Reward**. *J. Reward Defic. Syndr* (2015) **1** 20-23. PMID: 26306328
48. Tullis LM, Dupont R, Frost-Pineda K, Gold MS. **Marijuana and tobacco: A major connection?**. *J. Addict. Dis* (2003) **22** 51-62. PMID: 14621344
49. Small E, Shah HP, Davenport JJ, Geier JE, Yavarovich KR, Yamada H, Sabarinath SN, Derendorf H, Pauly JR, Gold MS. **Tobacco smoke exposure induces nicotine dependence in rats**. *Psychopharmacology* (2010) **208** 143-158. PMID: 19936715
50. Bruijnzeel AW, Gold MS. **The role of corticotropin-releasing factor-like peptides in cannabis, nicotine, and alcohol dependence**. *Brain Res. Rev* (2005) **49** 505-528. PMID: 16269317
|
---
title: How do living conditions affect the gut microbiota of endangered Père David’s
deer (Elaphurus davidianus)? Initial findings from the warm temperate zone
authors:
- Hongyu Yao
- Qiying Mo
- Hong Wu
- Dapeng Zhao
journal: PeerJ
year: 2023
pmcid: PMC9969852
doi: 10.7717/peerj.14897
license: CC BY 4.0
---
# How do living conditions affect the gut microbiota of endangered Père David’s deer (Elaphurus davidianus)? Initial findings from the warm temperate zone
## Abstract
Reintroduction is an effective strategy in the conservation of endangered species under scientific monitoring. Intestinal flora plays an important role in the envir onmental adaptation of endangered Père David’s deer (Elaphurus davidianus). In this study, 34 fecal samples from E. davidianus were collected from different habitats in Tianjin city of China to investigate differences in the intestinal flora under captive and semi-free-ranging conditions. Based on 16S rRNA high-throughput sequencing technology, a total of 23 phyla and 518 genera were obtained. Firmicutes was dominant in all individuals. At the genus level, UCG-005 ($13.05\%$) and Rikenellaceae_RC9_gut_group ($8.94\%$) were dominant in captive individuals, while Psychrobacillus ($26.53\%$) and Pseudomonas ($11.33\%$) were dominant in semi-free-ranging individuals. Alpha diversity results showed that the intestinal flora richness and diversity were significantly ($P \leq 0.001$) higher in captive individuals than in semi-free-ranging individuals. Beta diversity analysis also showed a significant difference ($$P \leq 0.001$$) between the two groups. In addition, some age- and sex-related genera such as Monoglobus were identified. In summary, the structure and diversity of intestinal flora showed significant habitat variation. This is the first time an analysis has been undertaken of the structural differences of the intestinal flora in Père David’s deer, under different habitats in the warm temperate zone, providing a reference basis for the conservation of endangered species.
## Introduction
Within conservation biology, reintroduction is a widespread technique that has helped many endangered or extinct wildlife species to recover their population size, including mammals, birds, and invertebrates (Seddon, Armstrong & Maloney, 2007; Sutton, 2015; Corlett, 2016). However, many factors affect the success rate of reintroduction, and poor performance is reported on reintroducing threatened or endangered species from captivity to the wild (Reading, Miller & Shepherdson, 2013; Seddon et al., 2014). Monitoring of released free-ranging animals is of utmost importance in improving the success rate of reintroduction (Seddon, Armstrong & Maloney, 2007; Yang et al., 2018). Previous research has shown that reintroduction has helped a number of species, such as giant pandas (Yang et al., 2018), peregrine falcons (Jacobsen et al., 2008), and Texas horned lizards (Williams, Rains & Hale, 2019) to successfully reappear and reproduce in their historic ranges. One representative success story is the reintroduction of Père David’s deer (Elaphurus davidianus) (Cheng et al., 2021). E. davidianus was endemic to China and ranged from Liaoning Province in northern China to Zhejiang Province in southern China, from 19 degrees North Latitude to 47 degrees North Latitude (Cheng et al., 2021). E. davidianus originated in the early Pleistocene period and reached their peak over 3,000 years ago. However, due to anthropogenic and natural pressures including human hunting, environmental destruction, climate change, and war, E. davidianus disappeared from their original habitats in the early 20th century (Cheng et al., 2021). The endangered species was reintroduced to China from England in 1985 and has bred in many areas since (e.g., Beijing, Jiangsu, Hubei, and Hunan), with a total population of nearly 10,000 to date (Cheng et al., 2021). Although the total number of E. davidianus in *China is* growing, the majority of deer live in captivity, and much work remains to be done for successful reintroduction and rewilding across their historic habitats (Sutton, 2015). Natural food selection and health of E. davidianus during the process of living from captivity to the wild are key aspects of reintroduction work (Sun et al., 2019).
In order to solve this problem effectively, intestinal microbiota monitoring, based on non-invasive sampling technology, can be used to detect the relationship between gut microbiome and health of the host animal, especially for threatened species. Research has shown that gut microbiota plays an essential role in contributing to food digestion and disease immunity of their host (Pan & Yu, 2014). Numerous factors can influence gut bacterial diversity, such as diet (Jiang et al., 2020), sex (Kim et al., 2020), and age (Jami et al., 2013). To date, investigations on intestinal microbiota of E. davidianus are still in their infancy, and only four related studies have been published (Zhang et al., 2018; Sun et al., 2019; Wang et al., 2019; Zhen et al., 2022). These studies have all focused on populations living in subtropical zones as well as those living in transition areas from warm temperate zones to subtropical zones. There is a great knowledge gap on the gut bacteria community of E. davidianus in the warm temperate zone, and no studies have been conducted to investigate the relationship between sex or age and gut microbiota establishment in E. davidianus.
The main purpose of this study was to analyze differences in gut microbiota composition and diversity of E. davidianus living in the warm temperate zone, under captive and semi-free-ranging living conditions, for the first time. We also analyzed the differences in gut microbiota among captive individuals of different sexes and ages. The results of this study could provide a scientific reference for the implementation of comprehensive reintroduction of E. davidianus in the future, as well as the scientific management and conservation of this endangered species in related protected areas.
## Study site and sample collection
In this study, fecal samples of E. davidianus were obtained using a non-invasive sampling technique from different rearing environments in Tianjin city, located in North China (38°34′–40°15′N, 116°43′–118°04′E) (Wu et al., 2021). It has a semi-humid monsoonal climate with an annual average temperature of about 14 °C. We collected a total of 34 fecal samples between October and December 2021. Among them, samples from six females (NO. CF01-CF06), seven males (NO. CM01-CM07), and six juveniles (NO. CJ01-CJ06) were collected from captive groups (C group) in Tianjin Zoo. Fifteen fecal samples (NO. SF01-SF15) were obtained from semi-free-ranging groups (S group) in Tianjin Qilihai Wetland. The main diet of E. davidianus was provided by keepers in Tianjin Zoo (Table S1), while the semi-free-ranging group foraged for plants by themselves.
In this study, we distinguished adults and juveniles by body size, and identified the sex of adult individuals based on the existence of antlers when sampling at Tianjin Zoo. The collected fecal samples were sterilized in 5 ml tubes. The samples were stored temporarily in a refrigerated insulated box, then brought back to the laboratory for freezing at –80 °C.
## DNA extraction, amplification, and sequencing
All samples were extracted using the TIANamp Stool DNA Kit (TIANGEN, Sichuan, China). Specific primers with barcodes were synthesized according to the specified sequencing region. PCR amplification was performed according to the manufacturer’s instructions, with three replicates per sample. PCR products from the same samples were mixed and detected by $2\%$ agarose gel electrophoresis. PCR products were recovered using AxyPrepDNA gel recovery kit (AXYGEN Corporation, Silicon Valley, CA, USA). Detection and quantification were performed by QuantiFluor-ST™ Blue Fluorescence Quantification System (Promega, Madison, WI, USA). Purified PCR amplicons were sequenced on the Illumina MiSeq platform at Shanghai Majorbio Bio-pharm Technology Co., Ltd.
## Bioinformatics, statistical analyses, and functional prediction
The raw data obtained from MiSeq sequencing were optimized using Qiime (version 1.9.1; http://qiime.org/). Sequences with at least $97\%$ identity were subjected to Operational Taxonomic Unit (OTU) clustering analysis using Uparse (version 7.0.1090; https://drive5.com/uparse/). The taxonomic analysis of I was performed by RDP Classifier (version 2.11; http://rdp.cme.msu.edu/classifier/classifier.jsp). Rarefaction curves were created to express the species richness of each sample and the reasonableness of the sequencing depth. The Wilcoxon test was applied to detect differences in the abundance of flora between different groups at the phylum and genus level. LEfSe analysis (Score >4) was used to seek biomarkers with significant differences between the groups. In this study, Chao1, ACE, Shannon, Simpson, and *Coverage alpha* diversity indexes were determined using mothur software (version v.1.30.1; https://mothur.org/wiki/mothur_v.1.9.0/), demonstrating the microbial community richness and diversity in each sample. T-test analysis based on alpha diversity index was used to identify any significant differences between the groups. Principal Coordinates Analysis (PCoA) of beta diversity was implemented to visualize similarities or dissimilarities of microbial community diversity between samples. Microbial functional prediction was executed by PICRUSt based on high-quality sequences.
## Sequencing data and microbiota composition
In this study, a total of 6,922,773 optimized sequences were obtained after denoising all 34 samples by Illumina MiSeq sequencing with an average sequence length of 416 bp, ranging from 43,036 to 306,943 sequences in all samples (Table S2). By performing clustering on all sequences, a total of 3,940 OTUs with a $97\%$ sequence similarity threshold were retrieved. The taxonomic analysis OTU showed that the gut microbiota of E. davidianus could be divided into 23 phyla, 48 classes, 127 orders, 241 families, and 518 genera.
## Microbiota composition and relative abundance of all samples
The rarefaction curve based on OTUs tended to gradually flatten, suggesting that the fecal samples collected in our study were enough to analyze and reflect the maximum level of bacterial diversity (Fig. S1). There were 1,799 OTUs found to be shared by all samples, while the number of OTUs shared by the C group and S group was 1,618 and 578 under the same sequencing depth, respectively (Fig. S1).
At the phylum level, a total of 23 prokaryotic phyla were identified based on the 16S rRNA sequencing. The gut microbiota from the C group were dominated by Firmicutes ($67.89\%$) and Bacteroidota ($28.43\%$), followed by Actinobacteriota ($0.85\%$) and Verrucomicrobiota ($0.78\%$). The gut microbiota of the S group were dominated by Firmicutes ($68.02\%$) and Proteobacteria ($17.74\%$), followed by Actinobacteriota ($10.09\%$) and Bacteroidota ($3.47\%$) (Fig. 1 and Table S3).
**Figure 1:** *The composition of the intestinal flora from Père David’s deer at phylum level.(A) Microbial structure of all fecal samples at phylum level. The pie diagram shows the most abundant phylum in the semi-free-ranging group (B) and captive group (C).*
In terms of genus level, there were 15 genera with more than $3\%$ abundance in all samples. The most abundant genera in the C group were UCG-005 ($13.05\%$), Rikenellaceae_RC9_gut_group ($8.94\%$), Christensenellaceae_R-7_group ($8.37\%$), norank_f_UCG-010 ($4.64\%$), Monoglobus ($4.20\%$), norank_f_Eubacterium_coprostanoligenes_group ($3.50\%$), Bacteroides ($3.22\%$), and Romboutsia ($3.20\%$). The most abundant genera in the S group were Psychrobacillus ($26.53\%$), Pseudomonas ($11.33\%$), UCG-005 ($7.98\%$), Arthrobacter ($6.22\%$), Paenisporosarcina ($4.58\%$), Sporosarcina ($3.71\%$), Acinetobacter ($3.42\%$), and norank_f_UCG-010 ($3.04\%$) (Fig. 2 and Table S4). The stacked percentage histograms of relative abundance at the phylum level (others <$1\%$) and genus level (others <$5\%$) were compared to visualize the relative abundance of intestinal flora in Figs. 1A and 2A, respectively.
**Figure 2:** *Composition of the intestinal flora from Père David’s deer at genus level.(A) Microbial structure of all fecal samples at genus level. The pie diagram shows the most abundant genus in the semi-free-ranging group (B) and captive group (C).*
Wilcoxon tests at the phylum level showed that the percentage of Proteobacteria and Actinobacteriota in the C group were significantly lower than the S group, while the percentage of Bacteroidota was significantly higher than the S group (Fig. 3A). At the genus level, Psychrobacter, Pseudomonas, Arthrobacter, and Paenisporosarcina in the S group were significantly higher than C group, while UCG-005, Christensenellaceae_R-7_group, Rikenellaceae_RC9_gut_group, norank_f_UCG-010, Monoglobus, and Bacteroides were significantly lower than the C group (Fig. 3B).
**Figure 3:** *Comparison of differences in intestinal flora between semi-free-ranging and captive individuals.Differential analysis of dominant bacterial phyla (A) and genera (B) between semi-free-ranging group and captive group based on Wilcoxon tests. LEfSe analysis based on characterizing discriminative features of OTUs (C). *P < 0.05, **P < 0.01, and ***P < 0.001.*
LEfSe analysis identified 26 and 19 taxa (LDA = 4.0) with discrepancies in relative abundance in the C group and S group, respectively. The results of LEsfe analysis indicated that biomarkers in the S group were Pseudomonas, Acinetobacter, Arthrobacter, Solibacillus, Paenisporosarcina, Psychrobacillus, and Sporosarcina. The biomarkers in the C group were Romboutsia, unclassified_f_Peptostreptococcaceae, UCG-005, Christensenellaceae_R-7_group, Monoglobus, Prevotellaceae_UCG-003, Bacteroides, and Rikenellaceae_RC9_gut_group (Fig. 3C).
## The alpha and beta diversity of gut microbiota from different habitats
The alpha diversity was calculated with a T-test using mothur in our study (Table 1 and Fig. S2). The alpha diversity index showed that Chao1, ACE, and Shannon indexes of the C group were significantly higher than the S group, reflecting the richness and diversity of gut microbiota in C group (Fig. 4).
Bacterial community clusters in all samples were visualized by PCoA plots under both weighted and unweighted UniFrac metrics, with each symbol representing an intestinal flora on the PCoA plot. Most samples within the S group and C group were close together and highly aggregated, so they could be distinguished from samples in the other group. The bacterial communities of the S group were separated from those of the C group along main axis 1 (PC1) using weighted UniFrac distances, with the greatest amount of variation ($76.84\%$). When using the unweighted UniFrac distance, the amount of variation reached $35.38\%$ (Fig. 5). This result indicated a high similarity in gut microbiota composition within each group at the OTU level, but differing significantly between the S group and C group.
**Figure 5:** *Analysis of PCoA plots of gut microbiome from Père David’s deer.(A) PCoA plots based on weighted UniFrac distances. (B) PCoA plots based on unweighted UniFrac distances.*
## Differences across age and sex in individuals in the captivity group
Compared with the gut microbiota from adult individuals, the abundance of Cyanobacteria at the phylum level and Bacteroides, Phascolarctobacterium, Clostridium_sensu_stricto_1 at the genus level were significantly higher in the juvenile individuals compared with adult individuals (Fig. 6A). However, the relative abundance of Monoglobus, Lachnospiraaceae_UGG-010, and norank_f_norank_o_norank_c_Clostridia in adult individuals were significantly higher than juveniles (Fig. 6B).
**Figure 6:** *Analysis of intestinal flora in individuals of different ages and genders.Differential analysis of dominant bacterial phyla (A) and genera (B) between adult and juvenile groups; Differential analysis of dominant bacterial phyla (C) and genera (D) between male and female groups; * P < 0.05, ** P < 0.01, *** P < 0.001. (E) Clustering analysis of the evolution of gut microbiota in the C group based on the Bray–Curtis distances generated by mothur; (F) gut microbial function prediction of total individuals based on KEGG databases.*
We analyzed the effect of sex on the composition of gut microbiota, and found that at the phylum level, the abundances of Verrucomicrobiota, Proteobateria, and Desulfobaterota were higher in female individuals (Fig. 6C). At the genus level, Rikenellaceae_RC9_gut_group, unclassified_f_Oscillospiraceae, Phascolarctobacterium, and norank_o_WCHB1-41 were more abundant in female than male individuals, whereas the males had a significantly higher relative abundance of Monoglobus and norank_f_norank_o_norank_c_Clostridia compared with females (Fig. 6D). The community heat map analysis between individuals of different ages and sex at the genus level is shown in Fig. 6E.
## Function prediction of gut microbiota communities
Based on high-throughput sequencing, a total of 46 KEGG pathways were mapped and then divided into secondary KEGG pathways (Fig. S3). The gut microbial function was predicted based on the Clusters of Orthologous Genes (COG) database (Fig. S4). The secondary KEGG pathways related to gut microbiota included metabolism, genetic information processing, and environmental information processing, which demonstrated that differences in gut microbiota had remarkable influence on the metabolism of E. davidianus (Fig. 6).
## Discussion
In this study, Firmicutes and Bacteroidota were the dominant phyla in Père David’s deer from different living conditions, consistent with previous conclusions in this endangered species (Sun et al., 2019). Previous studies have shown that the gut microbiota of ruminants e.g., white-lipped deer (Cervus albirostris) (Li et al., 2022; You et al., 2022), forest musks (Moschus berezovskii) (Zhao et al., 2021), alpine musks (Moschus sifanicus) (Jiang et al., 2021), and blue sheep (Pseudois nayaur) (Zhu et al., 2020), were predominantly made up by Firmicutes and Bacteroidota. This was mainly because ruminants are herbivores that need microorganisms to help them digest and absorb nutrients from plants (Zhu et al., 2020). The bacteria of Firmicutes could encode enzymes that promote energy metabolism by utilizing a variety of substances (Kaakoush, 2015), and many have the function of degrading carbohydrates including cellulose and starch, as well as fat (Jiang et al., 2021). For example, UCG-005 is a cellulose-degrading bacterium (Li et al., 2020), and norank_f_UCG-010 is important for energy uptake in ruminants (Guan et al., 2017; Hassan et al., 2021; Zhang et al., 2022). They were the dominant genera in both semi-free-ranging and captive Père David’s deer.
The dynamics of intestinal flora is an important mechanism by which the host adapts to environmental changes (Moeller & Sanders, 2020). One of the main factors affecting the intestinal flora is diet (Fernando et al., 2010; Zhen et al., 2022). The variety in microbial community composition caused by differences in forage has been demonstrated (Henderson et al., 2015). Christensenellaceae_R-7_group belonging to Firmicutes, which was enriched in captive individuals, was predominantly associated with carbohydrate metabolism and energy metabolism. The result was consistent with previous findings in the same species by Wang et al. [ 2019]. Monoglobus and norank_f_Eubacterium_coprostanoligenes_group, which also belong to Firmicutes, are intestinal microorganisms that specifically degrade pectinand contribute to the absorption and utilization of fat, respectively (Kim et al., 2019; Wei et al., 2021). The forage for captive Père David’s deer in Tianjin Zoo mainly consisted of wheat bran, corn, soybean, and sorghum which are wealthy in starch, protein, fat, and fiber (Maloiy et al., 1970). Thus, this diet structure led to the enrichment of Bacteroidota (Zhao et al., 2019), which are involved in the degradation of macromolecular compounds such as proteins and carbohydrates (Jami, White & Mizrahi, 2014; Hu et al., 2017). For example, Rikenellaceae_RC9_gut_group the dominant genera in captive individuals, is beneficial bacteria in the gut which promote host health and degradation of structural carbohydrates including lignin and cellulose (Zened et al., 2013; Qin et al., 2022).
Previous work had shown that the ratio of Firmicutes and Bacteroidota (F/B) was often associated with the digestion and absorption of carbohydrate-rich foods (Turnbaugh et al., 2006), increasing the ability to metabolize fat (Backhed et al., 2005). This ratio was higher in the S group compared with the C group, since the main foods were grasses, such as gramineous plants which have poor survival during winter for the semi-free-ranging individuals. A higher ratio usually implies a greater ability to absorb nutrients (Mariat et al., 2009; Jami, White & Mizrahi, 2014). Therefore, we presume that semi-free-ranging individuals need to improve the efficiency of energy extraction from ‘poor quality food’ (Fernando et al., 2010) in order to adapt to the harsh natural conditions (Zhao et al., 2019). The genus Paenisporosarcina survives in cold regions and is an important plant rhizosphere microorganism with anti-freeze functions (Zheng et al., 2018). Accordingly, we speculated that its enrichment within the intestine may come from natural foods eaten in the winter. In our study, the abundances of Proteobacteria and Actinobacteriota in the S group were significantly higher than the C group. *The* genera *Pseudomonas and* Acinetobacter, belonging to Proteobacteria, were enriched in the S group. They are conditionally pathogenic bacteria, and can cause an inflammatory response in the organism (Von Klitzing et al., 2017). It has been shown that the aggregation of Proteobacteria can be used as an indicator of dysbiosis (Shin, Whon & Bae, 2015; Zhao et al., 2019), and the abundance of Acinetobacter in forest musk individuals with pneumonia was significantly higher than that in healthy individuals (Zhao et al., 2021). The diet structure of Père David’s deer in a semi-free-ranging state was unstable since additional forage was provided due to limited plant resources in winter (Zhao et al., 2019; Zhen et al., 2022). The results showed that alpha diversity was significantly higher in the C group than the S group ($P \leq 0.01$), which was similar to many other studies on Cervidae species (Li et al., 2020; Minich et al., 2021). Captive feeding may increase alpha diversity due to the adequate high-fiber food provided by zoos in the winter (Guan et al., 2017).
In our study, we found that individuals of different ages and gender did not differ significantly in alpha diversity of gut microbiota, but there were significant differences in flora composition, which was consistent with the research on forest musks (Zhao et al., 2019). It has been experimentally demonstrated that there are changes in total food intake and energy acquisition in animals at different ages (Passadore et al., 2004), leading to variation in intestinal flora (Jiang et al., 2020). The composition of intestinal flora shows significant differences in many ruminants before and after weaning (Jami et al., 2013; Li et al., 2020). In our study, the abundance of Bacteroides, which has the function of digesting fat and protein, was significantly higher in the intestine of juvenile individuals than in adult individuals. We hypothesized that it facilitated the absorption of nutrients required during development in juvenile individuals. Since forage was the main food for adult individuals, the contents of the genus Monoglobus, which can degrade the pectin component of plant cell walls, was significantly richer in the gut of adult individuals than juveniles (Jewell et al., 2015). The enrichment of Lachnospiraceae_UGG-010 in adult individuals may be positively correlated with feed utilization. In addition, the results showed that the genus Phascolarctobacterium was also enriched in the gut of juvenile individuals. Since the decrease of Phascolarctobacterium could lead to an imbalance in human host immune homeostasis (Chen et al., 2021), we speculated that Phascolarctobacterium could protect young individuals from disease (Oikonomou et al., 2013), and thus the composition of gut microbiota in deer was strongly related to the immune function of the host (Moeller & Sanders, 2020).
Numerous studies on animals and humans have shown that gender affects the structure of the intestine flora (Kim et al., 2020; Minich et al., 2021). In our study, only captive individuals were analyzed to find the influence of gender on gut microbiota. It has been shown that adult females invest larger amounts of energy reserves and consume food to produce offspring (Moyes et al., 2006). In this study, the abundances of Rikenellaceae_RC9_gut_group and Phascolarctobacterium were significantly higher in females than in males, promoting host nutrient absorption and health as mentioned previously (Chen et al., 2021; Qin et al., 2022). Furthermore, the results showed that the abundances of both the phylum Verrucomicrobiota and the genus norank_o_WCHB1-41 belonging to this phylum were significantly higher in females than in males. Since Verrucomicrobiota can degrade many complex polysaccharides (Sichert et al., 2020), norank_o_WCHB1-41 might be the genus that plays a key role in this phylum (Hassan et al., 2021). At present, the influence of intestinal microorganisms on Père David’s deer is based on our analysis and previous reports, and further experiments are needed to verify this potential link.
## Conclusions
Reintroduction is an important behavioral measure for endangered species conservation, and the process of adaptation to the environment during species reintroduction needs scientific monitoring. The detection of intestinal microorganisms provides a new idea for the reintroduction and protection of wildlife. This study is the first report to compare the differences in gut microbiota composition of Père David’s deer in different habitats from the warm temperate zone. It was found that there are significant differences in gut microbiota composition and diversity between the captive group and semi-free-ranging group, and that age and gender also affect the composition of the gut microbiota under the same feeding condition. This research provides comparative information of gut microbiota from Père David’s deer in northern China, which will be useful to further understand how living conditions affect gut microbiota of this endangered species and provide a scientific reference for successful introductions.
## References
1. Backhed F, Ley RE, Sonnenburg JL, Peterson DA, Gordon JI. **Host-bacterial mutualism in the human intestine**. *Science* (2005) **307** 1915-1920. DOI: 10.1126/science.1104816
2. Chen J, Wang W, Guo Z, Huang S, Lei H, Zang P, Lu B, Shao J, Gu P. **Associations between gut microbiota and thyroidal function status in Chinese patients with Graves’ disease**. *Journal of Endocrinological Investigation* (2021) **44** 1913-1926. DOI: 10.1007/s40618-021-01507-6
3. Cheng Z, Tian X, Zhong Z, Li P, Sun D, Bai J, Meng Y, Zhang S, Zhang Y, Wang L, Liu D. **Reintroduction, distribution, population dynamics and conservation of a species formerly extinct in the wild: a review of thirty-five years of successful Milu (**. *Global Ecology and Conservation* (2021) **31**. DOI: 10.1016/j.gecco.2021.e01860
4. Corlett RT. **Restoration, reintroduction, and rewilding in a changing world**. *Trends in Ecology & Evolution* (2016) **31** 453-462. DOI: 10.1016/j.tree.2016.02.017
5. Fernando SC, Purvis HT, Najar FZ, Sukharnikov LO, Krehbiel CR, Nagaraja TG, Roe BA, Desilva U. **Rumen microbial population dynamics during adaptation to a high-grain diet**. *Applied and Environmental Microbiology* (2010) **76** 7482-7490. DOI: 10.1128/AEM.00388-10
6. Guan Y, Yang H, Han S, Feng L, Wang T, Ge J. **Comparison of the gut microbiota composition between wild and captive Sika deer (**. *AMB Express* (2017) **7** 212. DOI: 10.1186/s13568-017-0517-8
7. Hassan F-U, Guo Y, Li M, Tang Z, Peng L, Liang X, Yang C. **Effect of methionine supplementation on rumen microbiota, fermentation, and amino acid metabolism in**. *Microorganisms* (2021) **9** 1717. DOI: 10.3390/microorganisms9081717
8. Henderson G, Cox F, Ganesh S, Jonker A, Young W, Janssen PH. **Rumen microbial community composition varies with diet and host, but a core microbiome is found across a wide geographical range**. *Scientific Reports* (2015) **5** 14567. DOI: 10.1038/srep14567
9. Hu X, Liu G, Shafer ABA, Wei Y, Zhou J, Lin S, Wu H, Zhou M, Hu D, Liu S. **Comparative analysis of the gut microbial communities in forest and Alpine musk deer using high-throughput sequencing**. *Frontiers in Microbiology* (2017) **8** 572. DOI: 10.3389/fmicb.2017.0057
10. Jacobsen F, Nesje M, Bachmann L, Lifjeld JT. **Significant genetic admixture after reintroduction of peregrine falcon (**. *Conservation Genetics* (2008) **9** 581-591. DOI: 10.1007/s10592-007-9373-4
11. Jami E, Israel A, Kotser A, Mizrahi I. **Exploring the bovine rumen bacterial community from birth to adulthood**. *The International Society for Microbial Ecology Journal* (2013) **7** 1069-1079. DOI: 10.1038/ismej.2013.2
12. Jami E, White BA, Mizrahi I. **Potential role of the bovine rumen microbiome in modulating milk composition and feed efficiency**. *PLOS ONE* (2014) **9**. DOI: 10.1371/journal.pone.0085423
13. Jewell KA, McCormick CA, Odt CL, Weimer PJ, Suen G. **Ruminal bacterial community composition in dairy cows is dynamic over the course of two lactations and correlates with feed efficiency**. *Applied and Environmental Microbiology* (2015) **81** 4697-4710. DOI: 10.1128/AEM.00720-15
14. Jiang H, Chen W, Su L, Huang M, Lin L, Su Q, Li G, Ahmad HI, Li L, Zhang X, Li H, Chen J. **Impact of host intraspecies genetic variation, diet, and age on bacterial and fungal intestinal microbiota in tigers**. *MicrobiologyOpen* (2020) **9**. DOI: 10.1002/mbo3.1050
15. Jiang F, Gao H, Qin W, Song P, Wang H, Zhang J, Liu D, Wang D, Zhang T. **Marked seasonal variation in structure and function of gut microbiota in forest and Alpine musk deer**. *Frontiers in Microbiology* (2021) **12** 699797. DOI: 10.3389/fmicb.2021.699797
16. Kaakoush NO. **Insights into the role of**. *Frontiers in Cellular and Infection Microbiology* (2015) **5** 84. DOI: 10.3389/fcimb.2015.00084
17. Kim CC, Healey GR, Kelly WJ, Patchett ML, Jordens Z, Tannock GW, Sims IM, Bell TJ, Hedderley D, Henrissat B, Rosendale DI. **Genomic insights from**. *The ISME Journal* (2019) **13** 1437-1456. DOI: 10.1038/s41396-019-0363-6
18. Kim YS, Unno T, Kim BY, Park MS. **Sex differences in gut microbiota**. *The World Journal of Men’s Health* (2020) **38** 48-60. DOI: 10.5534/wjmh.190009
19. Li B, Gao H, Song P, Liang C, Jiang F, Xu B, Liu D, Zhang T. **Captivity shifts gut microbiota communities in white-lipped deer (**. *Animals* (2022) **12** 432. DOI: 10.3390/ani12040431
20. Li Y, Shi M, Zhang T, Hu X, Zhang B, Xu S, Ding J, Hu D, Liu S. **Dynamic changes in intestinal microbiota in young forest musk deer during weaning**. *PeerJ* (2020) **8**. DOI: 10.7717/peerj.8923
21. Maloiy GM, Kay RN, Goodall ED, Topps JH. **Digestion and nitrogen metabolism in sheep and red deer given large or small amounts of water and protein**. *British Journal of Nutrition* (1970) **24** 843-855. DOI: 10.1186/1471-2180-9-123
22. Mariat D, Firmesse O, Levenez F, Guimaraes V, Sokol H, Dore J, Corthier G, Furet JP. **The**. *BMC Microbiology* (2009) **9** 123. DOI: 10.1186/1471-2180-9-123
23. Minich D, Madden C, Evans MV, Ballash GA, Barr DJ, Poulsen KP, Dennis PM, Hale VL. **Alterations in gut microbiota linked to provenance, sex, and chronic wasting disease in white-tailed deer (**. *Scientific Reports* (2021) **11** 13218. DOI: 10.1038/s41598-021-89896-9
24. Moeller AH, Sanders JG. **Roles of the gut microbiota in the adaptive evolution of mammalian species**. *Philosophical Transactions of the Royal Society B: Biological Sciences* (2020) **375** 20190597. DOI: 10.1098/rstb.2019.0597
25. Moyes K, Coulson T, Morgan BJT, Donald A, Morris SJ, Clutton-Brock TH. **Cumulative reproduction and survival costs in female red deer**. *Oikos* (2006) **115** 241-252. DOI: 10.1111/j.2006.0030-1299.15200.x
26. Oikonomou G, Teixeira AG, Foditsch C, Bicalho ML, Machado VS, Bicalho RC. **Fecal microbial diversity in pre-weaned dairy calves as described by pyrosequencing of metagenomic 16S rDNA. Associations of**. *PLOS ONE* (2013) **8**. DOI: 10.1371/journal.pone.0063157
27. Pan D, Yu Z. **Intestinal microbiome of poultry and its interaction with host and diet**. *Gut Microbes* (2014) **5** 108-119. DOI: 10.4161/gmic.26945
28. Passadore MD, Griggio MA, Nunes MT, Luz J. **Effects of ageing on the energy balance of food-restricted rats**. *Acta Physiologica Scandinavica* (2004) **181** 193-198. DOI: 10.1111/j.1365-201X.2004.01281.x
29. Qin S, He Z, Wu Y, Zeng C, Zheng Z, Zhang H, Lv C, Yuan Y, Wu H, Ye J, Liu Z, Shi M. **Instant dark tea alleviates hyperlipidaemia in high-fat diet-fed rat: from molecular evidence to redox balance and beyond**. *Frontiers in Nutrition* (2022) **9** 819980. DOI: 10.3389/fnut.2022.819980
30. Reading RP, Miller B, Shepherdson D. **The value of enrichment to reintroduction success**. *Zoo Biology* (2013) **32** 332-341. DOI: 10.1002/zoo.21054
31. Seddon PJ, Armstrong DP, Maloney RF. **Developing the science of reintroduction biology**. *Conservation Biology* (2007) **21** 303-312. DOI: 10.1111/j.1523-1739.2006.00627.x
32. Seddon PJ, Griffiths CJ, Soorae PS, Armstrong DP. **Reversing defaunation: restoring species in a changing world**. *Science* (2014) **345** 406-412. DOI: 10.1126/science.1251818
33. Shin NR, Whon TW, Bae JW. **Proteobacteria: microbial signature of dysbiosis in gut microbiota**. *Trends in Biotechnology* (2015) **33** 496-503. DOI: 10.1016/j.tibtech.2015.06.011
34. Sichert A, Corzett CH, Schechter MS, Unfried F, Markert S, Becher D, Fernandez-Guerra A, Liebeke M, Schweder T, Polz MF, Hehemann JH. **Verrucomicrobia use hundreds of enzymes to digest the algal polysaccharide fucoidan**. *Nature Microbiology* (2020) **5** 1026-1039. DOI: 10.1038/s41564-020-0720-2
35. Sun CH, Liu HY, Liu B, Yuan BD, Lu CH. **Analysis of the gut microbiome of wild and captive Pere David’s deer**. *Frontiers in Microbiology* (2019) **10** 2331. DOI: 10.3389/fmicb.2019.02331
36. Sutton AE. **Leadership and management influences the outcome of wildlife reintroduction programs: findings from the sea eagle recovery project**. *PeerJ* (2015) **3**. DOI: 10.7717/peerj.1012
37. Turnbaugh PJ, Ley RE, Mahowald MA, Magrini V, Mardis ER, Gordon JI. **An obesity-associated gut microbiome with increased capacity for energy harvest**. *Nature* (2006) **444** 1027-1031. DOI: 10.1038/nature05414
38. Von Klitzing E, Ekmekciu I, Bereswill S, Heimesaat MM. **Intestinal and systemic immune responses upon multi-drug resistant**. *Frontiers in Microbiology* (2017) **8** 2590. DOI: 10.3389/fmicb.2017.02590
39. Wang L, Ding J, Yang Z, Chen H, Yao R, Dai Q, Ding Y, Zhu L. **Pere David’s deer gut microbiome changes across captive and translocated populations: implications for conservation**. *Evolutionary Applications* (2019) **12** 622-635. DOI: 10.1111/eva.12743
40. Wei W, Jiang W, Tian Z, Wu H, Ning H, Yan G, Zhang Z, Li Z, Dong F, Sun Y, Li Y, Han T, Wang M, Sun C. **Fecal g. Streptococcus and**. *Clinical Nutrition* (2021) **40** 4234-4245. DOI: 10.1016/j.clnu.2021.01.031
41. Williams DA, Rains ND, Hale AM. **Population genetic structure of Texas horned lizards: implications for reintroduction and captive breeding**. *PeerJ* (2019) **7**. DOI: 10.7717/peerj.7746
42. Wu H, Wu FT, Zhou QH, Zhao DP. **Comparative analysis of gut microbiota in captive and wild oriental white storks: implications for conservation biology**. *Frontiers in Microbiology* (2021) **12** 649466. DOI: 10.3389/fmicb.2021.649466
43. Yang ZS, Gu XD, Nie YG, Huang F, Huang Y, Dai Q, Hu YB, Yang Y, Zhou X, Zhang HM, Yang XY, Wei FW. **Reintroduction of the giant panda into the wild: a good start suggests a bright future**. *Biological Conservation* (2018) **217** 181-186. DOI: 10.1016/j.biocon.2017.08.012
44. You Z, Deng J, Liu J, Fu J, Xiong H, Luo W, Xiong J. **Seasonal variations in the composition and diversity of gut microbiota in white-lipped deer (**. *PeerJ* (2022) **10**. DOI: 10.7717/peerj.13753
45. Zened A, Combes S, Cauquil L, Mariette J, Klopp C, Bouchez O, Troegeler-Meynadier A, Enjalbert F. **Microbial ecology of the rumen evaluated by 454 GS FLX pyrosequencing is affected by starch and oil supplementation of diets**. *FEMS Microbiology Ecology* (2013) **83** 504-514. DOI: 10.1111/1574-6941.12011
46. Zhang M, Shi M, Fan M, Xu S, Li Y, Zhang T, Cha M, Liu Y, Guo X, Chen Q, Li Y, Zhang S, Hu D, Liu S. **Comparative analysis of gut microbiota changes in Pere David’s deer populations in Beijing Milu Park and Shishou, Hubei province in China**. *Frontiers in Microbiology* (2018) **9** 1258. DOI: 10.3389/fmicb.2018.01258
47. Zhang YP, Zhao HW, Li QQ, Tsechoe D, Yuan HL, Su GJ, Yang JS. **Environmental factors influence yak milk composition by modulating short-chain fatty acid metabolism in intestinal microorganisms**. *LWT-Food Science and Technology* (2022) **163** 113608. DOI: 10.1016/j.lwt.2022.113608
48. Zhao G, Ma T, Tang W, Li D, Mishra SK, Xu Z, Wang Q, Jie H. **Gut microbiome of Chinese forest musk deer examined across gender and age**. *BioMed Research International* (2019) **2019** 9291216. DOI: 10.1155/2019/9291216
49. Zhao W, Ren Z, Luo Y, Cheng J, Wang J, Wang Y, Yang Z, Yao X, Zhong Z, Yang W, Wu X. **Metagenomics analysis of the gut microbiome in healthy and bacterial pneumonia forest musk deer**. *Genes & Genomics* (2021) **43** 43-53. DOI: 10.1007/s13258-020-01029-0
50. Zhen J, Ren Y, Zhang H, Yuan X, Wang L, Shen H, Liu P, Chen Y. **Effect of different dietary regimes on the gut microbiota and fecal metabolites of Pere David’s deer**. *Animals* (2022) **12** 584. DOI: 10.3390/ani12050584
51. Zheng Y, Dzakpasu M, Wang X, Zhang L, Ngo HH, Guo W, Zhao Y. **Molecular characterization of long-term impacts of macrophytes harvest management in constructed wetlands**. *Bioresource Technology* (2018) **268** 514-522. DOI: 10.1016/j.biortech.2018.08.030
52. Zhu Z, Sun Y, Zhu F, Liu Z, Pan R, Teng L, Guo S. **Seasonal variation and sexual dimorphism of the microbiota in wild blue sheep (**. *Frontiers in Microbiology* (2020) **11** 1260. DOI: 10.3389/fmicb.2020.01260
|
---
title: Alpha-Mangosteen lessens high-fat/high-glucose diet and low-dose streptozotocin
induced-hepatic manifestations in the insulin resistance rat model
authors:
- Vivian Soetikno
- Prisma Andini
- Miskiyah Iskandar
- Clark Christensen Matheos
- Joshua Alward Herdiman
- Iqbal Kevin Kyle
- Muhammad Nur Imaduddin Suma
- Melva Louisa
- Ari Estuningtyas
journal: Pharmaceutical Biology
year: 2023
pmcid: PMC9969969
doi: 10.1080/13880209.2023.2166086
license: CC BY 4.0
---
# Alpha-Mangosteen lessens high-fat/high-glucose diet and low-dose streptozotocin induced-hepatic manifestations in the insulin resistance rat model
## Abstract
### Context
α-Mangosteen (α-MG) attenuates insulin resistance (IR). However, it is still unknown whether α-MG could alleviate hepatic manifestations in IR rats.
### Objective
To investigate the effect of α-MG on alleviating hepatic manifestations in IR rats through AMP-activated protein kinase (AMPK) and sterol-regulatory element-binding protein-1 (SREBP-1) pathway.
### Materials and methods
IR was induced by exposing male Sprague-Dawley rats (180–200 g) to high-fat/high-glucose diet and low-dose injection of streptozotocin (HF/HG/STZ), then treated with α-MG at a dose of 100 or 200 mg/kg/day for 8 weeks. At the end of the study (11 weeks), serum and liver were harvested for biochemical analysis, and the activity of AMPK, SREBP-1c, acetyl-CoA carboxylase (ACC), tumor necrosis factor (TNF)-α, interleukin (IL)-1β, IL-6, insulin receptor substrate (IRS)-1, Bax and liver histopathology were analyzed.
### Results
α-MG at both doses significantly lowered ALT, AST, triglyceride, and cholesterol total by 16.5, 15.7, 38, and $36\%$, respectively. These beneficial effects of α-MG are associated with the downregulation of the IR-induced inflammation in the liver. Furthermore, α-MG, at both doses, activated AMPK by 24–29 times and reduced SREBP-1c by 44–$50\%$ as well as ACC expression by 19–$31\%$ similar to metformin. All treatment groups showed liver histopathology improvement regarding fat deposition in the liver.
### Conclusions
Based on the findings demonstrated, α-MG protected against HF/HG/STZ-induced hepatic manifestations of the IR rats, at least in part via the modulation of the AMPK/SREBP-1c/ACC pathway and it could be a potential drug candidate to prevent IR-induced hepatic manifestations.
## Introduction
Modern lifestyles with excessive food intake, especially high-fat and high-carbohydrate foods that are not balanced with energy expenditure, will trigger the occurrence of a metabolic syndrome which is currently becoming a global epidemic (Saklayen 2018). In Indonesia, the prevalence of the metabolic syndrome is 28 and $46\%$ in men and women, respectively (Sigit et al. 2021). It has been proven that insulin resistance (IR) contributes to the pathophysiology of metabolic syndrome (Alwahsh et al. 2017). Many studies have also demonstrated that IR is a high-risk factor for the occurrence of several chronic diseases such as type 2 diabetes mellitus, non-alcoholic fatty liver disease, and cardiovascular disease (Manco 2017; Mechanick et al. 2020). Therefore, the development of targeted therapies to treat IR is urgently needed, so that metabolic syndrome with its complications can be prevented.
Previous studies have revealed that AMP-activated protein kinase (AMPK), which is a master regulator of energy metabolism, plays a vital role in several cellular events, namely, protein synthesis, lipid metabolism, and glucose metabolism (Hardie et al. 2012; Garcia and Shaw 2017; Entezari et al. 2022). AMPK is a kinase that directly targets sterol regulating element-binding proteins (SREBPs), inhibit cleavage and nuclear translocation of SREBPs, as well as suppresses SREBPs target genes, including acetyl-CoA carboxylase (ACC) in hepatocytes exposed to high glucose levels, thereby reducing lipogenesis (Li et al. 2011).
Metformin, an AMPK activator, may decrease systemic inflammation in metabolic syndrome by decreasing C-reactive protein and interleukin (IL)-6 levels (Akbar 2003). Metformin can also alleviate liver disease and dyslipidemia due to a high-fat diet (HFD) intake by stimulating antioxidant and anti-inflammatory pathways (Yasmin et al. 2021). A previous study has also revealed that metformin can help protect pancreatic β-cells from exhaustion and decompensation and improved lipid metabolism, though it did not significantly improve IR (Sun et al. 2019; Huang et al. 2021). Until now, metformin is widely used as a drug that can increase insulin sensitivity through various mechanisms, such as decreasing hepatic glucose production and increasing glucose disposal in peripheral tissues (Wróbel et al. 2017). Besides having beneficial effects, metformin also has adverse effects that are often complained of, such as gastrointestinal disorders, vitamin B12 deficiency, and lactic acidosis (Shurrab and Arafa 2020). Therefore, drug candidates are needed that can also be used to alleviate IR-induced hepatic manifestations with the hope of well-tolerated adverse effects.
α-Mangostin (α-MG) is the main active compound of the *Garcinia mangostana* L. (Clusiaceae) fruit which is widely found in Asian countries, including Indonesia; it has been shown that α-MG exhibits anti-inflammatory, antioxidant, antitumor, anti-aging, and other biological activities (Fang et al. 2016; Lee et al. 2018; Ratwita et al. 2018). Several studies have also revealed that α-MG has various beneficial effects, including improving IR by increasing the expression of GLUT-4 and PPARγ in cardiac muscle and adipocyte (Ratwita et al. 2018, 2019), and reduced hepatic steatosis by regulating mitochondria function and apoptosis (Tsai et al. 2016). Furthermore, α-MG has been reported to increase AMPK activity and improve pulmonary fibrosis (Li et al. 2019), has neuroprotection against rotenone-induced Parkinson’s disease (Parkhe et al. 2020), suppress de novo lipogenesis and increase the gemcitabine response in the gallbladder carcinoma cells (Shi et al. 2020), and reduce leptin levels in olanzapine-induced metabolic disorders (Ardakanian et al. 2022). In the current study, we evaluate the effect of α-MG on hepatic manifestations of IR in rats via the AMPK/SREBP-1/ACC pathway and compare it with metformin, a well-known AMPK activator.
## Drugs and chemicals
Unless otherwise stated, all reagents were of analytical grade and purchased from Sigma-Aldrich, Indonesia. α-MG was purchased from Aktin Chemicals, Inc., Chengdu, China (batch #AM-170622, and ≥$98\%$ purity with HPLC method), whereas streptozotocin (STZ) was purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA (catalog #sc-200719A)).
## Insulin resistance induction
IR was induced by giving a HFD (TestDiet, 58V8 rat chow, Richmond, USA) which contain by weight $46.1\%$ fat, $35.8\%$ carbohydrate, and $18.1\%$ protein, with a total energy of 4.60 kcal/g, combined with $20\%$ high-glucose (HG) drinking water for 3 weeks and a single intraperitoneal (i.p.) injection of STZ at a dose of 35 mg/kg body weight to 8- to 10-week-old male Wistar rats, which were obtained from the Laboratory Animal Center of Litbangkes, Indonesia. STZ was dissolved in 0.01 M citrate buffer (pH 4.5) and injected within 5 min of preparation. Age-matched male Wistar rats were injected with 10 µL of citrate buffer and used as non-IR-normal rats (Soetikno et al. 2020). The animals were maintained with free access to water and chow throughout the study period and were treated in accordance with the Institute of Animal Studies Ethics Committee regulations approved by our institute (ethical clearance #KET-1188/UN2.F1/ETIK/PPM$\frac{.00.02}{2020}$) and comply with the ARRIVE guideline. All efforts were made to minimize suffering.
## Experimental protocol
At 72 h after the STZ injection, the fasting blood glucose (FBG) of the rats was measured using a Glucometer 4 Accu-check, USA. The animals with FBG levels of ≥250 mg/dL were included in the study. If the FBG had not reached the target, then a second injection of STZ with a half-dose repeated once. As a result, 36 rats were divided into the following six groups ($$n = 6$$): [1] normal-control rats (N), [2] normal-control rats received α-MG 200 mg/kg/day (N + α-MG 200), [3] vehicle-treated HF/HG/STZ-induced IR rats (IR), [4] metformin 200 mg/kg/day-treated IR rats (IR + Met), [5] α-MG 100 mg/kg/day-treated IR rats (IR + α-MG 100), and [6] α-MG 200 mg/kg/day-treated IR rats (IR + α-MG 200). We used α-MG doses of 100 mg/kg and 200 mg/kg based on our previous study which proved that α-MG at both doses was able to increase antioxidant activity in the heart tissue of IR rats (Lazarus et al. 2020; -MG was dissolved in 1 mL of corn oil, whereas metformin was dissolved in 1 mL of $0.5\%$ carboxymethyl cellulose and both treatments were administered orally for 8 weeks. At week 11 after the HF/HG/STZ and α-MG administration, the rats were anesthetized with a single i.p. injection of ketamine/xylazine 0.15 mL/100 g body weight and euthanized by cervical dislocation. Their liver was excised and weighed and the blood was collected, then the serum was taken and centrifuged at 1000 g, 4 °C, 10 min and used for biochemical parameter analyses. Half of the liver was immediately snap-frozen in liquid nitrogen and stored at −80 °C until the subsequent protein analysis. The remaining excised liver was fixed in $10\%$ formalin and used for histopathological studies.
## Estimation of biochemical parameters
At the end of the experimental period (11 weeks), all rats have fasted for 12 h, then the blood was taken from the tail vein and collected into serum separator tubes. After the blood was allowed to sit for 30 min at room temperature was centrifuge at 1000 g, 10 min, 4 °C for separation of serum. Serum was stored at −80 °C until assays were performed. The serum used for the estimation of aspartate aminotransferase (AST), alanine aminotransferase (ALT), total cholesterol, and triglyceride was measured using the kits from DiaSys Diagnostic Systems GmbH (Holzheim Germany), then read using a UV/VIS spectrophotometer (PerkinElmer).
## Determination of lipid content in the liver
Hepatic lipid was extracted using a lipid extraction kit from Cell Biolabs, Inc., San Diego, CA, USA (catalog #STA-612) and were quantified using a lipid quantification kit from Cell Biolabs, Inc. (catalog #STA-613) (Seki et al. 2022).
## Gene expression analysis
Total RNA was extracted from liver tissue using a High Pure RNA Isolation kit (Roche Applied Science, Penzberg, Germany) according to the manufacturer’s instructions. RNA concentration was measured using a Nanodrop 1000 Spectrophotometer (Thermo Scientific) at a wavelength of 260 nm, followed by complementary DNA synthesis using 1 µg of RNA and the Transcriptor First Strand cDNA Synthesis kit (Roche Applied Science). The expression of genes of interest (Table 1) was analyzed using quantitative real-time PCR (qRT-PCR) using the LightCycler® 480 Instrument (Roche Applied Science) with FastStart Essential DNA Green Master Miz (Roche Life Science). All reactions were performed in the same manner: 95 °C for 10 s, followed by 45 cycles of 95 °C for 15 s and 60 °C for 1 min, and all values were normalized to the level of β-actin. The results were analyzed using the Livak method (Livak and Schmittgen 2001).
**Table 1.**
| Gene | Primers | Primers.1 |
| --- | --- | --- |
| β-actin | Forward | 5′-CTGGTCGTACCACAGGCATT-3′ |
| | Reverse | 5′-CTCTTTGATGTCACGCACGA-3′ |
| TNF-α | Forward | 5′-TCTACTCCCAGGTTCTCTTCA-3′ |
| | Reverse | 5′-CTCCTGGTATGAAATGGCAAATC-3′ |
| IL-1β | Forward | 5′-CTTGGGACTGATGCTGGTGA-3′ |
| | Reverse | 5′-TGCAAGTGCATCATCGTTGT-3′ |
| Bax | Forward | 5′-AGGGTGGCTGGGAAGGC-3′ |
| | Reverse | 5′-TGAGCGAGGCGGTGAGG-3′ |
## Estimation of acetyl-CoA carboxylase (ACC), sterol regulatory element-binding protein 1C (SREBP-1C), insulin receptor substrate 1 (IRS-1), and phosphoinositide-3-kinase (PI3K) in liver tissue
Sandwich-ELISA kits for ACC, SREBP-1c, IRS-1, and PI3K (MyBioSource catalog #MBS8303295, #MBS940898, #MBS774816, and #MBS260381, USA, respectively) were performed to analyze the effect of α-MG on liver tissues.
## Western blot analysis for AMPK in liver tissue
The total protein concentration from liver tissues were measured by the bicinchoninic acid method. To determine the protein levels of AMPKα and phospho-AMPKα, equal amounts of protein extract (50 µg) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Bio-Rad, CA, USA) and transferred electrophoretically to nitrocellulose membranes. The membranes were then blocked with $5\%$ bovine serum albumin in Tris-buffered saline Tween (20 mM Tris, (pH 7.6), 137 mM NaCl, and $0.1\%$ Tween 20). The primary antibodies against AMPKα and phospho-AMPKα were obtained from Cell Signalling Technology, Inc. (Beverly, MA, USA) with the catalog number #CST-2532 and #CST-2535, respectively. All of the antibodies were used at a dilution of 1:1000. The membrane was incubated overnight at 4 °C with the primary antibody, and the bound antibody was visualized using the respective horseradish peroxidase-conjugated secondary antibodies (Cell Signalling Technology, Inc. (Beverly, MA, USA) catalog #CST-5127)) and chemiluminescence developing agents (Amersham Biosciences, Buckinghamshire, UK).
## Light microscopic morphological study
The samples of liver tissue were fixed with $4\%$ paraformaldehyde for 24 h, then embedded in paraffin. Sections (5 µm thick) were stained with hematoxylin-eosin (H&E) to observe lipid accumulation in the liver.
Formalin-fixed, paraffin-embedded liver tissue sections were used for the immunohistochemical staining. After being deparaffinized and hydrated, the slides were washed in Tris-buffered saline (TBS, 10 mmol/L Tris; $0.85\%$ NaCl, pH 7.5) containing $10\%$ bovine serum albumin. Endogenous peroxidase activity was quenched by incubating the slides in methanol and $0.6\%$ H2O2/methanol. For antigen retrieval, the sections were pretreated with trypsin for 15 min at 37 °C. Blocking was performed with normal rabbit serum. After overnight incubation with an anti-4-hydroxynonenal (4-HNE) antibody (Bioss, Inc., USA catalog #bs-6313-R) (1:50 dilution) at 4 °C, the slides were washed in TBS buffer, and HRP conjugated secondary antibody (Bioss, Inc. catalog #bs-0295M)) was added and incubated in the room temperature for 45 min. The immunostaining was visualized using diaminobenzidine tetrahydrochloride, and the slides were counterstained with hematoxylin. Measurement of 4-HNE was made by counting the mean number of stained cells under 400-fold magnification using a light microscope. For all sections, 25 random fields were examined per section, and 3 animals were used per group.
## Statistical analysis
Data were represented as means ± standard error of the mean (SEM). Statistical analysis of the differences between groups were performed using the One-way ANOVA followed by post hoc Tukey’s test using GraphPad Prism 5.0 software. Differences were considered significant at $p \leq 0.05.$
## Roles of α-MG on the general appearance and serum parameters in HF/HG/STZ-induced IR rats
At 11 weeks after HF/HG/STZ administration, the body weights (BW) of vehicle-treated IR rats were decreased considerably as compared to normal rats. In contrast, the liver weights (LW) and the LW/BW ratio of vehicle-treated IR rats, which is a surrogate marker to assess the effects of xenobiotics on specific organs, were significantly increase when compared to respective controls. Interestingly, α-MG in both doses and metformin administration could decrease LW and LW/BW ratio significantly as compared to the vehicle-treated IR rats by 19.2–$24.4\%$ and 6.9–$21.6\%$, respectively (Table 2). In addition, administration of HF/HG/STZ leads to a pronounced increase in serum total cholesterol and triglyceride in the vehicle-treated IR rats as compared to the respective controls. On the other hand, α-MG in both doses and metformin treatment significantly suppressed the increase in both serum total cholesterol and triglyceride by 16.5–$32\%$ and 15.7–$40.3\%$, respectively (Figure 1(A,B)). To investigate the protective effects of α-MG on HF/HG/STZ-induced IR on liver function, we measured AST and ALT in serum. The α-MG-treated rats displayed reduced AST and ALT levels compared with the vehicle-treated IR rats (Figure 1(C,D)). Next, we measured the IRS1 and PI3K in the liver tissues in all groups of experimental animals. We showed that the protein expression of IRS1 was markedly reduced in vehicle-treated IR rats compared to normal and normal-treated α-MG rats. However, there was no significant difference in the protein expression of PI3K between vehicle-treated IR rats and normal as well as normal-treated α-MG rats. The administration of α-MG at both doses could increase the protein expression of IRS1 and PI3K (Figure 1(E,F)).
**Figure 1.:** *α-MG protected against high-fat/high-glucose/low-dose streptozotocin (HF/HG/STZ) induced-insulin resistance (IR) rats on hepatic manifestations such as the reduced liver lipid levels, improved liver function tests, and improved protein expression of insulin receptor substrate (IRS)-1 and phosphoinositide 3-kinase (PI3K) in the liver tissues of IR rats. The levels of liver cholesterol total (A), liver triglyceride (B), AST level in serum (C), ALT level in serum (D), protein expression of IRS-1 in the liver tissues (E), and protein expression of PI3K in the liver tissues (F). N: normal group, N + α-MG 200: normal-treated α-MG at a dose of 200 mg/kg/day group, IR: vehicle-treated insulin resistance (IR) group, IR + Met: metformin-treated insulin resistance group, IR + α-MG 100: α-MG at a dose of 100 mg/kg/day-treated insulin resistance group, IR + α-MG 200: α-MG at a dose of 200 mg/kg/day-treated insulin resistance group. Values are mean ± SD ($$n = 6$$). **$p \leq 0.01$ vs. N; §§$p \leq 0.01$ vs. N + α-MG 200; ††$p \leq 0.01$ vs. IR.* TABLE_PLACEHOLDER:Table 2.
## Effects of α-MG on the mRNA expression of inflammatory cytokines
Induction of inflammation has been implicated in IR-mediated hepatic disease, which was confirmed by the upregulated gene expression levels of Bax and IL-1β as well as upregulated protein expression levels of TNF-α and IL-6 in the vehicle-treated IR rats. The α-MG-treated IR rats markedly reduced the gene and protein expression of those inflammatory cytokines in their liver (Figure 2(A–E)).
**Figure 2.:** *α-MG regulates high-fat/high-glucose/low-dose streptozotocin (HF/HG/STZ) induced-IR rats on inflammation process in the liver tissues. Effect of α-MG on: (A) gene expression of Bax in the liver tissues, (B) gene expression of TNF- α in the liver tissues, (C) protein expression of TNF- α in the liver tissues, (D) gene expression of IL-1β in the liver tissues, and (E) protein expression of IL-6 in the liver tissues. Values are mean ± SD (n = 6). *p < 0.05 vs. N; **p < 0.01 vs. N; §p < 0.05 vs. N + α-MG 200; §§p < 0.01 vs. N + α-MG 200; †p < 0.05 vs. IR; ††p < 0.01 vs. IR.*
## α-MG abrogates the downregulation of p-AMPK and its downstream proteins
It has been reported that the downregulation AMPK causes an increase in the activity of SREBP-1c and ACC which in turn leads to enhanced production and impaired catabolism of lipid and glucose in the liver which contributes to IR and dyslipidemia. In this study, we also observed a downregulated AMPK phosphorylation in the liver of vehicle-treated IR rats (Figure 3(A)). Furthermore, the vehicle-treated IR rats displayed increased SREBP-1c and ACC protein expressions. The α-MG-treated IR rats at both doses demonstrated significantly attenuated phosphorylation of AMPK and decreased SREBP-1c as well as ACC protein expressions in their liver (Figure 3(A–C)).
**Figure 3.:** *α-MG increased the expression of AMPK and decreased the expression of SREBP-1 and ACC. (A) Western blots p-AMPK and its relative contents, (B) protein expression of SREBP-1c in the liver tissues, and (C) protein expression of ACC in the liver tissues. Values are mean ± SD (n = 6). *p < 0.05 vs. N; **p < 0.01 vs. N; §p < 0.05 vs. N + α-MG 200; §§p < 0.01 vs. N + α-MG 200; ††p < 0.01 vs. IR.*
## α-MG reduces lipid accumulation and lipid quantification in the liver of HF/HG/STZ-induced IR rats
HE staining of paraffinized sections showed that the lipid accumulation in the vehicle-treated IR liver rats were more pronounced than the normal liver tissues, while lipid accumulation was markedly reduced in the groups administered with α-MG (Figure 4(A–F)). In line with HE staining, lipid quantification also showed that α-MG administration (100 and 200 mg/kg) significantly reduced lipid accumulation in the liver tissues (Figure 4(G)).
**Figure 4.:** *(A–F) Hematoxylin & Eosin staining of liver tissue. Lipid accumulation was marked in the IR group compared to other groups. Treatment with α-MG at 100 and 200 mg/kg/day markedly suppressed lipid accumulation (×400 magnification), (G) lipid quantification analysis in the liver tissues, (H–M) immunohistochemical staining with 4-HNE show lipid peroxidation with stained positive for 4-HNE (arrows). Treatment with α-MG at 100 and 200 mg/kg/day reduced 4-HNE (×200 magnification).*
## α-MG reduces 4-HNE expression in the liver of HF/HG/STZ-induced IR rats
We performed IHC to measure the protein expression of 4-HNE, one of the stable and reliable markers of lipid peroxidation in the liver, in the normal, vehicle, and α-MG-treated IR rats. The vehicle-treated IR liver rats were found to have increased expression of 4-HNE (Figure 4(J)) in comparison to normal and α-MG-treated normal rats (Figure 4(H,I)). The metformin and α-MG administration at both doses attenuated the increased expression of 4-HNE in the liver of IR rats (Figure 4(K–M)).
## Discussion
In the Western diet, long-term consumption of high-fat and high-carbohydrate foods can lead to various diseases including obesity, IR, hypertriglyceridemia, NAFLD, and NASH, all of which are associated with inflammatory and increased oxidative stress conditions (Kim et al. 2021). Our previous study has indicated that the administration of a high-fat and high-carbohydrate diet which contains by weight $46.1\%$ fat, $35.8\%$ carbohydrate, and $18.1\%$ protein, can cause IR in rats, which is characterized by an increase in Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) and FBG levels (Soetikno et al. 2020). In the present study, we found that IR also causes lipid accumulation and lipid peroxidation in the liver. Treatment with α-MG at both doses significantly attenuated those alterations in the liver of rats as evidenced by decreased lipid droplets, decreased protein expression of 4-HNE, as well as attenuated inflammation in the liver, at least by modulating the AMPK/SREBP-1c/ACC cascade.
AMPK is the most important regulator of energy metabolism homeostasis both at the cellular and whole organism level, besides that AMPK also plays a crucial role in various physiological processes, including modulating inflammation and increasing insulin sensitivity in peripheral tissues (Herzig and Shaw 2018). The pharmacological function of α-MG is to activate AMPK has been revealed in various physiological processes. α-MG decreased oxidative stress in the lungs partly by activating the AMPK mediated signalling pathway (Li et al. 2019). In addition, it has been known that α-MG was able to restore leptin levels and oxidative stress in olanzapine-induced metabolic disorders in rats by increasing AMPK phosphorylation in liver (Ardakanian et al. 2022). A previous study has shown that under conditions of oxidative stress, lipid peroxidation of polyunsaturated fatty acids lead to the production of 4-HNE, which can further impair insulin action in muscle cells (Pillon et al. 2012). In accordance with previous studies, we demonstrated that α-MG at both doses increases AMPK activity and reduces oxidative stress, comparable to metformin, as shown by decreased protein expression of 4-HNE in the liver tissues of HF/HG/STZ-induced IR rats.
One of the important features of IR is the contribution of inflammation. It has been reported that increased activity of AMPK can prevent and/or ameliorate type 2 diabetes mellitus and IR by inhibiting inflammatory response (Behrouz et al. 2020; Yap et al. 2020). AMPK impedes the inflammatory process by directly inhibiting the NF-κB signalling pathway and inhibiting IKK phosphorylation (Chen et al. 2018). It has also been observed that metformin, an AMPK agonist, can diminish systemic inflammation by lessening the activity of PPARγ and normalized the SREBP-1c and fatty acid synthase mRNA in the liver of HFD fed rats (Yasmin et al. 2021). In this study, we found that α-MG as well as metformin can decrease the inflammatory response through AMPK activation and further decrease proinflammatory cytokines such as TNF-α, IL-1β, IL-6, and Bax.
It has been shown that AMPK downregulation leads to an augmentation of SREBP1c and ACC genes expression which will further increase cholesterol and fatty acid biosynthesis (Kobayashi et al. 2018). Our data demonstrated that HF/HG/STZ-mediated downregulated of AMPK and upregulated SREBP-1c and ACC protein expressions were significantly suppressed after α-MG treatment. Thus, in our study, the administration of α-MG can reduce cholesterol and triglyceride levels. Shi et al. [ 2020], found that in gallbladder carcinoma cells, α-MG was effective in inhibiting lipogenesis via targeting the AMPK/SREBP1. A previous study has also indicated that in HFD-fed rats, the increase in plasma cholesterol and triglyceride lead to lipid accumulation in the liver (Shin et al. 2020). In line with the previous study (Shin et al. 2020), we also demonstrated that lipid accumulation in the liver occurred in HFD-fed rats and administration of α-MG was able to reduce those conditions. In addition, we also showed that in HFD-fed rats there was an increase in AST and ALT levels; and interestingly, α-MG administration as well as metformin were able to reduce these levels to normal values. This result is in accordance with the previous study which proved that AST and ALT are associated with hyperinsulinemia and IR, independent of obesity (Esteghamati et al. 2011).
Next, we evaluated the insulin signalling pathway and its relationship to AMPK. A previous study has shown that insulin signalling and AMPK show vital roles in balancing intracellular energy levels and glucose uptake, in which both pathways stimulate energy conservation and survival of muscle exposed to severe glucose deprivation (Chopra et al. 2012). It has been demonstrated that insulin signaling is mediated by IRS protein. IRS1 is the most common and widely expressed in many tissues including the liver. Metabolic effects of insulin downstream of IRS proteins are mediated by the PI3K (Gallagher et al. 2012). In the present study, HF/HG/STZ administration was able to downregulate the level of IRS1 and PI3K as compared to that of normal rats, while the administration of α-MG at both doses and metformin showed an increase in IRS1 and PI3K. This result is consistent with evidence that α-MG activates insulin signalling and protects pancreatic β-cells against STZ-induced apoptotic damage (Lee et al. 2018).
This study demonstrates that α-MG improves the manifestations of hepatic dysfunction due to IR in rats, by modulating the AMPK/SREBP-1c/ACC signalling pathways. In this study, we also noticed that oral administration of α-MG doses of 100 mg/kg and 200 mg/kg did not show significantly different effectiveness, this may be due to the low oral bioavailability of α-MG because it undergoes extensive first-pass metabolism in the liver (Li et al. 2011). Though in this study α-MG showed promising results in preventing IR-induced hepatic dysfunction, there are several limitations. First, in the present study, the time to induce IR to lipid accumulation in the liver with a HFD was only 11 weeks, which is most likely if the time for induction is extended, lipid accumulation in the liver will be more extensive. Secondly, the use of experimental animals to induce IR by giving a HFD and STZ injection does not always resemble the conditions that actually occur in humans.
## Conclusions
This current study aimed to evaluate the effectiveness of α-MG as compared to metformin against impaired liver function in IR rats. The results of this study showed that α-MG at a dose of 100 mg/kg and 200 mg/kg is comparable to metformin and is able to prevent impaired liver function by at least increased AMPK activity and decreased expression of SREBP-1c and ACC, and as a result, attenuated IR-mediated inflammation, and lipid peroxidation, and lipid accumulation in liver tissues. Given the promising results of this study, α-MG may be a candidate for therapy to treat liver disorders caused by IR, and may be used as a substitute for metformin, although further subclinical and clinical studies are needed to prove this.
## Author contributions
Study design: Vivian Soetikno; Data collection: Prisma Andini, Miskiyah Iskandar, Clark Christensen Matheos, Joshua Alward Herdiman, Iqbal Kevin Kyle, Muhammad Nur Imaduddin Suma; Statistical analysis: Melva Louisa, Vivian Soetikno; Data Interpretation: Vivian Soetikno, Prisma Andini, Miskiyah Iskandar, Clark Christensen Matheos, Joshua Alward Herdiman, Iqbal Kevin Kyle, Muhammad Nur Imaduddin Suma, Melva Louisa, Ari Estuningtyas; Manuscript preparation: Vivian Soetikno; Literature search: Vivian Soetikno, Prisma Andini, Miskiyah Iskandar, Clark Christensen Matheos, Joshua Alward Herdiman, Iqbal Kevin Kyle, Muhammad Nur Imaduddin Suma, Melva Louisa, Ari Estuningtyas; Funds collection: Vivian Soetikno
## Disclosure statement
No potential conflict of interest was reported by the author(s).
## References
1. Akbar DH.. **Effect of metformin and sulfonylurea on C-reactive protein level in well-controlled type 2 diabetics with metabolic syndrome**. *Endocrine* (2003) **20** 215-218. PMID: 12721499
2. Alwahsh SM, Dwyer BJ, Forbes S, Thiel DH, Lewis PJ, Ramadori G.. **Insulin production and resistance in different models of diet-induced obesity and metabolic syndrome**. *IJMS* (2017) **18** 285. PMID: 28134848
3. Ardakanian A, Ghasemzadeh Rahbardar M, Omidkhoda F, Razavi B, Hosseinzadeh H.. **Effect of alpha-mangostin on olanzapine-induced metabolic disorders in rats**. *Iranian J Basic Med Sci* (2022) **25** 198-207
4. Behrouz V, Dastkhosh A, Hedayati M, Sedaghat M, Sharafkhah M, Sohrab G.. **The effect of crocin supplementation on glycemic control, insulin resistance and active AMPK levels in patients with type 2 diabetes: a pilot study**. *Diabetol Metab Syndr* (2020) **12** 59. PMID: 32670418
5. Chen X, Li X, Zhang W, He J, Xu B, Lei B, Wang Z, Cates C, Rousselle T, Li J.. **Activation of AMPK inhibits inflammatory response during hypoxia and reoxygenation through modulating JNK-mediated NF-κB pathway**. *Metabolism* (2018) **83** 256-270. PMID: 29526538
6. Chopra I, Li HF, Wang H, Webster KA.. **Phosphorylation of the insulin receptor by AMP-activated protein kinase (AMPK) promotes ligand-independent activation of the insulin signalling pathway in rodent muscle**. *Diabetologia* (2012) **55** 783-794. PMID: 22207502
7. Entezari M, Hashemi D, Taheriazam A, Zabolian A, Mohammadi S, Fakhri F, Hashemi M, Hushmandi K, Ashrafizadeh M, Zarrabi A. **AMPK signaling in diabetes mellitus, insulin resistance and diabetic complications: a pre-clinical and clinical investigation**. *Biomed Pharmacother* (2022) **146** 112563. PMID: 35062059
8. Esteghamati A, Noshad S, Khalilzadeh O, Khalili M, Zandieh A, Nakhjavani M.. **Insulin resistance is independently associated with liver aminotransferases in diabetic patients without ultrasound signs of nonalcoholic fatty liver disease**. *Metab Syndr Relat Disord* (2011) **9** 111-117. PMID: 21091060
9. Fang Y, Su T, Qiu X, Mao P, Xu Y, Hu Z, Zhang Y, Zheng X, Xie P, Liu Q.. **Protective effect of alpha-mangostin against oxidative stress induced-retinal cell death**. *Sci Rep* (2016) **6** 21018. PMID: 26888416
10. Gallagher EJ, Fierz Y, Vijayakumar A, Haddad N, Yakar S, LeRoith D.. **Inhibiting PI3K reduces mammary tumor growth and induces hyperglycemia in a mouse model of insulin resistance and hyperinsulinemia**. *Oncogene* (2012) **31** 3213-3222. PMID: 22037215
11. Garcia D, Shaw RJ.. **AMPK: mechanisms of cellular energy sensing and restoration of metabolic balance**. *Mol Cell* (2017) **66** 789-800. PMID: 28622524
12. Hardie DG, Ross FA, Hawley SA.. **AMPK: a nutrient and energy sensor that maintains energy homeostasis**. *Nat Rev Mol Cell Biol* (2012) **13** 251-262. PMID: 22436748
13. Herzig S, Shaw RJ.. **AMPK: guardian of metabolism and mitochondrial homeostasis**. *Nat Rev Mol Cell Biol* (2018) **19** 121-135. PMID: 28974774
14. Huang H, Lorenz BR, Zelmanovitz PH, Chan CB.. **Metformin preserves β-cell compensation in insulin secretion and mass expansion in prediabetic Nile rats**. *IJMS* (2021) **22** 421. PMID: 33401592
15. Kim JY, He F, Karin M.. **From liver fat to cancer: perils of the Western diet**. *Cancers* (2021) **13** 1095. PMID: 33806428
16. Kobayashi M, Fujii N, Narita T, Higami Y.. **SREBP-1c-dependent metabolic remodeling of white adipose tissue by caloric restriction**. *IJMS* (2018) **19** 3335. PMID: 30373107
17. Lazarus G, Alexander S, Kusuma GO, Wijaya K, Soetikno V.. **Antioxidative activities of alpha-mangostin in high-fat/high-glucose diet and streptozotocin-induced insulin-resistant rodents**. *J Appl Pharm Sci* (2020) **10** 035-039
18. Lee D, Kim YM, Jung K, Chin YW, Kang KS.. **Alpha-Mangostin improves insulin secretion and protects INS-1 cells from streptozotocin-induced damage**. *Int J Mol Sci* (2018) **19** 1484. PMID: 29772703
19. Li RS, Xu GH, Cao J, Liu B, Xie HF, Ishii Y, Zhang CF.. **Alpha-Mangostin ameliorates bleomycin-induced pulmonary fibrosis in mice partly through activating adenosine 5′-monophosphate-activated protein kinase**. *Front Pharmacol* (2019) **10** 1305. PMID: 31798444
20. Li L, Brunner I, Han AR, Hamburger M, Kinghorn AD, Frye R, Butterweck V.. **Pharmacokinetics of α-mangostin in rats after intravenous and oral application**. *Mol Nutr Food Res* (2011) **55** s-74
21. Li Y, Xu S, Mihaylova MM, Zheng B, Hou X, Jiang B, Park O, Luo Z, Lefai E, Shyy JY. **AMPK phosphorylates and inhibits SREBP activity to attenuate hepatic steatosis and atherosclerosis in diet-induced insulin-resistant mice**. *Cell Metab* (2011) **13** 376-388. PMID: 21459323
22. Livak KJ, Schmittgen TD.. **Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method**. *Methods* (2001) **25** 402-408. PMID: 11846609
23. Manco M.. **Insulin resistance and NAFLD: a dangerous liaison beyond the genetics**. *Children* (2017) **4** 74. PMID: 28805745
24. Mechanick JI, Farkouh ME, Newman JD, Garvey WT.. **Cardiometabolic-based chronic disease, adiposity and dysglycemia drivers: JACC state-of-the-art review**. *J Am Coll Cardiol* (2020) **75** 525-538. PMID: 32029136
25. Parkhe A, Parekh P, Nalla LV, Sharma N, Sharma M, Gadepalli A, Kate A, Khairnar A.. **Protective effect of alpha mangostin on rotenone induced toxicity in rat model of Parkinson’s disease**. *Neurosci Lett* (2020) **716** 134652. PMID: 31778768
26. Pillon NJ, Croze ML, Vella RE, Soulère L, Lagarde M, Soulage CO.. **The lipid peroxidation by-product 4-hydroxy-2-nonenal (4-HNE) induces insulin resistance in skeletal muscle through both carbonyl and oxidative stress**. *Endocrinology* (2012) **153** 2099-2111. PMID: 22396448
27. Ratwita W, Sukandar EY, Kurniati NF, Adnyana IK.. **Alpha mangostin and xanthone from mangosteen (**. *J Pharm Nutr Sci* (2018) **8** 83-90
28. Ratwita W, Sukandar EY, Adnyana IK, Kurniati NF.. **Alpha mangostin and xanthone activity on fasting blood glucose, insulin and Langerhans islet of Langerhans in alloxan-induced diabetic mice**. *PJ* (2019) **11** 64-68
29. Saklayen MG.. **The global epidemic of the metabolic syndrome**. *Curr Hypertens Rep* (2018) **20** 12. PMID: 29480368
30. Seki M, Miwa A, Ohsaka F, Karatsu Y, Tsuruta T, Hino S, Morita T, Sonoyama K.. **Local free fatty acids trigger the expression of lipopolysaccharide-binding protein in murine white adipose tissue**. *Biosci Microbiota Food Health* (2022) **41** 54-65. PMID: 35433160
31. Shi Y, Fan Y, Hu Y, Jing J, Wang C, Wu Y, Geng Q, Dong X, Li E, Dong D.. **α-Mangostin suppresses the**. *J Cell Mol Medi* (2020) **24** 760-771
32. Shin MK, Yang SM, Han IS.. **Capsaicin suppresses liver fat accumulation in high-fat diet-induced NAFLD mice**. *Anim Cells Syst* (2020) **24** 214-219
33. Shurrab NT, Arafa E-SA. **Metformin: a review of its therapeutic efficacy and adverse effects**. *Obesity Med* (2020) **17** 100186
34. Sigit FS, Trompet S, Tahapary DL, Sartono E, Willems van Dijk K, Yazdanbakhsh M, Supali T, Smit JWA, Rosendaal FR, de Mutsert R.. **The associations of leptin and adiponectin with the metabolic syndrome in an Indonesian and a Dutch population**. *Nutr Metab Cardiovasc Dis* (2021) **31** 2426-2435. PMID: 34154890
35. Soetikno V, Murwantara A, Andini P, Charlie F, Lazarus G, Louisa M, Arozal W.. **Alpha-mangostin improves cardiac hypertrophy and fibrosis and associated biochemical parameters in high-fat/high-glucose diet and low-dose streptozotocin injection-induced type 2 diabetic rats**. *J Exp Pharmacol* (2020) **12** 27-38. PMID: 32095085
36. Sun J, Wang Y, Zhang X, He H.. **The effects of metformin on insulin resistance in overweight or obese children and adolescents: A PRISMA-compliant systematic review and meta-analysis of randomized controlled trials**. *Medicine* (2019) **98** e14249. PMID: 30681616
37. Tsai SY, Chung PC, Owaga EE, Tsai IJ, Wang PY, Tsai JI, Yeh TS, Hsieh RH.. **Alpha-mangostin from mangosteen (**. *Nutr Metab* (2016) **13** 88
38. Wróbel MP, Marek B, Kajdaniuk D, Rokicka D, Szymborska-Kajanek A, Strojek K.. **Metformin- a new old drug**. *Endokrynol Pol* (2017) **68** 482-496. PMID: 28819951
39. Yap KH, Yee GS, Candasamy M, Tan SC, Md SA, Majeed AB, Bhattamisra SK.. **Catalpol ameliorates insulin sensitivity and mitochondrial respiration in skeletal muscle of type-2 diabetic mice through insulin signalling pathway and AMPK/SIRT1/PGC-1α/PPAR-γ activation**. *Biomolecules* (2020) **10** 1360. PMID: 32987623
40. Yasmin T, Rahman MM, Khan F, Kabir F, Nahar K, Lasker S, Islam MD, Hossain MM, Hasan R, Rana S. **Metformin treatment reverses high fat diet- induced non-alcoholic fatty liver diseases and dyslipidemia by stimulating multiple antioxidant and anti-inflammatory pathways**. *Biochem Biophys Rep* (2021) **28** 101168. PMID: 34825068
|
---
title: Exogenous Hydrogen Sulfide Mitigates Oxidative Stress and Mitochondrial Damages
Induced by Polystyrene Microplastics in Osteoblastic Cells of Mice
authors:
- Qingping Shi
- Feihong Chen
- Yuanyi Feng
- Yangxi Zheng
- Ximei Zhi
- Wen Wu
journal: Disease Markers
year: 2023
pmcid: PMC9969973
doi: 10.1155/2023/2516472
license: CC BY 4.0
---
# Exogenous Hydrogen Sulfide Mitigates Oxidative Stress and Mitochondrial Damages Induced by Polystyrene Microplastics in Osteoblastic Cells of Mice
## Abstract
Polystyrene microplastics (mic-PS) have become harmful pollutants that attracted substantial attention about their potential toxicity. Hydrogen sulfide (H2S) is the third reported endogenous gas transmitter with protective functions on numerous physiologic responses. Nevertheless, the roles for mic-PS on skeletal systems in mammals and the protective effects of exogenous H2S are still indistinct. Here, the proliferation of MC3T3-E1 cell was analyzed by CCK8. Gene changes between the control and mic-PS treatment groups were analyzed by RNA-seq. The mRNA expression of bone morphogenetic protein 4 (Bmp4), alpha cardiac muscle 1 (Actc1), and myosin heavy polypeptide 6 (Myh6) was analyzed by QPCR. ROS level was analyzed by 2′,7′-dichlorofluorescein (DCFH-DA). The mitochondrial membrane potential (MMP) was analyzed by Rh123. Our results indicated after exposure for 24 h, 100 mg/L mic-PS induced considerable cytotoxicity in the osteoblastic cells of mice. There were 147 differentially expressed genes (DEGs) including 103 downregulated genes and 44 upregulated genes in the mic-PS-treated group versus the control. The related signaling pathways were oxidative stress, energy metabolism, bone formation, and osteoblast differentiation. The results indicate that exogenous H2S may relieve mic-PS toxicity by altering Bmp4, Actc1, and Myh6 mRNA expressions associated with mitochondrial oxidative stress. Taken together, this study demonstrated that the bone toxicity effects of mic-PS along with exogenous H2S have protective function in mic-PS-mediated oxidative damage and mitochondrial dysfunction in osteoblastic cells of mice.
## 1. Introduction
Polystyrene microplastics (mic-PS) are plastic particles with diameter < 5 mm [1], originating from industrial products and plastics demoted into pieces by UV radiation, physical, or biodegradation [2]. Mic-PS contain high-density and low-density polyethylene (HD/LD-PE), polyethylene terephthalate (PET), polypropylene (PP), and polyvinylchloride (PVC), together with polystyrene microplastic (PS-MP) [3]. More recently, these minor plastic products have been widely detected in freshwater organisms, ranging from algae to fish, even in mammals. Therefore, pollution by mic-PS was classified as the second most crucial threat in ecological environment at the United Nations Environmental Conference in 2015 [4].
Mic-PS less than 20 μm can easily access the mammalian tissues, while mic-PS with particle size of 0.1~10 μm can effectively pass through the cytomembrane, intestinal mucosal barrier, and blood-brain barrier, even transmit through the placenta to the next generation [5, 6]. Further investigation substantiates the toxicity of mic-PS on index such as oxidative stress, enzymatic activity, quantity of egg laying, feeding rate, and growth rate [5–7]. Specific polystyrene mic-PS with size of 5 to 20 μm can accrue in the liver, lung, and kidney; additionally, they can evoke oxidative damage along with metabolic alterations [8, 9]. In contrast, it remains unknown the relationship among mic-PS intake and bone destruction of terrestrial mammals.
Hydrogen sulfide (H2S), a colorless indispensable endogenous gas, can subsequently enhance catalytic activity by attaching hydropersulfide group (-SSH) to relevant cysteine residues into targeted protein. Exogenous H2S has been reported to regulate numerous signaling pathways associated with biological processes, for instance, regulation of kinase, maintaining intracellular mitochondrial ATP generation, and scavenging reactive oxygen species (ROS). In osteoblastic cell, the scavenging ability to reduce oxidative stress and sustaining maintenance of mitochondrial membrane potential is a key signal for cells and is a crucial target of osteoporosis responsible for increased bone fracture threat. Further research on mic-PS stated that exogenous H2S increases the expression of heme oxygenase-1 and NAD(P)H: quinone oxidoreductase 1, consequently decreasing microplastics producing hepatic apoptosis and inflammation [10]. Therefore, exogenous H2S might be an innovative antioxidant medium under MP stress system. So far, the beneficial effect of exogenous H2S in mic-PS-induced bone toxicity remains undiscovered.
In this study, we aimed to study the cytotoxic effects of mic-PS in MC3T3-E1 cells, then investigating the toxicity of mic-PS in osteoblastic cell through RNA sequencing (RNA-seq). Finally, we aimed to explore whether H2S ameliorated mic-PS exposure induced damage by attenuating oxidative stress and mitochondrial damage.
## 2.1. Materials and Reagents
Mic-PS (100 nm) were purchased from the Tianjin DAE Scientific Co. Ltd (Tianjin, China). GYY4137 (as the donor of H2S) and rhodamine 123 (Rh123) were bought from Sigma (St. Louis, MO, USA). 2′,7′-Dichlorofluorescein diacetate (DCFH-DA) was obtained from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Cell counting kit-8 (CCK-8) was obtained from Dojindo Laboratories (Kumamoto, Japan). Fetal bovine serum (FBS) and *Gibco minimum* essential medium α (α-MEM) were purchased from Thermo Fisher (Waltham, MA, USA).
## 2.2. Cell Culture and Treatment
The mouse calvaria-derived MC3T3-E1 osteoblasts were bought commercially from the National Collection of Authenticated Cell Cultures (Shanghai, China). Osteoblasts were seeded at 1 × 105 cells/mL into 75 cm2 flasks, cultured in α-MEM supplemented with FBS ($10\%$). The basic medium was replaced every three days. The growing conditions were at 37°C with $5\%$ CO2. There were four groups, including the control group, mic-PS group, mic-PS+H2S group, and H2S group. The concentration of H2S released from GYY4137 was 100 μM.
## 2.3. Cell Viability Assay
MC3T3-E1 cells (1 × 104 cells/ml) were cultured in 96-well plates. Then they were incubated for 24 hours at 37°C. The cells were washed with phosphate-buffered saline (PBS), and the cell counting kit-8 (10 μl, at $10\%$ dilution) was added in each well. After incubation, the absorbance was measured with the Multiskan MK3 microplate reader (Thermo Fisher). The mean optical density (OD) was conducted to count the cell viability (%) following the equation (ODtreatment/ODcontrol) × 100. The cell viability assay in each group was repeated five times.
## 2.4. Measurement of Intracellular ROS Generation
MC3T3-E1 cells were incubated with 10 μ M of 2′,7′-dichlorofluorescein (DCFH-DA) for 30 minutes at 37°C. Then, the cells were washed with PBS. The DCF fluorescence was visualized through a fluorescence microscope. The mean fluorescence intensity (MFI) indicated the amount of ROS in the intracellular environment. The measurements were performed by using the ImageJ software (version 1.8.0, Bethesda, Maryland, USA). The experiment was performed three times.
## 2.5. Examination of the Mitochondrial Membrane Potential (MMP)
The MC3T3-E1 cells were incubated for 45 minutes at 37°C with Rh123 (2 μM). Then, the cells were washed with PBS. The fluorescence was then detected by a fluorescence microscope. The MFI of five random fields indicated the levels of MMP. The measurements were performed by using the ImageJ software (version 1.8.0).
## 2.6. mRNA Library Construction and Sequencing
The total RNA was isolated, purified, and next quantified by the NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, USA). The RNA integrity was assessed by Bioanalyzer 2100 (Agilent Technologies, CA) and later confirmed by gel electrophoresis. After purification from the total RNA (1 μg) with Dynabeads Oligo (dT)25 (Thermo Fisher), the poly(A) RNA was fragmented through the Magnesium RNA Fragmentation Module (NEB, Ipswich, MA, USA). Then, the cleaved fragments were reverse transcribed through the SuperScript Reverse Transcriptase (Invitrogen, USA) and used to synthetize U-labeled, second-stranded DNA. The AMPure XP bead was used to perform the size selection. Later, the ligated products were expanded through the polymerase chain reaction (PCR). The denaturation was initially performed for 3 minutes at 95°C. Then, 8 cycles of denaturation were performed for 15 seconds at 98°C. The annealing was performed for 15 seconds at 60°C, with a following extension for 30 seconds at 72°C. The final extension was conducted at 72°C for 5 minutes. Regarding the final cDNA library, the average insert size was 300 ± 50 bp. The 2 × 150 bp paired-end sequencing (PE150) was done with the NovaSeq 6000 sequencing system (Illumina).
## 2.7. Quantitative Real-Time PCR (QPCR) Analysis
The total RNA of MC3T3-E1 cells was extracted and amplified using a SYBR Green based real-time PCR assay (Eppendorf, Germany). The PCR reaction was performed holding for 3 minutes at 95°C, then for 10 seconds at 95°C, for 30 seconds at 60°C, and for further 35 seconds at 72°C. The comparative cross threshold method was used to quantify the mRNA expression. The QPCR in each group was repeated three times. For primer sequences, see Table 1.
## 2.8. Statistical Analysis
Continuous variables are described as mean ± standard error of the mean (SEM). Comparisons were tested by one-way analysis of variance (ANOVA) followed by Dunnett's test using GraphPad Prism 8.0.2 software. A $p \leq 0.05$ was considered statistically significant.
## 3.1. Cytotoxic Effects of Mic-PS in MC3T3-E1 Cells
To evaluate the cytotoxic effects of mic-PS in MC3T3-E1 cells, we treated the cells with different times and concentrations. As shown in Figure 1(a), MC3T3-E1 cells were exposed to increasing concentrations of mic-PS with the same 24 hours. Mic-PS at 50 mg/L showed no significant effects on cell viability ($p \leq 0.05$). Mic-PS at 100 and 150 mg/L displayed toxic effects on MC3T3-E1 cells ($p \leq 0.01$), with no statistical difference between the two concentrations ($p \leq 0.05$). Remarkably, both concentrations lead to a decline in cell viability to almost $50\%$ compared to the control group ($p \leq 0.01$). Therefore, mic-PS at 100 mg/L was used in the following time-response experiment (Figure 1(b)). Exposed MC3T3-E1 cells to mic-PS for 24 hours and 36 hours induced considerable cytotoxicity; the maximum decrease in cell viability was observed at 24 h ($p \leq 0.01$). In accordance with the above results, the MC3T3-E1 cells were cultured with mic-PS at 100 mg/L for 24 h in the following experiments.
## 3.2. Mic-PS Treatment Induced Gene Expression Change in MC3T3-E1 Cells
Next, to further investigate the different gene expressions that occur between the mic-PS and control groups, RNA-seq was conducted. There were 147 differentially expressed genes (DEGs) (absolute Log2 fold change ≥ 1, p ≤ 0.05). Among them, 103 genes were downregulated, while 44 genes were upregulated (Figure 1(c)). The top 100 genes with the lowest p value were further evaluated by hierarchical clustering analysis (Figure 1(d)). This analysis showed a clear distinction between the mic-PS and control groups.
## 3.3. The Gene Ontology (GO) Enrichment Analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) Enrichment Analysis
GO biological process (GO-BP), GO cellular component (GO-CC), and GO molecular function (GO-MF) were the three sections of the GO enrichment analysis (Figure 2(a)). According to the three parts of the GO analysis, the upregulated and downregulated DEGs were functionally categorized, especially regulation of transcription, inflammatory response, protein binding, oxidation-reduction process, and apoptotic. A biological pathway distribution was observed in KEGG enrichment analysis between the mic-PS and control groups (Figure 2(b)). There were several signaling pathways influenced by mic-PS, including peroxisome proliferator-activated receptors (PPAR), arginine and proline metabolism, thyroid hormone metabolism, estrogen metabolism, cyclic adenosine monophosphate (cAMP), phosphonate and phosphinate metabolism, and calcium signal transduction pathway. The signaling pathways are probably related to oxidative stress resistance, energy metabolism, osteoblast differentiation, and bone formation. Thus, the regulated oxidative stress resistance and mitochondrial ATP energy metabolism might be the determinative mechanism linking mic-PS to dysfunction in MC3T3-E1 cells.
## 3.4. Exogenous H2S Reduced Mic-PS-Induced Cytotoxicity in MC3T3-E1 Cells
As described in Figure 3(a), exposure of MC3T3-E1 cells to 100 mg/L mic-PS for 24 hours induced considerable cytotoxicity ($p \leq 0.01$). When cells were treated with H2S at 100 μM, 24 hours, they are markedly ameliorated from mic-PS injury ($p \leq 0.01$). We evaluated the effects of H2S and mic-PS on ROS levels in MC3T3-E1 cells. As expected, mic-PS elevated ROS levels, but the effect was reversed by H2S ($p \leq 0.01$) (Figures 3(b) and 3(g)). Similar findings were obtained with the MMP. The mic-PS exposure reduced MMP ($p \leq 0.01$). On the contrary, H2S treatment alleviated ($p \leq 0.01$) the reduced MMP mediated by mic-PS (Figures 3(c) and 3(h)). The above results revealed that exogenous H2S inhibits mic-PS-induced cytotoxicity, oxidative stress, and dissipation of MMP damage.
## 3.5. Exogenous H2S Increased the Mitochondrial Damage-Related Gene Expression
According to the bioinformatic analysis and above results, we chose bone morphogenetic protein 4 (Bmp4), alpha cardiac muscle 1 (Actc1), and myosin heavy polypeptide 6 (Myh6) as the potential candidate genes. To verify whether these three genes are involved in H2S signaling pathway, mRNA expression was tested by QPCR. Expression of these three gene expressions decreased in the mic-PS group, while exogenous H2S increased a certain extent of these three genes ($p \leq 0.01$, Figures 3(d)–3(f)). Besides, we observed that mic-PS treatment most affected the *Actc1* gene expression, while the effect of mic-PS on Bmp4 expression was comparable minimal. These results suggested that H2S increases Bmp4, Actc1, and Myh6 expressions to mitigate mic-PS-induced oxidative stress and mitochondrial damage in osteoblastic cell.
## 4. Discussion
Since the durability and indecomposable features of plastics as discarded pollutants, the plastic contaminants have risen dramatically worldwide. 322 million tons were produced in 2015 [11], while it will increase by 33 billion tons in 2050 as predicted [12]. Mic-PS can be detected everywhere, including human biological samples. Thus, it is meaningful to study the toxicity of mic-PS in mammals. Previous data have shown that mic-PS induces intestinal microbial growth, reproductive toxicity, metabolic disorders, and intestinal barrier dysfunction in mice [13, 14]. Nevertheless, little is known about mic-PS effects on mouse bone metabolism. In this study, we found that mouse osteoblastic cell activity declined after being exposed to mic-PS. RNA-seq analysis showed 147 differentially expressed genes between the mic-PS and control groups. Furthermore, we proved that exogenous H2S could increase the related gene expression to reduce mic-PS-induced oxidative stress and mitochondrial injury (Figure 4).
The bioinformatic analysis was subsequently performed to evaluate the toxicity of mic-PS in mouse osteoblastic cell through GO and KEGG databases. GO analysis indicates that exposure to mic-PS significantly affected biological processes such as transcription, inflammatory response, protein binding, oxidation-reduction process, and apoptotic. Currently studies suggest that inducing oxidative stress was the relevant effect of mic-PS toxicity [15, 16]. Xu et al. researched on human lung epithelial cells; they proved that mic-PS significantly affect the cell viability via inducing significant upregulation of proinflammatory and proapoptotic proteins, including TNF-α, IL-8, caspase-3, caspase-8, and caspase-9 [17]. Based on the KEGG databases in our study, involving peroxisome proliferator-activated receptors (PPAR), arginine and proline metabolism, thyroid hormone, estrogen, cyclic adenosine monophosphate(cAMP), phosphonate and phosphinate metabolism, and calcium pathway were significantly enriched pathways for DEGs. PPAR signaling pathway was also observed in exposure to MPs on grass carp through KEGG enrichment analysis [18]. PPARs are transcription factors that regulate the expression of genes involved in energy and lipid metabolism; it is interesting that activation of PPARδ improves mitochondrial function [19]. Presently, Sun et al. reported that Jak/Stat pathway, nicotinamide metabolism, and unsaturated fatty acids are associated with mic-PS-mediated toxicity in the mouse hematological system; furthermore, they found that decreased *Nnt is* possibly correlated with reduced antioxidant power and mitochondrial damage after mic-PS exposure [20]. Oxidative stress and mitochondrial metabolic changes are closely related to osteogenic capacity. Mitochondrial dysfunction and ROS rise induced osteoblast senescence and osteoblast activity [21] and led to type 2 diabetic osteoporosis [22]. Proanthocyanidins and notoginsenoside R1 treatments reduced ROS level and weakened mitochondrial dysfunction to improve osteoblast activity [23, 24]. These studies suggest that oxidative stress and disturbances in mitochondrial metabolism are targets for improving osteogenic capacity. Hence, the change of oxidative stress and mitochondrial metabolism was selected for further toxicity mechanism caused by exposure to mic-PS in this research.
To validate of the potential of the change of oxidative stress and mitochondrial metabolism after mic-PS exposure, we then performed experiments in vitro. The mic-PS exposure significantly elevated oxidative stress, as well as dissipated MMP, while exogenous H2S mitigated mic-PS-induced oxidative and mitochondrial injury. Presently, accumulating evidence has demonstrated that Bmp4, Actc1, and Myh6 are closely related to homeostasis of mitochondria and redox reactions [25–27]. Especially, BMP4 is a group of bone growth factors firstly identified because of their capability to enhance bone and cartilage formation. As a result, these three genes (Bmp4, Actc1, and Myh6) were selected. We next used QPCR to test our findings. We observed that the downward trend of QPCR results in the mic-PS group was essentially comparable with the sequencing analysis. A similar alteration of Bmp4 and Myh5 had also been observed in zebrafish exposed to mic-PS (100 μg/L) [28]. However, Bhagat et al. described upregulation of bmp4 in zebrafish embryos when exposed to polystyrene nanoplastics (1 mg/L) and azole fungicides (ketoconazole and fluconazole) [29]. Different size and concentration of mic-PS should be considered in making the results. Meanwhile, experimental results of Umamaheswari et al. indicated that mic-PS exposure upregulated the gstp1, hsp70l, and ptgs2a gene expressions, while it downregulated cat, sod1, gpx1a, and ache genes, illustrating the potential of mic-PS to mediate different degrees of toxic effects in aquatic animals through changing ROS medicated oxidative stress altering its metabolic process, histological architecture, and gene regulatory modes [30].
Of note, in a range of biological mechanisms, exogenous H2S has essential physiological and pathological impacts. According to a recent study, exogenous H2S protects osteoblastic cells from H2O2-induced cell oxidative stress injury [31]. In this study, the findings supported the inhibitory effect of exogenous H2S on mic-PS-induced mitochondrial damage and oxidative stress in mouse osteoblastic cells. In addition, exogenous H2S considerably increases the alleviated expression of Bmp4, Actc1, and Myh6 derived from mic-PS. The mechanisms underlying mic-PS-induced oxidative stress and mitochondrial damage could be complex and diverse. Mic-PS has been shown to reduce the activity of glutathione S-transferases, limiting detoxification and resulting in ROS generation [32]. Moreover, it is generally acknowledged that H2S is associated with increasing glutathione S-transferase transcript level. Recently, a study has demonstrated that H2S suppressed inflammation and oxidative stress induced by mic-PS in mouse liver via upregulated Keap1-Nrf2 pathway [10]. Notably, Bmp4, Actc1, and Myh6 were reported to be associated with glutathione S-transferases. Reportedly, exogenous H2S improved the BMP4 expression in rat pulmonary arterial smooth muscle cells [33]. Thus, these results indicate that exogenous H2S may relieve mic-PS toxicity by altering Bmp4, Actc1, and Myh6 expressions associated with mitochondrial oxidative stress.
The study has two main limitations. First, the effect of exogenous H2S and mic-PS needs to be explored in vivo. Second, the downstream signaling pathways regulated by H2S in mic-PS-induced injury need to find in further exploration.
## 5. Conclusion
In conclusion, we used RNA-seq and validation experiment at molecular and cellular levels to demonstrate that mic-PS (100 mg/L) caused a significant toxicity in osteoblastic cells in mice (an advanced mammal), while exogenous H2S may considerably mitigate mic-PS-induced oxidative stress and mitochondrial damage through increasing Bmp4, Actc1, and Myh6 expressions. Our research unveiled the bone toxicity effects of mic-PS and provides novel insights in the mechanism of exogenous H2S protective function in mic-PS-induced bone injury.
## Data Availability
The research data are available upon request.
## Conflicts of Interest
All authors declare they have no conflict of interest.
## Authors' Contributions
Qingping Shi and Feihong Chen contributed equally to this work.
## References
1. Moore C. J.. **Synthetic polymers in the marine environment: a rapidly increasing, long-term threat**. (2008) **108** 131-139. DOI: 10.1016/j.envres.2008.07.025
2. Law K. L., Thompson R. C.. **Microplastics in the seas**. (2014) **345** 144-145. DOI: 10.1126/science.1254065
3. Gigault J., Halle A. T., Baudrimont M.. **Current opinion: what is a nanoplastic?**. (2018) **235** 1030-1034. DOI: 10.1016/j.envpol.2018.01.024
4. Amaral-Zettler L. A., Zettler E. R., Slikas B.. **The biogeography of the plastisphere: implications for policy**. (2015) **13** 541-546. DOI: 10.1890/150017
5. Browne M. A., Dissanayake A., Galloway T. S., Lowe D. M., Thompson R. C.. **Ingested microscopic plastic translocates to the circulatory system of the mussel, Mytilus edulis (L)**. (2008) **42** 5026-5031. DOI: 10.1021/es800249a
6. von Moos N., Burkhardt-Holm P., Kohler A.. **Uptake and effects of microplastics on cells and tissue of the blue mussel**. (2012) **46** 11327-11335. DOI: 10.1021/es302332w
7. Sussarellu R., Suquet M., Thomas Y.. **Oyster reproduction is affected by exposure to polystyrene microplastics**. (2016) **113** 2430-2435. DOI: 10.1073/pnas.1519019113
8. Deng Y., Zhang Y., Qiao R.. **Evidence that microplastics aggravate the toxicity of organophosphorus flame retardants in mice (**. (2018) **357** 348-354. DOI: 10.1016/j.jhazmat.2018.06.017
9. Deng Y., Zhang Y., Lemos B., Ren H.. **Tissue accumulation of microplastics in mice and biomarker responses suggest widespread health risks of exposure**. (2017) **7**. DOI: 10.1038/srep46687
10. Li S., Shi M., Wang Y., Xiao Y., Cai D., Xiao F.. **Keap1-Nrf2 pathway up-regulation via hydrogen sulfide mitigates polystyrene microplastics induced-hepatotoxic effects**. (2021) **402, article 123933**. DOI: 10.1016/j.jhazmat.2020.123933
11. Landon-Lane M.. **Corporate social responsibility in marine plastic debris governance**. (2018) **127** 310-319. DOI: 10.1016/j.marpolbul.2017.11.054
12. Rochman C. M., Browne M. A., Halpern B. S.. **Classify plastic waste as hazardous**. (2013) **494** 169-171. DOI: 10.1038/494169a
13. Jin H., Ma T., Sha X.. **Polystyrene microplastics induced male reproductive toxicity in mice**. (2021) **401, article 123430**. DOI: 10.1016/j.jhazmat.2020.123430
14. Jin Y., Lu L., Tu W., Luo T., Fu Z.. **Impacts of polystyrene microplastic on the gut barrier, microbiota and metabolism of mice**. (2019) **649** 308-317. DOI: 10.1016/j.scitotenv.2018.08.353
15. Hamed M., Soliman H. A. M., Osman A. G. M., Sayed A. E. H.. **Antioxidants and molecular damage in Nile tilapia (Oreochromis niloticus) after exposure to microplastics**. (2020) **27** 14581-14588. DOI: 10.1007/s11356-020-07898-y
16. Zhu K., Jia H., Sun Y.. **Long-term phototransformation of microplastics under simulated sunlight irradiation in aquatic environments: roles of reactive oxygen species**. (2020) **173, article 115564**. DOI: 10.1016/j.watres.2020.115564
17. Xu M., Halimu G., Zhang Q.. **Internalization and toxicity: a preliminary study of effects of nanoplastic particles on human lung epithelial cell**. (2019) **694, article 133794**. DOI: 10.1016/j.scitotenv.2019.133794
18. Liu Y., Jia X., Zhu H.. **The effects of exposure to microplastics on grass carp (**. (2022) **286, Part 3, article 131831**. DOI: 10.1016/j.chemosphere.2021.131831
19. Takada I., Makishima M.. **Peroxisome proliferator-activated receptor agonists and antagonists: a patent review (2014-present)**. (2020) **30** 1-13. DOI: 10.1080/13543776.2020.1703952
20. Sun R., Xu K., Yu L.. **Preliminary study on impacts of polystyrene microplastics on the hematological system and gene expression in bone marrow cells of mice**. (2021) **218, article 112296**. DOI: 10.1016/j.ecoenv.2021.112296
21. Schoppa A. M., Chen X., Ramge J. M.. **Osteoblast lineage Sod2 deficiency leads to an osteoporosis-like phenotype in mice**. (2022) **15**. DOI: 10.1242/dmm.049392
22. Wang X., Ma H., Sun J.. **Mitochondrial ferritin deficiency promotes osteoblastic ferroptosis via mitophagy in type 2 diabetic osteoporosis**. (2022) **200** 298-307. DOI: 10.1007/s12011-021-02627-z
23. Chen L., Hu S. L., Xie J.. **Proanthocyanidins-mediated Nrf2 activation ameliorates glucocorticoid-induced oxidative stress and mitochondrial dysfunction in osteoblasts**. (2020) **2020** 14. DOI: 10.1155/2020/9102012
24. Li X., Lin H., Zhang X.. **Notoginsenoside R1 attenuates oxidative stress-induced osteoblast dysfunction through JNK signalling pathway**. (2021) **25** 11278-11289. DOI: 10.1111/jcmm.17054
25. Shahzadi S. K., Naidoo N., Alsheikh-Ali A.. **Reconnoitering the role of long-noncoding RNAs in hypertrophic cardiomyopathy: a descriptive review**. (2021) **22** p. 9378. DOI: 10.3390/ijms22179378
26. Yamamoto H., Kurachi M., Naruse M., Shibasaki K., Ishizaki Y.. **BMP4 signaling in NPCs upregulates Bcl-xL to promote their survival in the presence of FGF-2**. (2018) **496** 588-593. DOI: 10.1016/j.bbrc.2018.01.090
27. Zhou Y., Al-Saaidi R. A., Fernandez-Guerra P.. **Mitochondrial spare respiratory capacity is negatively correlated with nuclear reprogramming efficiency**. (2017) **26** 166-176. DOI: 10.1089/scd.2016.0162
28. Wu Q., Li G., Huo T.. **Mechanisms of parental co-exposure to polystyrene nanoplastics and microcystin-LR aggravated hatching inhibition of zebrafish offspring**. (2021) **774, article 145766**. DOI: 10.1016/j.scitotenv.2021.145766
29. Bhagat J., Zang L., Nakayama H., Nishimura N., Shimada Y.. **Effects of nanoplastic on toxicity of azole fungicides (ketoconazole and fluconazole) in zebrafish embryos**. (2021) **800, article 149463**. DOI: 10.1016/j.scitotenv.2021.149463
30. Umamaheswari S., Priyadarshinee S., Bhattacharjee M., Kadirvelu K., Ramesh M.. **Exposure to polystyrene microplastics induced gene modulated biological responses in zebrafish (**. (2021) **281, article 128592**. DOI: 10.1016/j.chemosphere.2020.128592
31. Yan X., Wu H., Wu Z.. **The new synthetic H2S-releasing SDSS protects MC3T3-E1 osteoblasts against H2O2-induced apoptosis by suppressing oxidative stress, inhibiting MAPKs, and activating the PI3K/Akt pathway**. (2017) **8** p. 7. DOI: 10.3389/fphar.2017.00007
32. Xie X., Deng T., Duan J., Xie J., Yuan J., Chen M.. **Exposure to polystyrene microplastics causes reproductive toxicity through oxidative stress and activation of the p38 MAPK signaling pathway**. (2020) **190, article 110133**. DOI: 10.1016/j.ecoenv.2019.110133
33. Yao Z., Wang C.. **A novel mechanism of sildenafil improving the excessive proliferation and H2S production in pulmonary arterial smooth muscle cells**. (2019) **74** 355-363. DOI: 10.1097/FJC.0000000000000714
|
---
title: Development of Prognostic Features of Hepatocellular Carcinoma Based on Metabolic
Gene Classification and Immune and Oxidative Stress Characteristic Analysis
authors:
- Shimeng Cui
- Minghao Zhang
- Shanshan Bai
- Yanfang Bi
- Shan Cong
- Shi Jin
- Shuang Li
- Hui He
- Jian Zhang
journal: Oxidative Medicine and Cellular Longevity
year: 2023
pmcid: PMC9969974
doi: 10.1155/2023/1847700
license: CC BY 4.0
---
# Development of Prognostic Features of Hepatocellular Carcinoma Based on Metabolic Gene Classification and Immune and Oxidative Stress Characteristic Analysis
## Abstract
### Background
The molecular classification of HCC premised on metabolic genes might give assistance for diagnosis, therapy, prognosis prediction, immune infiltration, and oxidative stress in addition to supplementing the limitations of the clinical staging system. This would help to better represent the deeper features of HCC.
### Methods
TCGA datasets combined with GSE14520 and HCCDB18 datasets were used to determine the metabolic subtype (MC) using ConsensusClusterPlus. ssGSEA method was used to calculate the IFNγ score, the oxidative stress pathway scores, and the score distribution of 22 distinct immune cells, and their differential expressions were assessed with the use of CIBERSORT. *To* generate a subtype classification feature index, LDA was utilized. Screening of the metabolic gene coexpression modules was done with the help of WGCNA.
### Results
Three MCs (MC1, MC2, and MC3) were identified and showed different prognoses (MC2-poor and MC1-better). Although MC2 had a high immune microenvironment infiltration, T cell exhaustion markers were expressed at a high level in MC2 in contrast with MC1. Most oxidative stress-related pathways are inhibited in the MC2 subtype and activated in the MC1 subtype. The immunophenotyping of pan-cancer showed that the C1 and C2 subtypes with poor prognosis accounted for significantly higher proportions of MC2 and MC3 subtypes than MC1, while the better prognostic C3 subtype accounted for significantly lower proportions of MC2 than MC1. As per the findings of the TIDE analysis, MC1 had a greater likelihood of benefiting from immunotherapeutic regimens. MC2 was found to have a greater sensitivity to traditional chemotherapy drugs. Finally, 7 potential gene markers indicate HCC prognosis.
### Conclusion
The difference (variation) in tumor microenvironment and oxidative stress among metabolic subtypes of HCC was compared from multiple angles and levels. A complete and thorough clarification of the molecular pathological properties of HCC, the exploration of reliable markers for diagnosis, the improvement of the cancer staging system, and the guiding of individualized treatment of HCC all gain benefit greatly from molecular classification associated with metabolism.
## 1. Introduction
Hepatocellular carcinoma (HCC) is ranked as the second major contributor to malignant tumors [1], and its etiology is closely related to viral infection and liver fibrosis [2]. Following surgery or ablation, around $70\%$ of cases of HCC will recur within the first 5 years, and the 5-year survival rate is only approximately 30-$40\%$ [3]. The main factors leading to the dismal prognosis of HCC patients include an elevated degree of malignancy, easy recurrence, insensitivity to radiotherapy and chemotherapy, and proneness to vascular, lymphatic, and distant metastasis [4].
The tumor microenvironment (TME), especially the tumor immune microenvironment, performs an integral function in HCC. The TME comprises cancer cells, signaling molecules, fibroblasts, infiltrating immune cells, surrounding blood vessels, and the extracellular matrix [5]. Previous research has demonstrated that the TME may influence the gene expression of tumor tissues in a variety of ways, which could in turn influence the onset and progression of cancers [6]. For instance, tumor cells may modulate the TME via the negative control mechanism established by the immune system of the body. A wide variety of immunosuppressive states may be employed to counter the antitumor immunity that the body naturally has [7]. The degree of immunosuppression present in the TME is intimately linked to the individual differences in the effectiveness of tumor immunotherapy [5]. The dynamic network that constitutes the TME is primarily made up of stromal and immune cells that have infiltrated the tumor tissues. It is generally accepted that the mutual metabolic needs of immune and tumor cells are the root cause of immunosuppression in the surrounding environment [8], and inflammation performs a fundamental function in the development, invasion, and metastasis of HCC [9]. Although earlier research examined the tumor immune microenvironment (TIME) of HCC, further studies are needed to identify diagnostic markers for HCC.
Currently, clinicopathological staging is a standard method that is frequently utilized for determining the prognosis of individuals who have HCC. Nevertheless, HCC often exhibits clinical heterogeneity, which diminishes the efficacy of prognostic evaluations that are routinely performed. To provide high-risk groups with more clinically meaningful treatment strategies, prolong survival, and improve quality of life, the findings of the clinical staging prediction need to be optimized by the development of an innovative prognostic prediction model. As a result of the advent of gene chips and high-throughput sequencing technologies, as well as the large amounts of data included in the GEO and the TCGA databases, the systematic and thorough investigation of genes associated with tumors and the modulatory mechanisms that control them based on bioinformatic approaches has become an integral aspect of the present tumor genomics group. In the past ten years, screening for genetic alterations at the genome level has seen extensive use of gene sequencing and bioinformatic analysis, elucidating part of the molecular mechanism of the onset and advancement of HCC and guiding the determination of differentially expressed genes (DEGs) that participate in the progression of HCC. Functional pathway research provides more options for HCC therapy [10]. Recently, the identification of genetic markers for the prognosis of HCC has become a hot spot in many studies [11, 12]. However, because the properties of tumor markers are dependent on the tumor burden, their significance in the early diagnosis of tumors is limited.
The high proliferation and persistent inflammation of HCC cells are related to oxidative stress. In HCC, HBV genome-encoded X protein (HBx) is associated with increased ROS in mitochondria [13]. Increased levels of oxidative DNA damage, 8-oxyoguanine (8-oxoG), were observed in human hepatocellular carcinoma cells infected with HCV and in the liver of transgenic mice expressing HCV core protein in vitro [14]. Site-specific epigenetic alterations in HCC cells include methylation of the E-cadherin promoter caused by H2O2 treatment and methylation of the cytokine signaling suppressor SOCS3 caused by HBV-induced mitochondrial ROS accumulation [15]. *In* general, increased ROS levels are one of the causes of HCC proliferative.
As a crucial aspect of malignancies, metabolic disorders are an important factor to consider [16], since they all influence various HCC biological behaviors, development, and transfer recurrence [17, 18]. Carcinogenic factors, on the one hand, disturb the body's delicate metabolic equilibrium and lead to the development of metabolic recombinant cell carcinoma; in addition, the metabolic system after recombination is mediated by various biological behaviors, participating in the proliferative, invasive, and metastatic abilities of cancer cells [19, 20]. Recently, researchers have investigated the development of the molecular pathological properties of HCC [21]. These studies have summarized many abnormal metabolic genes related to HCC prognosis from different cells, animals, or HCC [22, 23]. In summary, the goal of HCC metabolism research is to fully understand the molecular pathological characteristics of HCC, explore reliable markers of HCC diagnosis and transfer recurrence prediction, improve the hepatoma staging system, guide individualization, and improve the treatment of HCC [24].
Based on this purpose, we divided HCC into different metabolic subtypes and multidimensional differences between different metabolic subtypes. Different MCs were analyzed, and different response patterns were observed with immunotherapy. At the same time, the correlation between the immunological examination points and the distinct metabolic molecular types and the variation in molecular mutations were subjected to a comparison. The final screen was selected with the potential prognostic marker associated with the metabolic feature index. In summary, we have created numerous subtypes of an immunological festival index and developed a molecular categorization model premised on metabolic properties, thus supplementing the lack of clinical installment systems. Our findings will serve as a base of research directions as well as a theoretical foundation for the evaluation of prognosis and the tailored therapy of HCC patients.
## 2.1. Sources of Expression Profile Information and Genes Involved in Metabolism
We retrieved TCGA-HCC RNA-seq data as well as clinical survival and characteristic data by utilizing the TCGA GDC API. We downloaded datasets with survival time including the GSE14520 from the Gene Expression Omnibus (GEO) database and the HCCDB18 from the HCCDB database (http://lifeome.net/database/hccdb/home.html). The date of the data retrieval was the 7th of April 2021. We downloaded the metabolism-related genes corresponding to the keywords carbohydrates, oxidations, glycogen, glycogenolysis, glycolysis, pyruvate, citric acid, and fatty acid from the KEGG and Reactome databases. *Duplicate* genes between pathways were removed, and then, 619 genes were identified (Table 1).
## 2.2. TCGA-HCC, GSE14520, and HCCDB18 Data Preprocessing
TCGA-HCC's RNA-*Seq data* processing is as follows: [1] We eliminated any samples that lacked information about clinical follow-up, survival time, and status. [ 2] Ensembl was converted to gene symbol. [ 3] When comparing the expression of various gene symbols, we used the median value. [ 4] Genes whose expression in sample < 0.5 were filtered, which accounted for more than $50\%$ of the genes.
Dataset processing of GSE14520 is as follows: [1] We retained the samples from the GPL3921 platform. [ 2] Samples lacking information on clinical follow-up information, survival status, and time were removed. [ 3] HCC samples were retained. [ 4] Probes were converted to gene symbol. [ 5] A probe in response to multiple genes was removed. [ 6] The expression of multiple gene symbols was assessed by taking its midvalue.
Dataset processing of HCCDB18 is as follows: [1] We eliminated any samples that lacked information on clinical follow-up, survival time, and status. [ 2] HCC samples were retained. The following are the pretreatment samples: TCGA 365, GSE14520 221, and HCCDB18 203 (Table S1).
## 2.3. HCC Subtype Classification
First, the “metabolic genes” that were linked to prognosis were screened using a univariate Cox regression analysis. By employing the ConsensusClusterPlus technique, the 365 TCGA-HCC samples were clustered, and stable clustering findings were established based on the cumulative distribution function (CDF) as well as the CDF delta area curve. With the help of the specified metabolic genes, the HCC metabolic genes were constructed. The rationality for clustering is validated with the use of the resampling-based technique known as the ConsensusClusterPlus 1.52.0 method. km arithmetic and “1-Spearman correlation” distance were utilized to complete 500 bootstraps with every bootstrap having specimens (≥$80\%$) of TCGA-HCC sample dataset. Cluster number k was between 2 and 10, and the optimum k was identified as per cumulative distribution function (CDF) and consistency matrix. The process of resampling has the potential to disrupt the original dataset. In this way, cluster analysis was carried out on each of the resampled samples, and the results were then analyzed in detail. The results of the analysis of subclusters provided an evaluation of consistency (consensus).
## 2.4. Single-Sample Gene Set Enrichment Analysis (ssGSEA)
Single-sample GSEA (ssGSEA) is an extension of gene set enrichment analysis (GSEA). The absolute degree to which genes from a particular gene set are enriched in a sample is reflected by the ssGSEA enrichment score assigned to each gene. The levels of gene expression for a particular sample were categorized and then standardized. To compute an enrichment score, the empirical cumulative distribution function (ECDF) was applied to both the genes that were included in the signature and the genes that were retained. We employed the ssGSEA approach to determine each patient's IFNγ score to assess the Th1/IFNγ expression variations in metabolic subtypes and enrichment score of 14 oxidative stress-related biological pathways. The ssGSEA score was normalized to uniform distribution, for which the ssGSEA score is distributed between 0 and 1.
## 2.5. Immune Infiltration Characteristics
To contrast the immunological properties of the various metabolic groups, when analyzing the score distribution and differential expression of 22 different immune cells present in the TCGA-HCC sample, we made use of the CIBERSORT technique. CIBERSORT [22] is a technology that may be used to deconvolve the expression matrix of different types of immune cells using the linear support vector regression methodology as the foundation. The analysis of the expression pattern based on transcriptomic sequencing was performed with the CIBERSORT tool, and the denoising method and the unknown mixture content were removed by the deconvolution method to determine the relative proportion of each of the 22 distinct types of immune cells. The relative expression of certain genes was assessed based on the data from the expression patterns of each sample that was sequenced. This allowed for the proportions of 22 different types of immune cells to be predicted.
## 2.6. Prediction of Chemotherapeutic/Immunotherapeutic Response and Establishment of the Subtype Characteristic Index
To examine the similarities between patients' diverse metabolic subtypes and the GSE91061 dataset (melanoma dataset undergoing anti-CTLA-4 and anti-PD-1 therapy), we used a subclass mapping technique. When the p value is decreased, the degree of similarity increases. At the same time, we compared the degree of responsiveness between various subtypes and conventional chemotherapeutic agents (cisplatin, vinorelbine, embellin, and pyrimethamine) using the same methodology.
We employed linear discriminant analysis, also abbreviated as LDA, to create a subtype classification feature index so that we could more accurately measure the immunological features of patients who were represented by a variety of sample cohorts. In the TCGA dataset, we employed the LDA model to compute each patient's subtype feature index, and we examined the feature index of each of the distinct subtypes. Within the TCGA dataset, we assessed the characteristics that were linked to prognosis. Firstly, a z-score was done on each individual feature, and Fisher's LDA optimization standard was utilized to specify that the centroids of each group should be as dispersed as possible. It was discovered that a linear combination A maximized the between-class variance of A in comparison to the variance of the within-class measure. The properties of the model allow for the differentiation between samples of various subtypes analyzed.
## 2.7. Weighted Correlation Network Analysis (WGCNA)
We clustered the samples using the R software program WGCNA and then filtered the coexpression modules of metabolic genes after selecting the TCGA expression profile dataset with a MAD value of more than $50\%$. According to the findings of the research, the coexpression network conforms to the scale-free free network; i.e., the logarithm of the connection degree k (log(k)) of a node has an inverse correlation with the logarithm of the likelihood that the node occurs, which is denoted by log(p(k)), and the correlation coefficient is >0.85. Further conversion into an adjacency matrix was performed on the expression matrix, and after that, a topological matrix was derived from the adjacency matrix. To cluster genes, we utilized the TOM and a technique called average-linkage hierarchical clustering as per the criteria of the hybrid dynamic shearing tree and established 80 as the basic threshold for each gene count in the gene network module. After identifying the gene modules with the use of the dynamic shear approach, we carried out cluster analysis on the modules before calculating each module's eigengenes in turn. We created a new module by merging the modules that were physically located closer to one another and set minModuleSize = 80, DeepSplit = 3, and height = 0.25.
## 2.8. Statistical Analysis
R (version 3.6.0) was adopted to execute all analyses of statistical data. All statistical tests were bilateral. The statistical significance level was established at p value < 0.05.
## 3.1. Molecular Typing Based on Metabolic Gene Construction
To construct molecular subtypes, we performed the univariate Cox regression analysis of metabolic genes in the three datasets, and the results of the intersection with prognostic-related genes (TCGA: 214, GSE14520: 133, and HCCDB18: 169, Table S2-4) showed that there were only 30 prognostic-related genes (Figure 1(a)), which indicates that the consistency of metabolic genes between datasets on different platforms is poor, and the expression of a particular metabolic gene might vary remarkably depending on the cohort. As a consequence, to conduct the subsequent analysis, we based on 30 metabolic genes that were all recognized as prognostic-related genes ($p \leq 0.05$).
Consensus clustering (ConsensusClusterPlus) was performed on 365 TCGA-HCC samples to determine a stable clustering result (cluster = 3, Figure 1(b)), and lastly, three metabolic subtypes (metabolism cluster, MC$\frac{1}{2}$/3) were obtained (Figure 1(c)). Additional investigation into its prognostic qualities revealed that the prognosis for patients within MC2 was unfavorable, while the prognosis for those within MC1 was positive, and the disparity between the two was statistically remarkable (Figure 1(d)). In GSE14520 and HCCDB18, we also observed the same result (Figures 1(e) and 1(f)). These findings suggest that the 3 metabolic subtypes that were developed premised on metabolic genes are reproducible across multiple research groups.
## 3.2. The Link between Metabolic Subtypes, Oxidative Stress, and Common Gene Mutations
To study the link between metabolic subtypes and oxidative stress, we obtained the TCGA expression profile dataset and then assessed the enrichment scores of 14 oxidative stress-related biological pathways in each sample by using the ssGSEA method. Then, we calculated the distribution of these oxidative stress pathway enrichment scores in the three metabolic molecular subtypes, and we observed that there were significant differences in 9 ($64.3\%$) oxidative stress-related pathways. Among these significantly different biological pathways, except for “response to oxidative stress” and “cellular response to oxidative stress,” MC2 has lower scores in oxidative stress-related pathways (Figure 2(a)). These results indicated that changes in oxidative stress levels are related to different clinical outcomes of hepatocellular carcinoma. Following this, we filtered an additional 2484 genes (mutation frequency > 3) and utilized the chi-square test to search for genes exhibiting considerably greater mutation frequencies in each subtype (threshold value $p \leq 0.05$), ultimately yielding 133 genes (Figure 2(b)).
## 3.3. Expression of Chemokines in Metabolic Typing and Expression of Immune Checkpoint Genes
To examine the different ways in which chemokines are expressed among the 3 different metabolic subtypes, we evaluated the expression of chemokines as well as their corresponding receptor genes in the TCGA cohort. Among 41 chemokines, 34 ($82.93\%$) had substantial variations in subtypes (Figure 3(a)), while 16 of the 18 chemokine receptor genes ($88.89\%$) exhibited substantial variations in the expression of metabolic subtypes (Figure 3(b)). From these findings, it is convinced that various metabolic subtypes will have a varying degree of immune cell infiltration.
We obtained Th1/IFNγ gene signatures to assess the variations in expression levels of Th1/IFNγ that exist across the 3 metabolic categories [25], and by using ssGSEA methodology, we determined each patient's IFNγ score. According to the findings, the MC2 and MC3 subgroups both exhibited greater IFNγ scores, whereas the MC1 subgroup exhibited the smallest IFNγ scores (Figure 3(c)).
To examine the lytic activity of immune T cells in relation to the 3 metabolic subtypes, we used the average expression levels of GZMA and PRF1 [26] to determine the level of lytic activity exhibited by immune T cells in each patient's tumor. Interestingly, MC2 and MC3 possessed the greatest immune T cell lysis activity, whereas MC1 possessed the least, and there was a significant difference between subgroups (Figure 3(d)).
To examine the variances in angiogenesis score expression between the three metabolic subgroups, we obtained the angiogenesis-related gene set [27] to analyze each patient's angiogenesis score. The findings illustrated that MC1 had a considerably greater angiogenesis score as opposed to MC2 and MC3. The difference between subgroups was significant (Figure 3(e)).
47 immune checkpoint-related genes were analyzed to determine the expression variations across the 3 metabolic categories [25]. The results illustrated that 41 ($87.23\%$) genes were present in the 3 metabolic subgroups. There were remarkable variations among the subtypes. It was found that the expression of the majority of genes associated with immune checkpoints was much higher in MC2 cells in contrast with that in MC1 cells. The levels of expression of T cell exhaustion markers including HAVCR2, CD276, PDCD1, CTLA4, and LAG3 were much greater in MC2 cells as opposed to the levels in MC1 cells (Figure 3(f)). Based on these findings, it seems that distinct patient subgroups might exhibit varying degrees of immunotherapeutic response.
## 3.4. Immune and Metabolic Pathway Features in Various Metabolic Subgroups
We adopted CIBERSORT to assess the immune properties in distinct metabolic subtypes. According to the findings, there were remarkable variations in the immunological features shown by each of the subtypes (Figure 4(a)). Different CD8 T cells, M0, M1, and M2 macrophages as well as resting CD4 memory T cells were significantly and highly expressed in different subtypes (Figure 4(b)), indicating that they may have an instrumental function in HCC.
We examined 10 distinct oncogenic pathways to ascertain their characteristics across various metabolic subtypes [28]. According to the findings, eight of the ten pathways displayed substantially different characteristics depending on the subtype. Among these 8 pathways, MC2 scores were relatively high in the cell cycle, NOTCH, RAS, and TP53 pathways, while the remaining two pathways had low enrichment scores in MC2 (Figure 4(c)). The results of an analysis of immune infiltration revealed that MC2 and MC3 had greater levels of immune microenvironment infiltration compared to MC1, and MC1 was shown to have the least ImmuneScore (Figure 4(d)).
We acquired the data on molecular subtypes [29] from these samples to examine the link between these molecular subtypes and the previous six pan-cancer immunotypes. The findings demonstrated that there are substantial variations across the various pan-cancer immunotypes (Figure 4(e) and Figure S1). The C1 and C2 subtypes with unfavorable prognosis accounted for significantly higher proportions in our definition of MC3 and MC2 subtypes than MC1, and the proportion of C3 with better prognosis in MC2 was significantly lower than MC1 (Figure 4(e)), which is consistent with the poor prognosis of MC2 and MC3. Based on these outcomes, it appears that these 3 subtypes could be added to the established six subtypes that were employed in the prior research.
## 3.5. Analysis of the Variation in TIDE across Metabolic Subtypes
We utilized the TIDE (http://tide.dfci.harvard.edu/) tool to assess the possible clinical impacts of immunotherapy in the three metabolic subtypes. The greater the TIDE prediction score, the greater the likelihood of immune evasion, suggesting that immunotherapy is less likely to be effective for the patient. The results showed that the TIDE score of MC2 in the TCGA dataset was considerably elevated in contrast with that of MC1 (Figure 5(a)), indicating that MC1 is more likely to respond favorably to immunotherapy as opposed to MC$\frac{2}{3.}$ Moreover, by comparing the findings of the T cell dysfunction versus rejection scores, we discovered that MC2 displayed a lower T cell dysfunction score in contrast with MC$\frac{1}{3}$ (Figure 5(b)); MC2 was shown to have an elevated T cell rejection score, whereas MC1 had a higher score (Figure 5(c)). We observed the same results in the HCCDB18 (Figures 5(d)–5(f)) and GSE14520 (Figures 5(g)–5(i)) datasets. These findings could explain why the prognosis for MC2 is unfavorable while that of MC1 is satisfactory.
## 3.6. Comparative Analysis of Metabolic Subgroups and Chemotherapy/Immunotherapy
To assess the variations between immunotherapeutic and chemotherapeutic interventions in various metabolic subtypes, we compared the similarities across the 3 metabolic subgroups and immunotherapy patients in the GSE91061 dataset using subclass mapping. We found that in different datasets, the MC1 subtype showed a greater degree of sensitivity to CTLA4 inhibitors compared to the other two subtypes (Figures 6(a), 6(c), and 6(e)). When we compared the responses of various subtypes to standard chemotherapeutic agents (cisplatin, vinorelbine, embellin, and pyrimethamine), we discovered that the MC2 subtype was more responsive to these four medications (Figures 6(b), 6(d), and 6(f)).
## 3.7. Establishment of the Metabolic Subtype Characteristic Index
We utilized linear discriminant analysis (LDA) to construct a subtype classification feature index premised on 30 prognosis-related parameters which allowed for the accurate assessment of the immunological characteristics of patients in various sample cohorts. The first two characteristics of the model may be differentiated as samples of several subtypes (Figure 7(a)). We noticed statistically significant variations between the characteristic indices of the various subtypes when using the LDA model to generate each subtype's value for its subtype characteristic index (Figure 7(b)). ROC analysis illustrated that the characteristic index was in different subtypes. For the classification performance in the model, the multicategory comprehensive prediction AUC was 0.93 (Figure 7(e)). The metabolic subtype feature index was applied to the GSE14502 and HCCDB18 datasets, and the results were similar to the TCGA dataset: the feature indexes of distinct subtypes varied significantly (Figures 7(c) and 7(d)), and ROC analysis illustrated that the comprehensive AUC was 0.94 and 0.93, respectively (Figures 7(f) and 7(g)). Those data suggested that those 3 MCs had better predictive ability.
## 3.8. Determination of Metabolic Feature Index among Coexpressed Gene Modules
We clustered the samples with the help of the WGCNA package of the R tool and screened the coexpression modules (Figure 8(a), soft threshold = 12). To guarantee that the network was not susceptible to scaling problems, we chose β = 12 and obtained 12 modules (Figures 8(b)–8(d)). The gray module represents a set of genes that cannot be merged with those of other modules. Figure 8(e) displays the data on the transcripts of each individual module. Gene modules that cannot be assigned are described as gray modules. Each module's relationship to the patient's age, sex, M stage, N stage, T stage, stage, MC1, MC2, and MC3 was examined. The module had a substantial and negative correlation with MC1 and a significant and positive correlation with MC2, respectively. The correlation between the green, pink, and salmon modules and MC2 was greater than 0.4 (Figure 8(f)).
## 3.9. Prognostic Analysis of Metabolic Feature Index in Coexpressed Gene Modules
We determined the degree to which these 12 modules' feature vectors were correlated with the metabolic feature indices by performing correlation analysis. According to the findings, each of the 12 modules had a significant correlation with the immunological characteristic index (Figure 9(a)). To conduct a reliable and accurate prognostic analysis, we chose modules that had a strong correlation with the metabolic feature index, and we observed that green, purple, cyan, dark, salmon, and royal blue were significantly correlated (Figure 9(b), p ≤ 0.01). We further screened the green and dark modules based on the link between the module, metabolic molecular subtypes, and prognosis, with the module feature vector correlation coefficient > 0.9 and the prognostic significant gene ($p \leq 0.01$) as the hub genes for the module. Finally, seven key genes (RBM12, SENP1, SART3, DHX9, ARMC8, CREB1, and ZNF207) were identified in the dark module. At the same time, we classified the patients into groups with low and high expression premised on gene expression and evaluated whether there were prognostic differences between the genes in these two expression groups. The findings illustrated that except for the group survival curve of the ARMC8 gene, which was marginally significant, the survival curves of other genes had significant differences (Figure 9(c)).
Next, we enriched the genes of the dark and green modules (cluster profile package). The findings highlighted that the dark module was associated with tumor processes, including DNA replication, cell cycle, and autophagy-animal (Figure S2). Meanwhile, the green module was related to Huntington's disease, Parkinson's disease, Alzheimer's disease, and other related diseases (Figure S3).
## 4. Discussion
Currently, clinical decisions for HCC are often based on the disease staging system. It divides patients into different subgroups according to clinical factors associated with HCC prognosis, especially pathological factors, and determines the corresponding treatment plan [30]. Clinical factors associated with the prognosis of patients with HCC mainly reflect the degree to which tumors have spread (distant metastasis, vascular invasion, number of tumor nodules, tumor size, and others) and the severity of liver damage (symptoms of liver decompensation, protein synthesis, and detoxification function, among others) [31]. Although this classification method based on pathological diagnosis has effectively guided clinical practice and brought tangible benefits to patients with HCC, it can only describe the biological characteristics of tumors at the tissue level and fails to accurately represent the biological nature of tumors, especially the essential differences in molecular biology between different classifications [32]. In addition, considering that HCC is such a very heterogeneous illness, even individuals with the same TNM stage might have substantially different responses to therapeutic drugs and varying duration of survival time.
With the in-depth analysis of the molecular pathological characteristics of HCC, some studies have explored the link between the expression profiles of specific genes and the clinicopathological characteristics of HCC, opening new doors for clinical practice. However, it is difficult to attain a global comprehension of the molecular biological characteristics of HCC metastasis, recurrence, and prognosis based on the expression level of one or a few genes. With the development of molecular pathology research on tumors genome-wide, a new tumor typing method, tumor molecular typing, has emerged. It classifies tumors according to their pathological characteristics (mainly genomic characteristics) at the molecular level. Molecular typing can provide global features at the molecular level, such as the tumor genome/proteome/metabolome, enrich molecular pathological information on tumor occurrence and progression, and provide support for clinical diagnosis, staging, individualized treatment, and prognosis prediction [33].
In the current research, we attempted to molecularly categorize HCC at the metabolic level, and as a result, we discovered some novel insights. Our samples were classified into three different metabolic subtypes (MC1, MC2, and MC3) premised on the 619 metabolic genes that were used to classify HCC, where the subtypes showed significant differences in prognosis (Figure 1). The immunological properties of the various metabolic subtypes were distinct, and these subtypes might have varied responses to immunotherapy (Figure 3). In independent datasets (HCCDB18 and GSE14520), there was a significant degree of reproducibility across metabolic subtypes. To more accurately assess the immunological characteristics of patients and accurately represent their varying degrees of immune infiltration, an immune feature index was developed premised on the metabolic subtypes. There was a link between the metabolic feature index and the immune checkpoint. In the meantime, using the coexpression network analysis as a foundation, we conducted a screening of seven possible gene markers (RBM12, SENP1, SART3, DHX9, ARMC8, CREB1, and ZNF207) that are linked to the metabolic feature index.
The TME has been shown to have a significant modulatory function in the onset and advancement of tumors. The TME that is generated as a result of the process of dynamic alterations is controlled by a wide range of immunosuppression signals, and its diversity may help determine several factors, particularly patients' prognosis and how well they will respond to therapy [34, 35]. Recently, numerous studies have discovered that the onset and growth of malignant tumors are directly linked to the components that are present in the microenvironment that surrounds tumor cells. For instance, the chemokines and cytokines that are produced in the liver can enhance angiogenesis as well as immune evasion and antiapoptotic responses and may activate a range of immune cells inside the TME, assist T cells in entering the tumor, alter the immune response of the tumor, and mediate the therapeutic benefits of the treatment [36]. According to the findings of our investigation, the expression of chemokines and the genes that code for their receptors is significantly distinct across various metabolic subtypes. These differential expressions suggest that various metabolic subgroups have varying levels of immune infiltration, which might influence the progression of tumors and patients' immune response to them. Furthermore, tumor-associated chemokines and cytokines may mobilize and polarize immunological subpopulations as well as facilitate the differentiation of cells into protumor phenotypes, thus promoting the progression of tumors. Tumor-associated macrophages (TAMs) are among these immune subtypes and can be polarized to M2 by IL-13, IL-4, TGF-β, or IL-10, within the TME. Macrophages have a phenotype and are responsible for the growth of tumors. Additionally, they stimulate angiogenesis, which is necessary for the recruitment of regulatory cells (Tregs) [37]. In most cases, an unfavorable prognosis for a variety of cancers, including HCC, is linked to the accumulation of TAMs in the area of the tumor.
Long-term strong exposure to reactive oxygen species can induce chronic inflammatory disease progression and carcinogenesis [38, 39]. ROS has been shown to be associated with cancers of the digestive system, such as gastrointestinal cancer, cholangiocarcinoma, pancreatic cancer, and HCC [40, 41]. Representative mechanisms of HBV- and HCV-related chronic liver disease progression and hepatocellular carcinoma have been shown to involve the function of viral proteins, such as immune interference, tumor initiation or tumor suppression interference, and oxidative stress response induction [42]. In Figure 2(a), several oxidative stress-related pathway scores were different in 3 MCs.
We observed that CD4+ T cells, macrophages, and CD8+ T cells are expressed at a high level in various metabolic subtypes. Although the presence of cells such as Tem and *Trm is* linked to the prognosis of patients with HCC, these cells commonly express PD-1. Immune depletion markers including cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) and lymphocyte activation gene 3 (LAG-3) are inversely related to their function [43]. Therefore, these depletion markers have become the primary targets for the immune checkpoint blockade (ICB) to activate and restore CD8+ T cell function [44]. Moreover, CD8+ T cells within the TME may synthesize IFN-γ, thereby promoting the upmodulation of IDO1 and PD-1/PD-L1 gene expression [45]. Research has illustrated that PD-L1 upregulation in tumor cells, particularly when combined with PD-1 produced by tumor-infiltrating activated T cells, may cause exhaustion and suppress the antitumor immunological function of these effector cells, which allows tumor cells to escape the immune system [46]. Additionally, the density of CD8+ T cells with high inhibitory expression of PD-1 is linked to a grim clinical prognosis for HCC [47]. IDO1 upregulation has been shown to have a favorable correlation with not only a dismal prognosis but also tumor advancement and metastasis [48, 49].
These existing research results have also been confirmed in our study. Our research evaluated the Th1/IFNγ ratio, immune checkpoint-related genes, angiogenesis score, and immune T cell lytic activity in the 3 metabolic subtypes. Combined with the above findings, we discovered that MC2 had a high score for immune T cell lysis activity (Figure 3(d)), implying that this subtype possessed more potent immunogenicity and a suitable TME, which should bring better clinical results, but its prognosis was worse than that of MC1. In the comparative study of the expression of immune checkpoints in various subtypes, we discovered that compared to that in MC1, the level of expression of the majority of genes associated with immunological checkpoints (HAVCR2, CD276, PDCD1, CTLA4, and LAG3) was much higher in MC2, suggesting that MC2 may be exhausted by T cells in the subtypes, which might explain why MC2 had an elevated immune microenvironment infiltration level but also an unfavorable prognosis.
In 2017, the TCGA team analyzed the histopathological data of 196 cases of HCC and showed that $22\%$ of HCC exhibited moderate or high lymphocyte infiltration; the team further analyzed the gene expression results of 66 immune markers [3]. Unsupervised hierarchical clustering identified 6 tumor sample clusters, two of which showed high expression of immune markers, including the following immune checkpoint genes: cytotoxic T lymphocyte-associated protein 4 (CTLA4), procedural death receptor 1 (PD-1), and programmed death ligand 1 (PD-L1). This may indicate that ICI treatment can substantially affect HCC with moderate or high lymphocyte infiltration and high expression of immunosuppressive molecules [50]. Combined with our research, it is suggested that MC2 patients may have excellent responsiveness to ICI treatment during the treatment of HCC. Furthermore, we screened 7 potential genetic markers. To date, many studies have confirmed that multigene markers have good predictive power for the metastasis, progression, recurrence, and survival rate of HCC [51, 52]. Gene signatures are a potentially useful high-throughput molecular identification tool via their use in clinical practice, which is premised on gene expression profiling. In short, this sort of combination model based on clinical, pathological, and genetic traits has been proven by accumulating research as having the potential to be highly applicable in clinical settings [53].
There is accumulating evidence to suggest that epigenetic alterations perform an instrumental function in carcinogenesis. Epigenetic alterations have been linked to the clinical prognosis of HCC patients by several different research investigations. This also complicates the molecular classification of HCC. These findings expand the potential therapeutic targets of HCC and enable us to have deeper insights into the molecular classification of HCC by integrating this aspect of the molecular characteristics. In the future, we plan to conduct experimental verification of the results of this article and use more reasonable bioinformatic strategies to improve the model. In addition, the HCC population has undergone tremendous changes in the past few years. The proportion of HCC and early HCC caused by nonhepatitis viruses is gradually increasing [54]. Research on HCC should pay more attention to this part of the population's molecular characteristics and molecular typing.
In summary, the results of this research established a metabolic classification that is capable of independently functioning as an HCC prognostic indicator of HCC. It measured the HCC patients' prognosis risk by analyzing the differences in the features of the TIME across subtypes in order to help clinical diagnosis, staging, and personalized therapy. In addition, prognostic prediction can be used to provide support to HCC patients.
## Data Availability
The datasets analyzed in this study could be found in GSE14520 at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE14520, in GSE91061 at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE91061, and in GSE14502 at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE14502.
## Additional Points
Access to Resources and Materials. The dataset utilized and processed for this investigation is accessible upon valid request from the corresponding authors.
## Ethical Approval
The research was approved by the Ethics Committee of the First Affiliated Hospital of Dalian Medical University.
## Conflicts of Interest
The authors report that there are no conflicting interests.
## Authors' Contributions
All authors contributed to this present work: SMC and MHZ designed the study. SSB, YFB, and SJ acquired the data. SC and SL drafted the manuscript. HH and JZ revised the manuscript. All authors read and approved the manuscript. HH, SL, and JZ conceived and designed the study. SMC, MHZ, and SL contributed to acquisition and analysis of data. SMC, SSB, YFB, and SC wrote, reviewed, and revised the manuscript. HH, SL, and JZ contributed to study supervision. The final document was reviewed and approved by all authors. Shimeng Cui, Minghao Zhang, and Shanshan Bai contributed equally to this work.
## References
1. Forner A., Reig M., Bruix J.. **Hepatocellular carcinoma**. (2018) **391** 1301-1314. DOI: 10.1016/S0140-6736(18)30010-2
2. McGlynn K. A., Petrick J. L., London W. T.. **Global epidemiology of hepatocellular carcinoma: an emphasis on demographic and regional variability**. (2015) **19** 223-238. DOI: 10.1016/j.cld.2015.01.001
3. **Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma**. (2017) **169** 1327-1341.e23. DOI: 10.1016/j.cell.2017.05.046
4. Kulik L., El-Serag H. B.. **Epidemiology and management of hepatocellular carcinoma**. (2019) **156** 477-491.e1. DOI: 10.1053/j.gastro.2018.08.065
5. Roma-Rodrigues C., Mendes R., Baptista P. V., Fernandes A. R.. **Targeting tumor microenvironment for cancer therapy**. (2019) **20** p. 840. DOI: 10.3390/ijms20040840
6. Reina-Campos M., Moscat J., Diaz-Meco M.. **Metabolism shapes the tumor microenvironment**. (2017) **48** 47-53. DOI: 10.1016/j.ceb.2017.05.006
7. Bader J. E., Voss K., Rathmell J. C.. **Targeting metabolism to improve the tumor microenvironment for cancer immunotherapy**. (2020) **78** 1019-1033. DOI: 10.1016/j.molcel.2020.05.034
8. Tahmasebi Birgani M., Carloni V.. **Tumor microenvironment, a paradigm in hepatocellular carcinoma progression and therapy**. (2017) **18** p. 405. DOI: 10.3390/ijms18020405
9. Shi Y., Qin N., Zhou Q.. **Role of IQGAP3 in metastasis and epithelial-mesenchymal transition in human hepatocellular carcinoma**. (2017) **15** p. 176. DOI: 10.1186/s12967-017-1275-8
10. Zucman-Rossi J., Villanueva A., Nault J. C., Llovet J. M.. **Genetic landscape and biomarkers of hepatocellular carcinoma**. (2015) **149** 1226-1239.e4. DOI: 10.1053/j.gastro.2015.05.061
11. Li W., Lu J., Ma Z., Zhao J., Liu J.. **An integrated model based on a six-gene signature predicts overall survival in patients with hepatocellular carcinoma**. (2020) **10** p. 1323. DOI: 10.3389/fgene.2019.01323
12. Wang S., Zhang J. H., Wang H.. **A novel multidimensional signature predicts prognosis in hepatocellular carcinoma patients**. (2019) **234** 11610-11619. DOI: 10.1002/jcp.27818
13. Jung S. Y., Kim Y. J.. **C-terminal region of HBx is crucial for mitochondrial DNA damage**. (2013) **331** 76-83. DOI: 10.1016/j.canlet.2012.12.004
14. Pal S., Polyak S. J., Bano N.. **Hepatitis C virus induces oxidative stress, DNA damage and modulates the DNA repair enzyme NEIL1**. (2010) **25** 627-634. DOI: 10.1111/j.1440-1746.2009.06128.x
15. Lim S. O., Gu J. M., Kim M. S.. **Epigenetic changes induced by reactive oxygen species in hepatocellular carcinoma: methylation of the E-cadherin promoter**. (2008) **135** 2128-2140.e8. DOI: 10.1053/j.gastro.2008.07.027
16. Hanahan D., Weinberg R. A.. **Hallmarks of cancer: the next generation**. (2011) **144** 646-674. DOI: 10.1016/j.cell.2011.02.013
17. Pavlova N. N., Thompson C. B.. **The emerging hallmarks of cancer metabolism**. (2016) **23** 27-47. DOI: 10.1016/j.cmet.2015.12.006
18. Siegel A. B., Zhu A. X.. **Metabolic syndrome and hepatocellular carcinoma**. (2009) **115** 5651-5661. DOI: 10.1002/cncr.24687
19. DeBerardinis R. J., Lum J. J., Hatzivassiliou G., Thompson C. B.. **The biology of cancer: metabolic reprogramming fuels cell growth and proliferation**. (2008) **7** 11-20. DOI: 10.1016/j.cmet.2007.10.002
20. Yang G., Wang Y., Feng J.. **Aspirin suppresses the abnormal lipid metabolism in liver cancer cells via disrupting an NF**. (2017) **486** 827-832. DOI: 10.1016/j.bbrc.2017.03.139
21. Martinez-Outschoorn U. E., Peiris-Pages M., Pestell R. G., Sotgia F., Lisanti M. P.. **Erratum: Cancer metabolism: a therapeutic perspective**. (2017) **14** p. 113. DOI: 10.1038/nrclinonc.2017.1
22. Yu D., Shi X., Meng G.. **Kidney-type glutaminase (GLS1) is a biomarker for pathologic diagnosis and prognosis of hepatocellular carcinoma**. (2015) **6** 7619-7631. DOI: 10.18632/oncotarget.3196
23. Beyoğlu D., Imbeaud S., Maurhofer O.. **Tissue metabolomics of hepatocellular carcinoma: tumor energy metabolism and the role of transcriptomic classification**. (2013) **58** 229-238. DOI: 10.1002/hep.26350
24. Puszyk W. M., Trinh T. L., Chapple S. J., Liu C.. **Linking metabolism and epigenetic regulation in development of hepatocellular carcinoma**. (2013) **93** 983-990. DOI: 10.1038/labinvest.2013.94
25. Danilova L., Ho W. J., Zhu Q.. **Programmed cell death ligand-1 (PD-L1) and CD8 expression profiling identify an immunologic subtype of pancreatic ductal adenocarcinomas with favorable survival**. (2019) **7** 886-895. DOI: 10.1158/2326-6066.CIR-18-0822
26. Rooney M. S., Shukla S. A., Wu C. J., Getz G., Hacohen N.. **Molecular and genetic properties of tumors associated with local immune cytolytic activity**. (2015) **160** 48-61. DOI: 10.1016/j.cell.2014.12.033
27. Masiero M., Simões F. C., Han H. D.. **A core human primary tumor angiogenesis signature identifies the endothelial orphan receptor ELTD1 as a key regulator of angiogenesis**. (2013) **24** 229-241. DOI: 10.1016/j.ccr.2013.06.004
28. Sanchez-Vega F., Mina M., Armenia J.. **Oncogenic signaling pathways in the cancer genome atlas**. (2018) **173** 321-337.e10. DOI: 10.1016/j.cell.2018.03.035
29. Thorsson V., Gibbs D. L., Brown S. D.. **The immune landscape of cancer**. (2019) **51** 411-412. DOI: 10.1016/j.immuni.2019.08.004
30. Braillon A.. **Hepatocellular carcinoma**. (2012) **380** p. 469. DOI: 10.1016/S0140-6736(12)61282-3
31. Tandon P., Garcia-Tsao G.. **Prognostic indicators in hepatocellular carcinoma: a systematic review of 72 studies**. (2009) **29** 502-510. DOI: 10.1111/j.1478-3231.2008.01957.x
32. Liotta L., Petricoin E.. **Molecular profiling of human cancer**. (2000) **1** 48-56. DOI: 10.1038/35049567
33. Goossens N., Sun X., Hoshida Y.. **Molecular classification of hepatocellular carcinoma: potential therapeutic implications**. (2015) **2** 371-379. DOI: 10.2217/hep.15.26
34. Fridman W. H., Pages F., Sautes-Fridman C., Galon J.. **The immune contexture in human tumours: impact on clinical outcome**. (2012) **12** 298-306. DOI: 10.1038/nrc3245
35. Galon J., Mlecnik B., Bindea G.. **Towards the introduction of the ‘immunoscore’ in the classification of malignant tumours**. (2014) **232** 199-209. DOI: 10.1002/path.4287
36. Zhu A. X., Duda D. G., Sahani D. V., Jain R. K.. **HCC and angiogenesis: possible targets and future directions**. (2011) **8** 292-301. DOI: 10.1038/nrclinonc.2011.30
37. Noy R., Pollard J. W.. **Tumor-associated macrophages: from mechanisms to therapy**. (2014) **41** 49-61. DOI: 10.1016/j.immuni.2014.06.010
38. Muriel P.. **Role of free radicals in liver diseases**. (2009) **3** 526-536. DOI: 10.1007/s12072-009-9158-6
39. Yongvanit P., Pinlaor S., Bartsch H.. **Oxidative and nitrative DNA damage: key events in opisthorchiasis-induced carcinogenesis**. (2012) **61** 130-135. DOI: 10.1016/j.parint.2011.06.011
40. Uchida D., Takaki A., Ishikawa H.. **Oxidative stress balance is dysregulated and represents an additional target for treating cholangiocarcinoma**. (2016) **50** 732-743. DOI: 10.3109/10715762.2016.1172071
41. Kuang Y., Sechi M., Nurra S., Ljungman M., Neamati N.. **Design and synthesis of novel reactive oxygen species inducers for the treatment of pancreatic ductal adenocarcinoma**. (2018) **61** 1576-1594. DOI: 10.1021/acs.jmedchem.7b01463
42. Arzumanyan A., Reis H. M., Feitelson M. A.. **Pathogenic mechanisms in HBV- and HCV-associated hepatocellular carcinoma**. (2013) **13** 123-135. DOI: 10.1038/nrc3449
43. Barber D. L., Wherry E. J., Masopust D.. **Restoring function in exhausted CD8 T cells during chronic viral infection**. (2006) **439** 682-687. DOI: 10.1038/nature04444
44. Jacquelot N., Yamazaki T., Roberti M. P.. **Sustained type I interferon signaling as a mechanism of resistance to PD-1 blockade**. (2019) **29** 846-861. DOI: 10.1038/s41422-019-0224-x
45. Garcia-Diaz A., Shin D. S., Moreno B. H.. **Interferon receptor signaling pathways regulating PD-L1 and PD-L2 expression**. (2017) **19** 1189-1201. DOI: 10.1016/j.celrep.2017.04.031
46. Twyman-Saint Victor C., Rech A. J., Maity A.. **Radiation and dual checkpoint blockade activate non-redundant immune mechanisms in cancer**. (2015) **520** 373-377. DOI: 10.1038/nature14292
47. Ma J., Zheng B., Goswami S.. **PD1(Hi) CD8(+) T cells correlate with exhausted signature and poor clinical outcome in hepatocellular carcinoma**. (2019) **7** p. 331. DOI: 10.1186/s40425-019-0814-7
48. Zhang R., Liu H., Li F., Li H., Yu J., Ren X.. **The correlation between the subsets of tumor infiltrating memory T cells and the expression of indoleamine 2,3-dioxygenase in gastric cancer**. (2013) **58** 3494-3502. DOI: 10.1007/s10620-013-2837-0
49. Chen J. Y., Li C. F., Kuo C. C., Tsai K. K., Hou M. F., Hung W. C.. **Cancer/stroma interplay via cyclooxygenase-2 and indoleamine 2,3-dioxygenase promotes breast cancer progression**. (2014) **16** p. 410. DOI: 10.1186/s13058-014-0410-1
50. Prieto J., Melero I., Sangro B.. **Immunological landscape and immunotherapy of hepatocellular carcinoma**. (2015) **12** 681-700. DOI: 10.1038/nrgastro.2015.173
51. Villa E., Critelli R., Lei B.. **Neoangiogenesis-related genes are hallmarks of fast-growing hepatocellular carcinomas and worst survival. Results from a prospective study**. (2016) **65** 861-869. DOI: 10.1136/gutjnl-2014-308483
52. Hoshida Y., Villanueva A., Sangiovanni A.. **Prognostic gene expression signature for patients with hepatitis C-related early-stage cirrhosis**. (2013) **144** 1024-1030. DOI: 10.1053/j.gastro.2013.01.021
53. Villanueva A., Hoshida Y., Battiston C.. **Combining clinical, pathology, and gene expression data to predict recurrence of hepatocellular carcinoma**. (2011) **140** 1501-1512.e2. DOI: 10.1053/j.gastro.2011.02.006
54. Noureddin M., Rinella M. E.. **Nonalcoholic fatty liver disease, diabetes, obesity, and hepatocellular carcinoma**. (2015) **19** 361-379. DOI: 10.1016/j.cld.2015.01.012
|
---
title: 'Prevalence of hypertension in endemic and non-endemic areas of Keshan disease:
A cross-sectional study in rural areas of China'
authors:
- Jie Hou
- Lifang Zhu
- Shuran Jin
- Jinshu Li
- Zhifeng Xing
- Yanling Wang
- Xiaoyan Wan
- Xianni Guo
- Anwei Wang
- Xiuhong Wang
- Jinming Liu
- Jing Ma
- Shuang Zhou
- Xiangdong Zhang
- Heming Zheng
- Jianhui Wang
- Hongqi Feng
- Shuqiu Sun
- Tong Wang
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9969988
doi: 10.3389/fnut.2023.1086507
license: CC BY 4.0
---
# Prevalence of hypertension in endemic and non-endemic areas of Keshan disease: A cross-sectional study in rural areas of China
## Abstract
### Background
Hypertension is a major public health concern that strongly influences the quality of life of people worldwide. Keshan disease (KD) is an endemic cardiomyopathy related to low selenium, threatening residents in rural areas of 16 provinces in China. Furthermore, the prevalence of hypertension in the KD-endemic areas has been increasing annually. However, hypertension research associated with KD has only focused on endemic regions, and no studies have compared hypertension prevalence between endemic and non-endemic areas. Therefore, this study investigated the prevalence of hypertension to provide a basis for preventing and controlling hypertension in the KD-endemic areas, even in rural areas.
### Methods
We extracted blood pressure information from cardiomyopathy investigation data from a cross-sectional study of the KD-endemic and non-endemic areas. The hypertension prevalence between the two groups was compared using the Chi-square test or Fisher s exact test. Additionally, Pearson’s correlation coefficient was employed to evaluate the relationship between the per capita gross domestic product (GDP) and hypertension prevalence.
### Results
There was a statistically significant increase of hypertension prevalence in the KD-endemic areas ($22.79\%$, $95\%$ confidence interval [CI]: 22.30–$23.27\%$) over the non-endemic areas ($21.55\%$, $95\%$ CI: 21.09–$22.02\%$). In the KD-endemic areas, more men had hypertension than women ($23.90\%$ vs. $21.65\%$, $P \leq 0.001$). Furthermore, the hypertension prevalence was higher in the north than in the south in the KD-endemic areas ($27.52\%$ vs. $18.76\%$, $P \leq 0.001$), non-endemic areas ($24.86\%$ vs. $18.66\%$, $P \leq 0.001$), and overall ($26.17\%$ vs. $18.68\%$, $P \leq 0.001$). Finally, the prevalence of hypertension positively correlated with per capita GDP at province level.
### Conclusions
The increasing hypertension prevalence is a public health problem in the KD-endemic areas. Healthy diets, such as high consumption of vegetables and seafoods, and foods that are rich in selenium, might help prevent and control hypertension in the KD-endemic areas and other rural areas in China.
## 1. Introduction
More than 200 million adults have hypertension in China, accounting for over one-fifth of China’s adult population [1]. Moreover, hypertension is a major public health concern worldwide and is a risk factor for abundant diseases, such as cardiovascular [2, 3] and kidney diseases [4]. More than half of the global adult population have not been diagnosed with or treated for hypertension. Consequently, only approximately $20\%$ of adult patients have experienced hypertension control through medical care [4]. Moreover, residents of rural areas may be at higher risk than those in urban areas in low-and middle-income countries.
Keshan disease (KD) is an endemic cardiomyopathy characterized by degeneration, necrosis, and fibrosis of cardiomyocytes, and heart dilatation, threatening the residents of rural areas in 16 provinces of China [5]. In addition, hair and serum samples from individuals from these populations and soil and grain samples from the KD-endemic areas indicated low selenium levels [6, 7]. Moreover, Mihailović et al. [ 8] found that patients with arterial hypertension had significantly lower whole-blood and plasma selenium concentrations.
A 10-year follow-up study verified that hypertension is a risk factor for latent KD worsening into chronic KD [9]. Recently, with economic development and changes in diet, the prevalence of hypertension in the KD-endemic areas has increased annually [10], exceeding the national average [11]. Previous hypertension research associated with KD has focused on the endemic areas, whilst no studies have investigated hypertension disparities between the KD-endemic and non-endemic areas. Therefore, this study used the blood pressure data from a cardiomyopathy investigation of residents of the KD-endemic and non-endemic areas in 2011 to understand the prevalence of hypertension and provide a base for preventing and controlling hypertension in the KD-endemic areas, even in the rural areas in China.
## 2.1. Multistage cluster sampling
We extracted data from a cross-sectional study comprising KD surveillance in KD-endemic counties and dilated cardiomyopathy surveyed in non-endemic counties in 13 provinces. The provinces included Heilongjiang, Nei Mongol, Jilin, Gansu, Shaanxi, Liaoning, Shanxi, Shandong, Henan, Hebei, Yunnan, Sichuan, and Chongqing. There are more endemic counties and higher KD prevalence in the 13 provinces among 16 provinces affected by KD in China. KD surveillance has been gradually conducted in those provinces since 1990. The other three provinces, Hubei, Guizhou and Tibet, were excluded due to only one KD county and few KD cases occurred. The KD-endemic areas were determined using the Delimitation and Classification of Keshan Disease Areas (GB17020-2010) [12]. In this study, we used multistage cluster sampling. For each county, we performed the case search [6] to identify two townships with the most patients with KD or dilated cardiomyopathy. Then, we selected the village with the most patients in either of the two townships for the investigation. The included villages had populations greater than or equal to 500 people. The endemic and non-endemic counties were individually matched based on the geographical location and residents’ lifestyles. Finally, hypertension data were collected from the 49 KD-endemic and 49 non-endemic counties.
## 2.2. Participants
All village residents underwent medical examinations, including blood pressure and electrocardiograms. After, patients with suspected KD or dilated cardiomyopathy were examined using echocardiography and chest radiography. We required a response rate of $80\%$ or higher or at least 400 surveyed individuals. If the quantity did not meet these requirements, it would be supplemented by the neighboring village. All included participants had lived in the surveyed village for more than 6 consecutive months or had left for no more than 3 months in the past year. We examined 43,240 and 104,166 people in the KD-endemic and non-endemic areas, respectively. Then, we extracted blood pressure data for 58,994 participants aged 20 years or older for the analysis.
## 2.3. Blood pressure measurements
After sitting in a relaxed position comfortably and quietly for more than 5 min, blood pressure was measured using a mercury sphygmomanometer. The participants were informed that smoking, drinking, and other activities resulting in blood pressure instability were forbidden for at least 30 min before the measurement. During the measurement, the elbow and forearm were bent flush with the heart, and the cuff was placed on the right bare upper arm one inch above the bend of the elbow with appropriate tightness. The disk of the stethoscope was placed face down under the cuff, just to the inner side of the upper arm, where the brachial artery pulse could be felt. The cuff was rapidly inflated until the pulse voice disappeared and continued to be pressurized until it was slowly deflated after the gauge reading had risen by 20–30 mmHg. The first loud beat heard was the systolic blood pressure (SBP), and the last beat before it disappeared was the diastolic blood pressure (DBP).
## 2.4. Economic and demographic data
Demographic data including age and sex and the per capita gross domestic product (GDP) were collected for each province in 2011 from the 2012 China Statistical Yearbook [13].
## 2.5. Ethics
All participants signed an informed statement to give permission and indicate that they had no direct interest in the study’s results. This study conformed to the Declaration of Helsinki and has been authorized by the Medical Ethics Committee of Harbin Medical University.
## 2.6. Statistical analyses
The statistical analyses were executed with R Studio version 1.4.1717.1 Hypertension was defined as either a SBP of 140 mmHg or greater or a DBP of 90 mmHg or greater or the presence of both based on the 2019 Annual Report on Cardiovascular Health and Diseases in China [14]. Hypertension was classified into three categories: Grade 1 (140–$\frac{159}{90}$–99 mmHg), Grade 2 (160–$\frac{179}{100}$–109 mmHg), and Grade 3 (≥$\frac{180}{110}$ mmHg) [15]. We excluded SBP, DBP, or pulse pressure data outside a $99.73\%$ confidence interval (CI). We also screened for duplicated records and, if identified, randomly retained one of the duplicates. Repeat data were defined as consistent information, including province, county, township, age, sex, SBP, DBP, pinyin of name, and telephone number.
The distributions of the participant characteristics were described using the population pyramid. Bar charts were used to depict the prevalence of hypertension in the KD-endemic and non-endemic areas by age categories. Forest plots were employed to describe the hypertension prevalence rate at the province level, while error bars with $95\%$ CIs were used to demonstrate hypertension differences between the sexes. The prevalence was standardized by age and sex based on the 2012 China Statistical Yearbook [13]. The prevalence between the two groups was compared using the Chi-square test or Fisher’s exact test, and the relationship between the per capita GDP and hypertension prevalence at province level was analyzed using Pearson’s correlation coefficient. The statistical significance was delimited at $P \leq 0.05.$
## 3.1. Total hypertension prevalence
We recruited 58,994 participants, including 28,738 participants from the KD-endemic areas and 30,256 from the non-endemic areas. Figure 1 presents the age and sex distributions of the respondents. The prevalence of hypertension was higher in the KD-endemic areas ($22.79\%$, $95\%$ CI: 22.30–$23.27\%$) than in the non-endemic areas ($21.55\%$, $95\%$ CI: 21.09–$22.02\%$, $P \leq 0.001$, Table 1).
**FIGURE 1:** *Population pyramid of participants in KD-endemic and non-endemic areas. KD, Keshan disease; N, number.* TABLE_PLACEHOLDER:TABLE 1
## 3.2. Hypertension prevalence by age categories, sex, and grade
Hypertension prevalence increased with age; the highest prevalence was in the 75–79 and 80+ year age groups in the endemic and non-endemic areas, respectively. Furthermore, in four age groups (50–54, 55–59, 60–64, and 70–74 years), the prevalence of hypertension was lower in the non-endemic areas than in the KD-endemic areas ($P \leq 0.05$, Figure 2 and Supplementary Table 1).
**FIGURE 2:** *Hypertension prevalence in KD-endemic and non-endemic areas by age groups. The prevalence was standardized by age and sex based on the 2012 China Statistical Yearbook. KD. Keshan disease; *P < 0.05 compared with KD-endemic areas; **P < 0.01 compared with KD-endemic areas; ***P < 0.001 compared with KD-endemic areas.*
The prevalence of hypertension among men in the KD-endemic areas exceeded that among men in the non-endemic areas and women in the KD-endemic areas ($P \leq 0.001$, Figure 3). The prevalence of Grade 1 and 2 hypertension was higher in the KD-endemic areas than in the non-endemic areas, whereas the prevalence of Grade 3 hypertension was the opposite ($P \leq 0.001$, Table 2).
**FIGURE 3:** *Hypertension prevalence in KD-endemic and non-endemic areas by sex. Error bars with $95\%$ CIs were used to demonstrate hypertension differences between the sexes. The prevalence was standardized by age and sex based on the 2012 China Statistical Yearbook. KD, Keshan disease; ***$P \leq 0.001$ compared with women in the KD-endemic areas or men in the non-endemic areas.* TABLE_PLACEHOLDER:TABLE 2
## 3.3. Hypertension prevalence by region
The annual average temperature of the capital cities in the provinces included in this investigation was 11.7°C. Therefore, provinces with an annual average temperature above 11.7°C were classified as being in the south and included Shandong, Shaanxi, Henan, Hebei, Yunnan, Sichuan, and Chongqing. Conversely, provinces with an annual average temperature below 11.7°C were classified as being in the north and included Heilongjiang, Jilin, Liaoning, Nei Mongol, Shanxi, and Gansu.
In the north, the prevalence of hypertension was significantly higher in the KD-endemic areas than in the non-endemic areas ($P \leq 0.001$). Moreover, the prevalence of hypertension was significantly higher in the north than in the south, regardless of the endemic classification ($P \leq 0.001$, Table 3).
**TABLE 3**
| Region | KD areas | Participants | Patients with hypertension | Prevalence (95% CI) | Age, sex-standardized prevalence (95% CI) |
| --- | --- | --- | --- | --- | --- |
| South | Endemic | 15741 | 3727 | 23.68% (23.01–24.35%) | 18.76% (18.15–19.38%) |
| South | Non-endemic | 16043 | 3962 | 24.70% (24.03–25.37%) | 18.66% (18.06–19.27%) |
| South | Total | 31784 | 7689 | 24.19% (23.72–24.67%) | 18.68% (18.25–19.11%) |
| North | Endemic | 12997 | 4445 | 34.20% (33.38–35.02%) | 27.52% (26.76–28.30%)a,b |
| North | Non-endemic | 14213 | 4402 | 30.97% (30.21–31.74%) | 24.86% (24.15–25.58%)b |
| North | Total | 27210 | 8847 | 32.51% (31.96–33.07%) | 26.17% (25.65–26.70%)b |
## 3.4. Hypertension prevalence by province
In the Shanxi, Henan, Heilongjiang, and Chongqing provinces, the prevalence of hypertension was significantly lower in the non-endemic areas than in the endemic areas ($P \leq 0.001$, Figure 4 and Supplementary Table 2). However, the opposite was observed in the Sichuan ($P \leq 0.001$), Shandong ($P \leq 0.001$), and Shaanxi ($P \leq 0.05$) provinces.
**FIGURE 4:** *Hypertension prevalence in KD-endemic and non-endemic areas by province. Forest plots were employed to describe the hypertension prevalence rate at the province level. The prevalence was standardized by age and sex based on the 2012 China Statistical Yearbook. KD, Keshan disease; The number in the bracket represents the total number of participants in each province. The upright solid line indicates the hypertension prevalence in the KD-endemic areas in total. The upright dotted line indicates the hypertension prevalence in the non-endemic areas in total.*
## 3.5. Hypertension prevalence and per capita GDP
Pearson’s correlation coefficient was determined to be $r = 0.6672$ ($$P \leq 0.0127$$), suggesting that the prevalence of hypertension positively correlated with the per capita GDP by province (Figure 5, Supplementary Table 3 and Supplementary Figure 1).
**FIGURE 5:** *Hypertension prevalence and per capita GDP at province level. The relationship between the per capita GDP and hypertension prevalence at province level was analyzed using Pearson’s correlation coefficient (r = 0.6672, P = 0.0127). The prevalence was standardized by age and sex based on the 2012 China Statistical Yearbook. GDP, gross domestic product.*
## 4. Discussion
We conducted a large-scale and representative study, reporting for the first time that the prevalence of hypertension is higher in the KD-endemic areas than in the non-endemic areas. This may be related to the suboptimal selenium intake of residents in the KD-endemic areas. The Western European longitudinal population study demonstrated that a 20 μg/L or higher blood selenium level at baseline reduced the risk of hypertension by $37\%$ in men [16], and adults with low toenail selenium concentrations had an increased risk of hypertension [17]. Furthermore, Xie et al. [ 18] reported a negative correlation between selenium intake and hypertension in participants in northern provinces but a positive correlation in participants in southern provinces in China. Lower urinary selenium concentrations were also associated with higher SBP and DBP values in Asian countries [19], similar to the associations identified between the serum selenium level and SBP and DBP in pregnant women [20]. However, some studies have suggested a positive correlation between selenium levels and hypertension (21–23), which may be due to the presence of high levels of selenium in the areas investigated. About half of the Chinese population does not meet the recommended selenium intake defined by the Food and Agriculture Organization and the World Health Organization [24]. Therefore, increasing the intake of selenium-rich foods might be beneficial for the residents of low-selenium areas, reducing the prevalence of hypertension.
Men are more prone to hypertension than women (25–27), which is consistent with our study’s results (Supplementary Figures 2, 3). We found that the prevalence of hypertension was higher in men than in women in the KD-endemic areas. Everett and Zajacova [28] reported that among Americans aged 24–34 years, women were far less likely to be hypertensive because they were more aware of hypertension than men. Women had more advanced hypertension awareness than men in China [1], the United States [29], and Romania [30]. Meanwhile, men consumed alcohol in larger amounts and more frequently than women [31]. After stratifying by sex, daily drinking increased the risk of hypertension in men but did not affect women in Southwest China [32]. A J-shaped relationship between alcohol consumption and hypertension has been identified in women, while alcohol consumption was linearly correlated with the risk of hypertension in men [33, 34]. These studies indicate that alcohol consumption could be the reason for a higher hypertension prevalence in men.
We found the hypertension prevalence was higher in the north than in the south, perhaps highlighting the role of temperature. A previous study reported that the prevalence of hypertension and the average SBP and DBP in northern tourists in Hainan, located in one of the most southern regions of China, were significantly higher than those of local residents and northern residents living in Hainan for more than 5 years [35]. Moreover, Duranton et al. [ 36] collected data from 261 hemodialysis patients in different latitudes, discovering that the rising outdoor temperatures and prolonged sunshine hours were associated with decreased blood pressure before dialysis. When the temperature dropped by 1°C, the SBP and DBP for the total population rose by 0.55 and 0.26 mmHg, respectively [37]. Moreover, residents in the north of China had 2.32 g more sodium daily than those in the south of China [38]. One study reported that individuals with hypertension or normal blood pressure could lower their blood pressure by moderately reducing their salt intake for 4 weeks or more [39]. Another study reported a reduction in SBP and DBP by 1.10 mmHg and 0.33 mmHg, respectively, for every 50 mmol of sodium excretion in 24 h [40]. Thus, temperature and the amount of salt in the diet may explain the distinct hypertension prevalence in the northern and southern regions.
Income has been identified as a hypertension risk factor [41], and our study supports these findings. We identified a positive correlation between the prevalence of hypertension and per capita GDP by province. In Bengal, adults in richer household wealth quintiles had a significantly higher prevalence and odds of hypertension [42], and women in the highest wealth quantile were more prone to hypertension in Kenya [43]. In developing countries, generally, hypertension is positively correlated with economic status, but the opposite is true in many developed countries, such as the United States and Canada [44]. It was revealed that higher income, occupation, and the mother’s education level were protective factors for hypertension among African Americans [45]. Diet might also play a key role in influencing blood pressure by income. In developed countries, individuals with high socioeconomic status (based on occupation, education, and income) mainly consume foods abundant in fiber and protein and low in fat [46]. In China, dietary consumption patterns are changing; the consumption of vegetable oil, animal foods, and sweeteners is increasing, and the consumption of coarse grains and beans is decreasing [47], especially among wealthier individuals [48]. Vegetables might help reduce blood pressure [49, 50], and vegans and vegetarians have been shown to have lower SBP and DBP values than omnivores [51]. A national cross-sectional study among Chinese adolescents aged 13–17 years found that adolescents whose daily vegetable consumption was three or more servings (one serving is approximately one cup, approximately 200 g) had a lower hazard ratio for high blood pressure than those who consumed less than one serving daily [52]. Another study reported that both raw (tomatoes, carrots, and shallots) and cooked (tomatoes, peas, and celery) vegetable intake significantly affected blood pressure [53]. Conversely, a Korean study found that vegetable intake did not influence the risk of hypertension [54], which might be owing to the manner of cooking. The Dietary Approaches to Stop Hypertension Diet, comprising whole cereal, vegetables, fruits, and low-fat food, was as effective as some antihypertensive drugs and significantly reduced blood pressure [55]. Not only vegetables but also seafoods lessened the risk of hypertension. The inverse relationship was identified between high seafood intake and childhood hypertension in Iranian students aged 7–12 years [56]. Seafood is abundant in omega-3 polyunsaturated fatty acids, resulting in a small but significant decrease in blood pressure [57]. Moreover, one study found that people in the highest quarter of the Omega-3 Index had an SBP and DBP 4 mmHg and 2 mmHg lower, respectively, than those in the lowest quarter [58]. It was noteworthy that obtaining more omega-3 polyunsaturated fatty acids from the diet led to a clinically related decrease in DBP in a randomized controlled trial [59]. Since the KD-endemic areas all lie within the agricultural hinterland, increasing the intake of seafood is widely advocated for preventing hypertension in the affected population and may be an important control strategy.
This study has some limitations. First, only those aged 20 or older were included owing to the 2012 China Statistical Yearbook age group classifications; adults aged 18 and 19 years were not included. Second, the survey included many participants from several rural areas of China. Thus, we did not collect information on the participants’ hypertension drug use.
In conclusion, the prevalence of hypertension was higher in the KD-endemic areas than in the non-endemic areas. Therefore, healthy diets, such as high consumption of vegetables and seafoods, and foods that are rich in selenium, might help prevent and control hypertension in the KD-endemic areas. In addition, this study provides a better understanding of hypertension statuses in rural China, which may help with prevention.
## Data availability statement
The datasets presented in this article are not readily available because the data supporting the results of this study were obtained from the Center for Endemic Disease Control, Chinese Center for Disease Control and Prevention. The data were licensed to be used in the current study, but sharing of data was not allowed; therefore, the data resource is not publicly available. Nonetheless, the data can be available upon rational demand and with the approval of the Center for Endemic Disease Control. Requests to access the datasets should be directed to JH, [email protected].
## Ethics statement
The studies involving human participants were reviewed and approved by the Medical Ethics Committee of Harbin Medical University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JH and TW performed the design and concretization of the study. JH, LZ, and SJ performed the data analysis and participated in the writing of manuscript and revision and result interpretation. JH, JSL, ZX, YW, XYW, XG, AW, XHW, JML, JM, SZ, XZ, HZ, JW, HF, and SS contributed to the field investigation and data collection. TW contributed to the project funds. All authors read the final version and approved it.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1086507/full#supplementary-material
## References
1. Wang Z, Chen Z, Zhang L, Wang X, Hao G, Zhang Z. **Status of hypertension in China: results from the China hypertension survey, 2012-2015.**. (2018) **137** 2344-56. DOI: 10.1161/CIRCULATIONAHA.117.032380
2. Yusuf S, Rangarajan S, Teo K, Islam S, Li W, Liu L. **Cardiovascular risk and events in 17 low-, middle-, and high-income countries.**. (2014) **371** 818-27. DOI: 10.1056/NEJMoa1311890
3. Zhou M, Wang H, Zeng X, Yin P, Zhu J, Chen W. **Mortality, morbidity, and risk factors in China and its provinces,1990-2017:a systematic analysis for the Global Burden of Disease Study 2017.**. (2019) **394** 1145-58. DOI: 10.1016/S0140-6736(19)30427-1
4. 4.World Health Organization.
Hypertension. (2021). Available online at: https://www.who.int/news-room/fact-sheets/detail/hypertension
(accessed April 10, 2021).. (2021)
5. Sun D.. (2011) 182 p
6. Hou J, Zhu L, Chen C, Feng H, Li D, Sun S. **Association of selenium levels with the prevention and control of Keshan disease: a cross-sectional study.**. (2021) **68**. DOI: 10.1016/j.jtemb.2021.126832
7. Liu X, Wang Y, Han S, Zhang Y, Zou Y, Su S. **A spatial ecological study on serum selenium and Keshan disease in Heilongjiang Province, China.**. (2021) **199** 3253-61. DOI: 10.1007/s12011-020-02478-0
8. Mihailović M, Avramović D, Jovanović I, Pesut O, Matić D, Stojanov V. **Blood and plasma selenium levels and GSH-Px activities in patients with arterial hypertension and chronic heart disease.**. (1998) **17** 285-9. PMID: 9726804
9. Liu X, Yang G, Wang L, Wei J, Tan W, Zhu Y. **Progressive risks of latent Keshan disease: a long term follow-up study.**. (2016) **47** 398-401. DOI: 10.13464/j.scuxbyxb.2016.03.020
10. Pang R, Zhang J. **Dynamics of incident hypertension in Keshan disease area from 2011 to 2016.**. (2017) **5**. DOI: 10.16282/j.cnki.cn11-9336/r.2017.02.067
11. Wang Z, Wang T, Zhen R, Chen X, Xu C, Han X. **Prevalence, awareness, treatment, and control of hypertension among residents in the severe historical endemic areas of Keshan disease in Heilongjiang Province.**. (2017) **36** 434-9
12. 12.The Ministry of Health of the People’s Republic of China, The standardization Administration of the People’s Republic of China.
Delimitation and Classification of Keshan Disease Areas (GB17020-2010). (2011). Available online at: http://www.nhc.gov.cn/wjw/s9500/201106/51931/files/331adfa14f7e4f6a9a40c7e055fb4ba1.pdf
(accessed April 10, 2022).. (2011)
13. 13.National Bureau of Statistics.
2012 China Statistical Yearbook. (2013). Available online at: http://www.stats.gov.cn/tjsj/ndsj/2012/indexch.htm
(Accessed April 10, 2022).. (2013)
14. 14.National Center for Cardiovascular Diseases, China.
Annual Report on Cardiovascular Health and Diseases in China (2019). (2020). Available online at: https://www.nccd.org.cn/Sites/Uploaded/File/2021/5/%E4%B8%AD%E5%9B%BD%E5%BF%83%E8%A1%80%E7%AE%A1%E5%81%A5%E5%BA%B7%E4%B8%8E%E7%96%BE%E7%97%85%E6%8A%A5%E5%91%8A%EF%BC%882019%EF%BC%89%E5%90%88.pdf
(accessed April 10, 2022).. (2020)
15. Williams B, Mancia G, Spiering W, Agabiti Rosei E, Azizi M, Burnier M. **2018 ESC/ESH Guidelines for the management of arterial hypertension: the task force for the management of arterial hypertension of the European Society of Cardiology and the European Society of Hypertension: the task force for the management of arterial hypertension of the European Society of Cardiology and the European Society of Hypertension.**. (2018) **36** 1953-2041. DOI: 10.1097/HJH.0000000000001940
16. Nawrot T, Staessen J, Roels H, Den Hond E, Thijs L, Fagard R. **Blood pressure and blood selenium: a cross-sectional and longitudinal population study.**. (2007) **28** 628-33. DOI: 10.1093/eurheartj/ehl479
17. Liu L, Lin G, Wang H, Zhang B, Du S. **Selenium exposure and incident hypertension among Chinese adults (P24-020-19).**. (2019) **3**. DOI: 10.1093/cdn/nzz044.P24-020-19
18. Xie C, Xian J, Zeng M, Cai Z, Li S, Zhao Y. **Regional difference in the association between the trajectory of selenium intake and hypertension: a 20-year cohort study.**. (2021) **13**. DOI: 10.3390/nu13051501
19. Mizuno Y, Shimizu-Furusawa H, Konishi S, Inaoka T, Ahmad S, Sekiyama M. **Associations between urinary heavy metal concentrations and blood pressure in residents of Asian countries.**. (2021) **26**. DOI: 10.1186/s12199-021-01027-y
20. Eze S, Ododo N, Ugwu E, Enebe J, Onyegbule O, Eze I. **Serum selenium levels of pre-eclamptic and normal pregnant women in Nigeria: a comparative study.**. (2020) **15**. DOI: 10.1371/journal.pone.0238263
21. Vinceti M, Chawla R, Filippini T, Dutt C, Cilloni S, Loomba R. **Blood pressure levels and hypertension prevalence in a high selenium environment: results from a cross-sectional study.**. (2019) **29** 398-408. DOI: 10.1016/j.numecd.2019.01.004
22. Wu G, Li Z, Ju W, Yang X, Fu X, Gao X. **Cross-sectional study: relationship between serum selenium and hypertension in the Shandong Province of China.**. (2018) **185** 295-301. DOI: 10.1007/s12011-018-1272-7
23. Bastola M, Locatis C, Maisiak R, Fontelo P. **Selenium, copper, zinc and hypertension: an analysis of the National Health and Nutrition Examination Survey (2011-2016).**. (2020) **20**. DOI: 10.1186/s12872-020-01355-x
24. Dinh Q, Cui Z, Huang J, Tran T, Wang D, Yang W. **Selenium distribution in the Chinese environment and its relationship with human health: a review.**. (2018) **112** 294-309. DOI: 10.1016/j.envint.2017.12.035
25. Defianna S, Santosa A, Probandari A, Dewi F. **Gender differences in prevalence and risk factors for hypertension among adult populations: a cross-sectional study in Indonesia.**. (2021) **18**. DOI: 10.3390/ijerph18126259
26. Bachir Cherif A, Bouamra A, Taleb A, Nedjar R, Bouraghda A, Hamida F. **Differences in prevalence, treatment and control rates of hypertension between male and female in the area of Blida (Algeria).**. (2017) **66** 123-9. DOI: 10.1016/j.ancard.2017.04.009
27. Agho K, Osuagwu U, Ezeh O, Ghimire P, Chitekwe S, Ogbo F. **Gender differences in factors associated with prehypertension and hypertension in Nepal: a nationwide survey.**. (2018) **13**. DOI: 10.1371/journal.pone.0203278
28. Everett B, Zajacova A. **Gender differences in hypertension and hypertension awareness among young adults.**. (2015) **61** 1-17. DOI: 10.1080/19485565.2014.929488
29. Zhang Y, Moran A. **Trends in the prevalence, awareness, treatment, and control of hypertension among young adults in the United States, 1999 to 2014.**. (2017) **70** 736-42. DOI: 10.1161/HYPERTENSIONAHA.117.09801
30. Dorobantu M, Gheorghe-Fronea O, Scafa-Udriste A, Onciul S, Pop C, Dorobantu L. **Gender disparities in the prevalence, awareness, treatment, and control of high blood pressure.**. (2021) **27** 3172-9. DOI: 10.2174/1381612827666210125154749
31. Erol A, Karpyak V. **Sex and gender-related differences in alcohol use and its consequences: contemporary knowledge and future research considerations.**. (2015) **156** 1-13. DOI: 10.1016/j.drugalcdep.2015.08.023
32. Wang Y, Yao Y, Chen Y, Zhou J, Wu Y, Fu C. **Association between drinking patterns and incident hypertension in Southwest China.**. (2022) **19**. DOI: 10.3390/ijerph19073801
33. Briasoulis A, Agarwal V, Messerli F. **Alcohol consumption and the risk of hypertension in men and women: a systematic review and meta-analysis.**. (2012) **14** 792-8. DOI: 10.1111/jch.12008
34. Sesso H, Cook N, Buring J, Manson J, Gaziano J. **Alcohol consumption and the risk of hypertension in women and men.**. (2008) **51** 1080-7. DOI: 10.1161/HYPERTENSIONAHA.107.104968
35. Lin K, Lin L, He C, Pang M, Chen H. **Epidemic characteristics of hypertension in south and north China.**. (2014) **35** 114-7. DOI: 10.14188/j.1671-8852.2014.01.015
36. Duranton F, Palma A, Stegmayr B, Wauthier M, Torres A, Argilés À. **Blood pressure seasonality in hemodialysis patients from five European cities of different latitudes.**. (2018) **43** 1529-38. DOI: 10.1159/000494019
37. Hu J, He G, Luo J, Xu Y, Xu X, Song X. **Temperature-adjusted hypertension prevalence and control rate: a series of cross-sectional studies in Guangdong Province, China.**. (2021) **39** 911-8. DOI: 10.1097/HJH.0000000000002738
38. Wang M, Huang Y, Song Y, Chen J, Liu X. **Study on environmental and lifestyle factors for the North-South differential of cardiovascular disease in China.**. (2021) **9**. DOI: 10.3389/fpubh.2021.615152
39. He F, MacGregor G. **Effect of modest salt reduction on blood pressure: a meta-analysis of randomized trials. Implications for public health.**. (2002) **16** 761-70. DOI: 10.1038/sj.jhh.1001459
40. Huang L, Trieu K, Yoshimura S, Neal B, Woodward M, Campbell N. **Effect of dose and duration of reduction in dietary sodium on blood pressure levels: systematic review and meta-analysis of randomised trials.**. (2020) **368**. DOI: 10.1136/bmj.m315
41. Sharma J, Mabhida S, Myers B, Apalata T, Nicol E, Benjeddou M. **Prevalence of hypertension and its associated risk factors in a rural black population of Mthatha Town, South Africa.**. (2021) **18**. DOI: 10.3390/ijerph18031215
42. Kibria G, Hashan M, Hossain M, Zaman S, Stennett C. **Clustering of hypertension, diabetes and overweight/obesity according to socioeconomic status among Bangladeshi adults.**. (2021) **53** 157-66. DOI: 10.1017/S0021932020000085
43. Yaya S, El-Khatib Z, Ahinkorah B, Budu E, Bishwajit G. **Prevalence and socioeconomic factors of diabetes and high blood pressure among women in Kenya: a cross-sectional study.**. (2021) **11** 397-404. DOI: 10.1007/s44197-021-00004-6
44. Leng B, Jin Y, Li G, Chen L, Jin N. **Socioeconomic status and hypertension: a meta-analysis.**. (2015) **33** 221-9. DOI: 10.1097/HJH.0000000000000428
45. Glover L, Cain-Shields L, Wyatt S, Gebreab S, Diez-Roux A, Sims M. **Life course socioeconomic status and hypertension in African American adults: the Jackson Heart Study.**. (2020) **33** 84-91. DOI: 10.1093/ajh/hpz133
46. Darmon N, Drewnowski A. **Does social class predict diet quality?**. (2008) **87** 1107-17. DOI: 10.1093/ajcn/87.5.1107
47. Popkin B. **Synthesis and implications: China’s nutrition transition in the context of changes across other low- and middle-income countries.**. (2014) **15** 60-7. DOI: 10.1111/obr.12120
48. Popkin B. **The nutrition transition: an overview of world patterns of change.**. (2004) **62** S140-3. DOI: 10.1111/j.1753-4887.2004.tb00084.x
49. van der Avoort C, Jonvik K, Nyakayiru J, van Loon L, Hopman M, Verdijk LB. **A nitrate-rich vegetable intervention elevates plasma nitrate and nitrite concentrations and reduces blood pressure in healthy young adults.**. (2020) **120** 1305-17. DOI: 10.1016/j.jand.2020.02.014
50. He Z.. (2021)
51. Pettersen B, Anousheh R, Fan J, Jaceldo-Siegl K, Fraser G. **Vegetarian diets and blood pressure among white subjects: results from the Adventist Health Study-2 (AHS-2).**. (2012) **15** 1909-16. DOI: 10.1017/S1368980011003454
52. Yang Y, Dong B, Zou Z, Wang S, Dong Y, Wang Z. **Association between vegetable consumption and blood pressure, stratified by BMI, among Chinese adolescents aged 13-17 years: a national cross-sectional study.**. (2018) **10**. DOI: 10.3390/nu10040451
53. Chan Q, Stamler J, Brown I, Daviglus M, Van Horn L, Dyer A. **Relation of raw and cooked vegetable consumption to blood pressure: the INTERMAP Study.**. (2014) **28** 353-9. DOI: 10.1038/jhh.2013.115
54. Kim J, Kim J. **Association between fruit and vegetable consumption and risk of hypertension in middle-aged and older Korean adults.**. (2018) **118** 1438-49. DOI: 10.1016/j.jand.2017.08.122
55. Strilchuk L, Cincione R, Fogacci F, Cicero A. **Dietary interventions in blood pressure lowering: current evidence in 2020.**. (2020) **78** 659-66. DOI: 10.33963/KP.15468
56. Cabo J, Alonso R, Mata P. **Omega-3 fatty acids and blood pressure.**. (2012) **107** S195-200. DOI: 10.1017/S0007114512001584
57. Izadi A, Khedmat L, Tavakolizadeh R, Mojtahedi S. **The intake assessment of diverse dietary patterns on childhood hypertension: alleviating the blood pressure and lipidemic factors with low-sodium seafood rich in omega-3 fatty acids.**. (2020) **19**. DOI: 10.1186/s12944-020-01245-3
58. Filipovic M, Aeschbacher S, Reiner M, Stivala S, Gobbato S, Bonetti N. **Whole blood omega-3 fatty acid concentrations are inversely associated with blood pressure in young, healthy adults.**. (2018) **36** 1548-54. DOI: 10.1097/HJH.0000000000001728
59. Stanton A, James K, Brennan M, O’Donovan F, Buskandar F, Shortall K. **Omega-3 index and blood pressure responses to eating foods naturally enriched with omega-3 polyunsaturated fatty acids: a randomized controlled trial.**. (2020) **10**. DOI: 10.1038/s41598-020-71801-5
|
---
title: Identification and characterization of a novel molecular classification incorporating
oxidative stress and metabolism-related genes for stomach adenocarcinoma in the
framework of predictive, preventive, and personalized medicine
authors:
- Ying Dong
- Qihang Yuan
- Jie Ren
- Hanshuo Li
- Hui Guo
- Hewen Guan
- Xueyan Jiang
- Bing Qi
- Rongkuan Li
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9969989
doi: 10.3389/fendo.2023.1090906
license: CC BY 4.0
---
# Identification and characterization of a novel molecular classification incorporating oxidative stress and metabolism-related genes for stomach adenocarcinoma in the framework of predictive, preventive, and personalized medicine
## Abstract
### Background
Stomach adenocarcinoma (STAD) is one of the primary contributors to deaths that are due to cancer globally. At the moment, STAD does not have any universally acknowledged biological markers, and its predictive, preventive, and personalized medicine (PPPM) remains sufficient. Oxidative stress can promote cancer by increasing mutagenicity, genomic instability, cell survival, proliferation, and stress resistance pathways. As a direct and indirect result of oncogenic mutations, cancer depends on cellular metabolic reprogramming. However, their roles in STAD remain unclear.
### Method
743 STAD samples from GEO and TCGA platforms were selected. Oxidative stress and metabolism-related genes (OMRGs) were acquired from the GeneCard Database. A pan-cancer analysis of 22 OMRGs was first performed. We categorized STAD samples by OMRG mRNA levels. Additionally, we explored the link between oxidative metabolism scores and prognosis, immune checkpoints, immune cell infiltration, and sensitivity to targeted drugs. A series of bioinformatics technologies were employed to further construct the OMRG-based prognostic model and clinical-associated nomogram.
### Results
We identified 22 OMRGs that could evaluate the prognoses of patients with STAD. Pan-cancer analysis concluded and highlighted the crucial part of OMRGs in the appearance and development of STAD. Subsequently, 743 STAD samples were categorized into three clusters with the enrichment scores being C2 (upregulated) > C3 (normal) > C1 (downregulated). Patients in C2 had the lowest OS rate, while C1 had the opposite. Oxidative metabolic score significantly correlates with immune cells and immune checkpoints. Drug sensitivity results reveal that a more tailored treatment can be designed based on OMRG. The OMRG-based molecular signature and clinical nomogram have good accuracy for predicting the adverse events of patients with STAD. Both transcriptional and translational levels of ANXA5, APOD, and SLC25A15 exhibited significantly higher in STAD samples.
### Conclusion
The OMRG clusters and risk model accurately predicted prognosis and personalized medicine. Based on this model, high-risk patients might be identified in the early stage so that they can receive specialized care and preventative measures, and choose targeted drug beneficiaries to deliver individualized medical services. Our results showed oxidative metabolism in STAD and led to a new route for improving PPPM for STAD.
## Introduction
Stomach adenocarcinoma (STAD) is the most frequent histological form of gastric cancer (GC), which is ranked the third major contributor to cancer-caused deaths [1]. Nearly half ($47\%$) of the world’s new GC cases are diagnosed in China annually; over $60\%$ of these patients experienced advanced disease at the time of diagnosis and treatment with a 5-year survival rate below $30\%$ [2, 3]. Stomach cancer responds poorly to standard therapies including surgery, chemotherapy, and radiotherapy. Even though molecularly targeted drugs have been developed for stomach cancer, the targeted treatment for this illness is still years behind what it is for lung cancer, breast cancer, colon cancer, and other common forms of cancer [4, 5]. As a result, it is of the utmost importance to locate valid biomarkers to accurately anticipate the outcome (prognosis) of STAD patients and to provide individualized therapy.
Oxidative stress is characterized by increased levels of intracellular reactive oxygen species (ROS), which may be harmful to DNA as well as proteins and lipids [6]. According to evidence, excessive ROS production results in oxidative stress in tissues and cells, which results in several disorders, including cancer [7, 8]. ROS is a potent mutagen that has been linked to the onset and advancement of cancer [9]. Meanwhile, oxidative stress promotes tumor cell survival and develops tumor angiogenesis and metastasis [10, 11]. High-speed cell development in solid tumors causes insufficient blood flow, which creates hypoxic areas within the tumor and encourages the creation of ROS [12]. Metabolic reprogramming is a feature that is common in many different types of solid tumors and affects the availability of nutrients and the manner that cells utilize those resources to fulfill the material and energy needs of cancer (13–15). Oncogene-driven metabolic modifications provide cancerous cells with the ability to survive and thrive in the tumor microenvironment (TME) [16]. The prognosis of patients and their responsiveness to chemotherapy and immunotherapy are strongly correlated with metabolic heterogeneity [17, 18]. Previous research has shown that patients with GC have considerable metabolic remodeling [19, 20]. As per our current understanding, the prognostic capabilities of oxidative stress and metabolism in GC are yet to be clarified. In-depth research on oxidative stress and metabolism has not yet been conducted to determine how these factors influence GC in a global manner. As a result, additional research must be conducted into the relationship between oxidative stress and metabolism-related genes (OMRGs) and GC.
In this study, we evaluated the copy number variation (CNV), single nucleotide variation (SNV), methylation, mRNA expression, prognostic significance, and pathway regulation changes in the OMRGs of various cancer types using gene data and clinical data gleaned from The Cancer Genome Atlas (TCGA) database. Then, by using bioinformatics techniques, we comprehensively assessed the OMRGs and STAD prognosis. As per the total oxidative metabolism score as well as the expression levels of OMRGs, we classified the dataset of STAD patients into three clusters. We also examined how these three clusters were connected to prognosis and therapies for patients. Owing to the mounting evidence suggesting a pivotal function for immune cells and immune checkpoint genes (ICGs) in tumorigenesis [21, 22], We probed whether the oxidative metabolism score was linked to ICI and ICGs. Additionally, we developed a risk score model based on13 OMRGs to anticipate the prognosis of STAD patients. Finally, the development of a nomogram to forecast survival probabilities for patients with STAD could be employed to bolster clinical judgment and tailor treatment for each patient. Our findings provided promising insights into oxidative stress and metabolism in STAD and paved the way for directing clinical therapy for individualized treatment within the framework of predictive, preventive, and personalized medicine (PPPM).
## Data collection and processing
The TCGA database was searched to retrieve RNA-sequencing (RNA-seq) data and matched clinical features from the TCGA-STAD cohort. The GSE84437 cohort’s RNA-seq and clinical data were acquired from the GEO database. Following the exclusion of patients with a survival of 30 days or less, bioinformatics analysis was performed on 312 STAD and 32 normal samples from TCGA, and 431 STAD samples from GSE84437 [23]. Batch normalization was done with the aid of the “sva” package in R [24]. We also obtained SNV data, transcriptome profiles, CNV data, methylation data, and clinical variables of pan-cancer transcriptomes from the TCGA platform [25].
In addition, 3098 oxidative stress-related genes and 2803 metabolism-related genes (Relevance score>2.5) were acquired from the GeneCard Database (https://www.genecards.org/). GeneCards mines and integrates comprehensive information on human genetics from more than 80 data sources and provides concise genome, proteome, transcriptome, disease, and function data on all known and predicted human genes. After taking the intersection of the two groups of genes, 1520 OMRGs were obtained and displayed by a Venn diagram. The OMRGs in the TCGA and GEO cohorts with prognostic significance were obtained via univariate cox regression analysis. Finally, we obtained 22 OMRGs with prognostic significance and applied the “corrplot” package to visualize the co-expression correlation between any two OMRGs. The prognostic significance of 22 OMRGs was also validated using Kaplan-Meier (KM) analysis based on 743 STAD samples.
## Pan-cancer analysis
Currently, despite the fact that there may be a link between oxidative metabolism and cancer, very little research has been performed on the topic. Therefore, there is a lack of a comprehensive account of how OMRGs vary between cancer types. To offer a holistic view of OMRGs across cancer types, we examined and visually displayed data on SNVs, CNVs, methylation, and mRNA expression. To additionally evaluate the roles of OMRGs in cancer prognosis, we examined the link between mRNA expression and OS by a univariate Cox regression model. The R programming language was applied to perform all analyses [26, 27].
To uncover the unique role of pathways affected by OMRGs across a spectrum of human cancers, we employed single-sample gene set enrichment analysis (ssGSEA) to calculate OMRGs scores in each cancer sample. Both the top and bottom $30\%$ of OMRG scores were used to classify the samples into two groups. Differences (variations) in pathway activity across the two groups premised on the transcriptomes were studied via gene set enrichment analysis (GSEA).
## Analysis of clusters premised on oxidative metabolism score
*We* generated an oxidative metabolism score model as per mRNA expression to highlight the differential expression levels across the samples as a result of the huge discrepancies in gene expression patterns that were seen in the previously collected datasets. In short, the enrichment scores of OMRGs were initially assessed using a ssGSEA [28]. The RStudio “Gplots” package was utilized for performing the differential analysis, while the “pheatmap” program was employed to generate the heat map of the cluster analysis outcomes. Comparing the mRNA expression patterns of the genes in normal samples with those in cancer tissues allowed for the classification of the tumors into three categories (groups) based on their mRNA expression status: oxidative metabolism inactive (cluster 1 or C1), oxidative metabolism active (cluster 2 or C2), and normal oxidative metabolism (cluster 3 or C3). To additionally emphasize the links between the gene expression profiles of these 3 clusters, we employed the violin plot to illustrate the enrichment scores of the clusters, which were generated with the aid of the “ggpubr” package [29]. Afterward, Kaplan-Meier (KM) analysis was executed to evaluate the prognostic relevance of clusters. Finally, we downloaded and curated 50 typical hallmark pathways from the Molecular Signatures Database (MsigDB) [30]. Through ssGSEA analysis of tumor cells and normal cells in each sample, we obtained the enrichment score of each pathway. A heatmap was utilized to show the discrepancies between pathway enrichment scores and oxidative metabolism scores among three clusters. $P \leq 0.05$ signifies statistical significance.
## Association between the oxidative metabolism score and immune status
As part of the ESTIMATE (Estimation of Stromal and Immune cells in Malignant Tumor tissues using Expression data) study, the immune and stroma scores were derived utilizing the “estimate” in the R program. In addition, the algorithm enabled the assessment of tumor purity. The 29 TCGA-retrieved immune-associated gene sets were quantified using ssGSEA, yielding 707 genes in total that reflect distinct immune cell types, functions, and pathways (31–33). ssGSEA allows for the study of gene signals produced by multiple cell groups within the immune system in a single sample. To assess and visualize the link between the oxidative metabolism score and immunological components, we utilized the RStudio packages “ggstatsplot,” “data.table,” “dplyr,” “tidyr,” and “ggplot2.” *The area* of each sphere in the mapped figure stands for the degree of correlation, whereas the color stands for the p-value. Furthermore, the “ggscatterstats” program was employed to create a scatter plot showing the relationships among the six standard immune cell groups (viz., Macrophages, CCR, mast cells, type II interferon (IFN) response, Dendritic cells (DCs), and T follicular helper (Tfh) cells), and the oxidative metabolism score. In addition, When comparing the levels of expression of common ICGs across low- and high-risk groupings, we only displayed statistically significant findings. $p \leq 0.05$: statistical significance.
## Drug sensitivity analysis
Each STAD patient’s medication responsiveness was predicted utilizing the “pRRophetic” R package [34, 35], and the possibly sensitive medications were screened for high- and low-risk categories with the “Wilcox.test” function in R. It should be noted that drug sensitivity improves with decreasing IC50 values. $p \leq 0.05$:statistical significance.
## Construction and verification of a prognostic signature based on OMRGs
Then, LASSO-Cox regression analysis was applied to 22 OMRGs linked to patients’ prognoses, with minimal criteria determining the penalty parameter (λ) [36]. The equation for computing the risk score was as follows: Risk score = ∑$k = 1$nexpk*βk. We used the median risk score to designate STAD patients in the TCGA and GSE84437 cohorts as either high- or low-risk. For further analysis, the TCGA cohort served as the training cohort, whereas the GSE84437 cohort served as the test cohort.
For the formulation and verification of the signature, the following stages were performed on the train and test cohorts [1]: visualizing sample categorization was done using principal component analysis (PCA) [2]; a survival study using the KM technique was carried out to determine whether the signature could be used to forecast survival [3]; time-dependent receiver operating characteristic (ROC) curves and AUC values were created with the aid of the “survival ROC” R package to evaluate the sensitivity and specificity of the risk score.
## Variations in ICG expression and immune function between low- and high-risk populations
First, we evaluated the variations in the expression of common ICGs between high- and low-risk groups, and we only showed findings that were statistically significant ($p \leq 0.05$). The correlation between immune function and risk levels was then investigated. The above analysis was performed in both the train and test cohorts.
## Creating a predictive nomogram that incorporates clinical characteristics and risk score
Each patient’s clinical information (age, gender, grade, and stage) was retrieved from TCGA cohorts along with their risk scores. The indicators that showed statistical significance ($p \leq 0.05$) in the univariate Cox survival analysis were subsequently incorporated into the multivariate Cox survival analysis. These indicators independently functioned as prognostic variables as per the results of the multivariate analysis ($p \leq 0.05$). The aforementioned clinical characteristics and risk score were used to develop a nomogram. Thereafter, the nomogram’s discriminating power and prediction accuracy were assessed using calibration curves. The prediction performance was also assessed using the time-dependent ROC curve.
## Verification of thirteen model genes by GEPIA and HPA platforms
GEPIA is a web-based data management system for the systematic study of enormous volumes of RNA-seq data from the TCGA and GTEx datasets (http://gepia.cancer-pku.cn/) [37]. We used the GEPIA database to compare the expression of 13 OMRGs between cancer and paired normal tissues. The Human Protein Atlas (https://www.proteinatlas.org/) is a database of immunohistochemistry-based protein expression profles in cancer tissues, normal tissues, and cell lines [38]. We used it to compare the protein expression levels of 13 OMRGs between cancer and normal samples. However, CTLA4 and GRP protein expression data could not be located in the HPA database, thus we reported the protein expression levels of the remaining 11 OMRGs. In addition, antibody staining in various forms of cancer found in human tissue was classified as not observed, low, medium, or high in the HPA dataset. This score was computed by considering both the intensity of staining and the proportion of total stained cells. Similarly, the HPA database was utilized to demonstrate the immunofluorescence localization of cells.
## Data procession
Figure 1 is a flowchart that demonstrates the processes involved in the present research. First, we took the intersection of oxidative stress-related genes and metabolic-related genes to obtain 1520 OMRGs (Figure 2A). Then, the intersection was taken after univariate analysis in TCGA and GEO cohorts respectively, and 22 OMRGs with prognostic significance were obtained for analysis in this study (Figure 2B). KM analysis based on 743 STAD samples validated the prognostic significance of 22 OMRGs (Supplementary Figure 1). To explore the relationship between OMRGs, a co-expression analysis of the 22 OMRGs was carried out. As per the findings, the majority of genes had a positive link to one another across both the TCGA and the GEO cohorts. However, SLC25A15 was negatively associated with most genes (Figure 2C).
**Figure 1:** *The flowchart of the research methodology.* **Figure 2:** *22 oxidative metabolism-related genes (OMRGs) with prognostic significance were obtained for subsequent analysis. (A) Venn diagram to find 1520 OMRGs. (B) Venn diagram depicting 22 prognostic OMRGs in both TCGA and GEO datasets. (C) The plot shows the result of co-expression relationships of 22 OMRGs in STAD. The size of the dot reflects the value of the p.*
## Pan-cancer analysis of OMRGs
The CNV, methylation, SNV, mRNA expression patterns, and survival data for 22 OMRGs across a diverse range of cancer types for the pan-cancer investigation were derived from TCGA. To identify the frequency and types of variants in each cancer subtype, we evaluated SNV data linked to OMRGs. Supplementary Figure 2A demonstrates that SKCM, UCEC, LUSC, and COAD all displayed remarkable SNV frequency of OMRGs. The OMRGs have a $73.59\%$ (1510 of 2052 tumors) frequency of SNVs. Based on the results of the variant analysis, missense mutations were shown to be the most common form of SNP. As per the percentage of SNVs, the five most mutated genes were identified as follows: VWF, FBN1, NOS3, NOX4, and GAD1, of which the mutation percentages were $22\%$, $22\%$, $10\%$, $8\%$, and $7\%$, respectively (Supplementary Figure 2B). To delve deeper into the genetic alterations of OMRGs that occur in cancer, the proportion of CNV was analyzed. The findings revealed that while general CNV appeared at high frequencies in the majority of cancers (Figure 3A), all of the OMRGs in THCA demonstrated a low frequency of CNV. OMRGs displayed a wide variety of CNV characteristics. For instance, SERPINE1, APOD, CAV1, NOS3, and CAV2 were more likely to achieve CNV gain than a loss in almost all cancers, but GSTO1 and ST3GAL4 had the reverse profile. In addition to CNV, promoter methylation can control gene expression, and abnormal promoter DNA methylation is linked to cancer [39]. We noticed that among the 20 cancer types, the majority of OMRGs displayed intricate methylation patterns. However, FBN1 and GRP consistently showed hypermethylation in several tumors, while NOS3 and CTLA4 showed the opposite (Figure 3B).
**Figure 3:** *Panoramic depiction of OMRGs in pan-cancer. (A) The frequency of copy number variation (CNV) for each OMRG in each type of cancer is displayed in a histogram. (B) Genes that are hypermethylated and hypomethylated are indicated in red and blue, respectively, on a heatmap that displays the differential methylation of OMRGs in malignancies (Wilcoxon rank-sum test). (C) The histogram in the top panel shows the total number of substantially differentially expressed genes, whereas the heatmap depicts the fold change and FDR of OMRGs in each tumor. (D) OMRGs’ survival profiles across cancers. (E) Comparison of tumor samples with high and low OMRGs scores by performing enrichment analysis for cancer pathway signaling.*
For every cancer type, differential expression analysis was done to evaluate the variations in the gene expression patterns of OMRGs in addition to genetic changes between the tumor and nearby normal tissues. The findings illustrated that most of the gene expression profiles in cancer tissues were distinct from those observed in healthy tissues, except for those found in pancreatic cancer tissues. Significantly elevated expression profiles of LOX and NOX4 were seen in a variety of malignancies (Figure 3C). Subsequently, as depicted in Figure 3D, we applied univariate Cox regression of mRNA expression and OS to discover risky OMRGs with HR > 1 and p-value of< 0.05 as well as protective OMRGs with HR< 1 and p-Value of< 0.05. We found that most of the genes were risk factors in several cancers, except for CTLA4, KCNQ1, and GLS2. Since it is currently unknown how oxidative metabolism regulates pathways connected to cancer, it is imperative to examine any possible connections between these pathways and OMRGs. This will establish the groundwork for future research into how OMRGs regulate pathways related to pan-cancer. We discovered that TNFA signaling through NFKB, KRAS signaling, epithelial-mesenchymal transition (EMT), the inflammatory response, hypoxia, and the interferon-gamma response, were all remarkably correlated with OMRGs in pan-cancer (Figure 3E).
## Cluster analysis centered on the oxidative metabolism score
We clustered the OMRGs and separated the 743 STAD patients into three categories predicated on their final oxidative metabolism score and gene expression patterns for the purpose of delving deeper into the connection between oxidative metabolism and STAD. C1 contained tumor tissues with inactive oxidative metabolism, C2 featured tumor tissues with active oxidative metabolism, and C3 contained tumor tissues with normal oxidative metabolism (Figure 4A). As per the violin plot, the following is the sequence in which the enrichment scores for the 3 clusters appeared: C2 > C3 > C1 (Figure 4B). The survival curves for the three groupings were then plotted to see if the clustering made sense. Patients in group C2 had the highest OS while those in group C1 had the opposite (Figure 4C), indicating that the oxidative metabolism score represents a risky factor. The heatmap illustrated the expression profiles of 22 OMRGs in the three subgroups (Figure 4D). Finally, the links between cancer hallmarks and oxidative metabolism scores were evaluated, and the findings revealed that more than half of the hallmarks were frequently remarkably linked to oxidative metabolism scores (Figure 4E).
**Figure 4:** *Oxidative metabolism scores-based cluster analysis. (A) Heat map cluster analysis of TCGA and GEO gene data illustrating three distinct clusters: inactive oxidative metabolism (cluster 1/C1), active oxidative metabolism (cluster 2/C2), and normal oxidative metabolism (cluster 3/C3). (B) C2 has the highest enrichment score, followed by C3 and C1 in descending order, as observed by the “ggpubr” package’s violin plot. (C) The survival curve of the three clusters. Years (depicted by the abscissa) are plotted against a survival rate (denoted by the ordinate). (D) The results of differentially expressed OMRGs between three clusters. (E) The relationship between cancer signaling and OMRGs (red color to blue color signifies high to low). *P < 0.05; ***P < 0.001; ****P < 0.0001.*
## Correlations between the oxidative metabolism score and ICI
The tumor microenvironment (TME) encompasses stromal, tumor, and immune cells, as well as secreted chemokines and cytokines [40], which modulate the onset and advancement of cancer [41]. The clinical outcome of cancer is tightly tied to immune cells, which are an important part of the TME and useful anticancer therapeutic targets [42]. We evaluated TME components in C1, C2, and C3 to probe into the link between oxidative metabolism and immune response among STAD patients. The TME components in the different oxidative metabolic clusters were as follows: ESTIMATEScore: C1 > C3 > C2; ImmuneScore: C1 > C3 > C2; StromalScore: C1 > C3 > C2; tumor purity: C2 > C3 > C1 (Figures 5A–D). Next, we examined how the oxidative metabolic score interacted with ICI (Figures 5E–K) and discovered that it was positively linked to the infiltration levels of macrophages, CCR, mast cells, type II IFN response, DCs, and T helper cells. As a compensatory strategy, a higher incidence of ICI may have occurred due to a deficient local immune response. As shown in Figure 5L, most ICGs expression was enhanced in the C2 subtype. Increased ICG expression inhibited effective anti-cancer immune responses, thereby inducing the migration of immunocytes into the TME to enhance compensatory responses.
**Figure 5:** *Comparative immunological status examination of two molecular groups. (A-D) Comparative analysis of the TME components. (E) The plot illustrates the link between the score for oxidative metabolism and the infiltration of immune cells. (F-K) The scatter figure illustrates the link between the oxidative metabolism score and six substances associated with immune infiltration. A favorable correlation was identified between the oxidative metabolism score and the infiltration of macrophages, CCR, mast cells, type II IFN response, DCs, and T helper cells. (L) Comparison of the three subtypes’ immune checkpoint gene expression using a differential analysis method. *P < 0.05; **P < 0.01; ***P < 0.001.*
## Association of drug sensitivity with the oxidative metabolism clusters
Given that molecularly targeted treatment is used to treat STAD at present, we tested these 3 oxidative metabolism clusters against 12 different medications. The majority of these drugs are either frequently used targeted therapies, especially STAD, or standard pharmacological treatments utilized in the study of tumors. As per the results of the ridge regression model, the various angiogenesis clusters showed the following patterns of drug sensitivity: Sunitinib: C2 > C3 > C1; Dasatinib: C2 > C3 > C1; Imatinib: C2 > C3 > C1; Midostaurin: C2 > C3 > C1; Bexarotene: C2 > C3 > C1; Pazopanib: C2 > C3 > C1; Lapatinib: C3 > C2 > C1; Sorafenib: C1 > C3 > C2; Paclitaxel: C1 > C3 > C2; Methotrexate: C1 > C3 > C2; Tipifarnib: C1 > C3 > C2; and Vinorelbine: C1 > C3 > C2 (Figures 6A–L).
**Figure 6:** *The link between sensitivity to pharmaceuticals and clusters of oxidative metabolism. The box plots of the predicted IC50 values for twelve different kinds of commonly used chemotherapy drugs are displayed in (A–L) for cluster 1 (blue), cluster 2 (yellow), and cluster 3 (green). The following are the 12 different kinds of chemotherapy-related drugs: Sunitinib, Dasatinib, Imatinib, Midostaurin, Bexarotene, Pazopanib, Lapatinib, Sorafenib, Paclitaxel, Methotrexate, Tipifarnib, and Vinorelbine.*
## Determination and verification of an OMRGs-based prognostic signature
To determine if the OMRGs might be used to generate a signature for anticipating the therapeutic outcomes of STAD patients, LASSO-Cox regression was applied to analyze the 22 genes. Ultimately, 13 genes were chosen to create the risk score model (Supplementary Figure 3). Risk score = (-0.054980970832357) * expression of SLC25A15 + 0.332441495029009 * expression of GSTO1 + 0.0711757410696192 * expression of VWF+ 0.08897393300098 * expression of ANXA5 + 0.144637581187946 * expression of SERPINE1 + 0.140239358545706 * expression of GRP + (-0.0683192836732951) * expression of COX10 + 0.0057794428801345 * expression of APOD + (-0.104333942341152) * expression of GAD1 + 0.140768401184509 * expression of NOS3 + (-0.19273718582153) * expression of CTLA4 + (-0.0301338387672) * expression of KCNQ1 + (-0.0404719679140591) * expression of GLS2. Then, samples from the training cohort were stratified into high- and low-risk categories premised on the median risk score (Figure 7A). Patients who had high-risk scores exhibited a dismal chance of survival, as measured by the probability distributions of risk scores (Figure 7B). Figure 7C shows a PCA that can be used to easily differentiate between high- and low-risk groups (categories) The high-risk category exhibited a substantially more unfavorable OS, disease-specific survival (DSS), disease-free interval (DFI), and progression-free interval (PFI) ($p \leq 0.05$), as depicted in Figures 7D-G. In addition, the 1-, 3-, 5-, and 7-year AUC values for the survival rate of the ROC curves of risk score were 0.679, 0.751, 0.798, and 0.827, correspondingly (Figure 7H), indicating that the risk score is a significant factor in the survival prediction of patients with STAD. In addition, to further verify the accuracy of the signatures, we compared our signatures with seven other published signatures (43–49). C-index and ROC curves illustrated that our signature prediction ability was significantly better than the other seven published signatures (Supplementary Figures 4A-I).
**Figure 7:** *Construction of OMRG-related signature in the training cohort. (A) Various groups of the training cohort were created premised on the median risk score. (B) The survival rate of the training cohort as well as the distributions of risk scores (C) PCA for the training cohort. (D-G) KM analyses (OS, DSS, DFI, and PFI) of the training cohort. (H) Values of the AUC for the training cohort.*
## Verifying the prediction value of the risk signature in the test population
To verify the reproducibility of the risk score in another patient population, OMRG expression was measured in 431 HCC patients with complete survival data from the GEO cohort (GSE84437). The train cohort’s median risk score was used to classify the GEO dataset into high- and low-risk subsets (Figure 8A). Figure 8B shows that the low-risk population had a higher survival rate relative to the high-risk category, which had more deaths overall. The PCA method was used to further classify patients in the two risk categories into two groups (Figure 8C). KM curves for OS depicted in Figure 8D illustrate that the high-risk patients had a poorer prognosis in contrast with those with low risk, with the high-risk population experiencing a shorter OS duration. A time-dependent ROC curve was examined to further gauge the predictive risk signature’s accuracy. The risk score’s outstanding diagnostic utility was demonstrated by the AUC of its ROC curves (Figure 8E).
**Figure 8:** *Internal validation of the OMRG-related signature in the test cohort (A) The test cohort was subdivided into various subgroups. (B) The survival rate of the test cohort as well as the distribution of risk scores (C) Principal component analysis was performed on the test1 cohort. (D) Survival curve of the testing cohort. (E) AUC values of ROC curves in the test cohort.*
## ICG expression and immune function differences between low- and high-risk categories
The influence that varying levels of ICG expression had on the tumor immune milieu of the tumor was studied. In both the training and testing cohorts, YTHDF1, CD160, TNFRSF25, CTLA4, TNFRSF14, JAK2, and CD244 were more highly expressed in the low-risk subgroup, whereas TNFSF4, NRP1, CD276, and CD244 were overexpressed in the high-risk population (Figures 9A, B). Also, a heatmap was generated to ascertain the link between the risk score and immune function. Mast cell function, MHC class I expression, parainflammation, Th2 cells, and Type II IFN response all varied substantially between the high- and low-risk categories in both the training and testing sets (Figures 9C, D).
**Figure 9:** *Immunological function abundance and immune checkpoint gene expression variations in groups with low and high risk. (A-B) Analysis of the differential expression of ICGs between the training and testing cohorts. (C-D) The state of immune function across the training and the testing cohorts. *P < 0.05; **P < 0.01; ***P < 0.001. ns, no significance.*
## Nomogram development and verification
The clinical parameters (sex, grade, age, and stage) and risk score in the training cohort were then examined utilizing univariate and multivariate Cox regression analyses. By analyzing the risk score, stage, and age of the training cohort, we discovered that they independently functioned as prognostic indicators in both univariate and multivariate Cox regression models (Figures 10A, B). After that, a nomogram was produced by combining the aforementioned parameters (Figure 10C). Additionally, survival rates predicted by the nomogram and those observed were found to be in good agreement when calibration curves were established to verify the nomogram’s predictive capacity (Figure 10D). Values of the nomogram’s AUC were recorded to be 0.711, 0.731, and 0.730 over 1, 3, and 5 years, correspondingly (Figure 10E).
**Figure 10:** *The development and validation of a risk score-based nomogram. (A, B). Cox regression analysis, both univariate and multivariate, were performed on the training cohort. (D) Calibration curves for validating the nomogram’s capacity to predict outcomes over 1, 3, and 5 years. (E) The ROC curves’ AUC values offer a better assessment of the nomogram’s prognostic capacity. ***P < 0.001.*
## Verification of thirteen model genes by GEPIA and HPA platform
Finally, we examined the expression levels of the thirteen OMRGs using the GEPIA and HPA databases. The GEPIA database is a merger of the TCGA database and the GTEx database. The TCGA database offered all 408 of the tumor tissue samples for STAD, while the GTEx database gave 175 of the 211 normal gastric tissue samples and the TCGA database contributed 36. According to the GEPIA database, ANXA5, COX10, GAD1, GLS2, GSTO1, NOS3, SERPINE1, SLC25A15, VWF, and CTLA4 are all strongly expressed in STAD relative to normal tissue, whereas APOD, GRP, and KCNQ1 showed an opposite trend (Figure 11).
**Figure 11:** *Expression validation of thirteen models OMRGs from the GEPIA database. (Box above represents the expression difference between STAD and normal samples derived from TCGA; the box below represents the expression difference between STAD and normal samples derived from TCGA and GTEX). *P < 0.05.*
Then, we examined the immunohistochemistry (IHC) results of tumor and normal gastric tissues utilizing the HPA database to examine the levels of protein expressions of 13 genes (Figure 12). Because the HPA database did not provide information on the protein expression of CTLA4 and GRP, we performed the other 11 genes’ proteins. We found ANXA5, COX10, GLS2, GSTO1, and SLC25A15 protein levels were elevated in STAD as opposed to the normal tissues, while APOD and KCNQ1 were the opposite, which was in line with mRNA expression results from GEPIA. And GAD1, NOS3, SERPINE1, and VWF protein levels were not significantly different in tumor and normal samples. In addition, we explored the cellular localization of these genes, only ANXA5, APOD, COX10, GAD1, and KCNQ1 were found in the HPA database. The expression product of ANXA5, APOD, COX10, GAD1, and KCNQ1 was mainly located on the nuclear membrane, plasma membrane, mitochondria, nucleoli and cytosol, vesicles, plasma membrane and cytosol, and endoplasmic reticulum and plasma membrane, respectively (Figure 12).
**Figure 12:** *Immunohistochemistry and immunofluorescence of clinical samples (tumor tissues vs normal nearby tissues).*
## Discussion
Oxidative stress has been linked to the onset and progression of cancer via diverse pathways, including the activation of cell survival, proliferation, and stress resistance systems. as well as enhancing genomic instability and mutagenicity. Meanwhile, the process of carcinogenesis is based on the reprogramming of cellular metabolism, which occurs either directly or indirectly as a result of carcinogenic mutations. For example, altered glucose metabolism is a hallmark of GC, and upregulated aerobic glycolysis in gastric cancer to meet the demands of cell proliferation is associated with genetic mutations, epigenetic modifications, and proteomic alterations [50]. Likewise, abnormal lipid metabolism affects the development of GC. Low serum HDL levels have been found to predict higher risk of gastric cancer, higher rates of lymphovascular and vascular infiltration, advanced lymph node metastasis, and poor prognosis [51]. In addition, dysregulated metabolism of amino acids has been identified as a metabolic regulator that supports cancer cell growth [52]. Meanwhile, some studies have reported that the molecular pathways of oxidative stress are related to glucose metabolism or lipid metabolism [53]. Within the context of cancer, the primary goals of PPPM are early detection, targeted prevention, prognosis, along with individualized management [54, 55]. Even though a variety of treatments exists for STAD, including surgical intervention, immunotherapy, adjuvant chemotherapy, and endoscopic resection [56], the prognosis for patients with advanced STAD continues to be very dismal because there are no prognostic markers available for early diagnosis. Target therapy has emerged as a new therapeutic approach in recent years, and several tests have demonstrated its efficacy [57, 58]. However, the molecular basis of STAD’s pathogenesis is still unknown. Thus, OMRG-based risk stratification of STAD is a promising strategy for prognosis assessment and personalized medicine.
As the study of oxidative metabolism has improved, researchers have uncovered the expanding roles that oxidative metabolism plays in the onset and advancement of cancer. Before looking at the effects of abnormal oxidative metabolism in STAD, we, therefore, describe the changes of OMRGs in a variety of malignancies. In actuality, partial OMRGs had predictive values for several malignancies, and OMRG variants occurred more or less frequently. In addition, the genetic mutations as well as the modifications of OMRGs were recognized in a variety of malignancies. OMRGs were positively linked to TNFA signaling via NFKB, KRAS signaling, inflammatory response, hypoxia, interferon Gamma response, and EMT in most types of tumors. Through its role in the regulation of these pathways, oxidative metabolism has been hypothesized to have a role in carcinogenesis.
After conducting further research into the connection between OMRGs and STAD, we clustered the samples into 3 clusters predicated on the scores assigned to their oxidative metabolism and the patterns of OMRG expression. The OS rates of patients whose oxidative metabolism was active were shown to be considerably lower compared to the rates of patients whose oxidative metabolism was inactive, indicating that the genes involved in oxidative metabolism were mostly risky. Considering that the intersecting metabolic reprogramming of tumor and immune cells is a potential mechanism via which antitumor immune response occurs in cancer, we additionally elucidated the link between OMRGs and immunological function, which could serve as a conceptual foundation for STAD immunotherapy in the long run. To estimate the stromal and immune composition of each patient, the ImmuneScore, StromalScore, and ESTIMATEScore were computed. TME is a niche made up of stromal cells, chemokines, and cytokines that support tumor tissues [59].
Greater ImmuneScore and StromalScore values correspond to greater TME components, respectively. These findings reveal that C2 subtypes linked to a worse outcome have a more robust immune abundance. In addition, the majority of immune-infiltrating agents were shown to have a positive correlation with OMRGs; this was especially true of macrophages, CCR, mast cells, type II IFN response, DCs, and T helper cells. The C2 subgroup’s prognosis was worse when there was a higher proportion of immunological components, indicating that immune checkpoint pathways were active. In three clusters, ICGs are subjected to differential expression. We discovered that ICGs are highly expressed in the C2 subtype, and these differentially expressed ICGs may be intrinsic to the differential prognosis of STAD and may be potential targets for treatment.
Currently, the first line of therapy for patients with advanced STAD is targeted drug therapy, however, its efficacy remains unsatisfactory and there is a need to identify a method to better predict the response to targeted drugs in STAD patients. Therefore, we explored whether there were discrepancies in the sensitivity of patients with three subtypes based on OMRG to commonly used chemotherapeutic agents. We discovered that the three patient groups had various medication sensitivity profiles, indicating that patients’ OMRG expression profiles might be used to tailor their treatment plans. For example, the use of Sunitinib, Dasatinib, Imatinib, Midostaurin, Bexarotene, and Pazopanib could be effective in treating patients whose oxidative metabolism is highly active, whereas Sorafenib, Paclitaxel, Methotrexate, Tipifarnib, and Vinorelbine could be effective in treating patients whose oxidative metabolism is inactive.
After that, to achieve an optimal signature that has clinical relevance, we screened 22 OMRGs by conducting a LASSO-cox regression analysis and assessed the optimal putative genes for signature creation. Following the completion of the validation, a novel OMRG-related prognostic signature that is comprised of 13 genes was developed (i.e., SLC25A15, GSTO1, VWF, ANXA5, SERPINE1, GRP, COX10, APOD, GAD1, NOS3, CTLA4, KCNQ1, and GLS2). Using the signature, STAD patients could be classified into the high-risk category, with a dismal prognosis, and the low-risk category, which has a favorable prognosis, in the train, test1, test2, and test3 cohorts. The AUC values of ROC curves demonstrated that the signature in question had an outstanding predictive performance. We investigated the difference in ICGs and immune function that exists between high- and low-risk categories of STAD patients because of the possible effect that the immune function and ICGs of the tumor might have on the treatment of the tumor. ICGs exhibit varied expressions in both groups. The upregulation of YTHDF1, CD160, TNFRSF25, CTLA4, TNFRSF14, JAK2, and CD244 and knockdown of TNFSF4, NRP1, CD276, and CD244 could be viable targets in STAD. Meanwhile, mast cells, MHC class I, parainflammation, Type II IFN response, and Th2 cells were statistically different in the high- and low-risk categories To make the most of the signature’s capacity for prediction, a nomogram was designed using the risk score and several other clinical data, and then a quantitative analysis was performed on the survival rate of patients suffering from STAD. Evaluation of the nomogram’s capacity to predict with a high degree of accuracy was carried out by means of calibration curves and ROC curves. Finally, we validated the thirteen model genes against the GEPIA and HPA databases.
There are some flaws in our research as well. First off, we only used retrospective data from the GEO and TCGA databases to validate the OMRG-based signature; going forward, we should conduct more prospective studies to assess its therapeutic implications. Meanwhile, additional sizable prospective clinical trials are required to evaluate its efficacy and usefulness.
## Conclusion
In this study, patients with STAD could be classified into three clusters with different prognoses, immune characteristics, and drug sensitivity premised on OMRG scores. For the first time, an OMRG-related signature was developed and confirmed to accurately predict the prognosis of STAD patients. After that, utilizing this signature as well as other clinical parameters, a nomogram was generated as a quantitative tool to assist in predicting the survival rate for STAD patients. In summary, the present research has the potential to assist in the identification of prognostic predictions, targeted prevention, and individualized treatments for patients, hence proposing a new route to enhance PPPM for STAD.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors.
## Author contributions
YD, QY, and JR contributed to this article equally. The content and authorship of the manuscript are entirely the responsibility of all authors. The study’s design, data collection and analysis, article preparation, and manuscript revision all benefited greatly from the efforts of all authors. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1090906/full#supplementary-material
## References
1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. **Global cancer statistics 2018: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA Cancer J Clin* (2018) **68** 394-424. DOI: 10.3322/caac.21492
2. Chen W, Zheng R, Baade PD, Zhang S, Zeng H, Bray F. **Cancer statistics in China, 2015**. *CA Cancer J Clin* (2016) **66**. DOI: 10.3322/caac.21338
3. Chugunov VV. **Influence of antidepressants on the convulsive action of corazole and strychnine**. *Farmakol Toksikol* (1976) **39**. DOI: 10.1002/cam4.3946
4. Kim HS, Seo HK. **Immune checkpoint inhibitors for urothelial carcinoma**. *Investig Clin Urol* (2018) **59**. DOI: 10.4111/icu.2018.59.5.285
5. Tabernero J, Hoff PM, Shen L, Ohtsu A, Shah MA, Cheng K. **Pertuzumab plus trastuzumab and chemotherapy for Her2-positive metastatic gastric or gastro-oesophageal junction cancer (Jacob): Final analysis of a double-blind, randomised, placebo-controlled phase 3 study**. *Lancet Oncol* (2018) **19**. DOI: 10.1016/S1470-2045(18)30481-9
6. Ebrahimi SO, Reiisi S, Shareef S. **Mirnas, oxidative stress, and cancer: A comprehensive and updated review**. *J Cell Physiol* (2020) **235**. DOI: 10.1002/jcp.29724
7. Doppler H, Storz P. **Mitochondrial and oxidative stress-mediated activation of protein kinase D1 and its importance in pancreatic cancer**. *Front Oncol* (2017) **7**. DOI: 10.3389/fonc.2017.00041
8. Morry J, Ngamcherdtrakul W, Yantasee W. **Oxidative stress in cancer and fibrosis: Opportunity for therapeutic intervention with antioxidant compounds, enzymes, and nanoparticles**. *Redox Biol* (2017) **11**. DOI: 10.1016/j.redox.2016.12.011
9. Ogrunc M, Di Micco R, Liontos M, Bombardelli L, Mione M, Fumagalli M. **Oncogene-induced reactive oxygen species fuel hyperproliferation and DNA damage response activation**. *Cell Death Differ* (2014) **21** 998-1012. DOI: 10.1038/cdd.2014.16
10. Jeon SM, Chandel NS, Hay N. **Ampk regulates nadph homeostasis to promote tumour cell survival during energy stress**. *Nature* (2012) **485**. DOI: 10.1038/nature11066
11. Morgan MJ, Liu ZG. **Crosstalk of reactive oxygen species and nf-kappab signaling**. *Cell Res* (2011) **21**. DOI: 10.1038/cr.2010.178
12. Ye J, Fan J, Venneti S, Wan YW, Pawel BR, Zhang J. **Serine catabolism regulates mitochondrial redox control during hypoxia**. *Cancer Discovery* (2014) **4**. DOI: 10.1158/2159-8290.CD-14-0250
13. Xia L, Oyang L, Lin J, Tan S, Han Y, Wu N. **The cancer metabolic reprogramming and immune response**. *Mol Cancer* (2021) **20** 28. DOI: 10.1186/s12943-021-01316-8
14. Sun L, Suo C, Li ST, Zhang H, Gao P. **Metabolic reprogramming for cancer cells and their microenvironment: Beyond the warburg effect**. *Biochim Biophys Acta Rev Cancer* (2018) **1870** 51-66. DOI: 10.1016/j.bbcan.2018.06.005
15. Hanahan D. **Hallmarks of cancer: New dimensions**. *Cancer Discovery* (2022) **12** 31-46. DOI: 10.1158/2159-8290.CD-21-1059
16. Boroughs LK, DeBerardinis RJ. **Metabolic pathways promoting cancer cell survival and growth**. *Nat Cell Biol* (2015) **17**. DOI: 10.1038/ncb3124
17. Yoshida GJ. **Metabolic reprogramming: The emerging concept and associated therapeutic strategies**. *J Exp Clin Cancer Res* (2015) **34** 111. DOI: 10.1186/s13046-015-0221-y
18. Punt CJ, Koopman M, Vermeulen L. **From tumour heterogeneity to advances in precision treatment of colorectal cancer**. *Nat Rev Clin Oncol* (2017) **14**. DOI: 10.1038/nrclinonc.2016.171
19. Kaji S, Irino T, Kusuhara M, Makuuchi R, Yamakawa Y, Tokunaga M. **Metabolomic profiling of gastric cancer tissues identified potential biomarkers for predicting peritoneal recurrence**. *Gastric Cancer* (2020) **23**. DOI: 10.1007/s10120-020-01065-5
20. Yu L, Lai Q, Feng Q, Li Y, Feng J, Xu B. **Serum metabolic profiling analysis of chronic gastritis and gastric cancer by untargeted metabolomics**. *Front Oncol* (2021) **11**. DOI: 10.3389/fonc.2021.636917
21. Dunn GP, Bruce AT, Ikeda H, Old LJ, Schreiber RD. **Cancer immunoediting: From immunosurveillance to tumor escape**. *Nat Immunol* (2002) **3**. DOI: 10.1038/ni1102-991
22. Zitvogel L, Tesniere A, Kroemer G. **Cancer despite immunosurveillance: Immunoselection and immunosubversion**. *Nat Rev Immunol* (2006) **6**. DOI: 10.1038/nri1936
23. Yoon SJ, Park J, Shin Y, Choi Y, Park SW, Kang SG. **Deconvolution of diffuse gastric cancer and the suppression of Cd34 on the Balb/C nude mice model**. *BMC Cancer* (2020) **20** 314. DOI: 10.1186/s12885-020-06814-4
24. Ren J, Yuan Q, Liu J, Zhong L, Li H, Wu G. **Identifying the role of transient receptor potential channels (Trps) in kidney renal clear cell carcinoma and their potential therapeutic significances using genomic and transcriptome analyses**. *BMC Med Genomics* (2022) **15** 156. DOI: 10.1186/s12920-022-01312-x
25. Chen X, Yuan Q, Liu J, Xia S, Shi X, Su Y. **Comprehensive characterization of extracellular matrix-related genes in paad identified a novel prognostic panel related to clinical outcomes and immune microenvironment: A silico analysis with in vivo and vitro validation**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.985911
26. Wang J, Ren J, Liu J, Zhang L, Yuan Q, Dong B. **Identification and verification of the ferroptosis- and pyroptosis-associated prognostic signature for low-grade glioma**. *Bosn J Basic Med Sci* (2022) **22**. DOI: 10.17305/bjbms.2021.6888
27. Yuan Q, Deng D, Pan C, Ren J, Wei T, Wu Z. **Integration of transcriptomics, proteomics, and metabolomics data to reveal Her2-associated metabolic heterogeneity in gastric cancer with response to immunotherapy and neoadjuvant chemotherapy**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.951137
28. Chi H, Xie X, Yan Y, Peng G, Strohmer DF, Lai G. **Natural killer cell-related prognosis signature characterizes immune landscape and predicts prognosis of hnscc**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.1018685
29. Cheng Q, Chen X, Wu H, Du Y. **Three Hematologic/Immune system-specific expressed genes are considered as the potential biomarkers for the diagnosis of early rheumatoid arthritis through bioinformatics analysis**. *J Transl Med* (2021) **19**. DOI: 10.1186/s12967-020-02689-y
30. Liberzon A, Birger C, Thorvaldsdottir H, Ghandi M, Mesirov JP, Tamayo P. **The molecular signatures database (Msigdb) hallmark gene set collection**. *Cell Syst* (2015) **1**. DOI: 10.1016/j.cels.2015.12.004
31. Wei Y, Ou T, Lu Y, Wu G, Long Y, Pan X. **Classification of ovarian cancer associated with Brca1 mutations, immune checkpoints, and tumor microenvironment based on immunogenomic profiling**. *PeerJ* (2020) **8**. DOI: 10.7717/peerj.10414
32. He Y, Jiang Z, Chen C, Wang X. **Classification of triple-negative breast cancers based on immunogenomic profiling**. *J Exp Clin Cancer Res* (2018) **37** 327. DOI: 10.1186/s13046-018-1002-1
33. Yue C, Ma H, Zhou Y. **Identification of prognostic gene signature associated with microenvironment of lung adenocarcinoma**. *PeerJ* (2019) **7**. DOI: 10.7717/peerj.8128
34. Geeleher P, Cox N, Huang RS. **Prrophetic: An r package for prediction of clinical chemotherapeutic response from tumor gene expression levels**. *PloS One* (2014) **9**. DOI: 10.1371/journal.pone.0107468
35. Chi H, Jiang P, Xu K, Zhao Y, Song B, Peng G. **A novel anoikis-related gene signature predicts prognosis in patients with head and neck squamous cell carcinoma and reveals immune infiltration**. *Front Genet* (2022) **13**. DOI: 10.3389/fgene.2022.984273
36. Wang X, Zhao Y, Strohmer DF, Yang W, Xia Z, Yu C. **The prognostic value of micrornas associated with fatty acid metabolism in head and neck squamous cell carcinoma**. *Front Genet* (2022) **13**. DOI: 10.3389/fgene.2022.983672
37. Tang Z, Li C, Kang B, Gao G, Li C, Zhang Z. **Gepia: A web server for cancer and normal gene expression profiling and interactive analyses**. *Nucleic Acids Res* (2017) **45** W98-W102. DOI: 10.1093/nar/gkx247
38. Uhlen M, Fagerberg L, Hallstrom BM, Lindskog C, Oksvold P, Mardinoglu A. **Proteomics. tissue-based map of the human proteome**. *Science* (2015) **347**. DOI: 10.1126/science.1260419
39. Shen H, Laird PW. **Interplay between the cancer genome and epigenome**. *Cell* (2013) **153** 38-55. DOI: 10.1016/j.cell.2013.03.008
40. Junttila MR, de Sauvage FJ. **Influence of tumour micro-environment heterogeneity on therapeutic response**. *Nature* (2013) **501**. DOI: 10.1038/nature12626
41. Grivennikov SI, Greten FR, Karin M. **Immunity, inflammation, and cancer**. *Cell* (2010) **140**. DOI: 10.1016/j.cell.2010.01.025
42. Binnewies M, Roberts EW, Kersten K, Chan V, Fearon DF, Merad M. **Understanding the tumor immune microenvironment (Time) for effective therapy**. *Nat Med* (2018) **24**. DOI: 10.1038/s41591-018-0014-x
43. Jiang X, Yan Q, Xie L, Xu S, Jiang K, Huang J. **Construction and validation of a ferroptosis-related prognostic model for gastric cancer**. *J Oncol* (2021) **2021**. DOI: 10.1155/2021/6635526
44. Shao W, Yang Z, Fu Y, Zheng L, Liu F, Chai L. **The pyroptosis-related signature predicts prognosis and indicates immune microenvironment infiltration in gastric cancer**. *Front Cell Dev Biol* (2021) **9**. DOI: 10.3389/fcell.2021.676485
45. Liu Y, Wu J, Huang W, Weng S, Wang B, Chen Y. **Development and validation of a hypoxia-Immune-Based microenvironment gene signature for risk stratification in gastric cancer**. *J Transl Med* (2020) **18** 201. DOI: 10.1186/s12967-020-02366-0
46. Guan K, Liu X, Li J, Ding Y, Li J, Cui G. **Expression status and prognostic value of M6a-associated genes in gastric cancer**. *J Cancer* (2020) **11**. DOI: 10.7150/jca.40866
47. Zheng H, Liu H, Li H, Dou W, Wang X. **Weighted gene Co-expression network analysis identifies a cancer-associated fibroblast signature for predicting prognosis and therapeutic responses in gastric cancer**. *Front Mol Biosci* (2021) **8**. DOI: 10.3389/fmolb.2021.744677
48. Wu Z, Wang L, Wen Z, Yao J. **Integrated analysis identifies oxidative stress genes associated with progression and prognosis in gastric cancer**. *Sci Rep* (2021) **11** 3292. DOI: 10.1038/s41598-021-82976-w
49. Qin M, Liang Z, Qin H, Huo Y, Wu Q, Yang H. **Novel prognostic biomarkers in gastric cancer: Cgb5, Mknk2, and Pappa2**. *Front Oncol* (2021) **11**. DOI: 10.3389/fonc.2021.683582
50. Yuan LW, Yamashita H, Seto Y. **Glucose metabolism in gastric cancer: The cutting-edge**. *World J Gastroenterol* (2016) **22**. DOI: 10.3748/wjg.v22.i6.2046
51. Cui MY, Yi X, Zhu DX, Wu J. **The role of lipid metabolism in gastric cancer**. *Front Oncol* (2022) **12**. DOI: 10.3389/fonc.2022.916661
52. Li Z, Zhang H. **Reprogramming of glucose, fatty acid and amino acid metabolism for cancer progression**. *Cell Mol Life Sci* (2016) **73**. DOI: 10.1007/s00018-015-2070-4
53. Ighodaro OM. **Molecular pathways associated with oxidative stress in diabetes mellitus**. *BioMed Pharmacother* (2018) **108**. DOI: 10.1016/j.biopha.2018.09.058
54. Grech G, Zhan X, Yoo BC, Bubnov R, Hagan S, Danesi R. **Epma position paper in cancer: Current overview and future perspectives**. *EPMA J* (2015) **6** 9. DOI: 10.1186/s13167-015-0030-6
55. Golubnitschaja O, Baban B, Boniolo G, Wang W, Bubnov R, Kapalla M. **Medicine in the early twenty-first century: Paradigm and anticipation - epma position paper 2016**. *EPMA J* (2016) **7** 23. DOI: 10.1186/s13167-016-0072-4
56. Joshi SS, Badgwell BD. **Current treatment and recent progress in gastric cancer**. *CA Cancer J Clin* (2021) **71**. DOI: 10.3322/caac.21657
57. Okines AF, Ashley SE, Cunningham D, Oates J, Turner A, Webb J. **Epirubicin, oxaliplatin, and capecitabine with or without panitumumab for advanced esophagogastric cancer: Dose-finding study for the prospective multicenter, randomized, phase Ii/Iii real-3 trial**. *J Clin Oncol* (2010) **28**. DOI: 10.1200/JCO.2010.29.2847
58. Dutton SJ, Ferry DR, Blazeby JM, Abbas H, Dahle-Smith A, Mansoor W. **Gefitinib for oesophageal cancer progressing after chemotherapy (Cog): A phase 3, multicentre, double-blind, placebo-controlled randomised trial**. *Lancet Oncol* (2014) **15** 894-904. DOI: 10.1016/S1470-2045(14)70024-5
59. Belli C, Trapani D, Viale G, D'Amico P, Duso BA, Della Vigna P. **Targeting the microenvironment in solid tumors**. *Cancer Treat Rev* (2018) **65** 22-32. DOI: 10.1016/j.ctrv.2018.02.004
|
---
title: Clinical measures of balance and gait cannot differentiate somatosensory impairments
in people with lower-limb amputation
authors:
- BA. Petersen
- PJ. Sparto
- LE. Fisher
journal: Gait & posture
year: 2022
pmcid: PMC9970031
doi: 10.1016/j.gaitpost.2022.10.018
license: CC BY 4.0
---
# Clinical measures of balance and gait cannot differentiate somatosensory impairments in people with lower-limb amputation
## Abstract
### Background:
In addition to a range of functional impairments seen in individuals with a lower-limb amputation, this population is at a substantially elevated risk of falls. Studies postulate that the lack of sensory feedback from the prosthetic limb contributes heavily to these impairments, but the extent to which sensation affects functional measures remains unclear.
### Research question:
The purpose of this study is to determine how sensory impairments in the lower extremities relate to performance with common clinical functional measures of balance and gait in individuals with a lower-limb amputation. Here we evaluate the effects of somatosensory integrity to clinical and lab measures of static, reactive and dynamic balance, and gait stability.
### Methods:
In 20 individuals with lower-limb amputation (AMP) and 20 age and gender-matched able-bodied controls (CON), we evaluated the effects of sensory integrity (pressure, proprioception, and vibration) on measures of balance and gait. Static, reactive, and dynamic balance were assessed using the Sensory Organization Test (SOT), Motor Control Test (MCT), and Functional Gait Assessment (FGA), respectively. Gait stability was assessed through measures of step length asymmetry and step width variability. Sensation was categorized into intact or impaired sensation by pressure thresholds and differences across groups were analyzed.
### Results:
There were significant differences between AMP and CON groups for reliance on vision for static balance in the SOT, MCT, and FGA ($p \leq 0.01$). Despite differences across groups, there were no significant differences within the AMP group based on intact or impaired sensation across all functional measures.
### Significance:
Despite being able to detect differences between able-bodied individuals and individuals with an amputation, these functional measures cannot distinguish between levels of impairment within participants with an amputation. These findings suggest that more challenging and robust metrics are needed to evaluate the effects of sensation and function in individuals with an amputation.
## Introduction
By the year 2050, an estimated 3.6 million Americans will be living with limb loss, with lower-limb amputations accounting for approximately $65\%$ of this population [1]. People with lower-limb amputation can suffer from a range of functional impairments and have a substantially higher risk of falls and fear of falling than the average adult [2,3]. Over $50\%$ of community-dwelling adults with lower-limb amputation reported at least one fall within the past year [2]. In comparison, only $27.5\%$ of able-bodied, community-dwelling adults over 65 years old reported a fall last in the last year [4]. In addition, people with lower-limb amputation can exhibit a wide range of gait and balance impairments compared to their able-bodied counterparts [5]. Determining the factors that influence functional impairments is critical to designing and implementing interventions to improve mobility. Studies suggest that impairments in gait and balance may be due to lack of sensory feedback from the prosthetic limb, but the extent to which sensation relates to various functional measures is not fully understood [6].
In animal models and humans, tactile and proprioceptive inputs to the spinal cord drive gait phase transitions and contribute to healthy balance mechanics and muscle activation [7,8]. Further, individuals with sensory loss exhibit a wide array of functional deficits, including balance and gait impairments. For example, diabetic peripheral neuropathy is associated with a five-fold increase in fall risk [9,10]. Studies have found direct correlations between measures of sensory loss and balance impairments, suggesting that sensation in individuals with intact limbs is crucial for balance [11,12]. In a small sample ($$n = 4$$) of individuals with an amputation, sensory integrity was correlated with a functional reach task [13]. However, the nature of any relationship between sensory integrity and functional outcome measures and assessments has not been established in people with lower-limb amputation.
Studies have consistently shown that individuals with a lower-limb amputation rely more heavily on visual feedback for static balance than able-bodied controls, likely as a compensatory mechanism for a lack of sensory feedback [14]. For example, a review conducted on the relative contributions of the amputated and intact limbs to balance control after amputation found that the intact limb contributed more to postural stability in quiet standing [14]. The authors postulated that this is likely due to disruption of the cutaneous and proprioceptive systems that occurs with an amputation [14]. More recently, another study on the contributions of sensory feedback in each limb found that individuals with transfemoral amputation relied more heavily on proprioceptive feedback in the intact limb for balance [6]. Notably, the participants in this study also did not have any dysvascular disorders that affected the intact contralateral limb. Without visual feedback, the reliance on the intact limb was notably increased [6]. These findings suggest that individuals with an amputation rely more on their contralateral limb with intact sensation and compensate using vision for balance in lieu of sensory feedback. Again, these studies did not evaluate tindividuals with sensory impairments on their intact limb, excluding the large group of people with dysvascular amputations and concomitant contralateral sensory impairments.
With existing outcome measures and assessment, there is conflicting evidence on whether individuals with dysvascular amputations (i.e. those with sensory impairments that often affect the intact limb) have more severe functional impairments than individuals with traumatic amputations. In individuals with a transtibial amputation, Jayakaran et al. used the Sensory Organization Test (SOT) to study postural control in individuals with traumatic and dysvascular transtibial amputations compared to controls with and without dysvascular conditions [15]. The SOT is a widely used clinical standard for measuring reliance on visual, somatosensory and vestibular systems for balance [16]. In the SOT, participants stand on a force platform while either visual (eyes closed, sway-referenced surround rotation) or somatosensory feedback (sway-referenced platform rotation) are altered across six conditions. By altering visual or somatosensory feedback, the SOT forces the participant to rely on their other systems for balance. In that study, there were no significant differences based on cause of amputation (i.e. traumatic vs. dysvascular), with both groups with amputation showing less anteroposterior stability than able-bodied groups [15]. Similar studies have been performed evaluating reactive balance. Impaired somatosensation likely plays a role in altered balance responses to these perturbations, given the reflexive pathways based on tactile and proprioceptive inputs that mediate these postural corrections and gait stability [17]. Molina-Rueda et al. evaluated reactive balance using the motor control test (MCT) across groups with traumatic and dysvascular transtibial amputations. The MCT is a test of involuntary, reactive balance in response to anteroposterior translations of the support surface. The MCT evaluates reaction latency, amplitude, and symmetry to assess subjects’ ability to respond to an external translation. They found that the dysvascular group had slower responses on their sound limb than the traumatic group. Additionally, several studies have shown key differences in gait kinematics and kinetics in people with dysvascular versus traumatic amputations, although there is evidence that these differences may be due primarily to differences in gait speed [18]. Though both of these studies included participants with sensory loss, they both excluded individuals with phantom limb sensation or pain, which may include up to $80\%$ of individuals with an amputation [19]. Thus, determining the impact of sensation across the full spectrum of individuals with an amputation is critical to evaluating whether current clinical outcome measures can detect functional differences due to sensory impairments.
The effects of clinical measures of somatosensation on clinical measures of function in individuals with a lower-limb amputation needs clarification. The purpose of this study is to determine whether some current clinical outcome measures can detect the effects of sensory impairments on function. To test this, we evaluated the difference in measures of static, reactive, and dynamic balance, and gait stability based on differences in somatosensory integrity across a wide range of individuals with a lower-limb amputation, regardless of level or nature of amputation. The relationship of quantitative measures of somatosensation to these outcomes can elucidate how well these outcome measures can detect differences in function due to sensory impairments.
## Methods
We collected measures of balance, gait, and sensation from 20 individuals with lower-limb amputation (AMP) and 20 age- and gender-matched able-bodied individuals (CON). The CON group was added to confirm these metrics were exhibiting previously established differences between CON and AMP groups in this population. Inclusion criteria for individuals with an amputation included: [1] amputation of one lower limb, [2] between the ages of 18 and 70, [3] ability to stand unassisted for 10 min, and [4] ability to ambulate. To evaluate these metrics across a wide range of individuals, our exclusion criteria were intentionally broad and included both transtibial and transfemoral amputation. Participants were excluded if they had a known balance disorder or were pregnant. All experiments were performed under the supervision of the University of Pittsburgh’s Institutional Review Board. For more detail on specific outcome measures, see Supplementary 1.
## Sensory measures
Measures of sensory integrity included somatosensory monofilament pressure thresholds, light touch sensation, protective sensation, lower extremity reflexes, proprioception, and vibration sense. All sensory tests were performed with the participant’s eyes closed. Somatosensory pressure thresholds were assessed using Semmes-Weinstein monofilaments, which include varying grades of monofilament thickness, ranging from 0.01 g to 300 g. The filament was applied perpendicular to the plantar aspect of the feet until the filament bent, three times per site. The plantar aspect of the hallux, first metatarsal head, fifth metatarsal head, and heel were tested. If the subject reported sensation for at least 2 of 3 trials, the next monofilament was tested, until the subject could no longer detect the filament. For the residual limb (AMP only), we tested the distal-most aspect of the residual limb with the limb stabilized to avoid excessive skin movement. Light touch, protective (pin prick), reflexes, proprioception and vibration sense were assessed bilaterally, as well (Supplementary 1).
## Performance measures
To quantify static, reactive, and dynamic balance and gait, we used the Sensory Organization Test (SOT), Motor Control Test (MCT), Functional Gait Assessment (FGA), and gait kinematics, respectively. These clinical assessments were selected because they have been used to assess changes with somatosensory feedback [20,21]. The SOT and MCT were both performed using the NeuroCom Equitest system (Supplementary 1). By altering the visual or somatosensory feedback participants receive, the test provides a method for measuring the reliance on the somatosensory, visual, and vestibular systems to maintain balance. Three, 20-second trials were completed per condition. Center of pressure (COP) traces were recorded from the force plates (100 Hz), filtered with a low-pass fourth-order Butterworth filter, and analyzed for standard measures of posturography, in addition to clinical measures. Standard posturography measures were calculated, including excursion, sway velocity, $95\%$ confidence interval ellipse of sway area, sample, and approximate entropy (Supplementary 1). Clinical measures, including equilibrium scores and somatosensory ability were also recorded. Equilibrium scores indicate a participant’s ability to stay within a normative 12.5° anteroposterior sway envelope (Supplementary 1). Somatosensory ability (ratio of equilibrium scores in static conditions without vision, condition 2, to equilibrium scores with normal vision, condition 1) indicates a participant’s ability to utilize somatosensation for balance when vision is impaired.
In the MCT, participants must maintain balance in the Equitest system following translational perturbations in both anterior and posterior directions. The perturbations in this task included three grades (small, medium, large) with random time delays ranging 1–3 s. The medium and large translational trials were assessed for latency of onset of active response and weight-bearing symmetry (Supplementary 1). For the AMP group, the active response was typically too small to be detected on the prosthetic side, so the latency and strength of responses was determined only on the intact limb [22].
Dynamic balance was assessed using the FGA, a clinical gait and dynamic balance assessment that involves ambulating 6 m down a hallway [23]. There are 10 items which are scored from 0 (severe impairment) to 3 (no impairment). This gait assessment has been validated in community-dwelling adults and individuals with neurological and balance disorders [24,25].
Gait kinematics during walking on a level surface were recorded using a 16-camera OptiTrack motion analysis system (Natural Point, OR, USA). Participants were instructed to walk at their self-selected speed for six trials across a 6-meter walkway. Sixteen reflective markers were placed on anatomical landmarks according to the OptiTrack “Conventional Lower Body” model [26]. Kinematic marker data was collected at 100 Hz and filtered using a 4th order low-pass Butterworth filter at 12 Hz. Step length asymmetry (normalized to stride length, SLA), step length variability, and step width variability (standard deviation and coefficient of variation) were calculated as measures of gait stability (Supplementary 1). Gait assessments were only collected from 12 of the 20 AMP participants, as our motion capture lab was only available midway through data collection.
## Statistical analysis
A Pearson correlation was performed between all measures of balance and clinical sensory scores. However, because we observed a bimodal distribution of sensory impairment in individuals with AMP (Fig. 1), monofilament threshold was categorized as intact (<10 g threshold, $$n = 10$$) or impaired sensation (>10 g threshold, $$n = 10$$) based on the clinical standard for diagnosing peripheral neuropathy [27]. Due to this distribution of sensory loss, the non-parametric Mann Whitney U test was used to determine significant differences between the participants with intact and impaired sensation. Comparisons between AMP and CON groups were completed using the Wilcoxon signed rank test for pair-wise comparisons. Significance level was a priori set at 0.05, with Bonferroni corrections completed to account for multiple comparisons. See Supplementary Tables 1–4 for all comparisons and p-values. Effect sizes as partial eta squared (η2) are reported.
## Participant characteristics
Twenty people with lower-limb amputation and twenty age- and gender-matched controls were included in the study. Age and gender across both AMP and CON groups are comparable (Table 1). The majority of amputations were transtibial ($80\%$) and the average use of the prosthesis exceeded 12 h per day. In addition, 10 participants had full sensation bilaterally, while 10 participants had impaired sensation (3 had impaired sensation bilaterally, 5 had impaired sensation on the contralateral limb only and 2 participants had impaired residual limb sensation only). In comparison to the CON group, the AMP group had a significantly slower self-selected gait speed (0.88 +0.18 m/s AMP, 1.12 +0.18 m/s CON, $p \leq 0.001$).
## Static Balance
For the SOT, the AMP group had a greater increase in sway area in the condition without vision than the CON group (Fig. 2A, 16.05 +19.77 cm2 AMP, 3.03 +3.05 cm2 CON, $p \leq 0.0063$, η2 = 0.878), with no significant differences in sway area between AMP group based on sensation ($p \leq 0.0063$, Fig. 2B). Similar, the CON group had significantly greater SOT somatosensory ability, the ratio of anteroposterior sway in the static condition without vision to the condition with vision, than the AMP group (Fig. 2C, 0.90 +0.08 AMP, 0.95 +0.03 CON, $p \leq 0.0075$, η2=0.213). However, SOT somatosensory ability was not significantly different between amputees with full or impaired sensation ($p \leq 0.007$). The distribution of scores in the impaired sensation group is larger and more skewed than the full sensation group for change in area without vision or SOT somatosensory ability (Fig. 2).
There were no significant differences in equilibrium scores in any conditions or in composite equilibrium scores between AMP groups with full or impaired sensation. No significant differences were found across groups based on other sensory measures (proprioception, reflexes, vibration), as well ($p \leq 0.007$).
Within the CON group, there were no significant relationships ($p \leq 0.0107$) between monofilament thresholds or other sensory measures and clinical or posturography measures of balance across all conditions of the SOT.
## Reactive balance
The latencies of responses on the intact limb in the AMP group were significantly slower than for the dominant limb in the CON group (150 +18 ms AMP, 132 +12 ms CON, $p \leq 0.025$, η2 = 0.82, Fig. 3A), however latencies on the intact limb were not significantly different based on sensation ($p \leq 0.025$, Fig. 3B). The AMP group were significantly less symmetrical than the CON group, bearing more weight on the intact limb (−16.54 +13.68 AMP, −0.01 +6.55 CON, $p \leq 0.025$, η2 = 0.150, Fig. 3C), with no significant differences based on sensation ($p \leq 0.025$, Fig. 3D).
## Dynamic balance and gait stability
Total FGA score was significantly lower in the AMP group compared to controls (19 +5 AMP, 29 +2 CON, $p \leq 0.0013$, η2 = 0.0213, Fig. 4A). The FGA showed no significant differences between AMP participants with full or impaired sensation on either residual or contralateral limbs ($p \leq 0.0013$, Fig. 4B). There were no significant differences in SLA between AMP and CON groups or between full and impaired sensation groups (Fig. 4C–D) or for step width variability (Fig. 4E–F).
## Discussion
In this study, we explored the relationship between sensory impairments in the residual and contralateral limbs and performance on a variety of clinical outcome measures for people with lower-limb amputation. Consistent with previous literature, the SOT, MCT, and FGA can detect differences in functional abilities between individuals with a lower-limb amputation and able-bodied individuals. However, these measures are not able to detect even substantial differences in somatosensory integrity within populations with a lower-limb amputation. These findings are surprising, given the critical role sensation and spinal reflexes play in balance and gait. While one possible interpretation of this result is that somatosensory impairments do not make a difference in function, the reflexive pathways and role of tactile feedback in balance and gait have been characterized extensively [7,8,31] and preliminary evidence in a small sample of individuals with an amputation ($$n = 4$$) has suggested plantar sensation plays a significant role in balance control [13]. Thus, these disturbances in somatosensory input in the AMP group still likely have a functional impact. Instead, these findings more likely suggest that the current battery of tests we have for this population are not able to distinguish between these differences in somatosensory integrity.
The lack of significant differences within the AMP group for the SOT are consistent with those seen by Jayakaran, who found no difference in SOT measures between groups with dysvascular and traumatic amputations [15]. Our reactive balance findings are inconsistent with previous studies, which found significant differences in latencies between individuals with dysvascular and traumatic amputations [22]. Again, these responses were only measured for the intact limb and an alternative measure of reactive balance utilizing the residual limb may be necessary to detect differences in sensation for the amputated side.
The lack of significant differences between AMP and CON groups for kinematic measures of gait stability has also been reported in previous literature. Keklicek et al. found that despite differences in step length variability, individuals with transtibial amputation demonstrated step lengths on both intact and residual limbs, though not normalized, similar to those in able-bodied subjects [32]. These findings differ from studies reporting SLA in individuals with amputation, however these studies did not normalize SLA to stride length [33], which is now recommended in use of SLA to avoid accentuating asymmetries in individuals with shorter stride lengths [34]. The dilemma posed here is that these gait measures (i.e., FGA, gait kinematics) are used clinically across the full range of individuals with an amputation. However, our results indicate that these outcomes should not be used for many subgroups of individuals with amputations. Together these findings suggest that more challenging and robust measures of gait analysis are necessary to capture differences between groups across the wide variety of impairments seen in this population. Newer measures are being studied to address this issue. For example, Thies et al. found that walking on an irregular surface can detect differences across subgroups of individuals with sensory impairments and Sawers et al. developed the narrowing beam walking test as a more robust outcome measure for this population [28,35]. Future work should explore whether these measures can detect differences in somatosensory function among people with limb amputation.
Notably, the measures of sensation used in this study are crude instruments designed to be used in clinics. This may account for the inability to characterize the more mild-moderate range of somatosensory impairments. Thus, future studies with more robust measures of sensation may further elucidate these findings. In addition, this was a small sample of 20 individuals with an amputation, only four of which had a transfemoral amputation. A larger sample size would be necessary to perform multiple regression to determine how factors like level of amputation, prosthetic usage, or time since amputation impact functional measures, in addition to sensation. Additionally, we did not control for fall history or medications, which could affect balance. Assessing fall risk more directly, as well as balance confidence, phantom pain, or prosthesis comfort should be evaluated in future studies to determine what other effects sensation may have for this population and what other factors are driving the differences in performance for this population when compared to healthy controls.
In conclusion, while these clinical measures detect differences between able-bodied individuals and individuals with an amputation, they are not able to distinguish between levels of somatosensory impairments within groups with an amputation. These findings, in addition to other recent work evaluating current outcomes for this population, suggest that more challenging and robust metrics are necessary to evaluate the role of sensation and other factors on functional impairments in people with lower-limb amputation.
## References
1. Ziegler-Graham K, MacKenzie E, Ephraim P, Travison TG, Brookmeyer R. **Estimating the prevalence of limb loss in the United States: 2005 to 2050**. *Arch. Phys. Med. Rehabil* (2008.0) **89** 422-429. DOI: 10.1016/j.apmr.2007.11.005
2. Wong CK, Chihuri ST, Li G. **Risk of fall-related injury in people with lower limb amputations: a prospective cohort study**. *J. Rehabil. Med* (2016.0) **48** 80-85. DOI: 10.2340/16501977-2042
3. Miller WC, Speechley M, Deathe B, Wc AM, Speechley M, The DB. **The prevalence and risk factors of falling and fear of falling among lower extremity amputees**. *Arch. Phys. Med. Rehabil* (2001.0) **82** 1031-1037. DOI: 10.1053/apmr.2001.24295
4. Moreland B, Kakara R, Henry A. **Trends in nonfatal falls and fall-related injuries among adults aged ≥65 Years - United States, 2012–2018**. *MMWR Morb. Mortal. Wkly. Rep* (2020.0) **69** 875-881. DOI: 10.15585/mmwr.mm6927a5
5. Vanicek N, Strike S, McNaughton L, Polman R. **Postural responses to dynamic perturbations in amputee fallers versus nonfallers: a comparative study with able-bodied subjects**. *Arch. Phys. Med. Rehabil* (2009.0) **90** 1018-1025. DOI: 10.1016/j.apmr.2008.12.024
6. Claret CR, Herget GW, Kouba L, Wiest D, Adler J, von Tscharner V, Stieglitz T, Pasluosta C. **Neuromuscular adaptations and sensorimotor integration following a unilateral transfemoral amputation**. *J. Neuroeng. Rehabil* (2019.0) **16** 115. DOI: 10.1186/s12984-019-0586-9
7. Prochazka A, Gorassini M. **Models of ensemble firing of muscle spindle afferents recorded during normal locomotion in cats**. *J. Physiol* (1998.0) **507** 277-291. DOI: 10.1111/j.1469-7793.1998.277bu.x
8. Pearson KG. **Generating the walking gait: role of sensory feedback**. *Prog. Brain Res* (2004.0) **143** 123-129. PMID: 14653157
9. Timar B, Timar R, Gaiță L, Oancea C, Levai C, Lungeanu D. **The impact of diabetic neuropathy on balance and on the risk of falls in patients with type 2 diabetes mellitus: a cross-sectional study**. *PLoS One* (2016.0) **11** e0154654. DOI: 10.1371/journal.pone.0154654
10. DeMott TK, Richardson JK, Thies SB, Ashton-Miller JA. **Falls and gait characteristics among older persons with peripheral neuropathy**. *Am. J. Phys. Med. Rehabil* (2007.0) **86** 125-132. DOI: 10.1097/PHM.0b013e31802ee1d1
11. Ducic I, Short KW, Dellon AL. **Relationship between loss of pedal sensibility, balance, and falls in patients with peripheral neuropathy**. *Ann. Plast. Surg* (2004.0) **52**
12. Wang T-Y, Lin S-I. **Sensitivity of plantar cutaneous sensation and postural stability**. *Clin. Biomech* (2008.0) **23** 493-499. DOI: 10.1016/j.clinbiomech.2007.11.014
13. Templeton CA, Strzalkowski NDJ, Galvin P, Bent LR. **Cutaneous sensitivity in unilateral trans-tibial amputees**. *PLoS One* (2018.0) **13** e0197557. DOI: 10.1371/journal.pone.0197557
14. Ku PX, Abu Osman NA, Wan Abas WAB. **Balance control in lower extremity amputees during quiet standing: a systematic review**. *Gait Posture* (2014.0) **39** 672-682. DOI: 10.1016/j.gaitpost.2013.07.006
15. Jayakaran P, Johnson GM, Sullivan SJ. **Postural control in response to altered sensory conditions in persons with dysvascular and traumatic transtibial amputation**. *Arch. Phys. Med. Rehabil* (2015.0) **96** 331-339. DOI: 10.1016/j.apmr.2014.09.037
16. Mancini M, Horak FB. **The relevance of clinical balance assessment tools to differentiate balance deficits**. *Eur. J. Phys. Rehabil. Med* (2010.0) **46** 239-248. PMID: 20485226
17. Honeycutt CF, Nichols TR. **Disruption of cutaneous feedback alters magnitude but not direction of muscle responses to postural perturbations in the decerebrate cat**. *Exp. Brain Res* (2010.0) **203** 765-771. DOI: 10.1007/s00221-010-2281-8
18. Nakajima H, Yamamoto S, Katsuhira J. **Effects of diabetic peripheral neuropathy on gait in vascular trans-tibial amputees**. *Clin. Biomech* (2018.0) **56** 84-89. DOI: 10.1016/j.clinbiomech.2018.05.005
19. Ephraim PL, Wegener ST, MacKenzie EJ, Dillingham TR, Pezzin LE. **Phantom pain, residual limb pain, and back pain in amputees: results of a national survey**. *Arch. Phys. Med. Rehabil* (2005.0) **86** 1910-1919. DOI: 10.1016/j.apmr.2005.03.031
20. Charkhkar H, Christie BP, Triolo RJ. **Sensory neuroprosthesis improves postural stability during sensory organization test in lower-limb amputees**. *Sci. Rep* (2020.0) **10** 6984. DOI: 10.1038/s41598-020-63936-2
21. Koehler-McNicholas SR, Danzl L, Cataldo AY, Oddsson LIE. **Neuromodulation to improve gait and balance function using a sensory neuroprosthesis in people who report insensate feet – a randomized control cross-over study**. *PLoS One* (2019.0) **14** e0216212. DOI: 10.1371/journal.pone.0216212
22. Molina-Rueda F, Molero-Sánchez A, Alguacil-Diego IM, Carratalá-Tejada M, Cuesta-Gómez A, Miangolarra-Page JC. **Weight symmetr and latency scores for unexpected surface perturbations in subjects with traumatic and vascular unilateral transtibial amputation**. *PM&R* (2016.0) **8** 235-240. DOI: 10.1016/j.pmrj.2015.05.024
23. Wrisley DM, Marchetti GF, Kuharsky DK, Whitney SL. **Reliability, internal consistency, and validity of data obtained with the functional gait assessment**. *Phys. Ther* (2004.0) **84** 906-918. DOI: 10.1093/ptj/84.10.906
24. Wrisley DM, Kumar NA. **Functional gait assessment: concurrent, discriminative, and predictive validity in community-dwelling older adults**. *Phys. Ther* (2010.0) **90** 761-773. DOI: 10.2522/ptj.20090069
25. Leddy AL, Crowner BE, Earhart GM. **Functional gait assessment and balance evaluation system test: reliability, validity, sensitivity, and specificity for identifying individuals with parkinson disease who fall**. *Phys. Ther* (2011.0) **91** 102-113. DOI: 10.2522/ptj.20100113
26. [26]Skeleton Tracking—NaturalPoint Product Documentation Ver 2.2, (n.d.).
https://v22.wiki.optitrack.com/index.php.. *Skeleton Tracking—NaturalPoint Product Documentation Ver 2.2*
27. Feng Y, Schlösser FJ, Sumpio BE. **The Semmes Weinstein monofilament examination as a screening tool for diabetic peripheral neuropathy, e1**. *J. Vasc. Surg* (2009.0) **50** 675-682. DOI: 10.1016/j.jvs.2009.05.017
28. Sawers A, Kim J, Balkman G, Hafner BJ. **Interrater and test-retest reliability of performance-based clinical tests administered to established users of lower limb prostheses**. *Phys. Ther* (2020.0) **100** 1206-1216. DOI: 10.1093/ptj/pzaa063
29. Cripps A, Livingston SC, DeSantis BM. *The Test- Retest Reliability and Minimal Detectable Change of the Sensory Organization Test and Head- Shake Sensory Organization Test* (2016.0)
30. Harro CC, Garascia C. **Reliability and validity of computerized force platform measures of balance function in healthy older adults**. *J. Geriatr. Phys. Ther* (2019.0) **42**
31. Miller S, Ruit JB, der Meche F.G. Van. **Reversal of sign of long spinal reflexes dependent on the phase of the step cycle in the high decerebrate cat**. *Brain Res* (1977.0) **128** 447-459. PMID: 884493
32. Keklicek H, Kirdi E, Yalcin A, Topuz S, Ulger O, Erbahceci F, Sener G. **Comparison of gait variability and symmetry in trained individuals with transtibial and transfemoral limb loss**. *J. Orthop. Surg. (Hong. Kong)* (2019.0) **27**. DOI: 10.1177/2309499019832665
33. Pröbsting E, Bellmann M, Schmalz T, Hahn A. **Gait characteristics of transtibial amputees on level ground in a cohort of 53 amputees-comparison of kinetics and kinematics with non-amputees**. *Can. PROSTHETICS Orthot. J* (2019.0) **2**
34. Stenum J, Choi JT. **Disentangling the energetic costs of step time asymmetry and step length asymmetry in human walking**. *J. Exp. Biol* (2021.0) **224** jeb242258. DOI: 10.1242/jeb.242258
35. Thies SB, Richardson JK, DeMott T, Ashton-Miller JA. **Influence of an irregular surface and low light on the step variability of patients with peripheral neuropathy during level gait**. *Gait Posture* (2005.0) **22** 40-46. DOI: 10.1016/j.gaitpost.2004.06.004
|
---
title: 'Association of coffee consumption pattern and metabolic syndrome among middle-aged
and older adults: A cross-sectional study'
authors:
- Ren Nina
- Huang Lingling
- Li Qiushuang
- Guo Honglin
- Sun Liyuan
- Zhang Yuting
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9970040
doi: 10.3389/fpubh.2023.1022616
license: CC BY 4.0
---
# Association of coffee consumption pattern and metabolic syndrome among middle-aged and older adults: A cross-sectional study
## Abstract
### Objectives
The association between coffee consumption and the risk of metabolic syndrome (MetS) remains inconsistent. The aim of this study was to evaluate the association between coffee intake and components of MetS.
### Method
A cross-sectional survey including 1,719 adults was conducted in Guangdong, China. Data on age, gender, education level, marriage status, body mass index (BMI), current smoking and drinking status and breakfast habit, coffee consumption type, and daily servings were derived based on 2-day, 24-h recall. MetS were assessed according to the International Diabetes Federation definition. Multivariable logistic regression was conducted to examine the association between the coffee consumption type, daily servings, and the components of MetS.
### Results
Regardless of the coffee type, compared with non-coffee consumers, coffee consumers had higher odds ratios (ORs) of the elevated fasting blood glucose (FBG) in both men [OR: 3.590; $95\%$ confidence intervals (CI): 2.891–4.457] and women (OR: 3.590; $95\%$ CI: 2.891–4.457). In women, the risk of elevated blood pressure (BP) was 0.553 times (OR: 0.553; $95\%$ CI: 0.372–0.821, $$P \leq 0.004$$) for people who drank total coffee > 1 serving/day than for non-coffee drinkers.
### Conclusion
In conclusion, regardless of type, coffee intake is associated with an increased prevalence of FBG in both men and women, but has a protective effect on hypertension only in women.
## Introduction
Metabolic syndrome (MetS), defined as the presence of physiologically related cardiovascular risk factors, including dyslipidemia, abdominal obesity, hyperglycemia, and hypertension, is closely correlated with increased cardiovascular risk and common cancers (1–3). The prevalence of MetS has considerably increased over recent decades and is now at epidemic proportions worldwide [4, 5]. Insulin resistance is a key hallmark feature of MetS and a critical risk factor for diabetes and other cardiovascular diseases (CVD) [6]. Recently, accumulating epidemiological and experimental evidence point out that MetS is affected by genetic (7–9) and lifestyle factors [10, 11], including smoking, alcohol consumption, sugar-sweetened beverage consumption physical activity, and sedentary behaviors. Indeed, MetS have been inversely affected by dietary intakes, such as vegetables, fruits, red wine, and green tea [12]. Therefore, experts emphasize dietary intakes for the primary interventions on MetS prevention [13, 14].
Coffee, which has antioxidant properties and a distinctive smell and taste, is now one of the world's most popular beverages [15]. With the far-reaching development of industry and rapid changes in dietary lifestyles, coffee consumption has been considerably increasing in Shenzhen. The constituents in coffee, including polyphenols, antioxidant properties, caffeine, potassium, niacin, vitamin E, and magnesium, have been proposed to be beneficial for potential health. Experimental studies revealed that caffeine might protect against type 2 diabetes mellitus (T2DM) by stimulating free fatty acid and fat oxidation release from peripheral tissues, increasing metabolic rate and thermogenesis, and mobilizing glycogen in muscles [16]. Therefore, epidemiologic studies reported a significant association between higher coffee consumption and decreased incidence of new-onset hypertension [17, 18], arterial stiffness [19, 20], T2DM [21], and promote weight loss [22]. However, another study conducted in the Japanese setting demonstrated that certain types of coffee led to an increase in all-cause mortality [23]. Also, other investigations reported that the intake of coffee with creamer or sugar was significantly associated with increased abdominal obesity [24] and risk of MetS [25].
The above inconsistent findings might be caused by different research designs, that is, some focused on the effect of daily coffee consumption volume, while others focused on the habitual coffee pattern. However, daily consumption patterns of coffee containing both quantitative and qualitative information are still lacking. For this reason, we performed a cross-sectional study to examine the association between coffee consumption patterns and MetS components among middle-aged and older adults.
## Study population
This cross-sectional survey was based on a large-scale, community-based routine health examination for the middle-aged and elderly. In total, 2,200 participants aged 40 years and above were recruited from January 2021 to March 2022 in Guangdong province. All individuals received a routine health check-up, including venous blood sampling and anthropometry. Among these, a subset of the individuals ($$n = 2$$,066) completed the 24-h food recall. Furthermore, we excluded individuals with a history of ischemic heart disease ($$n = 12$$) or stroke ($$n = 21$$), and those who take drugs to treat hyperlipidemia, diabetes, or hypertension ($$n = 314$$). Finally, 1,719 participants (800 men and 919 women) were included in the present study. The study was approved by the Ethics Committee of the Health Science Centre, Shenzhen University. All individuals signed written informed consent before participation.
## Diagnosis of mets
According to the guidance of the updated National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) [26], individuals who met at least three of the following criteria were diagnosed with MetS: [1] waist circumference (WC) ≥ 90 cm in men, ≥ 80 cm in women; [2] systolic blood pressure (SBP) ≥ 130 mmHg or diastolic blood pressure (DBP) ≥ 85 mmHg; [3] fasting blood glucose (FBG) ≥ 5.60 mmol/L; [4] blood high-density lipoprotein cholesterol (HDL-C) level < 40 mg/dL in men, < 50 mg/dL in women; [5] blood triglyceride (TG) ≥ 1.70 mmol/L.
## Anthropometric and biochemical measurement
At the mobile physical examination centers, anthropometric variables including weight, height, and blood pressure (BP) were measured using standardized calibrated equipment under the guidance of professional medical staff. Body mass index (BMI, kg/m2) was calculated as weight divided by the height squared. WC was measured at the narrowest between the iliac crest and the bottom of the ribs. BP was measured by a sphygmomanometer (Yu yue, YJ100002, Jiangsu, China) on the right arm after the individual had been supine for at least 20 min, and the mean value of three times record was used. Biochemical assessment variables including FBG, HDL-C, TG, and 2 h post-load glucose (2hPG) were assessed using a semi-automated analyzer (Sysmex 100 XN-3000, Tokyo, Japan) enzymatically after fasting for at least 8 h. Furthermore, a detailed data collection process was reported elsewhere [27].
## Coffee consumption measurement
Information regarding coffee consumption was obtained based on a 2-day, 24-h recall. Individuals who drank coffee at least three times per week were described as coffee drinkers [28]. The habitual coffee consumption questionnaire included habitual coffee type and daily coffee serving frequency. Black coffee was described as coffee powder or extracts without other ingredients. Coffee with creamer, milk, or sugar was defined as instant coffee. Other coffee is a collective name, covering a series of other types of coffee, excluding black coffee, and instant coffee. Based on the type of coffee they consumed, individuals were classified into the following five categories: non-coffee consumers, black coffee consumers, instant coffee consumers, other coffee consumers, and coffee consumers. If only black coffee or instant coffee was in a person's 2-day, 24-h food recall, the individual was determined as a black coffee or instant coffee consumer, respectively. Meanwhile, individuals who consumed any coffee type that appeared at least once were classified as coffee consumers. Other coffee consumers referred to participants who consumed other type's coffee.
## Demographic measurement
Demographic information of participants including age, gender, education level, marital status, current smoking and drinking status, breakfast habits, physical activity, and sitting time was collected through questionnaires. The level of education was categorized as up to junior high school, high school or secondary specialized school, and college and above. Marital status categories included unmarried, married or cohabiting, and others (divorced, separated, or widowed). Current smoking and drinking status were categorized as yes or no. Breakfast categories included none, 1–3 times/week, 4–5 times/week, and every day. Physical activity divided into four categories: < 0.5, 0.5–1, 1–2, and > 2 h/day. Sitting time divided into four categories: < 6, 6–8, 8–10, and > 10 h/day.
## Statistical analysis
All data are presented as mean (standard deviation) for continuous data and as percentages for categorical data according to the Shapiro–Wilk test of normality. Participants' demographic characteristics including age, education level, marital status, BMI, body weight status, current smoking and drinking status, breakfast habit, physical activity, sitting time, and Mets parameters according to coffee consumption type by gender, were compared using the Chi-square test for categorical variables and generalized linear model for continuous variables. A multivariate-adjust logistic regression model was conducted to explore the association between coffee consumption patterns and MetS components. We assigned the median of daily servings of coffee as a continuous variable and performed stratified analysis across coffee consumption categories. We adjusted covariates including BMI, education level, alcohol status, and physical activity for all the regression models, and the $95\%$ confidence intervals (CIs) of odds ratios (ORs) were estimated. A two-sided P-value of < 0.05 was considered statistical significance, and SAS software (version 9.4) was used to conduct all analyses.
## Results
Table 1 presented the participants' demographic characteristics according to coffee consumption categories by gender. In men, the proportion of participants in high school or secondary specialized school was the largest for all coffee consumption categories ($P \leq 0.05$). In men, mean BMI and FBG levels were significantly higher in instant coffee consumers than in other groups (all $P \leq 0.05$). In women, mean BMI, SBP, and FBG levels were significantly higher in other coffee consumers than in other groups (all $P \leq 0.05$). In both men and women, the proportion of participants with normal weight status was the largest for all coffee consumption categories (all $P \leq 0.05$). In men, non-alcohol drinkers were more likely to be non-coffee consumers, while compared with non-alcohol drinkers, the proportion of alcohol consumers was higher than nondrinkers in the other three types of coffee pattern groups (all $P \leq 0.05$). In men, the duration of physical activity was higher in black coffee consumers than in the other three types of coffee pattern groups ($P \leq 0.05$).
**Table 1**
| Variables | Men (n = 800) | Men (n = 800).1 | Men (n = 800).2 | Men (n = 800).3 | Men (n = 800).4 | Women (n = 919) | Women (n = 919).1 | Women (n = 919).2 | Women (n = 919).3 | Women (n = 919).4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Non-Coffee Consumer (n = 475) | Black-Coffee Consumer (n = 117) | Instant-Coffee Consumer (n = 66) | Other Coffee Consumer (n = 142) | P | Non-Coffee Consumer (n = 549) | Black-Coffee Consumer (n = 143) | Instant-Coffee Consumer (n = 80) | Other Coffee Consumer (n = 147) | P |
| Age (years) | 62.34 ± 15.62 | 61.24 ± 15.53 | 63.00 ± 16.28 | 62.44 ± 15.95 | 0.881 | 61.32 ± 14.83 | 62.16 ± 14.75 | 61.31 ± 15.42 | 62.46 ± 14.32 | 0.813 |
| Educational level | Educational level | Educational level | Educational level | Educational level | Educational level | Educational level | Educational level | Educational level | Educational level | Educational level |
| ~Junior high school | 90 (18.95) | 16 (13.68) | 8 (12.12) | 10 ( 7.04) | 0.016 | 94 (17.12) | 20 (13.99) | 13 (16.25) | 26 (17.69) | 0.943 |
| High school/Secondary specialized school | 262 (55.16) | 62 (52.99) | 39 (59.09) | 94 (66.20) | | 322 (58.65) | 83 (58.04) | 49 (61.25) | 85 (57.82) | |
| College~ | 123 (25.89) | 39 (33.33) | 19 (28.79) | 38 (26.76) | | 133 (24.23) | 40 (27.97) | 18 (22.50) | 36 (24.49) | |
| Marriage | Marriage | Marriage | Marriage | Marriage | Marriage | Marriage | Marriage | Marriage | Marriage | Marriage |
| Unmarried | 155 (32.63) | 41 (35.04) | 25 (37.88) | 59 (41.55) | 0.233 | 193 (35.15) | 55 (38.46) | 26 (32.50) | 55 (37.41) | 0.359 |
| Married/cohabiting | 289 (60.84) | 63 (53.85) | 35 (53.03) | 74 (52.11) | | 319 (58.11) | 85 (59.44) | 51 (63.75) | 82 (55.78) | |
| Divorced/Separated/Widowed | 31 ( 6.53) | 13 (11.11) | 6 ( 9.09) | 9 ( 6.34) | | 37 ( 6.74) | 3 ( 2.10) | 3 ( 3.75) | 10 ( 6.80) | |
| Height (cm) | 163.77 ± 8.11 | 164.38 ± 7.59 | 162.91 ± 7.96 | 163.96 ± 7.99 | 0.669 | 158.37 ± 6.39 | 157.68 ± 5.89 | 158.31 ± 6.08 | 157.80 ± 6.85 | 0.569 |
| Weight (kg) | 61.03 ± 11.05 | 63.90 ± 11.11 | 63.74 ± 10.41 | 62.08 ± 9.53 | 0.034 | 55.69 ± 10.29 | 56.85 ± 8.47 | 56.36 ± 9.22 | 57.47 ± 8.76 | 0.167 |
| Body mass index | 22.68 ± 3.37 | 23.59 ± 3.40 | 24.01 ± 3.35 | 23.07 ± 2.97 | 0.004 | 22.18 ± 3.82 | 22.89 ± 3.43 | 22.45 ± 3.18 | 23.09 ± 3.43 | 0.020 |
| Body weight status | Body weight status | Body weight status | Body weight status | Body weight status | Body weight status | Body weight status | Body weight status | Body weight status | Body weight status | Body weight status |
| Underweight | 36 ( 7.58) | 8 ( 6.84) | 2 ( 3.03) | 11 ( 7.75) | 0.024 | 86 (15.66) | 12 ( 8.39) | 4 ( 5.00) | 3 ( 2.05) | <0.001 |
| Normal | 234 (49.26) | 45 (38.46) | 23 (34.85) | 61 (42.96) | | 265 (48.27) | 66 (46.15) | 44 (55.00) | 86 (58.90) | |
| Overweight | 117 (24.63) | 31 (26.50) | 16 (24.24) | 33 (23.24) | | 102 (18.58) | 34 (23.78) | 15 (18.75) | 24 (16.44) | |
| Obese | 88 (18.53) | 33 (28.21) | 25 (37.88) | 37 (26.06) | | 96 (17.49) | 31 (21.68) | 17 (21.25) | 33 (22.60) | |
| Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status | Current smoking status |
| Yes | 223 (46.95) | 59 (50.43) | 26 (39.39) | 70 (49.30) | 0.500 | 236 (42.99) | 67 (46.85) | 34 (42.50) | 71 (48.30) | 0.612 |
| No | 252 (53.05) | 58 (49.57) | 40 (60.61) | 72 (50.70) | | 313 (57.01) | 76 (53.15) | 46 (57.50) | 76 (51.70) | |
| Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status | Current alcohol status |
| Yes | 216 (45.47) | 69 (58.97) | 37 (56.06) | 74 (52.11) | 0.032 | 247 (44.99) | 71 (49.65) | 35 (43.75) | 72 (48.98) | 0.647 |
| No | 259 (54.53) | 48 (41.03) | 29 (43.94) | 68 (47.89) | | 302 (55.01) | 72 (50.35) | 45 (56.25) | 75 (51.02) | |
| Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit | Breakfast habit |
| No | 24 ( 5.05) | 6 ( 5.13) | 1 ( 1.52) | 8 ( 5.63) | 0.661 | 29 ( 5.28) | 6 ( 4.20) | 3 ( 3.75) | 1 ( 0.68) | 0.334 |
| 1-3 times/week | 146 (30.74) | 46 (39.32) | 20 (30.30) | 44 (30.99) | | 166 (30.24) | 44 (30.77) | 27 (33.75) | 45 (30.61) | |
| 4-5 times/week | 159 (33.47) | 32 (27.35) | 25 (37.88) | 42 (29.58) | | 179 (32.60) | 53 (37.06) | 20 (25.00) | 51 (34.69) | |
| Every day | 146 (30.74) | 33 (28.21) | 20 (30.30) | 48 (33.80) | | 175 (31.88) | 40 (27.97) | 30 (37.50) | 50 (34.01) | |
| Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity |
| <0.5 h/day | 185 (38.95) | 44 (37.61) | 23 (34.85) | 66 (46.48) | 0.025 | 195 (35.52) | 57 (39.86) | 25 (31.25) | 55 (37.41) | 0.258 |
| 0.5-1 h/day | 139 (29.26) | 46 (39.32) | 19 (28.79) | 50 (35.21) | | 186 (33.88) | 36 (25.17) | 36 (45.00) | 54 (36.73) | |
| 1-2 h/day | 96 (20.21) | 19 (16.24) | 18 (27.27) | 13 ( 9.15) | | 101 (18.40) | 31 (21.68) | 12 (15.00) | 21 (14.29) | |
| ≥2 h/day | 55 (11.58) | 8 ( 6.84) | 6 ( 9.09) | 13 ( 9.15) | | 67 (12.20) | 19 (13.29) | 7 ( 8.75) | 17 (11.56) | |
| Sitting time | Sitting time | Sitting time | Sitting time | Sitting time | Sitting time | Sitting time | Sitting time | Sitting time | Sitting time | Sitting time |
| <6 h/day | 127 (26.74) | 31 (26.50) | 14 (21.21) | 32 (22.54) | 0.679 | 134 (24.41) | 40 (27.97) | 17 (21.25) | 37 (25.17) | 0.650 |
| 6-8 h/day | 182 (38.32) | 50 (42.74) | 28 (42.42) | 48 (33.80) | | 205 (37.34) | 48 (33.57) | 33 (41.25) | 51 (34.69) | |
| 8-10 h/day | 98 (20.63) | 20 (17.09) | 14 (21.21) | 38 (26.76) | | 144 (26.23) | 32 (22.38) | 22 (27.50) | 34 (23.13) | |
| ≥10 h/day | 68 (14.32) | 16 (13.68) | 10 (15.15) | 24 (16.90) | | 66 (12.02) | 23 (16.08) | 8 (10.00) | 25 (17.01) | |
| Systolic blood pressure (mmHg) | 134.89 ± 20.89 | 137.80 ± 19.39 | 131.80 ± 20.51 | 135.68 ± 20.19 | 0.257 | 134.92 ± 22.24 | 128.54 ± 20.29 | 131.69 ± 22.81 | 135.09 ± 22.47 | 0.008 |
| Diastolic blood pressure (mmHg) | 86.32 ± 13.35 | 87.13 ± 12.80 | 85.58 ± 14.68 | 85.51 ± 11.77 | 0.742 | 85.13 ± 13.51 | 82.17 ± 14.12 | 82.96 ± 14.67 | 84.62 ± 13.64 | 0.119 |
| Triglycerides (mmol/L) | 1.64 ± 0.51 | 1.53 ± 0.49 | 1.69 ± 0.47 | 1.64 ± 0.50 | 0.141 | 1.65 ± 0.51 | 1.71 ± 0.50 | 1.58 ± 0.51 | 1.63 ± 0.51 | 0.288 |
| High-density lipoprotein cholesterol (mmol/L) | 1.25 ± 0.24 | 1.29 ± 0.30 | 1.21 ± 0.23 | 1.26 ± 0.27 | 0.304 | 1.30 ± 0.28 | 1.28 ± 0.24 | 1.29 ± 0.26 | 1.29 ± 0.25 | 0.811 |
| Fasting blood glucose (mmol/L) | 5.72 ± 0.46 | 6.01 ± 0.47 | 6.02 ± 0.45 | 5.99 ± 0.47 | <0.001 | 5.73 ± 0.48 | 6.02 ± 0.42 | 6.00 ± 0.51 | 6.12 ± 0.53 | <0.001 |
| Blood glucose two hours after meals (mmol/L) | 9.11 ± 0.64 | 9.10 ± 0.66 | 8.92 ± 0.63 | 9.07 ± 0.64 | 0.146 | 9.05 ± 0.68 | 9.09 ± 0.67 | 9.06 ± 0.67 | 9.20 ± 0.69 | 0.115 |
Table 2 summarized the multivariable-adjusted OR and $95\%$ CI of MetS components across the type of coffee by gender. Regardless of the coffee type, compared with non-coffee consumers, coffee consumers had higher ORs of the elevated FBG in both men (OR: 3.590; $95\%$ CI: 2.891–4.457) and women (OR: 3.590; $95\%$ CI: 2.891–4.457). In women, the prevalence of elevated blood pressure (OR: 0.661; $95\%$ CI: 0.454–0.963) was significantly lower in black coffee consumers than in non-coffee consumers. The same inverse association can be also found in other types of coffee consumption.
**Table 2**
| Variable | Non-coffee consumer (reference) | Black-coffee consumer | Instant-coffee consumer | Other coffee consumer | Coffee consumer |
| --- | --- | --- | --- | --- | --- |
| Men | Men | Men | Men | Men | Men |
| Elevated TG | Elevated TG | Elevated TG | Elevated TG | Elevated TG | Elevated TG |
| Model 1 | Ref | 0.878 (0.582, 1.324) | 0.782 (0.451, 1.355) | 0.988 (0.674, 1.447) | 0.903 (0.678, 1.203) |
| Model 2 | 1.139 (0.755, 1.717) | Ref | 0.891 (0.471, 1.686) | 1.125 (0.682, 1.857) | - |
| Reduced HDL-C | Reduced HDL-C | Reduced HDL-C | Reduced HDL-C | Reduced HDL-C | Reduced HDL-C |
| Model 1 | Ref | 0.933 (0.706, 1.234) | 1.211 (0.854, 1.716) | 0.927 (0.709, 1.211) | 0.984 (0.809, 1.198) |
| Model 2 | 1.071 (0.811, 1.416) | Ref | 1.297 (0.861, 1.955) | 0.993 (0.704, 1.401) | - |
| Elevated BP | Elevated BP | Elevated BP | Elevated BP | Elevated BP | Elevated BP |
| Model 1 | Ref | 0.867 (0.651, 1.155) | 0.729 (0.510, 1.043) | 0.984 (0.744, 1.302) | 0.880 (0.717, 1.079) |
| Model 2 | 1.153 (0.865, 1.536) | Ref | 0.841 (0.553, 1.278) | 1.135 (0.796, 1.618) | - |
| Elevated FBG | Elevated FBG | Elevated FBG | Elevated FBG | Elevated FBG | Elevated FBG |
| Model 1 | Ref | 3.523 (2.635, 4.709) | 3.268 (2.273, 4.699) | 3.827 (2.895, 5.059) | 3.590 (2.891, 4.457) |
| Model 2 | 0.284 (0.212, 0.379) | Ref | 0.928 (0.617, 1.396) | 1.086 (0.776, 1.521) | - |
| Women | Women | Women | Women | Women | Women |
| Elevated TG | Elevated TG | Elevated TG | Elevated TG | Elevated TG | Elevated TG |
| Model 1 | Ref | 1.317 (0.795, 2.179) | 0.736 (0.340, 1.591) | 0.874 (0.497, 1.535) | 1.009 (0.683, 1.490) |
| Model 2 | 0.760 (0.459, 1.257) | Ref | 0.559 (0.238, 1.310) | 0.664 (0.340, 1.297) | - |
| Reduced HDL-C | Reduced HDL-C | Reduced HDL-C | Reduced HDL-C | Reduced HDL-C | Reduced HDL-C |
| Model 1 | Ref | 0.983 (0.676, 1.430) | 0.932 (0.579, 1.501) | 0.911 (0.628, 1.321) | 0.943 (0.721, 1.233) |
| Model 2 | 1.017 (0.699, 1.478) | Ref | 0.948 (0.543, 1.653) | 0.926 (0.579, 1.482) | - |
| Elevated BP | Elevated BP | Elevated BP | Elevated BP | Elevated BP | Elevated BP |
| Model 1 | Ref | 0.661 (0.454, 0.963) | 0.738 (0.456, 1.194) | 0.984 (0.668, 1.448) | 0.790 (0.600, 1.040) |
| Model 2 | 1.512 (1.038, 2.203) | Ref | 1.116 (0.640, 1.947) | 1.488 (0.924, 2.397) | - |
| Elevated FBG | Elevated FBG | Elevated FBG | Elevated FBG | Elevated FBG | Elevated FBG |
| Model 1 | Ref | 2.987 (2.021, 4.413) | 3.192 (1.962, 5.195) | 4.177 (2.844, 6.134) | 3.464 (2.589, 4.634) |
| Model 2 | 0.335 (0.227, 0.495) | Ref | 1.069 (0.616, 1.854) | 1.398 (0.880, 2.221) | - |
We further conducted stratified analyses to explore multivariable-adjusted OR and $95\%$ CI for MetS according to daily servings of coffee by gender, as presented in Table 3. In male black coffee drinkers, there was a linear trend between the increase of TG and the decrease in coffee consumption ($P \leq 0.05$). In addition, men who drank coffee > 1 serving/day had an increased risk of elevated FBG (OR: 4.112; $95\%$ CI: 2.537–6.666; $P \leq 0.05$). The same results were also observed in men who drank black coffee (OR: 3.835; $95\%$ CI: 2.009–7.319; $P \leq 0.001$) and instant coffee (OR: 3.651; $95\%$ CI: 1.329–10.031; $P \leq 0.001$). In women, there was a positive correlation between coffee consumption and elevated FBG. The risk of elevated FBG in people who drink > 1 serving/day is 3.798 times higher than that in people who drink ≤ 1 serving/day (OR: 3.798; $95\%$ CI: 2.555–5.647, $P \leq 0.001$), and the same results can be found in women who drink black coffee and instant coffee. The risk of elevated FBG was 3.109 times (OR: 3.109; $95\%$ CI: 1.799–5.371, $P \leq 0.001$) in women who drank black coffee and 3.514 times (OR: 3.109; $95\%$ CI: 1.503–8.218, $P \leq 0.001$) in women who drank instant coffee. In women, compared with non-coffee consumers, there was a negative correlation between coffee consumption and elevated BP. The risk of elevated BP was 0.553 times (OR: 0.553; $95\%$ CI: 0.372–0.821, $$P \leq 0.004$$) for people who drank coffee > 1 serving/day than for non-coffee drinkers in total coffee. The risk of elevated BP was 0.516 times that of non-coffee drinkers (OR: 0.516; $95\%$ CI: 0.296–0.898, $$P \leq 0.005$$) in black coffee. Among instant coffee drinkers, the risk of elevated BP was 0.276 times that of non-coffee drinkers (OR: 0.276; $95\%$ CI: 0.112–0.68, $$P \leq 0.037$$).
**Table 3**
| Variable | Non-coffee consumer (reference) | Non-coffee consumer (reference).1 | Total coffee | Total coffee.1 | Total coffee.2 | Total coffee.3 | Total coffee.4 | Black-coffee | Black-coffee.1 | Black-coffee.2 | Black-coffee.3 | Black-coffee.4 | Instant-coffee | Instant-coffee.1 | Instant-coffee.2 | Instant-coffee.3 | Instant-coffee.4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | | ≤ 1 Serving/day | ≤ 1 Serving/day | > 1 Serving/day | > 1 Serving/day | P | ≤ 1 Serving/day | ≤ 1 Serving/day | > 1 Serving/day | > 1 Serving/day | P | ≤ 1 Serving/day | ≤ 1 Serving/day | > 1 Serving/day | > 1 Serving/day | P |
| Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men | Men |
| n | | 475 | | 232 | | 93 | | | 70 | | 47 | | | 49 | | 17 | |
| Elevated TG | 1 | | 0.876 (0.545, 1.408) | | 0.486 (0.213, 1.108) | | 0.103 | 0.400 (0.152, 1.053) | | 0.348 (0.103, 1.174) | | 0.021 | 1.012 (0.443, 2.312) | | 0 (0, Inf) | | 0.240 |
| Reduced HDL-C | 1 | | 1.032 (0.743, 1.435) | | 1.071 (0.678, 1.693) | | 0.751 | 0.770 (0.448, 1.323) | | 1.025 (0.551, 1.908) | | 0.731 | 1.588 (0.869, 2.901) | | 1.728 (0.650, 4.594) | | 0.084 |
| Elevated BP | 1 | | 1.004 (0.704, 1.431) | | 1.182 (0.716, 1.953) | | 0.593 | 1.144 (0.638, 2.054) | | 1.637 (0.793, 3.378) | | 0.180 | 0.609 (0.323, 1.148) | | 0.777 (0.274, 2.208) | | 0.204 |
| Elevated FBG | 1 | | 3.928 (2.726, 5.659) | | 4.112 (2.537, 6.666) | | <0.001 | 4.937 (2.863, 8.513) | | 3.835 (2.009, 7.319) | | <0.001 | 3.658 (1.925, 6.951) | | 3.651 (1.329, 10.031) | | <0.001 |
| Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women | Women |
| n | | 549 | | 231 | | 139 | | | 80 | | 63 | | | 57 | | 23 | |
| Elevated TG | 1 | | 0.873 (0.543, 1.404) | | 1.044 (0.610, 1.786) | | 0.958 | 1.490 (0.793, 2.799) | | 0.917 (0.425, 1.980) | | 0.767 | 0.753 (0.305, 1.856) | | 0.603 (0.135, 2.698) | | 0.380 |
| Reduced HDL-C | 1 | | 0.895 (0.652, 1.228) | | 0.934 (0.636, 1.371) | | 0.587 | 0.860 (0.526, 1.406) | | 1.061 (0.622, 1.812) | | 0.958 | 0.922 (0.526, 1.617) | | 0.886 (0.373, 2.102) | | 0.704 |
| Elevated BP | 1 | | 0.839 (0.601, 1.171) | | 0.553 (0.372, 0.821) | | 0.004 | 0.602 (0.366, 0.992) | | 0.516 (0.296, 0.898) | | 0.005 | 1.134 (0.614, 2.092) | | 0.276 (0.112, 0.680) | | 0.037 |
| Elevated FBG | 1 | | 3.316 (2.376, 4.627) | | 3.798 (2.555, 5.647) | | <0.001 | 2.964 (1.807, 4.863) | | 3.109 (1.799, 5.371) | | <0.001 | 3.077 (1.745, 5.425) | | 3.514 (1.503, 8.218) | | <0.001 |
## Discussion
The current study examined the associations between coffee consumption patterns and MetS components among middle-aged and older adults in Guangdong. We found that both black coffee and instant coffee had positive associations with elevated FBG. Furthermore, these positive associations were robust in the stratified analyses among participants who consumed ≤ 1 vs. >1 Serving daily. In addition, according to gender-stratified analysis, regardless of the coffee type, women who drank a good amount of coffee were significantly associated with a lower prevalence of elevated BP than non-coffee consumers. Whereas, the same results could not be found in men. Our results revealed that habitual coffee drinking could prevent women from hypertension in a certain sense.
Our study suggested that most of the participants were of normal weight regardless of the coffee consumption type. But, in both men and women, compared with non-coffee consumers, coffee consumers were more likely to have a higher BMI. These findings are not in line with previous epidemiology studies. In a national wide cross-sectional study conducted in 2003–2004, coffee consumption was not significantly associated with BMI or waist circumference in either men or women [29]. However, another cross-sectional study in Poland revealed that lower coffee consumption was significantly associated with a higher risk of obesity [30]. The inconsistency may be caused by the differences in diet assessment measures and study population. The previous survey collected the data by the means of a validated food frequency questionnaire (FFQ) to measure coffee consumption, which may cause non-differential misclassification, leading to biased study results [31]. Furthermore, the response categories of FFQ were close-ended, which may lead to an underestimation of coffee consumption [32]. In the current survey, we adapted a 2-day, 24-h food recall to assess participants' habitual coffee consumption, which could avoid the mentioned above biases.
Unexpectedly, we found that compared to non-coffee consumers, participants who consumed ≤ 1 serving/day or >1 serving/day of any coffee were more likely to have increased FBG levels in both men and women. The results are not consistent with findings from previous epidemiologic surveys. In a cross-sectional prospective study in Dutch, higher coffee consumption tended to be significantly associated with a lower risk of T2DM [33]. While numerous prospective cohort studies indicated the inverse relationship between habitual coffee consumption and the incidence of T2DM [34]. This inconsistency could be attributed to various factors, such as coffee consumption type, dose, and other constitutional and environmental factors.
The mechanism of the association between coffee consumption and plasma glucose remains unclear yet. Caffeine, one of the main bioactive compounds in coffee, has numerous biological impacts on all aspects of human health [35]. A previous experimental study indicated that short-term coffee consumption could impair glucose tolerance and reduce insulin sensitivity due to the A1 attenuating aortic dissection affected by the caffeine-blocking; however, this effect will not last long [36]. Long-term coffee consumption could prevent the incidence of T2DM by affecting post-load rather than fasting glucose metabolism [37]. On the other hand, the effect of caffeine on plasma glucose is determined by the glycemic index of food [38]. From a genetic point of view, Robertson et al. revealed that the plasma glucose level might be affected by the plasma glucose level, such as rs762551 single-nucleotide polymorphism in the CYP1A2 gene, which can directly affect the rate of the body's metabolism of caffeine [39]. In this aspect, CYP1A2 activity can be effected by numerous environmental factors, including race, gender, smoking, and drinking status. Another experimental study suggested that compared to baseline, fasting glucose concentrations were higher after consuming 1 L of coffee daily for 2 weeks, but not after 4 weeks, indicating that caffeine is substantially influenced by the development of tolerance [40]. Therefore, the inconsistent results in the current study may be attributed to the ignorance of caffeine dose, and the participants were old, so as to affect the metabolism of caffeine.
In the current study, we found that the protective effect of habitual coffee drinking on BP was significant only in women. In line with this finding, Grosso et al. [ 41] reported that higher coffee consumption was associated with a decreased risk of hypertension appeared to be significant only in women [41]. Actually, numerous epidemiological studies on the influence of coffee or caffeine on the incidence of CVD system have provided controversial and inconsistent findings. A systematic review and meta-analysis of randomized controlled clinical trials indicated that habitual coffee consumption can slightly increases SBP and DBP [42]. In this regard, some previous studies reported a negative association between habitual coffee consumption and the risk of CVD [43, 44], while others revealed a positive association [45], or no significant association [46]. Another recent meta-analysis revealed that BP elevations tended to be associated only with caffeine but not coffee [47]. Thus, these conflicting findings may be due to the different types of brewing coffee, various confounding dietary factors, and the daily consuming amount.
Overall, the caffeine acute effects on BP are well-known, but the mechanism underlying the effect of chronic coffee consumption remains unclear [48]. There is experimental evidence that an acute raise in BP due to coffee intake develops increasing tolerance, and intravenous caffeine led to a rise in muscle sympathetic activity and increased BP among both non-habitual and habitual coffee consumers, while coffee dietary consumption led to elevated BP on existed in non-habitual coffee consumers [49]. This may be the reason that, compared to non-coffee consumers, habitual coffee consumers are less likely to show an average BP response after coffee intake. Moreover, phenolic, the main compound of coffee, can play a key role in regulating the cellular processes that lead to inflammatory responses [50]. Oxidative stress has a great impact on the process that causes metabolism impairment and chronic conditions development, including hypertension [51]. In this aspect, women have more antioxidants than men in natural differences [52], this may explain the gender difference in coffee consumption effect. From the point of view of genetics, lifestyle habits (such as drinking or smoking status) or genetics may influence the activity of enzymes so as to affect metabolize caffeine and BP levels. Taking into account all variables mentioned earlier, it may explain the significant protective effect of coffee intake for women but not men.
To the best of our knowledge, the current study is the first to discuss the association between coffee consumption patterns and MetS among middle-aged and older adults in Shenzhen. We assessed individuals' coffee consumption patterns upon 2-day, 24-h recall data, which can relatively obtain accurate information about habitual coffee consumption. Moreover, we estimated both coffee consumption type and daily serving times of each type, so as to provide not only qualitative but also quantitative information regarding coffee consumption patterns. However, several limitations should be noted. First, the causal associations between coffee consumption patterns and MetS could not be confirmed due to the cross-sectional nature. Second, we did not include actual consumption volumes, brewing method, sugar in coffee or other coffee ingredient consumption, total energy intake, and presence of caffeine were not obtained. Third, the study only included healthy residents, it may be potential for residual confounding factors and other lifestyle factors. Multilateral studies considering coffee consumption timing, volumes, frequency, ingredients, and other behavioral factors by gender are needed to address the association of coffee consumption patterns on MetS in a more expanded population.
## Conclusion
In conclusion, a significant positive association between coffee consumption patterns and elevated FBG in both men and women was found, whereas consumption was inversely associated with elevated BP only in women. Our findings reinforce the hypothesis on the possible benefits of hypertension for women. Due to methodological limitations, further research prospective studies or well-designed randomized controlled trials are needed to confirm the causal association.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethics Committee of the Health Science Centre, Shenzhen University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
RN and ZY conceived the study. RN, HL, LQ, GH, SL, and ZY collected data. RN and GH provided the recruitment resources. RN completed the original draft preparation. ZY reviewed, edited the final draft, and received the funding. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Reynolds K, He J. **Epidemiology of the metabolic syndrome**. *Am J Med Sci.* (2005) **330** 273-79. DOI: 10.1097/00000441-200512000-00004
2. Eckel RH, Grundy SM, Zimmet PZ. **The metabolic syndrome**. *Lancet.* (2005) **365** 1415-28. DOI: 10.1016/S0140-6736(05)66378-7
3. Huang J, Huang J, Wang J, Chung V, Wong M. **Metabolic syndrome and risk of cancer in Chinese populations: a systematic review and meta-analysis in 57 260 individuals**. *Lancet.* (2018) **392** S15. DOI: 10.1016/S0140-6736(18)32644-8
4. Ford ES, Giles WH, Dietz WH. **Prevalence of the metabolic syndrome among US adults - findings from the third national health and nutrition examination survey**. *J Am Med Assoc.* (2002) **287** 356-59. DOI: 10.1001/jama.287.3.356
5. Mozumdar A, Liguori G. **Persistent increase of prevalence of metabolic syndrome among US Adults: NHANES III to NHANES 1999-2006**. *Diabetes Care.* (2011) **34** 216-19. DOI: 10.2337/dc10-0879
6. Lu M, Cheng H, Lai JC, Chen S. **The relationship between habitual coffee drinking and the prevalence of metabolic syndrome in Taiwanese adults: evidence from the Taiwan biobank database**. *Nutrients.* (2022) **14** 1867. DOI: 10.3390/nu14091867
7. Song Y, Sung J, Lee K. **Genetic and environmental relationships of metabolic and weight phenotypes to metabolic syndrome and diabetes: the healthy twin study**. *Metab Syndr Relat D.* (2015) **13** 36-44. DOI: 10.1089/met.2014.0087
8. Brown AE, Walker M. **Genetics of insulin resistance and the metabolic syndrome**. *Curr Cardiol Rep.* (2016) **18** 1-8. DOI: 10.1007/s11886-016-0755-4
9. Gao M, Ding D, Huang J, Qu Y, Wang Y, Huang Q. **Association of genetic variants in the adiponectin gene with metabolic syndrome: a case-control study and a systematic meta-analysis in the Chinese population**. *Plos ONE.* (2013) **8** 58412. DOI: 10.1371/journal.pone.0058412
10. Kim U, Choi E, Park H, Lee H, Park B, Kim H, Hong Y, Jung S, Park H. **The mediating effect of inflammation between the dietary and health-related behaviors and metabolic syndrome in adolescence**. *Nutrients.* (2022) **14** 2339. DOI: 10.3390/nu14112339
11. Da Silva Moura AR, Rebelo Sampaio Da Paz SM, Goncalves Frota KDM, Resende Goncalves De Carvalho CM. **Lifestyle associated with risk for metabolic syndrome in adults and the elderly**. *Nutrition.* (2022) **99** 111647. DOI: 10.1016/j.nut.2022.111647
12. Grosso G, Pajak A, Mistretta A, Marventano S, Raciti T, Buscemi S. **Protective role of the Mediterranean diet on several cardiovascular risk factors: Evidence from Sicily, southern Italy**. *Nutr Metab Cardiovas.* (2014) **24** 370-77. DOI: 10.1016/j.numecd.2013.09.020
13. Julibert A, Del Mar Bibiloni M, Mateos D, Angullo E, Tur JA. **Dietary fat intake and metabolic syndrome in older adults**. *Nutrients.* (2019) **11** 1901. DOI: 10.3390/nu11081901
14. Mazidi M, Pennathur S, Afshinnia F. **Link of dietary patterns with metabolic syndrome: analysis of the national health and nutrition examination survey**. *Nutr Diabetes.* (2017) **7** 11. DOI: 10.1038/nutd.2017.11
15. Samoggia A, Riedel B. **Consumers perceptions of coffee health benefits and motives for coffee consumption and purchasing**. *Nutrients.* (2019) **11** 653. DOI: 10.3390/nu11030653
16. Jiang X, Zhang D, Jiang W. **Coffee and caffeine intake and incidence of type 2 diabetes mellitus: a meta-analysis of prospective studies**. *Eur J Nutr.* (2014) **53** 25-38. DOI: 10.1007/s00394-013-0603-x
17. Grosso G, Micek A, Godos J, Pajak A, Sciacca S, Bes-Rastrollo M, Galvano F, Martinez-Gonzalez MA. **Long-term coffee consumption is associated with decreased incidence of new-onset hypertension: a dose-response meta-analysis**. *Nutrients.* (2017) **9** 890. DOI: 10.3390/nu9080890
18. Saeed M, Naveed M, BiBi J, Ali Kamboh A, Phil L, Chao S. **Potential nutraceutical and food additive properties and risks of coffee: a comprehensive overview**. *Crit Rev Food Sci.* (2019) **59** 3293-319. DOI: 10.1080/10408398.2018.1489368
19. Suzuki A, Nomura T, Jokura H, Kitamura N, Saiki A, Fujii A. **Chlorogenic acid-enriched green coffee bean extract affects arterial stiffness assessed by the cardio-ankle vascular index in healthy men: a pilot study**. *Int J Food Sci Nutr.* (2019) **70** 901-08. DOI: 10.1080/09637486.2019.1585763
20. Echeverri D, Pizano A, Montes FR, Forcada P. **Acute effect of coffee consumption on arterial stiffness, evaluated using an oscillometric method**. *Artery Res.* (2017) **17** 16-32. DOI: 10.1016/j.artres.2017.01.001
21. Mattias C, Larsson SC. **Coffee consumption and reduced risk of developing type 2 diabetes: a systematic review with meta-analysis**. *Nutr Rev.* (2018) **76** 395-417. DOI: 10.1093/nutrit/nuy014
22. Moran-Lev H, Cohen S, Zelber-Sagi S, Mazkeret Mayer E, Anafy A, Yerushalmy-Feler A. **Effect of coffee and tea consumption on adolescent weight control: an interventional pilot study**. *Child Obes.* (2022). DOI: 10.1089/chi.2022.0032
23. Abe SK, Saito E, Sawada N, Tsugane S, Ito H, Lin Y. **Coffee consumption and mortality in Japanese men and women: a pooled analysis of eight population-based cohort studies in Japan (Japan cohort consortium)**. *Prev Med.* (2019) **123** 270-77. DOI: 10.1016/j.ypmed.2019.04.002
24. Lee J, Kim HY, Kim J. **Coffee consumption and the risk of obesity in Korean women**. *Nutrients.* (2017) **9** 1340. DOI: 10.20944/preprints201709.0165.v1
25. Shang F, Li X, Jiang X. **Coffee consumption and risk of the metabolic syndrome: a meta-analysis**. *Diabetes Metab.* (2016) **42** 80-7. DOI: 10.1016/j.diabet.2015.09.001
26. Grundy SM, Cleeman JI, Daniels SR, Donato KA, Eckel RH, Franklin BA. **Diagnosis and management of the metabolic syndrome - an american heart association/national heart, lung, and blood institute scientific statement**. *Curr Opin Cardiol.* (2006) **21** 1-06. DOI: 10.1097/01.hco.0000200416.65370.a0
27. Zhang Y, Xie Y, Huang L, Zhang Y, Li X, Fang Q. **Association of sleep duration and self-reported insomnia symptoms with metabolic syndrome components among middle-aged and older adults**. *Int J Environ Res Public Health.* (2022) **18** 11637. DOI: 10.3390/ijerph191811637
28. Shih Y, Wu H, Pan W, Chang H. **Increased coffee, tea, or other sugar-sweetened beverage consumption in adolescents is associated with less satisfactory dietary quality, body fatness and serum uric acid profiles over the past 18 years in Taiwan**. *Asia Pac J Clin Nutr.* (2019) **28** 371-82. DOI: 10.6133/apjcn.201906_28(2).0020
29. Bouchard DR, Ross R, Janssen I. **Coffee, tea and their additives: association with bmi and waist circumference**. *Obes Facts.* (2010) **3** 345-52. DOI: 10.1159/000322915
30. Suliga E, Koziel D, Ciesla E, Rebak D, Gluszek S. **Coffee consumption and the occurrence and intensity of metabolic syndrome: a cross-sectional study**. *Int J Food Sci Nutr.* (2017) **68** 507-13. DOI: 10.1080/09637486.2016.1256381
31. Olga P, Athanasia S, Nikolaos R, Charalampia D, Antonios K. **Development and validation of a functional foods frequency questionnaire for Greek adults**. *Curr Top Nutraceut R.* (2021) **19** 373-81. DOI: 10.37290/ctnr2641-452X.19:373-381
32. Watson EJ, Kohler M, Banks S, Coates AM. **Validation and reproducibility of an Australian caffeine food frequency questionnaire**. *Int J Food Sci Nutr.* (2017) **68** 617-26. DOI: 10.1080/09637486.2016.1268102
33. Van Dam RM, Dekker JM, Nijpels G, Stehouwer C, Bouter LM, Heine RJ. **Coffee consumption and incidence of impaired fasting glucose, impaired glucose tolerance, and type 2 diabetes: the Hoorn Study**. *Diabetologia.* (2004) **47** 2152-59. DOI: 10.1007/s00125-004-1573-6
34. Ding M, Bhupathiraju SN, Chen M, Van Dam RM, Hu FB. **Caffeinated and decaffeinated coffee consumption and risk of type 2 diabetes: a systematic review and a dose-response meta-analysis**. *Diabetes Care.* (2014) **37** 569-86. DOI: 10.2337/dc13-1203
35. Ricci E, Vigano P, Cipriani S, Somigliana E, Chiaffarino F, Bulfoni A, Parazzini F. **Coffee and caffeine intake and male infertility: a systematic review**. *Nutr J.* (2017) **16** 1-14. DOI: 10.1186/s12937-017-0257-2
36. Zhang Z, Jiang Y, Zhou Z, Huang J, Chen S, Zhou W. **Scavenger receptor A1 attenuates aortic dissection via promoting efferocytosis in macrophages**. *Biochem Pharmacol.* (2019) **168** 392-403. DOI: 10.1016/j.bcp.2019.07.027
37. Mirmiran R, Carlstrom M, Bahadoran Z, Azizi F. **Long-term effects of coffee and caffeine intake on the risk of pre-diabetes and type 2 diabetes: Findings from a population with low coffee consumption**. *Nutr Metab Cardiovas.* (2018) **28** 1261-66. DOI: 10.1016/j.numecd.2018.09.001
38. Moisey LL, Kacker S, Bickerton AC, Robinson LE, Graham TE. **Caffeinated coffee consumption impairs blood glucose homeostasis in response to high and low glycemic index meals in healthy men**. *Am J Clin Nutr.* (2008) **87** 1254-61. DOI: 10.1093/ajcn/87.5.1254
39. Robertson TM, Clifford MN, Penson S, Williams P, Robertson MD. **Postprandial glycaemic and lipaemic responses to chronic coffee consumption may be modulated by CYP1A2 polymorphisms**. *Brit J Nutr.* (2018) **119** 792-800. DOI: 10.1017/S0007114518000260
40. Van Dam RM, Pasman WJ, Verhoef P. **Effects of coffee consumption on fasting blood glucose and insulin concentrations – Randomized controlled trials in healthy volunteers**. *Diabetes Care.* (2004) **27** 2990-92. DOI: 10.2337/diacare.27.12.2990
41. Grosso G, Stepaniak U, Polak M, Micek A, Topor-Madry R, Stefler D. **Coffee consumption and risk of hypertension in the Polish arm of the HAPIEE cohort study**. *Eur J Clin Nutr.* (2016) **70** 109-15. DOI: 10.1038/ejcn.2015.119
42. Noordzij M, Uiterwaal C, Arends LR, Kok FJ, Grobbee DE, Geleijnse JM. **Blood pressure response to chronic intake of coffee and caffeine: a meta-analysis of randomized controlled trials**. *J Hypertens.* (2005) **23** 921-28. DOI: 10.1097/01.hjh.0000166828.94699.1d
43. Schaefer SM, Kaiser A, Behrendt I, Eichner G, Fasshauer M. **Association of alcohol types, coffee and tea intake with mortality: prospective cohort study of UK Biobank participants**. *Brit J Nutr.* (2022). DOI: 10.3390/brainsci12030360
44. Ding M, Satija A, Bhupathiraju SN, Hu Y, Sun Q, Han J, Lopez-Garcia E, Willett W, van Dam RM, Hu FB. **Association of coffee consumption with total and cause-specific mortality in 3 large prospective cohorts**. *Circulation.* (2015) **132** 2305. DOI: 10.1161/CIRCULATIONAHA.115.017341
45. Grioni S, Agnoli C, Sieri S, Pala V, Ricceri F, Masala G. **Espresso coffee consumption and risk of coronary heart disease in a large Italian cohort**. *Plos ONE.* (2015). DOI: 10.1371/journal.pone.0126550
46. Floegel A, Pischon T, Bergmann MM, Teucher B, Kaaks R, Boeing H. **Coffee consumption and risk of chronic disease in the European prospective investigation into cancer and nutrition (EPIC)-germany study**. *Am J Clin Nutr.* (2012) **95** 901-08. DOI: 10.3945/ajcn.111.023648
47. Eumann Mesas A, Leon-Munoz LM, Rodriguez-Artalejo F, Lopez-Garcia E. **The effect of coffee on blood pressure and cardiovascular disease in hypertensive individuals: a systematic review and meta-analysis**. *Am J Clin Nutr.* (2011) **94** 1113-26. DOI: 10.3945/ajcn.111.016667
48. Miranda AM, Steluti J, Fisberg RM, Marchioni DM. **Association between coffee consumption and its polyphenols with cardiovascular risk factors: a population-based Study**. *Nutrients.* (2017) **9** 276. DOI: 10.3390/nu9030276
49. Maserejian NN, Wager CG, Giovannucci EL, Curto TM, McVary KT, McKinlay JB. **Intake of caffeinated, carbonated, or citrus beverage types and development of lower urinary tract symptoms in men and women**. *Am J Epidemiol.* (2013) **177** 1399-410. DOI: 10.1093/aje/kws411
50. Okur I, Soyler B, Sezer P, Oztop MH, Alpas H. **Improving the recovery of phenolic compounds from spent coffee grounds (SCG) by environmentally friendly extraction techniques**. *Molecules.* (2021) **26** 613. DOI: 10.3390/molecules26030613
51. Griendling KK, Camargo LL, Rios FJ, Alves-Lopes R, Touyz RM. **Oxidative stress and hypertension**. *Circ Res.* (2021) **128** 993-1020. DOI: 10.1161/CIRCRESAHA.121.318063
52. Viña J, Borrás C. **Women live longer than men: understanding molecular mechanisms offers opportunities to intervene by using estrogenic compounds**. *Antioxid Redox Sign.* (2010) **13** 269-78. DOI: 10.1089/ars.2009.2952
|
---
title: Effect of remdesivir on adverse kidney outcomes in hospitalized patients with
COVID-19 and impaired kidney function
authors:
- Rituvanthikaa Seethapathy
- Qiyu Wang
- Sophia Zhao
- Ian A. Strohbehn
- Joshua D. Long
- James E. Dinulos
- Destiny Harden
- Vinay B. Kadiyala
- Daiana Moreno
- Meghan E. Sise
journal: PLOS ONE
year: 2023
pmcid: PMC9970064
doi: 10.1371/journal.pone.0279765
license: CC BY 4.0
---
# Effect of remdesivir on adverse kidney outcomes in hospitalized patients with COVID-19 and impaired kidney function
## Abstract
### Background
Chronic kidney disease (CKD) is an important risk factor for mortality from COVID-19. Remdesivir has been shown to shorten time to recovery in patients with severe COVID-19. However, exclusion of patients with severe kidney function impairment in clinical trials has led to concerns about kidney safety of remdesivir in patients with pre-existing kidney disease.
### Methods
Retrospective propensity score matched cohort study of hospitalized patients with COVID-19 admitted with estimated glomerular filtration rate (eGFR) between 15 − 60 mL/min/1.73m2. Remdesivir-treated patients were 1:1 matched to historical comparators admitted during the first wave of COVID-19 (between March-April 2020) prior to emergency use authorization of remdesivir using propensity scores accounting for factors predicting treatment assignment. Dependent outcomes included in-hospital peak creatinine, incidence of doubling of creatine, rate of kidney replacement therapy initiation and eGFR among surviving patients at day 90.
### Results
175 remdesivir-treated patients were 1:1 matched to untreated historical comparators. Mean age was 74.1 (SD 12.8), $56.9\%$ were male, $59\%$ patients were white, and the majority ($83.1\%$) had at least one co-morbidity. There were no statistically significant differences in peak creatinine during hospitalization (2.3mg/dL vs. 2.5 mg/dL, $$P \leq 0.34$$), incidence of doubling of creatinine ($10.3\%$ vs. $13.1\%$, $$P \leq 0.48$$), and rate of kidney replacement therapy initiation ($4.6\%$ vs. $6.3\%$, $$P \leq 0.49$$) in remdesivir-treated patients versus matched untreated historical comparators, respectively. Among surviving patients, there was no difference of the average eGFR at day 90 (54.7 ± 20.0 mL/min/1.73m2 for remdesivir-treated patients vs. 51.7 ± 19.5 mL/min/1.73m2 for untreated comparators, $$P \leq 0.41$$).
### Conclusions
Remdesivir use in patients with impaired kidney function (eGFR between 15 − 60 mL/min/1.73m2) who present to the hospital with COVID-19 is not associated with increased risk of adverse kidney outcomes.
## Introduction
Remdesivir was the first antiviral therapy approved by the Food and Drug Administration for treatment of coronavirus disease 2019 (COVID-19). It has been shown to shorten time to recovery in hospitalized adults with pneumonia [1], and a recent randomized clinical trial also demonstrated its efficacy in the outpatient setting to prevent hospitalization [2]. Chronic kidney disease (CKD) has been recognized as a key risk factor for severe COVID-19 and mortality [3–5]. However, patients with estimated glomerular filtration rate (eGFR) < 30mL/min/1.73m2 were excluded from registrational trials of remdesivir, and there has been ongoing concern that remdesivir may be nephrotoxic, especially in patients with pre-existing CKD. Initially developed as an antiviral therapy during the 2014 Ebola outbreak, remdesivir has limited water solubility and requires a sulfobutylether-beta-cyclodextrin (SBECD) excipient for intravenous delivery. SBECD is a large oligosaccharide that is eliminated by glomerular filtration. In animal studies, high doses (approximately 50-fold greater than the dose administered in clinical use) can lead to renal tubular vacuolization and foamy macrophages in the liver [6]. For hospitalized patients, remdesivir is typically prescribed for five doses (200mg intravenous loading dose on day 1, followed by 100mg daily on days 2 through 5). Real world data regarding nephrotoxicity of remdesivir remains controversial: multiple case series have suggested that remdesivir may be safe in patients with advanced kidney disease [7–12], however, analysis of pharmacovigilance databases demonstrated higher reports of kidney injury associated with remdesivir compared to other medications used to treat COVID-19 [13, 14].
Thus, there remains clinical equipoise on whether remdesivir should be used in patient with kidney function impairment who present with moderate to severe COVID-19. To provide further evidence on kidney safety, we designed a propensity-score matched study to compare four adverse kidney outcomes (peak creatinine value, rate of doubling of creatinine, rate of kidney replacement therapy initiation, and eGFR at day 90) among hospitalized patients with COVID-19 who were treated with remdesivir and historical comparators who did not receive remdesivir.
## Patients and propensity score matching
We included adult patients (≥18 years old) who were admitted to Mass General Brigham (MGB) healthcare system with COVID-19 whose admission creatinine corresponded to an eGFR value between 15 − 60mL/min/1.73m2 [15]. Remdesivir-treated patients received at least one dose of remdesivir, and had at least one repeat creatinine measurement after receiving remdesivir. Remdesivir-treated patients were excluded if they received remdesivir >72 hours after admission, or if they received remdesivir at an outside hospital prior to being transferred to our healthcare network. Historical comparators were adult patients admitted to MGB during the first wave of COVID-19 in Boston, MA between March-April 2020 (prior to the emergency use authorization of remdesivir) and had an admission creatinine that corresponded to an eGFR value between 15 − 60mL/min/1.73m2. Comparators were excluded if they had end-stage kidney disease (ESKD) or if they were placed into hospice/comfort care prior to receiving therapy for COVID-19 (Fig 1).
**Fig 1:** *Patient flow.Historical comparators were obtained from the first wave of COVID-19 in Boston, Massachusetts and had been admitted between March 17, 2020 and April 30, 2020. Remdesivir treated patients were admitted between April 21, 2020 and November 29, 2020. Abbreviations: eGFR = estimated glomerular filtration rate, EUA = emergency use authorization.*
In order to control for confounding factors that are associated with both treatment assignment and outcome [16], we performed propensity score matching between the remdesivir-treated cohort and the untreated historical comparators. Patients were matched on independent variables associated with the administration of remdesivir, including age, sex, race/ethnicity, diabetes, hypertension, CKD (defined by eGFR < 60mL/min/1.73m2 sustained for at least 90 days), kidney transplantation, admission creatinine, and markers of disease severity on hospital admission, including need for mechanical ventilation and the highest sequential organ failure assessment (SOFA) score within the first 12 hours of admission [17]. We used logistic regression to estimate the propensity of receiving remdesivir, and performed 1:1 nearest-neighbor matching without replacement with a caliper of 0.1 standard deviation of the propensity score [18]. Standardized mean difference was calculated to evaluate the quality of the match [16].
## Study outcomes
The primary dependent outcome was the peak creatinine level during hospitalization in remdesivir-treated patients compared to untreated historical comparators. The secondary dependent outcomes were 1) incidence of doubling of creatinine from admission; 2) initiation of kidney replacement therapy (either continuous or intermittent dialysis modality); and the 3) average eGFR among patients who were alive at day 90. To determine the eGFR at day 90, we evaluated all creatinine values between 75 to 180 days post-hospital admission and selected the value closest to day 90.
## Statistical analysis
Paired t-test or Wilcoxon signed-rank test was used to compare peak creatinine, hospital length of stay, creatinine measurements between remdesivir-treated patients and their matched untreated historical comparators, as appropriate. Among patients who initiated kidney replacement therapy, a peak creatinine of 10mg/dL was imputed on the day of kidney replacement therapy initiation if pre-dialysis creatinine level was less than 10mg/dL. McNemar test was used to compare the incidence of doubling of creatinine from admission and rate of kidney replacement therapy initiation between remdesivir-treated patients and matched untreated historical comparators. Among surviving patients, we compared the average eGFR at day 90 by remdesivir treatment status using independent sample t-test, as not all matched pairs had available day 90 creatinine levels.
To evaluate the robustness of our findings we performed the following sensitivity analyses: using a paired t-test, we compared the peak creatinine among remdesivir-treated patients who received full course of therapy (≥ 5 doses) and among remdesivir-treated patients who had ≥ 5 creatinine measurements after starting remdesivir. Finally, we stratified the analysis of peak creatinine by admission eGFR (eGFR 15–29 mL/min/1.73m2 and 30–60 mL/min/1.73m2).
## Results
There were 203 individuals with admission eGFR between 15 − 60 mL/min/1.73m2 who received remdesivir; 16 patients were excluded due to initiating remdesivir more than 72 hours after hospital admission and 4 patients were excluded due to having received remdesivir at an outside hospital prior to transfer to our healthcare system, leaving 183 remdesivir-treated patients included. There were 556 potential historical comparators; after applying the exclusion criteria, we included 460 potential historical comparators hospitalized with COVID-19 between March and April 2020 in propensity score matching (Fig 1).
Prior to matching, historical comparators were older and had higher SOFA scores, reflecting the high acuity of COVID-19 in patients hospitalized in the first wave in Boston, Massachusetts. A sufficiently close match was found for 175 of the 184 ($95\%$) remdesivir-treated patients. All patient characteristics achieved good balance (standardized mean differences < 0.1) (Table 1).
**Table 1**
| Unnamed: 0 | Before propensity score matching | Before propensity score matching.1 | Before propensity score matching.2 | Before propensity score matching.3 | After propensity score matching | After propensity score matching.1 | After propensity score matching.2 | After propensity score matching.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Characteristic | Remdesivir-treated N = 183 | Untreated historical comparators N = 460 | P value | SD* | Remdesivir-treated N = 175 | Untreated historical comparators N = 175 | P value | SD* |
| Age (mean (SD)) | 73.2 (12) | 76.9 (13) | 0.001 | -0.31 | 73.7 (12) | 74.5 (14) | 0.534 | -0.06 |
| Female (%) | 76 (42) | 234 (51) | 0.040 | 0.09 | 75 (43) | 76 (43) | 1.000 | 0.006 |
| White Race (%) | 105 (57) | 279 (61) | 0.500 | -0.03 | 101 (58) | 107 (61) | 0.576 | -0.03 |
| Comorbidities | | | | | | | | |
| Hypertension (%) | 147 (80) | 364 (79) | 0.812 | 0.01 | 140 (80) | 144 (82) | 0.677 | -0.02 |
| Diabetes (%) | 108 (59) | 216 (47) | 0.008 | 0.12 | 100 (57) | 100 (57) | 1.000 | 0.00 |
| Preexisting CKD (%) | 86 (47) | 207 (45) | 0.71 | 0.02 | 82 (47) | 73 (42) | 0.380 | 0.05 |
| Kidney transplant (%) | 3 (1.6) | 7 (1.5) | 1.000 | 0.001 | 2 (1) | 3 (2) | 1.000 | -0.0057 |
| Disease severity at presentation | | | | | | | | |
| Invasive ventilation (%) | 20 (11) | 78 (17) | 0.072 | -0.06 | 20 (11) | 17 (10) | 0.735 | 0.08 |
| SOFA Score (mean (SD)) | 3.42 (2.5) | 4.09 (3.0) | 0.007 | -0.27 | 3.50 (3) | 3.31 (2) | 0.459 | 0.07 |
| Admission creatinine (mean (SD))** | 1.61 (0.54) | 1.62 (0.57) | 0.79 | -0.02 | 1.61 (0.5) | 1.58 (0.6) | 0.673 | 0.05 |
In the matched cohort, mean age was 74 (SD 13), $56.9\%$ were male, and $59.0\%$ patients were white. The majority of patients had at least one co-morbidity ($83.1\%$), and $44.3\%$ had pre-existing CKD. A total of $10.6\%$ received mechanical ventilation within 12 hours of presentation. There were no significant differences in hospital length of stay or total number of creatinine measurements between the groups: median length of stay was 8 days (IQR 6–17) and number of creatinine measurements was 11 (IQR 7–21) in remdesivir-treated patients; while median length of stay was 9 days (IQR 5–16), and number of creatinine measurements was 10 (IQR 6–23) in untreated historical comparators (Wilcoxon signed-rank test, $$P \leq 0.38$$ and $$P \leq 0.1$$, respectively). The majority of patients (141 patients, [$81\%$]) received a full course of remdesivir (≥ 5 days). Among the remaining 34 patients who received shorter course of remdesivir (<5 doses), the reasons for early discontinuation included: rapid recovery [19], worsening kidney function [4], elevated transaminases [2], anaphylaxis [1], transition to comfort care [6] and uncertain causes [2].
## Peak creatinine level during hospitalization
Mean peak creatinine in the remdesivir-treated group was 2.3 mg/dL ($95\%$ confidence interval [CI] 1.98–2.57) compared to 2.5 mg/dL ($95\%$ CI 2.13–2.89) in matched untreated historical comparators (Paired t test, $$P \leq 0.34$$) (Fig 2A). Sensitivity analyses including only the subset of patients who received full course of remdesivir and those with at least 5 post-treatment creatinine measurements and their matched historical comparators did not identify differences in peak creatinine (S1A and S1B Fig). Stratification of remdesivir-treated patients by their admission eGFR also did not affect the comparison of peak creatinine during hospitalization (S2A and S2B Fig).
**Fig 2:** *Kidney outcomes in remdesivir-treated patients and propensity score matched untreated historical comparators.There was no statistically significant difference in any of the adverse kidney outcomes. A) Peak creatinine during hospitalization B) Percentage of patients developed doubling of creatinine C) Percentage of patients requiring initiation of kidney replacement therapy D) eGFR at day 90 among surviving patients. Boxplot (2A, 2D) showed the 1st quartile, median and 3rd quartile of peak creatinine and day 90 eGFR, respectively. Abbreviation: KRT: kidney replacement therapy; eGFR: estimated glomerular filtration rate.*
## Incidence of doubling of creatinine
Eighteen of the remdesivir-treated patients ($10.3\%$) versus 23 of the matched untreated historical comparators ($13.1\%$) experienced a doubling of serum creatinine during their hospitalization (McNemar test, $$P \leq 0.48$$) (Fig 2B).
## Incidence of renal replacement therapy initiation
Eight ($4.6\%$) of the remdesivir-treated patients and 11 ($6.3\%$) of the matched untreated historical comparators received kidney replacement therapy during hospitalization, respectively (McNemar test, $$P \leq 0.49$$) (Fig 2C).
## Day 90 eGFR among surviving patients
Among surviving patients to who were followed for at least 90 days post admission ($$n = 120$$), the average eGFR at day 90 was 54.7 mL/min/1.73m2 (SD 20.0) in remdesivir-treated patients ($$n = 66$$) compared to 51.7 mL/min/1.73m2 (SD 19.5) among untreated historical comparators ($$n = 54$$) (Independent sample t test, $$P \leq 0.41$$) (Fig 2D).
## Discussion
In this propensity score matched retrospective cohort study, we did not detect significant differences in adverse kidney outcomes including peak serum creatinine, rate of doubling of serum creatinine, need for kidney replacement therapy, and eGFR at day 90 between patients who received remdesivir and those who did not. Our findings provide additional safety signal and are complementary to previous studies suggesting that off-label remdesivir use in patients with ESKD and eGFR < 30mL/min/1.73m2 is safe and well tolerated [7–12]. Prior studies that identified a signal for remdesivir nephrotoxicity used pharmacovigilance databases, which may be limited by reporting bias making it challenging to generalize causal association [13, 14]. Thus, it is important to include untreated comparators, particularly given the high baseline rate of acute kidney injury among patients hospitalized for COVID-19 (ranging from 17~$56\%$) [19–23].
Exclusion and under-representation of patients with kidney disease in clinical trials of life-saving treatments is a major problem that has been magnified by the COVID-19 pandemic [24–26]. Studies have noted that kidney disease was an exclusion criteria in approximately half of the registered trials evaluating therapeutics for COVID-19, including all of the major trials that led to the approval of the antiviral agents for COVID-19 (remdesivir, nirmaltrelvir/ritonavir, and molnupiravir) [1, 25–28]. Thus, there is an urgent need to address this knowledge gap using real-world data, as patients with kidney disease suffer significantly higher morbidity and mortality from COVID-19. Numerous studies have shown ESKD and advanced CKD are among the top comorbid medical conditions associated with death in patients with COVID-19 [3, 5, 29, 30]. Recent studies show that even after vaccination, the risk of hospitalization and death is substantial in patients with ESKD who develop COVID-19; up to $45\%$ of patients with breakthrough infection may be hospitalized [31], and up to $7\%$ may die within 28 days [32]. This highlights the importance of including this particularly vulnerable population into clinical trials for evaluation of antiviral and immunomodulating treatments for COVID-19.
Our study has several limitations: First, we chose historical controls from the first wave of COVID-19 prior to emergency use authorization of remdesivir to avoid confounding by indication and minimize selection bias. We performed detailed chart review to ensure accurate information on disease severity and other baseline covariates that are associated with the likelihood of remdesivir administration and achieved good balance between the remdesivir-treated patients and matched untreated comparators after propensity score matching. However, the rapidly changing standard of care could have impacted the disease outcome and kidney injury rate [20], which could not be fully adjusted for by propensity score matching. For example, the effect of dexamethasone use could not be controlled for as all historical comparators were admitted prior to routine use of dexamethasone at our center [33]. Second, our outcome of eGFR at day 90 was limited by the fact that many surviving patients did not have creatinine values within our healthcare system available within the 75 to 180 days window, raising concern for informative censoring. Fortunately, loss to follow-up was similar in each group. Third, creatinine-based GFR estimating equations have low sensitivity for detecting renal dysfunction in critically ill patients who have low creatinine production, thus the rate of acute kidney injury and more importantly, the development of CKD in survivors may be underestimated [34].
## Conclusions
Our report adds to a growing body of data suggesting that in patients presenting to the hospital with impaired kidney function (eGFR 15–60mL/min/1.73m2), remdesivir is not associated with short-term adverse kidney outcomes.
## References
1. Beigel JH, Tomashek KM, Dodd LE. **Remdesivir for the Treatment of Covid-19—Final Report**. *New England Journal of Medicine* (2020.0) **383** 1813-1826. DOI: 10.1056/NEJMoa2007764
2. Gottlieb RL, Vaca CE, Paredes R. **Early Remdesivir to Prevent Progression to Severe Covid-19 in Outpatients**. *New England Journal of Medicine* (2021.0) **386** 305-315. DOI: 10.1056/NEJMoa2116846
3. **Chronic kidney disease is a key risk factor for severe COVID-19: a call to action by the ERA-EDTA**. *Nephrol Dial Transplant* (2021.0) **36** 87-94. DOI: 10.1093/ndt/gfaa314
4. Ozturk S, Turgutalp K, Arici M. **Mortality analysis of COVID-19 infection in chronic kidney disease, haemodialysis and renal transplant patients compared with patients without kidney disease: a nationwide analysis from Turkey**. *Nephrol Dial Transplant* (2020.0) **35** 2083-2095. PMID: 33275763
5. Flythe JE, Assimon MM, Tugman MJ. **Characteristics and Outcomes of Individuals With Pre-existing Kidney Disease and COVID-19 Admitted to Intensive Care Units in the United States**. *Am J Kidney Dis* (2021.0) **77**. DOI: 10.1053/j.ajkd.2020.09.003
6. Luke DR, Tomaszewski K, Damle B. **Review of the basic and clinical pharmacology of sulfobutylether-beta-cyclodextrin (SBECD)**. *J Pharm Sci* (2010.0) **99** 3291-3301. DOI: 10.1002/jps.22109
7. Thakare S, Gandhi C, Modi T. **Safety of Remdesivir in Patients With Acute Kidney Injury or CKD**. *Kidney Int Rep* (2021.0) **6** 206-210. DOI: 10.1016/j.ekir.2020.10.005
8. Pettit NN, Pisano J, Nguyen CT. **Remdesivir Use in the Setting of Severe Renal Impairment: A Theoretical Concern or Real Risk**. *Clin Infect Dis* (2021.0) **73** e3990-e3995. DOI: 10.1093/cid/ciaa1851
9. Ackley TW, McManus D, Topal JE. **A Valid Warning or Clinical Lore: an Evaluation of Safety Outcomes of Remdesivir in Patients with Impaired Renal Function from a Multicenter Matched Cohort**. *Antimicrob Agents Chemother* (2021.0) **65**
10. Estiverne C, Strohbehn IA, Mithani Z. **Remdesivir in Patients With Estimated GFR <30 ml/min per 1.73 m(2) or on Renal Replacement Therapy**. *Kidney Int Rep* (2021.0) **6** 835-838. PMID: 33263094
11. Seethapathy R, Zhao S, Long JD. **A propensity-score matched observational study of remdesivir in patients with COVID-19 and severe kidney disease**. *Kidney360* (2021.0) **2**. DOI: 10.34067/KID.0006152021
12. Biancalana E, Chiriaco M, Sciarrone P. **Remdesivir, Renal Function and Short-Term Clinical Outcomes in Elderly COVID-19 Pneumonia Patients: A Single-Centre Study**. *Clin Interv Aging* (2021.0) **16** 1037-1046. DOI: 10.2147/CIA.S313028
13. Chouchana L, Preta LH, Tisseyre M. **Kidney disorders as serious adverse drug reactions of remdesivir in coronavirus disease 2019: a retrospective case-noncase study**. *Kidney Int* (2021.0) **99** 1235-1236. DOI: 10.1016/j.kint.2021.02.015
14. Gérard AO, Laurain A, Fresse A. **Remdesivir and Acute Renal Failure: A Potential Safety Signal From Disproportionality Analysis of the WHO Safety Database**. *Clin Pharmacol Ther* (2021.0) **109** 1021-1024. DOI: 10.1002/cpt.2145
15. Levey AS, Stevens LA, Schmid CH. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med* (2009.0) **150** 604-612. DOI: 10.7326/0003-4819-150-9-200905050-00006
16. Austin PC. **An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies**. *Multivariate Behav Res* (2011.0) **46** 399-424. DOI: 10.1080/00273171.2011.568786
17. 17KDIGO 2012 CKD guideline. https://kdigo.org/guidelines/ckd-evaluation-and-management/.
18. Austin PC. **A comparison of 12 algorithms for matching on the propensity score**. *Stat Med* (2014.0) **33** 1057-1069. DOI: 10.1002/sim.6004
19. Chan L, Chaudhary K, Saha A. **AKI in Hospitalized Patients with COVID-19**. *J Am Soc Nephrol* (2021.0) **32** 151-160. DOI: 10.1681/ASN.2020050615
20. Charytan DM, Parnia S, Khatri M. **Decreasing Incidence of Acute Kidney Injury in Patients with COVID-19 Critical Illness in New York City**. *Kidney Int Rep* (2021.0) **6** 916-927. DOI: 10.1016/j.ekir.2021.01.036
21. Alenezi FK, Almeshari MA, Mahida R. **Incidence and risk factors of acute kidney injury in COVID-19 patients with and without acute respiratory distress syndrome (ARDS) during the first wave of COVID-19: a systematic review and Meta-Analysis**. *Ren Fail* (2021.0) **43** 1621-1633. PMID: 34882508
22. Fisher M, Neugarten J, Bellin E. **AKI in Hospitalized Patients with and without COVID-19: A Comparison Study**. *J Am Soc Nephrol* (2020.0) **31** 2145-2157. DOI: 10.1681/ASN.2020040509
23. Strohbehn IA, Zhao S, Seethapathy H. **Acute Kidney Injury Incidence, Recovery, and Long-term Kidney Outcomes Among Hospitalized Patients With COVID-19 and Influenza**. *Kidney International Reports* (2021.0) **6** 2565-2574. DOI: 10.1016/j.ekir.2021.07.008
24. Zoccali C, Blankestijn PJ, Bruchfeld A. **Children of a lesser god: exclusion of chronic kidney disease patients from clinical trials**. *Nephrol Dial Transplant* (2019.0) **34** 1112-1114. DOI: 10.1093/ndt/gfz023
25. Chewcharat A, Chang YT, Sise ME. **Phase-3 Randomized Controlled Trials on Exclusion of Participants With Kidney Disease in COVID-19**. *Kidney Int Rep* (2021.0) **6** 196-199. DOI: 10.1016/j.ekir.2020.10.010
26. Major R, Selvaskandan H, Makkeyah YM. **The Exclusion of Patients with CKD in Prospectively Registered Interventional Trials for COVID-19-a Rapid Review of International Registry Data**. *J Am Soc Nephrol* (2020.0) **31** 2250-2252. DOI: 10.1681/ASN.2020060877
27. Hammond J, Leister-Tebbe H, Gardner A. **Oral Nirmatrelvir for High-Risk, Nonhospitalized Adults with Covid-19**. *New England Journal of Medicine* (2022.0)
28. Jayk Bernal A, Gomes da Silva MM, Musungaie DB. **Molnupiravir for Oral Treatment of Covid-19 in Nonhospitalized Patients**. *New England Journal of Medicine* (2021.0) **386** 509-520. DOI: 10.1056/NEJMoa2116044
29. Williamson EJ, Walker AJ, Bhaskaran K. **Factors associated with COVID-19-related death using OpenSAFELY**. *Nature* (2020.0) **584** 430-436. DOI: 10.1038/s41586-020-2521-4
30. Holman N, Knighton P, Kar P. **Risk factors for COVID-19-related mortality in people with type 1 and type 2 diabetes in England: a population-based cohort study**. *The Lancet Diabetes & Endocrinology* (2020.0) **8** 823-833. PMID: 32798471
31. Anand S, Montez-Rath ME, Han J. **SARS-CoV-2 vaccine antibody response and breakthrough infection in dialysis**. *medRxiv* (2021.0)
32. Bell S, Campbell J, Lambourg E. **The Impact of Vaccination on Incidence and Outcomes of SARS-CoV-2 Infection in Patients with Kidney Failure in Scotland**. *Journal of the American Society of Nephrology* (2022.0)
33. **Dexamethasone in Hospitalized Patients with Covid-19**. *New England Journal of Medicine* (2020.0) **384** 693-704. DOI: 10.1056/NEJMoa2021436
34. Hoste EAJ, Damen J, Vanholder RC. **Assessment of renal function in recently admitted critically ill patients with normal serum creatinine**. *Nephrology Dialysis Transplantation* (2005.0) **20** 747-753. DOI: 10.1093/ndt/gfh707
|
---
title: 'Referral to the NHS Diabetes Prevention Programme and conversion from nondiabetic
hyperglycaemia to type 2 diabetes mellitus in England: A matched cohort analysis'
authors:
- Rathi Ravindrarajah
- Matt Sutton
- David Reeves
- Sarah Cotterill
- Emma Mcmanus
- Rachel Meacock
- William Whittaker
- Claudia Soiland-Reyes
- Simon Heller
- Peter Bower
- Evangelos Kontopantelis
journal: PLOS Medicine
year: 2023
pmcid: PMC9970065
doi: 10.1371/journal.pmed.1004177
license: CC BY 4.0
---
# Referral to the NHS Diabetes Prevention Programme and conversion from nondiabetic hyperglycaemia to type 2 diabetes mellitus in England: A matched cohort analysis
## Abstract
Rathi Ravindrarajah and colleagues investigate whether referral to the NHS Healthier You Diabetes Prevention Programme was associated with reduced risk of developing type 2 diabetes in at-risk patients attending primary care in England.
### Background
The NHS Diabetes Prevention Programme (NDPP) is a behaviour change programme for adults who are at risk of developing type 2 diabetes mellitus (T2DM): people with raised blood glucose levels, but not in the diabetic range, diagnosed with nondiabetic hyperglycaemia (NDH). We examined the association between referral to the programme and reducing conversion of NDH to T2DM.
### Methods and findings
Cohort study of patients attending primary care in England using clinical Practice Research *Datalink data* from 1 April 2016 (NDPP introduction) to 31 March 2020 was used. To minimise confounding, we matched patients referred to the programme in referring practices to patients in nonreferring practices. Patients were matched based on age (≥3 years), sex, and ≥365 days of NDH diagnosis. Random-effects parametric survival models evaluated the intervention, controlling for numerous covariates. Our primary analysis was selected a priori: complete case analysis, 1-to-1 practice matching, up to 5 controls sampled with replacement. Various sensitivity analyses were conducted, including multiple imputation approaches. Analysis was adjusted for age (at index date), sex, time from NDH diagnosis to index date, BMI, HbA1c, total serum cholesterol, systolic blood pressure, diastolic blood pressure, prescription of metformin, smoking status, socioeconomic status, a diagnosis of depression, and comorbidities.
A total of 18,470 patients referred to NDPP were matched to 51,331 patients not referred to NDPP in the main analysis. Mean follow-up from referral was 482.0 (SD = 317.3) and 472.4 (SD = 309.1) days, for referred to NDPP and not referred to NDPP, respectively. Baseline characteristics in the 2 groups were similar, except referred to NDPP were more likely to have higher BMI and be ever-smokers. The adjusted HR for referred to NDPP, compared to not referred to NDPP, was 0.80 ($95\%$ CI: 0.73 to 0.87) ($p \leq 0.001$). The probability of not converting to T2DM at 36 months since referral was $87.3\%$ ($95\%$ CI: $86.5\%$ to $88.2\%$) for referred to NDPP and $84.6\%$ ($95\%$ CI: $83.9\%$ to $85.4\%$) for not referred to NDPP. Associations were broadly consistent in the sensitivity analyses, but often smaller in magnitude. As this is an observational study, we cannot conclusively address causality. Other limitations include the inclusion of controls from the other 3 UK countries, data not allowing the evaluation of the association between attendance (rather than referral) and conversion.
### Conclusions
The NDPP was associated with reduced conversion rates from NDH to T2DM. Although we observed smaller associations with risk reduction, compared to what has been observed in RCTs, this is unsurprising since we examined the impact of referral, rather than attendance or completion of the intervention.
## Author summary
## Introduction
Type 2 diabetes mellitus (T2DM) is a major public health concern that has been rising globally, with over 3 million people in the United Kingdom currently diagnosed with T2DM. T2DM is an impairment in the way the body controls and regulates blood sugar levels, and it is usually developed due to a genetic predisposition to the condition as well as behavioural and environmental factors. Previous studies have shown that both lifestyle modifications through diet and physical activity and medication can prevent progression to T2DM [1]. Nondiabetic hyperglycaemia (NDH or “prediabetes”) is a condition with increased blood glucose levels but not in the range to be diagnosed as having T2DM. Although the definition of NDH has changed over time [2], diagnosis of NDH has been consistently associated with increased risk of developing T2DM and of developing other diabetes-related conditions [3].
Thus, the NDH population has been seen as an important group to target, in T2DM prevention.
Lifestyle interventions have also been shown to be as effective as pharmaceutical interventions, if not more, in several large randomised controlled trials (RCTs) in various settings and countries, by facilitating weight loss through diet adjustments and increasing exercise activity. The Da Qing IGT and Diabetes Study was the first large RCT to evaluate a lifestyle intervention, with the cumulative incidence of T2DM at 6 years being $68\%$ in the control group compared to $40\%$ in a diet-plus-exercise group, or a $42\%$ risk reduction [4]. The Finnish Diabetes Prevention Study was one of the first large-scale RCTs to evaluate an intensive lifestyle intervention, finding that cumulative incidence of T2DM after 4 years was $11\%$ ($95\%$ CI: 6, 15) in the intervention group and $23\%$ ($95\%$ CI: 17, 29) in the control group, with an overall reduction risk during the trial of $58\%$ in the intervention group [5]. The US Diabetes Prevention Program also achieved a $58\%$ relative risk reduction through lifestyle measures (individuals coached one-to-one and followed up intensely by telephone), compared with standard advice [6]. The Indian Diabetes Prevention Programme further confirmed the effectiveness lifestyle interventions to prevent progression to T2DM, although the effect size was smaller, with an observed relative risk reduction of $28.5\%$ for the lifestyle modification group, compared to standard care controls [7]. The evidence on effectiveness has been summarised in 2 large meta-analyses, one in 2007 that showed a risk reduction of $51\%$ [8] and a more recent one [2019] with a pooled overall risk reduction of $53\%$ [9]. This overwhelming body of evidence has driven large-scale interventions through national or regional Diabetes Prevention Programmes, like the Finnish national DPP “FIN-D2D” [10], and the Victoria-Australia “Life!” programme [11].
From a pharmacological point of view, metformin has been a widely studied medication in T2DM prevention, and it has been found to reduce risk by $31\%$, and being particularly effective for those who were more obese, had higher HbA1c levels, or were younger [6]. Although metformin is the main pharmaceutical intervention, it has also been shown that glitazones, acarbose, or orlistat can also prevent T2DM [12–14].
In the UK, the National Health Services (NHS) Diabetes Prevention Programme (NDPP) is a behavioural intervention programme led by a partnership of NHS England, Public Health England, and Diabetes UK. The programme was primarily offered through primary care practices ($99\%$) to NDH diagnosed adults aged 18 years and over. Identified individuals were either offered referral while in consultation or sent letters through which they could self-refer. The intervention was carried out by 5 commercial providers, and a phased rollout of the programme was introduced across England in 3 waves, from 2016 to 2019 and full population coverage was obtained by mid-2018 [15]. Informed consent was needed for practitioners to make a referral prior to passing the details to a DPP provider. Participants in the programme had an initial assessment followed by regular group education on nutrition and exercise for at least 16 hours over a period of 9 to 12 months. Data from December 2018 showed that 324,699 individuals had been referred of which $53\%$ attended the initial assessment. Approximately 32,665 individuals had at least one intervention session of which $53\%$ completed [16]. Although the NDPP is based on a strong international evidence base [17], justifying the commissioning of such a large and complex programme requires rigorous evidence that the programme is achieving benefits beyond those delivered by current prevention services.
The rollout of the programme makes formal randomised evaluation problematic. Most cases of T2DM are managed through primary care, and primary care administrative data are increasingly used to study diseases and their management [18]. We use primary care administrative data and a range of complex statistical techniques to provide an estimate of the impact of the DPP in reducing conversion of NDH to T2DM (incidence) and reducing the overall numbers of cases of diabetes.
## Data source
UK healthcare is free, publicly funded via the NHS, and over $98\%$ of the population is registered to general practices. We used electronic health records from patients in general practice records from the Clinical Practice Research Datalink (CPRD), collecting detailed and anonymised electronic health records from UK general practices using the VISION and EMIS clinical computer systems, collated in 2 separate databases, CPRD GOLD and CPRD AURUM, respectively. We combined data from these databases and analysed them as a single dataset to maximise sample size in the practice-matching designs.
CPRD GOLD captures approximately $7\%$ of the total UK population, whereas AURUM currently covers $13\%$ of the UK population. Data are mostly recorded by GP staff using version 2 Read codes, a semi-hierarchical clinical classification system containing over 100,000 clinical terms that record a patient’s details. Additional patient-level information from secondary care, disease registries and death registration records [19,20], can be linked for approximately $60\%$ of GOLD practices and all AURUM practices. The study period was from 1 April 2016 (the start of the DPP’s phased rollout) to 31 March 2020. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist).
## Cohort, exposure to the DPP, and outcome
The study had a proposed analysis plan, which is provided in S1 Analysis Plan.
Participants with NDH were identified using Read codes, which were actively used by GPs in the study: 44v2.00 (Glucose Tolerance Test impaired), C11y200 (Impaired glucose tolerance), C11y300 (Impaired fasting glycaemia), C11y500 (Prediabetes), C317.00 (Nondiabetic Hyperglycaemia), R102.00 ([D] Glucose Tolerance Test abnormal), R102.11 ([D] Prediabetes), R102.12 ([D] Impaired glucose tolerance test), R10D000 ([D] Impaired fasting glycaemia), R10D011 ([D] Impaired fasting glucose), R10E.00 ([D] Impaired glucose tolerance). Our previous work explored codes used to identify individuals, in the context of the changing definition of NDH as well as the conversion of NDH to T2DM in the population prior to the NDPP rollout [2].
The cohort comprised of people diagnosed with NDH during the study period, with the prevalence reported in Table A in S1 Appendix. We extracted data from a total of 2,209 practices (GOLD: 723; AURUM: 1,486). At least 1 referral was recorded in 1,359 practices (GOLD: 64; AURUM: 1,295). Participants were considered as referred to the programme if they were associated with one of the following codes: 679m000/EMISNQDI236-NHS DPP not completed, 679m100 NHS DPP completed, 679m200 NHS DPP started, 679m400/EMISNQRE591 Referral to NHS DPP, EMISNQAT50-NHS DPP attended, EMISNQDI236-NHS DPP not attended.
The outcome of interest was conversion of NDH to T2DM, during the study period. Individuals diagnosed with T2DM following the NDH diagnosis were considered to have progressed to T2DM during the study period. Patients with a previous record of type 1 diabetes were excluded. Code lists were uploaded to the clinical code lists website https://clinicalcodes.rss.mhs.man.ac.uk/medcodes/article/194/ [21] and are also provided in S1 Codelist.
## Matching
We explored different matching approaches to compare conversion rates from NDH to T2DM, between people who were referred to the NDPP and those who were not referred to the NDPP. Although participation to the programme was very high, at the practice level, it was introduced in waves and that allowed us to use a between-practice matching approach to reduce the risk of unmeasured confounding in referrals within a practice. We matched referring practices to nonreferring practices over the study period, before matching referred people (from the referring practice) to nonreferred people (from the matched nonreferring practice). We defined practices with none or one referral as nonreferring, and those with 20 or more referrals as referring, excluding other practices. Our main analysis used a 1-to-1 propensity score matching approach for practices, nearest neighbour with no replacement. We included practices from the UK, not just England, since our nonreferring practice pool drew heavily from non-English practices, where a similar intervention was not implemented during the study period. However, we conducted sensitivity analyses, with English-only practices (all participating in the programme either when contributing data to our study or later). The variables included in the model were the NDH registers (NDH registers are the number of patients who were identified as having NDH) of each practice, for 2016, 2017, 2018, and 2019, to ensure practices of similar counts in terms of the population of interest were matched (Fig A in S1 Appendix).
The next step involved matching referred people in a referring practice to nonreferred people in the matched nonreferring practice. We did this with replacement, to increase the sample size. This step is widely used in analyses of these databases, and the aim is to refine the selection process and reduce the population pool to relevant controls. To achieve this, exact matching is commonly used on certain key and complete covariates, usually age, sex, and general practice [22–26]. This approach was used to match 1 case up to 5 controls, to increase power, using age (within 3 years), sex, and date of NDH diagnosis (within 365 days). Age and sex are used as standard in such matching approaches, since they are complete variables that are likely to be linked to effect heterogeneity. We also included date of NDH diagnosis, since in previous work we identified a decreasing trend in conversion rates over time [2], and we considered it important to ensure that start dates and lengths of subsequent exposure to NDH were balanced between groups. Following this, we controlled for all other relevant recorded covariates in multivariable analyses (for example, biological parameters), with the cohort size allowing for numerous covariates to be included.
We also carried out various sensitivity analyses on the matching approach, with combinations of different strategies for between-practice, and within-practice matching. All these analyses were replicated in a multiple imputation framework where additional covariates with high levels of missingness were included, like HbA1c levels. Full details of all these alternative approaches are provided in Tables B-E and Figs B and C in S1 Appendix.
## Covariates
Index date was defined as the referral date for referred to NDPP and the matched controls. We extracted information on the covariates relevant to NDH and T2DM: age (at index date), sex, time from NDH diagnosis to index date, BMI, HbA1c, total serum cholesterol, systolic blood pressure, diastolic blood pressure, prescription of metformin, smoking status, socioeconomic status (location based, for practice and patients), and a diagnosis of depression. We also quantified the multimorbidity burden, using the Charlson Comorbidity Index (CCI), excluding diabetes with complications [27,28]. Biological parameters were categorised and included a “missing” category to allow a complete case analysis with no records dropped. Where levels of missingness were considered to be high for a biological parameter (above $50\%$), it was not included in the main analyses (HbA1c). Socioeconomic information was not included in the complete case analysis due it not being available for patients registered in non-English practices. In the multiple imputation analyses, all covariates were included irrespective of levels of missingness. More information on extraction of the covariates relative to the index date and categorisation, where relevant, is provided in S1 Appendix.
## Statistical analysis
We describe the characteristics of the matched cohort. A parametric survival model with a Weibull survival distribution and shared frailty for practice (random effects) was employed to examine associations between the covariates previously described and conversion to T2DM. Standard errors were obtained through 1,000 bootstrap replications of matched clusters. In addition, custom bootstraps of 1,000 replications were used to obtain average survival curves (and their variability) for referred and nonreferred patients, focusing on 12, 24, 30, and 36 months from referral. These estimates were used to quantify the associations of the intervention into numbers of prevented conversions to T2DM at a particular time point, per 1,000 referrals to the programme. We also used a parametric survival command (stpm2) to obtain average survival curves for the analysed individuals within each group, rather than survival curves for the average person [29]. Stata v16 was used for all analyses.
## Sensitivity analyses
Numerous sensitivity analyses were employed, and combinations of them, on top of the between-practice matching sensitivity analyses:
## Patient and public involvement
There was no direct patient involvement in the context of this evaluation, since stakeholders’ perceptions and experiences of the programme have been evaluated elsewhere [30].
## Ethical approval
This study is based on data from the CPRD obtained under license from the Medicines and Healthcare Products Regulatory Agency (MHRA). The data are provided by patients and collected by the NHS as part of their care and support. The interpretation and conclusions contained in this study are those of the authors alone, and not necessarily those of the MHRA, the National Institute for Health Research (NIHR), NHS, or the Department of Health. Approval to conduct this study using the CPRD was granted by the Independent Scientific Advisory Committee (ISAC) of the MHRA (protocol 18_101). We thank the contributing patients and practices to the CPRD who have allowed their data to be used for research purposes.
## Results
The final cohort, with 1-to-1 matching of practices and individual patients matched up to 5 controls with replacement, included 69,801 participants from 1,084 practices: 18,470 referred to NDPP and 51,331 not referred to NDPP (diagnosed with NDH but not referred to the programme) (Fig 1). Baseline characteristics are shown in Table 1. The mean age of referred to NDPP and not referred to NDPP were fairly similar with referred to NDPP being 61.9 (SD = 11.6) years and not referred to NDPP being 62.6 (SD = 11.0). Although the mean BMI of the referred to NDPP and not referred to NDPP were similar, at 30.8 (SD = 6.4) and 31.2 (SD = 6.7), respectively, those who were referred were more likely to be obese, with $34\%$ of referred to NDPP and $29\%$ of not referred to NDPP with a measurement of BMI equal to or above 30 kg/m2. Results showed that $27\%$ of the cohort had depression. Around $27\%$ of the cohort had at least one comorbidity according to the CCI. The prevalence of comorbidities according to the CCI score was higher in those referred to NDPP compared to not referred to NDPP, with $10\%$ of those referred to NDPP having more than 4 comorbidities, compared to $7\%$ of not referred to NDPP. Current smokers were more likely to be referred to the programme, with those referred to NDPP having $49\%$ and $32\%$ not referred to NDPP being current smokers. Metformin was prescribed in $3.4\%$ of those referred to NDPP and $2.8\%$ of those not referred to NDPP.
**Fig 1:** *Flowchart on final matching numbers CPRD AURUM and GOLD.[Matching: Patients referred to the scheme versus matched patients not referred within the same practice, based on age (up to 3 years), sex, and within 365 days of NDH diagnosis: Cases- Referred to NDPP; Controls- Not Referred to NDPP].* TABLE_PLACEHOLDER:Table 1 A total of 4,432 ($6.4\%$) participants developed T2DM during the study period, of which 1,152 ($6.2\%$) were referred to NDPP and 3,280 ($6.4\%$) were not referred to NDPP. The mean days from NDH diagnosis to index date was 405 days (SD = 742), and it was higher in those referred to NDPP with a mean of 498 days (SD = 835), compared to 371 days (SD = 702) in those not referred to NDPP. The rate of conversion to T2DM was lower in those referred to NDPP compared to those not referred to NDPP with a HR of 0.80 ($95\%$ CI: 0.73, 0.87) (Table 2). Nonconversion (“survival”) at 24 months since referral was estimated at $89.9\%$ ($95\%$ CI: $89.5\%$ to $90.4\%$) for not referred to NDPP and $91.8\%$ ($95\%$ CI: $91.2\%$ to $92.3\%$) for referred to NDPP. At 36 months, it was $84.6\%$ for not referred to NDPP ($95\%$ CI: $83.9\%$ to $85.4\%$) and $87.3\%$ for referred to NDPP ($95\%$ CI: $86.5\%$ to 88.2), respectively. The difference in conversion rates at 36 months was −$2.7\%$ ($95\%$ CI: −$3.7\%$ to −$1.7\%$). Average survival plots across the 2 groups are presented in Fig 2.
**Fig 2:** *Conversion of patients referred to the programme and those who received normal care in primary care in the study period.(Since the flexible parametric model command we used to obtain average survival curves (stpm2) does not allow for random-effects components, which we found to be important in the analyses, we obtained the average of the survival curves within each group for the 1-to-1 patient matched sensitivity analysis. Referrals and nonreferrals were, respectively, 18,470 and 18,470 for t > = 0; 10,466 and 10,444 for t > = 12; 4,148 and 4,122 for t > = 24; 919 and 896 for t > = 36; 0 and 0 for t > = 48. Model adjusted for age (at index date), sex, time from NDH diagnosis to index date, BMI, total serum cholesterol, systolic blood pressure, diastolic blood pressure, prescription of metformin, smoking status, depression, and multimorbidity burden (using the Charlson Comorbidity Index)).* TABLE_PLACEHOLDER:Table 2 NDH, nondiabetic hyperglycaemia; NDPP, NHS Diabetes Prevention Programme; T2DM, type 2 diabetes mellitus.
Females were less likely to convert to T2DM with a HR of 0.80 ($95\%$ CI: 0.75, 0.86) ($p \leq 0.001$) compared to men. People aged 45 to 54 had a higher risk of conversion to T2DM, compared to those aged 18 to 34, with a HR of 2.13 ($95\%$ CI: 1.44, 3.15) ($p \leq 0.001$). Cholesterol categories did not appear to be strongly associated with conversion to T2DM. People with high BMI had a higher risk of conversion to T2DM, with those classed overweight (BMI 25 to 30) having an HR of 1.54 ($95\%$ CI: 1.30, 1.82) ($p \leq 0.001$), and those classed obese (BMI > = 30) having an HR of 2.34 ($95\%$ CI: 2.00, 2.74) ($p \leq 0.001$), compared to individuals with a normal BMI (18.5 to 25). Having depression at baseline may have increased the risk of conversion (HR = 1.19, $95\%$ CI 1.11, 1.27) ($p \leq 0.001$). Those who had a prescription for metformin were at a much higher risk of developing T2DM, with an HR of 9.89 ($95\%$ CI: 9.03, 10.82) ($p \leq 0.001$). Those with a missing smoking status and those who had never smoked were also at a slightly higher risk of developing T2DM, compared to current smokers. Blood pressure was not significantly associated with risk of conversion to T2DM.
## Sensitivity analysis results
Results from the sensitivity analyses are presented in Table D in S1 Appendix, and there was broad agreement with the main analysis, although in most cases the association between referral and nonconversion to T2DM was smaller. The bespoke linkage analyses, where 130 of the 1,084 ($12.0\%$) practices were dropped due to potential data inconsistency, were in agreement with our main analyses. Estimates of associations with risk reduction were lower in within-practice matching analyses, and in between-practice matching analyses that controlled for region.
## Discussion
Our findings suggest that individuals who were referred to the programme were less likely to develop T2DM during the study period compared to those who were not referred to the programme. Assuming 1,000 referred to NDPP and 1,000 not referred to NDPP, by 36 months since referral, we would expect 127 ($95\%$ CI: 118 to 135) conversions to T2DM in those referred to NDPP and 154 ($95\%$CI: 146 to 161) in those not referred to NDPP. As this is an observational study, we cannot be conclusive on causality.
## Findings in context
Diabetes prevention trials have shown reductions in the incidence of T2DM. Analysis of the DPP provider dataset has shown that individuals who attended the NHS DPP were associated with a significant reduction in weight (“mean weight loss of 2.3 kg [$95\%$ CI 2.2, 2.3]”) and HbA1c (“an HbA1c reduction of 1.26 mmol/mol (1.20, 1.31) ($0.12\%$ [0.11, 0.12]”) in an intention-to-treat analysis, which could be a marker of reduction in risk of T2DM [16]. Most of the trial results showed that weight loss was the key factor in reducing risk of T2DM, and our results also suggest that increased BMI to be one of the key factors in developing diabetes.
In our study, metformin prescription was associated with higher risk of developing T2DM, reflecting the fact that metformin is prescribed to people deemed at very high risk. However, our aim was to control our analyses for metformin prescription, not examine its association with developing T2DM. Our data also suggest that metformin prescription was higher in women, which might be related to polycystic syndrome, which carries a higher risk of developing T2DM.
The NHS DPP was primarily based on a systematic review and meta-analysis by Public Health England, which assessed the effectiveness of lifestyle interventions for prevention of T2DM in routine practice over a period of 12 to 18 months. The pooled results from the review showed that those attending a DPP reduced the risk of progression to T2DM by $26\%$ compared to those receiving usual care [17]. These findings could be comparable to our results, which suggested those who were referred to the program had a $20\%$ lower risk of conversion to T2DM compared to those who were not referred, during the study period. However, we were able to evaluate the intervention over a longer period of time, when only a few of the clinical trials included in the systematic review used for the NHS DPP design reported outcomes beyond 12 months. Thus, we were in a better position to explore longer-term benefits, despite the observational nature of the design. A recent 30-year follow-up data from the Da Qing Diabetes Prevention Outcome Study showed that lifestyle intervention in individuals who were at risk of T2DM not only delayed the onset of T2DM but also reduced the incidence of other comorbidities related to T2DM, death, and increased life expectancy [31].
Our risk reduction estimates for conversion to T2DM are considerably lower to what has been reported in the largest RCTs, which reported risk reduction rates between $42\%$ and $58\%$ [4–6]. However, these RCTs recruited people at higher risk, and who can benefit more from an intervention, while the interventions were intensive, with rigorous and persistent follow-up regimes. In addition, the RCTs evaluated participation and completion, while we could only evaluate referral to the programme, irrespective of whether the patient completed or even attended the diabetes prevention programme. One should also bear in mind that there exists variation that we could not account for or model, due to practice anonymity in the CPRD, driven by differences in how the programme was delivered by the 4 contracted different providers to deliver the NDPP, and differences in the characteristics of the people who attended the 4 programmes [15,32,33].
## Strengths and limitations
Our study has several strengths. The data were based on a large, longitudinal sample that is generalizable to the UK population [20,34]. Using CPRD data, we were able to access a complete medical history of the patient including other comorbidities and biological measures. However, several limitations also exist. The matching process involved a lot of complicated decisions. We had to deal with different types of confounding (for example, potential selection bias within a referring practice) and a drop in NDH conversion rates over time (so need to match on NDH diagnosis date). We a priori considered the between-practice matching approach, where we compare referred patients in English referring practices to nonreferred patients in nonreferring English, Scottish, Northern Irish, and Welsh practices, as the least likely to be affected by selection bias. The disadvantage is that the sample size after matching was smaller, and we also were limited in the number of English practices that could serve as controls in the early years (because of the national coverage of the NHS DPP). On the other hand, if $50\%$ of referred to NDPP are not in the GP records, this is the biggest risk with the within-practice approach. However, in analyses where we controlled for region and database type, the observed association was smaller. Another limitation of the dataset is that a large number of individuals referred to NDPP being dropped during the matching process. The restrictions in the matching process were there to protect against unmeasured confounding, as much as possible under these designs. Thus, we accepted dropping individuals referred to NDPP and not referred to NDPP, to ensure that analyses will be more robust. However, further sensitivity analyses were carried out that included referrals without an NDH code, or an NDH code following referral—cases (referred to NDPP) that were excluded from the main analysis. We allowed declined referrals to enter the control (not referred to NDPP) pool (81 unique controls selected 179 times in our main analyses), but excluding them did not affect our results. Our estimates could also be affected by underrecoding of referrals in GP practices and the sensitivity using the bespoke CPRD linkage aimed to examine that. Our results are based on an intention to treat analysis; hence, we were not able to confirm whether all individuals who were referred to the programme actually attended and completed the programme (quality of recording is very poor for these 2 categories). However, as we had a code for those who declined referral, we excluded these individuals from our analysis. Another possible limitation is not being able to include HbA1c in our model due to high levels of data missingness—although we included it in the multiple imputation analyses. HbA1c completeness levels also varied between those referred to the programme and those who did not, with more missing data for those referred, which, on the surface, appears counterintuitive, but that was primarily driven by variation in recording across countries, with better recording in Scotland and Wales. There is also no standard definition for NDH in primary care, the thresholds used to define NDH varied across the years, and this could also have an effect on the cases and controls used in the study [35]. Finally, referred and nonreferred people may have different health-seeking behaviours, with referred people diagnosed with T2DM earlier due to higher levels of interaction with primary care. In that case, we would be underestimating the impact of the intervention.
## Implications for policy and future research
Our findings indicate that the NDPP appears to be successful in reducing progression to T2DM, even when we were only able to examine referral to the programme, rather than attendance or completion, in an observational setting. This supports the decision of the rapid large-scale implementation of the programme in England, rather than a slower or regional introduction, but also supports the continuation of the programme. Our findings also support the introduction of similar programmes to the rest of the UK (i.e., Northern Ireland, Scotland, and Wales), but we would expect our findings to be generalisable to similarly organised healthcare systems.
Future work should quantify the potential benefits of attending or completing the programme, examine longer-term outcomes, and investigate whether the programme delays or prevents progression to T2DM. In addition, it would be important to examine whether the programme attenuates or increases health inequalities, and also whether certain population groups appear to be benefitting more from the programme (i.e., whether there exists effect/association heterogeneity).
## Conclusions
The NDPP was associated with risk reduction in conversion to T2DM, at least in the short to medium term. Individuals who were referred to the NHS DPP by primary care physicians were less likely to develop T2DM compared to those who received usual care. Although we observed lower levels of associations compared to what has been observed in randomised clinical trials, we examined the impact of referral, rather than attendance or completion of the intervention. It will be important to explore whether the association of the programme varies across population subgroups, as well as the long-term effects of the programme. Finally, further work is needed to examine the long-term effects of the programme and to examine if the programme delays or prevents T2DM.
## References
1. Haw JS, Galaviz KI, Straus AN, Kowalski AJ, Magee MJ, Weber MB. **Long-term Sustainability of Diabetes Prevention Approaches: A Systematic Review and Meta-analysis of Randomized Clinical Trials.**. *JAMA Intern Med* (2017) **177** 1808-1817. DOI: 10.1001/jamainternmed.2017.6040
2. Ravindrarajah R, Reeves D, Howarth E, Meacock R, Soiland-Reyes C, Cotterill S. **Epidemiology and determinants of non-diabetic hyperglycaemia and its conversion to type 2 diabetes mellitus, 2000–2015: cohort population study using UK electronic health records.**. *BMJ Open.* (2020) **10** e040201. DOI: 10.1136/bmjopen-2020-040201
3. Punthakee Z, Goldenberg R, Katz P. **Definition, Classification and Diagnosis of Diabetes, Prediabetes and Metabolic Syndrome**. *Can J Diabetes* (2018) **42** S10-S15. DOI: 10.1016/j.jcjd.2017.10.003
4. Pan XR, Li GW, Hu YH, Wang JX, Yang WY, An ZX. **Effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance. The Da Qing IGT and Diabetes Study**. *Diabetes Care* (1997) **20** 537-544. DOI: 10.2337/diacare.20.4.537
5. Tuomilehto J, Lindström J, Eriksson JG, Valle TT, Hämäläinen H, Ilanne-Parikka P. **Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance**. *N Engl J Med* (2001) **344** 1343-1350. DOI: 10.1056/NEJM200105033441801
6. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *N Engl J Med* (2002) **346** 393-403. DOI: 10.1056/NEJMoa012512
7. Ramachandran A, Snehalatha C, Mary S, Mukesh B, Bhaskar AD, Vijay V. **The Indian Diabetes Prevention Programme shows that lifestyle modification and metformin prevent type 2 diabetes in Asian Indian subjects with impaired glucose tolerance (IDPP-1).**. *Diabetologia* (2006) **49** 289-297. DOI: 10.1007/s00125-005-0097-z
8. Gillies CL, Abrams KR, Lambert PC, Cooper NJ, Sutton AJ, Hsu RT. **Pharmacological and lifestyle interventions to prevent or delay type 2 diabetes in people with impaired glucose tolerance: systematic review and meta-analysis**. *BMJ* (2007) **334** 299. DOI: 10.1136/bmj.39063.689375.55
9. Uusitupa M, Khan TA, Viguiliouk E, Kahleova H, Rivellese AA, Hermansen K. **Prevention of Type 2 Diabetes by Lifestyle Changes: A Systematic Review and Meta-Analysis.**. *Nutrients* (2019) **11**. DOI: 10.3390/nu11112611
10. Saaristo T, Moilanen L, Korpi-Hyovalti E, Vanhala M, Saltevo J, Niskanen L. **Lifestyle intervention for prevention of type 2 diabetes in primary health care: one-year follow-up of the Finnish National Diabetes Prevention Program (FIN-D2D).**. *Diabetes Care* (2010) **33** 2146-2151. DOI: 10.2337/dc10-0410
11. Dunbar JA, Jayawardena A, Johnson G, Roger K, Timoshanko A, Versace VL. **Scaling Up Diabetes Prevention in Victoria, Australia: Policy Development, Implementation, and Evaluation**. *Diabetes Care* (2014) **37** 934-942. DOI: 10.2337/dc12-2647
12. Investigators DT, Gerstein HC, Yusuf S, Bosch J, Pogue J, Sheridan P. **Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: a randomised controlled trial**. *Lancet* (2006) **368** 1096-1105. DOI: 10.1016/S0140-6736(06)69420-8
13. Chiasson JL, Josse RG, Gomis R, Hanefeld M, Karasik A, Laakso M. **Acarbose for prevention of type 2 diabetes mellitus: the STOP-NIDDM randomised trial**. *Lancet* (2002) **359** 2072-2077. DOI: 10.1016/S0140-6736(02)08905-5
14. Torgerson JS, Hauptman J, Boldrin MN, Sjostrom L. **XENical in the prevention of diabetes in obese subjects (XENDOS) study: a randomized study of orlistat as an adjunct to lifestyle changes for the prevention of type 2 diabetes in obese patients.**. *Diabetes Care* (2004) **27** 155-161. DOI: 10.2337/diacare.27.1.155
15. Stokes J, Gellatly J, Bower P, Meacock R, Cotterill S, Sutton M. **Implementing a national diabetes prevention programme in England: lessons learned**. *BMC Health Serv Res* (2019) **19** 991. DOI: 10.1186/s12913-019-4809-3
16. Valabhji J, Barron E, Bradley D, Bakhai C, Fagg J, O’Neill S. **Early Outcomes From the English National Health Service Diabetes Prevention Programme**. *Diabetes Care* (2020) **43** 152-160. DOI: 10.2337/dc19-1425
17. Ashra NB, Spong R, Carter P, Davies MJ, Dunkley A, Gillies C. **A systematic review and metaanalysis assessing the effectiveness of pragmatic lifestyle interventions for the prevention of type 2 diabetes mellitus in routine practice**. *In: England PH* (2015)
18. Sharma M, Nazareth I, Petersen I. **Trends in incidence, prevalence and prescribing in type 2 diabetes mellitus between 2000 and 2013 in primary care: a retrospective cohort study.**. *BMJ Open.* (2016) **6** e010210. DOI: 10.1136/bmjopen-2015-010210
19. Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, van Staa T. **Data Resource Profile: Clinical Practice Research Datalink (CPRD).**. *Int J Epidemiol.* (2015) **44** 827-836. DOI: 10.1093/ije/dyv098
20. Wolf A, Dedman D, Campbell J, Booth H, Lunn D, Chapman J. **Data resource profile: Clinical Practice Research Datalink (CPRD) Aurum.**. *Int J Epidemiol.* (2019) **48** 1740-g. DOI: 10.1093/ije/dyz034
21. Springate DA, Kontopantelis E, Ashcroft DM, Olier I, Parisi R, Chamapiwa E. **ClinicalCodes: an online clinical codes repository to improve the validity and reproducibility of research using electronic medical records.**. *PLoS ONE.* (2014) **9** e99825. DOI: 10.1371/journal.pone.0099825
22. Price SJ, Stapley SA, Shephard E, Barraclough K, Hamilton WT. **Is omission of free text records a possible source of data loss and bias in Clinical Practice Research Datalink studies? A case–control study**. *BMJ Open* (2016) **6** e011664. DOI: 10.1136/bmjopen-2016-011664
23. Nicholson BD, Aveyard P, Hobbs FDR, Smith M, Fuller A, Perera R. **Weight loss as a predictor of cancer and serious disease in primary care: an ISAC-approved CPRD protocol for a retrospective cohort study using routinely collected primary care data from the UK**. *Diagn Progn Res* (2018) **2** 1. DOI: 10.1186/s41512-017-0019-9
24. Søgaard KK, Sørensen HT, Smeeth L, Bhaskaran K. **Acute Pericarditis and Cancer Risk: A Matched Cohort Study Using Linked UK Primary and Secondary Care Data.**. *J Am Heart Assoc* (2018) **7** e009428-e. DOI: 10.1161/JAHA.118.009428
25. Vinogradova Y, Coupland C, Hippisley-Cox J. **Exposure to bisphosphonates and risk of gastrointestinal cancers: series of nested case-control studies with QResearch and CPRD data**. *BMJ* (2013) **346** f114. DOI: 10.1136/bmj.f114
26. Doyle M, While D, Mok PLH, Windfuhr K, Ashcroft DM, Kontopantelis E. **Suicide risk in primary care patients diagnosed with a personality disorder: a nested case control study.**. *BMC Fam Pract* (2016) **17** 106. DOI: 10.1186/s12875-016-0479-y
27. Khan NF, Perera R, Harper S, Rose PW. **Adaptation and validation of the Charlson Index for Read/OXMIS coded databases.**. *BMC Fam Pract* (2010) **11** 1. DOI: 10.1186/1471-2296-11-1
28. Charlson ME, Pompei P, Ales KL, MacKenzie CR. **A new method of classifying prognostic comorbidity in longitudinal studies: development and validation**. *J Chronic Dis* (1987) **40** 373-383. DOI: 10.1016/0021-9681(87)90171-8
29. Lambert P.. *STPM2: Stata module to estimate flexible parametric survival models* (2020)
30. Rodrigues AM, Haste A, Penn L, Bell R, Summerbell C, White M. **Stakeholders’ perceptions and experiences of the National Health Service diabetes prevention programme in England: qualitative study with service users, intervention providers and deliverers, commissioners and referrers.**. *BMC Health Serv Res.* (2020) **20** 1-13. DOI: 10.1186/s12913-020-05160-2
31. Gong Q, Zhang P, Wang J, Ma J, An Y, Chen Y. **Morbidity and mortality after lifestyle intervention for people with impaired glucose tolerance: 30-year results of the Da Qing Diabetes Prevention Outcome Study**. *Lancet Diabetes Endocrinol* (2019) **7** 452-461. DOI: 10.1016/S2213-8587(19)30093-2
32. Hawkes RE, Cameron E, Cotterill S, Bower P, French DP. **The NHS Diabetes Prevention Programme: an observational study of service delivery and patient experience**. *BMC Health Serv Res* (2020) **20** 1098. DOI: 10.1186/s12913-020-05951-7
33. Howarth E, Bower PJ, Kontopantelis E, Soiland-Reyes C, Meacock R, Whittaker W. **‘Going the distance’: an independent cohort study of engagement and dropout among the first 100 000 referrals into a large-scale diabetes prevention program.**. *BMJ Open Diabetes Res Care* (2020) **8** e001835. DOI: 10.1136/bmjdrc-2020-001835
34. Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, van Staa T. **Data Resource Profile: Clinical Practice Research Datalink (CPRD).**. *Int J Epidemiol.* (2015) **44** 827-836. DOI: 10.1093/ije/dyv098
35. van Herpt TTW, Ligthart S, Leening MJG, van Hoek M, Lieverse AG, Ikram MA. **Lifetime risk to progress from pre-diabetes to type 2 diabetes among women and men: comparison between American Diabetes Association and World Health Organization diagnostic criteria.**. *BMJ Open Diabetes Res Care.* (2020) **8** e001529. DOI: 10.1136/bmjdrc-2020-001529
|
---
title: 'Assessing the impact of community-based interventions on hypertension and
diabetes management in three Minnesota communities: Findings from the prospective
evaluation of US HealthRise programs'
authors:
- Nancy Fullman
- Krycia Cowling
- Luisa S. Flor
- Shelley Wilson
- Paurvi Bhatt
- Miranda F. Bryant
- Joseph N. Camarda
- Danny V. Colombara
- Jessica Daly
- Rose K. Gabert
- Katie Panhorst Harris
- Casey K. Johanns
- Charlie Mandile
- Susan Marshall
- Claire R. McNellan
- Vasudha Mulakaluri
- Bryan K. Phillips
- Marissa B. Reitsma
- Naomi Sadighi
- Tsega Tamene
- Blake Thomson
- Alexandra Wollum
- Emmanuela Gakidou
journal: PLOS ONE
year: 2023
pmcid: PMC9970068
doi: 10.1371/journal.pone.0279230
license: CC0 1.0
---
# Assessing the impact of community-based interventions on hypertension and diabetes management in three Minnesota communities: Findings from the prospective evaluation of US HealthRise programs
## Abstract
### Background
Community-based health interventions are increasingly viewed as models of care that can bridge healthcare gaps experienced by underserved communities in the United States (US). With this study, we sought to assess the impact of such interventions, as implemented through the US HealthRise program, on hypertension and diabetes among underserved communities in Hennepin, Ramsey, and Rice Counties, Minnesota.
### Methods and findings
HealthRise patient data from June 2016 to October 2018 were assessed relative to comparison patients in a difference-in-difference analysis, quantifying program impact on reducing systolic blood pressure (SBP) and hemoglobin A1c, as well as meeting clinical targets (< 140 mmHg for hypertension, < $8\%$ Al1c for diabetes), beyond routine care. For hypertension, HealthRise participation was associated with SBP reductions in Rice (6.9 mmHg [$95\%$ confidence interval: 0.9–12.9]) and higher clinical target achievement in Hennepin (27.3 percentage-points [9.8–44.9]) and Rice (17.1 percentage-points [0.9 to 33.3]). For diabetes, HealthRise was associated with A1c decreases in Ramsey (1.3 [0.4–2.2]). Qualitative data showed the value of home visits alongside clinic-based services; however, challenges remained, including community health worker retention and program sustainability.
### Conclusions
HealthRise participation had positive effects on improving hypertension and diabetes outcomes at some sites. While community-based health programs can help bridge healthcare gaps, they alone cannot fully address structural inequalities experienced by many underserved communities.
## Introduction
Longstanding health disparities occur throughout the US [1], with differences often manifesting across multiple environmental and societal factors (e.g., geography, sex or gender, race or ethnicity, socioeconomic status) [2]. Non-communicable diseases (NCDs) and NCD-related risks like hypertension can uniquely affect underserved communities, as a constellation of structural factors–from service affordability and inadequate insurance to more entrenched socioeconomic obstacles like low access to nutritional food–can easily give way to high rates of chronic, debilitating conditions experienced in settings without good access to care or appropriate services. Community-based interventions with integrated care have emerged as approaches to bridge gaps in NCD care, particularly for underserved areas in the US [3–9]. Nonetheless, the effectiveness of these interventions vary across contexts; interventions implemented; roles of community health workers (CHWs)(e.g., direct involvement in patient care [10] versus assisting health providers who are then responsible for service provision [11]); and NCDs targeted. While CHWs and community-based interventions appear to be promising models of care for hypertension and diabetes in underserved communities [5–7], more rigorous evaluations across more local contexts and populations are needed.
To strengthen this evidence base, the HealthRise program was developed to implement and pilot locally-tailored programs for improving screening, diagnosis, management and control of hypertension and diabetes among underserved communities [12–14]. HealthRise took place in nine communities in four countries–Brazil, India, South Africa, and the US–from 2014 to 2018. In the US, Minnesota was selected, a state which generally surpasses national averages and ranks among the healthiest across many health measures [1, 15]. Nonetheless, county- and sub-county level health disparities remain in Minnesota, particularly among populations that face compounding barriers to care and improved health outcomes [15, 16]. To maximize the potential impact of HealthRise programs, especially within a relatively short time span (i.e., the earliest US program began in 2016), HealthRise targeted geographic areas with the greatest need and highest disease burden. A 2014–2015 needs assessment identified communities within three Minnesota counties–Hennepin, Ramsey, and Rice–as potential candidates for HealthRise programs based on a combination of quantitative indicators and key informant interviews [15, 17]. Such data pointed to high NCD burdens and risk profiles; widespread access challenges for healthcare, education, and nutritious foods; and sociodemographic characteristics often associated with greater healthcare barriers and worse outcomes stemming from structural racism (e.g., large proportions of populations identifying as Hispanic or Latinx, Black or African American; as well as immigrants or refugees). Three main recommendations emerged for onward HealthRise programming: [1] focus on people with the highest need and poorest clinical outcomes [2] use multi-faceted interventions to address multiple risks and comorbid conditions; and [3] identify opportunities to integrate CHWs within formal health system functions [15]. Drawing from the needs assessment and HealthRise grantee applications, three implementing partners were selected for each county–Pillsbury United Communities (PUC) in Hennepin, Regions Hospital Foundation in Ramsey, and HealthFinders Collaborative, Inc (HFC) in Rice–and HealthRise grantees developed their locally tailored community-based interventions and corresponding activities (as summarized Table 1) [14]. US HealthRise program implementation then took place from June 2016 to October 2018.
**Table 1**
| Unnamed: 0 | Unnamed: 1 | Hennepin County | Ramsey County | Rice County |
| --- | --- | --- | --- | --- |
| HealthRise grantee and/or local partners | HealthRise grantee and/or local partners | Pillsbury United Communities (PUC) | Regions Hospital Foundation | HealthFinders Collaborative, Inc (HFC) |
| HealthRise grantee and/or local partners | HealthRise grantee and/or local partners | North Rising partnership comprised of PUC and North Memorial Health, a network of hospitals and clinics | Minnesota Community Care (MCC) (formerly named West Side Community Health Services, abbreviated WSCHS) | |
| HealthRise implementation location | HealthRise implementation location | North Minneapolis, Minnesota | Saint Paul, Minnesota | Northfield, Minnesota |
| HealthRise program period | HealthRise program period | July 2016 to September 2018 | June 2016 to September 2018 | September 2016 to October 2018 |
| Key characteristics and/or challenges for communities served by HealthRise | Shared across sites | • High levels of poverty and/or unemployment (e.g., 40% at or below 200% poverty line in Hennepin; 97% of MCC clinic patients in Saint Paul are below 200% poverty line) | • High levels of poverty and/or unemployment (e.g., 40% at or below 200% poverty line in Hennepin; 97% of MCC clinic patients in Saint Paul are below 200% poverty line) | • High levels of poverty and/or unemployment (e.g., 40% at or below 200% poverty line in Hennepin; 97% of MCC clinic patients in Saint Paul are below 200% poverty line) |
| Key characteristics and/or challenges for communities served by HealthRise | Shared across sites | • High proportion of population are non-white and/or immigrants or refugees (e.g., past HFC patients were 60% Latino immigrants and 25% Somali refugees in Rice County; 30–65% of MCC clinic patients do not speak English as their primary language; North Minneapolis population is about 50% African American, 20% Asian, 15% Caucasian, and 15% Hispanic/Other) | • High proportion of population are non-white and/or immigrants or refugees (e.g., past HFC patients were 60% Latino immigrants and 25% Somali refugees in Rice County; 30–65% of MCC clinic patients do not speak English as their primary language; North Minneapolis population is about 50% African American, 20% Asian, 15% Caucasian, and 15% Hispanic/Other) | • High proportion of population are non-white and/or immigrants or refugees (e.g., past HFC patients were 60% Latino immigrants and 25% Somali refugees in Rice County; 30–65% of MCC clinic patients do not speak English as their primary language; North Minneapolis population is about 50% African American, 20% Asian, 15% Caucasian, and 15% Hispanic/Other) |
| Key characteristics and/or challenges for communities served by HealthRise | Shared across sites | • Insufficient support of disease management and behavioral changes to improve health within formal health system structures or communities (e.g., poor access to high-quality education, healthcare, and nutritious foods in Hennepin County; HFC specifically targeted uninsured or individuals with public insurance plans in Rice County) | • Insufficient support of disease management and behavioral changes to improve health within formal health system structures or communities (e.g., poor access to high-quality education, healthcare, and nutritious foods in Hennepin County; HFC specifically targeted uninsured or individuals with public insurance plans in Rice County) | • Insufficient support of disease management and behavioral changes to improve health within formal health system structures or communities (e.g., poor access to high-quality education, healthcare, and nutritious foods in Hennepin County; HFC specifically targeted uninsured or individuals with public insurance plans in Rice County) |
| Key characteristics and/or challenges for communities served by HealthRise | Shared across sites | • Inadequate or poor integration of community health care systems, including data systems (e.g., minimal integration of electronic health records across community health care sites in Hennepin; data systems between MCC clinics and hospitals in Ramsey were not integrated) | • Inadequate or poor integration of community health care systems, including data systems (e.g., minimal integration of electronic health records across community health care sites in Hennepin; data systems between MCC clinics and hospitals in Ramsey were not integrated) | • Inadequate or poor integration of community health care systems, including data systems (e.g., minimal integration of electronic health records across community health care sites in Hennepin; data systems between MCC clinics and hospitals in Ramsey were not integrated) |
| Key characteristics and/or challenges for communities served by HealthRise | Site-specific | • Urban setting (Twin Cities area) | • Urban setting (Twin Cities area) | • Primarily rural |
| Key characteristics and/or challenges for communities served by HealthRise | Site-specific | • Reported low trust in local health systems | • Reported high levels of emergency department repeat users | • Reported high cultural and language barriers among immigrants and undocumented migrant worker population |
| Key HealthRise interventions and activities | Shared across sites | • Community-based programs and training: hired and trained community health workers (CHWs) and community paramedics (CPs) to provide home-based care and linkages to clinic-based provider teams (e.g., doctors, nurses, pharmacists, clinical care coordinators, and diabetes education in Hennepin County); for Rice County, where HFC already had pre-existing CHW/CP care teams and networks, additional training and expanded services occurred (e.g., mental health, on-site lab for easier access to diagnostic tests) | • Community-based programs and training: hired and trained community health workers (CHWs) and community paramedics (CPs) to provide home-based care and linkages to clinic-based provider teams (e.g., doctors, nurses, pharmacists, clinical care coordinators, and diabetes education in Hennepin County); for Rice County, where HFC already had pre-existing CHW/CP care teams and networks, additional training and expanded services occurred (e.g., mental health, on-site lab for easier access to diagnostic tests) | • Community-based programs and training: hired and trained community health workers (CHWs) and community paramedics (CPs) to provide home-based care and linkages to clinic-based provider teams (e.g., doctors, nurses, pharmacists, clinical care coordinators, and diabetes education in Hennepin County); for Rice County, where HFC already had pre-existing CHW/CP care teams and networks, additional training and expanded services occurred (e.g., mental health, on-site lab for easier access to diagnostic tests) |
| Key HealthRise interventions and activities | Shared across sites | • Home-based care: CHWs and CPs visited patients for disease management (i.e., monitor health status, medications), health education (e.g., health education, healthy food cooking demonstrations), and support for social needs or social determinants of health (e.g., insurance, housing, transportation); often tailored frequency of in-home visits to patient care plans and based on trends in clinical targets | • Home-based care: CHWs and CPs visited patients for disease management (i.e., monitor health status, medications), health education (e.g., health education, healthy food cooking demonstrations), and support for social needs or social determinants of health (e.g., insurance, housing, transportation); often tailored frequency of in-home visits to patient care plans and based on trends in clinical targets | • Home-based care: CHWs and CPs visited patients for disease management (i.e., monitor health status, medications), health education (e.g., health education, healthy food cooking demonstrations), and support for social needs or social determinants of health (e.g., insurance, housing, transportation); often tailored frequency of in-home visits to patient care plans and based on trends in clinical targets |
| Key HealthRise interventions and activities | Shared across sites | • Technologies for care coordination: implemented tools to better coordinate care between CHW/CP teams and clinic-based teams (e.g., Pathways from Care Coordination Systems for Ramsey County) or incorporated home visit information into electronic medical record (EMR) systems (e.g., HFC designed EMRs to include in-home information into patient medical records at clinics) | • Technologies for care coordination: implemented tools to better coordinate care between CHW/CP teams and clinic-based teams (e.g., Pathways from Care Coordination Systems for Ramsey County) or incorporated home visit information into electronic medical record (EMR) systems (e.g., HFC designed EMRs to include in-home information into patient medical records at clinics) | • Technologies for care coordination: implemented tools to better coordinate care between CHW/CP teams and clinic-based teams (e.g., Pathways from Care Coordination Systems for Ramsey County) or incorporated home visit information into electronic medical record (EMR) systems (e.g., HFC designed EMRs to include in-home information into patient medical records at clinics) |
| Key HealthRise interventions and activities | Shared across sites | • Community activities and wellness programs: led via CHWs or supported via community centers to support nutrition education and resource connection (e.g., healthy eating demonstrations in Hennepin County and Opportunity grant nutrition-focused program in Ramsey County); disease management tailored to cultural and linguistic needs (e.g., monthly and quarterly diabetes management classes and Somali Health series in Rice County); and exercise/wellness programs (e.g., Pura Vida which included exercise classes, cooking and nutrition classes, etc. in Rice County). | • Community activities and wellness programs: led via CHWs or supported via community centers to support nutrition education and resource connection (e.g., healthy eating demonstrations in Hennepin County and Opportunity grant nutrition-focused program in Ramsey County); disease management tailored to cultural and linguistic needs (e.g., monthly and quarterly diabetes management classes and Somali Health series in Rice County); and exercise/wellness programs (e.g., Pura Vida which included exercise classes, cooking and nutrition classes, etc. in Rice County). | • Community activities and wellness programs: led via CHWs or supported via community centers to support nutrition education and resource connection (e.g., healthy eating demonstrations in Hennepin County and Opportunity grant nutrition-focused program in Ramsey County); disease management tailored to cultural and linguistic needs (e.g., monthly and quarterly diabetes management classes and Somali Health series in Rice County); and exercise/wellness programs (e.g., Pura Vida which included exercise classes, cooking and nutrition classes, etc. in Rice County). |
| Key HealthRise interventions and activities | Site-specific | • Established community-care teams of CHWs and CPs linked to clinic-based care teams: this model of care was relatively new to PUC and Hennepin County partners, so recruitment of CHWs/CPs and training with care teams occurred alongside other HealthRise-supported activities | • Established community-care teams of CHWs and CPs linked to clinic-based care teams: this model of care was relatively new to Regions and Ramsey County partners, so recruitment of CHWs/CPs and training with care teams occurred alongside other HealthRise-supported activities | • Developed new partnerships to expand community-based care: partnered with Northfield Hospital and Clinics to expand CP program; collaborating with the Mayo Clinic and Alaian Health System to extend model beyond NCDs (e.g., ob/gym care for Somali populations) |
| Key HealthRise interventions and activities | Site-specific | • Implemented interdisciplinary approaches to improving health: established a full-service grocery store (North Market) with linkages to an interdisciplinary wellness team (e.g., CHWs, nutritionist, pharmacy liaison, coordinator) and a Wellness Resource Center with North Memorial Health | • Focused on community-based nutrition programs: used Opportunity Grant to develop and implement a nutrition-focused program, in both English and Spanish, wherein sessions focused on nutrition education, effects of non-nutrition factors on blood sugar (e.g., physical activity, stress management), and grocery store tours highlighting ways to shop for healthy and affordable foods | • Employed several electronic tools for improving contact with patients: developed SMS/text-based appointment reminders and education programs (i.e., Care Message) |
With this study, we provide key findings from HealthRise programs implemented by grantees in Hennepin, Ramsey, and Rice Counties, Minnesota. Based on quantitative and qualitative data collected over the course of program implementation, we evaluated the potential impact of these community-based interventions on improving clinical and health outcomes for hypertension and diabetes patients. We conducted difference-in-difference analyses in relation to comparison patients to quantify this impact above and beyond what might be expected for demographically similar patients under routine care in the same communities. This study contributes to the science supporting the role of community-based programs in elevating the health of underserved communities, in the US and elsewhere.
## Study overview, design, and interventions
This analysis follows the global HealthRise prospective evaluation framework, which was established in 2014 and agreed upon by all partners; greater detail on the global team structure, interventions, and analyses are provided elsewhere [12–14]. In sum, the HealthRise program had global and US implementation partners coordinated by Abt Associates and evaluation activities overseen by the Institute for Health Metrics and Evaluation (IHME). Though ongoing coordination and collaboration occurred across implementation and evaluation organizations, these grant streams were purposefully structured and funded separately to support an independent assessment of the HealthRise programs.
For US HealthRise, patient-level monitoring data were routinely collected and collated by grantees during program implementation (June 2016 to October 2018). Using a mixed-methods quasi-experimental design, we synthesized qualitative and quantitative data from HealthRise and comparison patients and stakeholders (e.g., service providers, administrators, and policymakers) to inform its endline evaluation.
Table 1 summarizes key information on each US HealthRise site and interventions implemented by grantee, as interventions were tailored to address key challenges or structural drivers of inequalities identified for each site during the 2014–2015 needs assessment [15, 17]. Per the needs assessment [15, 17], populations across sites experienced high levels of poverty and/or unemployment; the majority of populations identified as Hispanic or Latinx, Black or African American, as well as immigrants or refugees; and substantive challenges around access to sufficient disease management and health promotion support occurred within formal health system structures and broader communities. HealthRise programming was designed to address both cross-cutting and site-specific needs or challenges; additional descriptions for each site’s interventions and activities as part of HealthRise are available elsewhere [14].
## Definitions
We used following case definitions for hypertension and diabetes at each time point: [1] prevalent cases were patients with documented diagnoses, or patients without prior diagnoses but with clinical readings that would qualify for diagnosis (i.e., systolic blood pressure [SBP] ≥ 140 mmHg or diastolic blood pressure [DBP] ≥ 90 mmHg for hypertension; hemoglobin A1c ≥ $6.5\%$ for diabetes); [2] diagnosed cases were patients with documented diagnoses; and [3] patients meeting treatment targets were prevalent cases with SBP < 140 mmHg and DBP < 90 mmHg for hypertension, and A1c < $8\%$ for diabetes. If DBP measures were not available for a given patient, then only SBP readings were used.
## Endline evaluation data collection
HealthRise patient data. Each US grantee collected patient-level data from existing sources and provided de-identified data over time. To best capture potential program impact, analyses were limited to HealthRise patients who [1] remained enrolled in HealthRise at endline (i.e., never withdrew from programs); and [2] had at least two separate biometric data points for blood pressure (i.e., ideally both SBP and DBP, but at minimum, SBP) or A1c (Table 2). Subsequently, evaluation results reflected potential effects from HealthRise participation, and not “intention to treat,” which would have included patients who enrolled but then withdrew from the program at some point. Rates of any program withdrawal varied by site, ranging from $16.7\%$ ($$n = 19$$) for Hennepin to $32.5\%$ ($$n = 25$$) for Ramsey. For the Rice HealthRise site, $3.2\%$ of patients ($$n = 5$$) lacked a clinic visit or biometric data since baseline, and thus were considered withdrawn. In Ramsey, most patients who withdrew did so after a few months of enrollment and within the first year of HealthRise implementation.
**Table 2**
| Data collection | Hennepin County | Hennepin County.1 | Ramsey County | Ramsey County.1 | Rice County | Rice County.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Data collection | HealthRise | Comparison | HealthRise | Comparison | HealthRise | Comparison |
| Quantitative | | | | | | |
| Total patients | 121 | 135 | 78 | 104 | 217 | 311 |
| Patients with baseline data and meeting inclusion criteria | 114 | 113 | 77 | 95 | 157 | 311 |
| Patients enrolled at endline | 95 | 113 | 52 | 95 | 152 | 311 |
| Patients with hypertension | 85 | 83 | 32 | 66 | 84 | 182 |
| Patients with hypertension and ≥ 2 biometric readings | 80 | 83 | 32 | 66 | 80 | 170 |
| Patients with diabetes | 76 | 75 | 48 | 78 | 125 | 303 |
| Patients with diabetes and ≥ 2 biometric readings | 37 | 39 | 43 | 72 | 96 | 296 |
| Qualitative | | | | | | |
| Total interviews and focus groups | 5 | - | 9 | - | 6 | - |
| Community health workers and frontline health workers | 2 | - | 3 | - | 2 | - |
| Facility- or clinic-based providers | 1 | - | 3 | - | 2 | - |
| Facility or clinic managers and administrators | 2 | - | 3 | - | 2 | - |
| Policymakers | 3 | 3 | 3 | 3 | 3 | 3 |
For baseline measures, we used biometric data collected at HealthRise program enrollment. If such data were not available at the precise enrollment date, then biometric data were used from the data closest to that of enrollment. For endline measures, we used patients’ most recent biometric measurements.
S1 Table provides additional descriptive statistics, including counts and percentages by age, reported sex, and self-identified race or ethnicity, for HealthRise patients at baseline and for those who remained enrolled through endline. Due to already small sample sizes for each site’s HealthRise program and concerns about potential differences or inconsistencies in race or ethnicity response options available across data sources, further analysis disaggregated by age, reported sex, and reported race or ethnicity was not conducted.
Comparison patient data. Grantees provided comparison data drawn from patient populations similar to those enrolled in HealthRise. Upon receiving each site’s dataset, we sought to reconstruct samples of comparison patients that were similar demographically and in terms of baseline health conditions to HealthRise patients (i.e., excluding comparison patients younger than 30 years and 90 years or older, and those without a diagnosis of hypertension or diabetes and had baseline biometric data that fell within disease control categories). As necessary, comparison patient data were censored to correspond with each site’s HealthRise program implementation period (Table 1), and thus better approximate similar follow-up times for comparison patients. After this censoring step, we then excluded any comparison patients who lacked more than one measurement of A1c and systolic blood pressure and therefore could not contribute to baseline versus endline comparisons. Included comparison patients, by site, are provided in Table 2.
Comparison patient data selection occurred between October 2018 and January 2019, with criteria determined by each HealthRise site. For Hennepin, data for patients who formed the comparison group were extracted from clinics associated with North Memorial but had not enrolled in HealthRise. Selection criteria included having at least two biometric readings for A1c or SBP–one in 2016 and one in 2018 –to approximate baseline and endline measures for HealthRise; and being between the ages of 30 and 89 years at “baseline.” For Ramsey, comparison patient data were extracted through MCC (formerly named West Side Community Health Services); eligible individuals were patients who had not enrolled in HealthRise and had similar baseline levels of A1c or SBP as HealthRise patients. For Rice, data were extracted from a partner clinic where HealthRise interventions were not implemented. Unlike other comparison patient datasets, International Classification of Disease codes for diabetes and hypertension were not available for patient diagnosis; instead, the diagnosis variable for Rice comparison patients listed active diagnoses. Consequently, a text-matching algorithm was applied to assign diabetes and/or hypertension diagnosis based on the text data in this variable.
Qualitative data. Twenty-three key informant interviews (KIIs) were conducted with local policymakers (non-site specific) and with different types of staff at each site (Table 2). Interviews were not conducted with patients or with staff at clinics from which comparison patient data were selected.
Initial potential interviewees were identified via leadership from HealthRise grantees and partner organizations, and then additional staff (e.g., clinic-based providers, community paramedics [CPs], CHWs) were contacted via snowball sampling. Of the original individuals identified, $79\%$ completed one-hour interviews via telephone with an IHME evaluation team member.
All interviews were audio-recorded and listened to multiple times by a single researcher. Key components were transcribed in an Excel template, with thematic coding applied to identify both site-specific and overarching themes across sites.
## Endline evaluation analyses
To quantify potential effects of HealthRise participation, we used two outcome indicators to measure patient-level changes from baseline to endline: [1] the proportion of patients meeting treatment targets (i.e., SBP < 140 mmHg and DBP < 90 mmHg for hypertension; < $8\%$ A1c for diabetes); and [2] patient biometric measures (i.e., SBP for hypertension, A1c for diabetes). All analyses were limited to patients who were prevalent cases at baseline and had corresponding biometric data for each time point.
We conducted difference-in-difference analyses in two steps for each site and by condition. First, we ran an unadjusted model, only including binary variables for HealthRise status and timing (i.e., baseline or endline) and an interaction term for HealthRise at endline to capture the effect of HealthRise participation over time. We then ran an adjusted model, including the following covariates to account for potential systematic differences in HealthRise and comparison patients: sex (female, male); age (< 50 years, ≥ 50 years); time elapsed from baseline to endline (< 12 months, ≥ 12 months); and comorbidities at baseline (prevalent case of only hypertension or diabetes; prevalent case of both hypertension and diabetes). We specified robust standard errors for each model, and used Welch’s t-tests (i.e., assuming unequal variance between each group) to evaluate statistically significant differences between HealthRise and comparison patients. All analyses were conducted in Stata version 15 and R version 3.6.2 [18, 19].
## Ethical approval
Ethical approval for this study was obtained from the University of Washington’s institutional review board, as well as the local data collection agencies and government entities for each site. All personal identifiers were removed prior to data sharing with IHME; only de-identified data were analyzed.
## Quantitative results
Overall, hypertension and diabetes indicators generally improved for HealthRise patients compared with their baseline measures (Fig 1, Table 3). Yet clinical improvements were heterogeneous since program enrollment (Fig 1), emphasizing potential challenges in effective case management among underserved populations.
**Fig 1:** *HealthRise patient shifts in disease severity categories between baseline and endline based on biometric readings for hypertension (A) and diabetes (B).The height of each column reflects $100\%$ of patients at each time point (baseline and endline), while the categories within each column represents the percentage of patients in each category at baseline and endline. Patient groups are color-coded by their categorization at endline (right column per site) and flow from their categorization at baseline (left column per site). By category percentages, for each site, are available in S1 Data.* TABLE_PLACEHOLDER:Table 3 Across sites, a considerable percentage of hypertension patients with baseline SBP measures exceeding 140 mmHg improved endline levels to below 140 mmHg (Fig 1A); this trend was particularly pronounced for Hennepin and Ramsey. In Hennepin, $76.8\%$ ($95\%$ confidence interval: 65.6 to $84.4\%$) of hypertension patients enrolled in HealthRise recorded endline SBP measures below 140 mmHg and $50.0\%$ (32.3 to $67.3\%$) of HealthRise patients with hypertension in Ramsey met this threshold. Nonetheless, some percentage of hypertension patients shifted into worse SBP categories by endline: $17.5\%$ in Hennepin, $9.4\%$ in Ramsey, and $13.8\%$ in Rice (Fig 1A; S1 Data).
Sizeable improvements occurred for diabetes patients meeting clinical targets since enrollment (Fig 1B), especially for Ramsey. Compared with baseline, where fewer than $5\%$ of patients with diabetes were meeting treatment targets, $25.6\%$ (14.5 to $41.0\%$) of Ramsey HealthRise patients with diabetes had A1c levels lower than $8\%$. Yet many HealthRise patients with diabetes still had A1c levels of $8\%$ or higher by endline across sites: $75.7\%$ in Hennepin, $74.4\%$ in Ramsey, and $49.0\%$ in Rice. A1c category shifts between baseline and endline were especially varied for Hennepin and Rice; for nearly every A1c category at baseline (i.e., < $7\%$, 7–$7.9\%$, 8–$9.9\%$, ≥$10\%$), some portion of patients moved to one of the other three A1c categories by endline.
Unadjusted and adjusted difference-in-difference model results were nearly identical for the effect of HealthRise (Table 4); accordingly, we report on the adjusted model results here. Overall, HealthRise patients trended toward greater progress in reducing biometric measures and meeting treatment targets than comparison patients; however, these differences were not consistently significant across indicators and sites. For hypertension patients, HealthRise participation was associated with statistically significant SBP reductions relative to comparison patients in Rice (6.9 mmHg decrease [0.9 to 13.0; $p \leq 0.05$]). Relative to comparison patients, HealthRise participation was also associated with a statistically significant increase in the percentage of hypertension patients meeting treatment targets in Hennepin (27.3 percentage-point rise [9.7 to 45.0; $p \leq 0.01$]) and Rice (17.1 percentage-point increase [0.9 to 33.4; $p \leq 0.05$]). In Ramsey, changes in hypertension indicators were not statistically different between the HealthRise and comparison groups, though program participation trended toward improvement: a 10.7 mmHg decrease (-0.3 to 21.8; $$p \leq 0.057$$) in SBP and 22.3 percentage-point increase (-0.01 to 44.4 = 5; $$p \leq 0.054$$) in meeting treatment targets since baseline.
**Table 4**
| A) Hypertension | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | Unnamed: 6 |
| --- | --- | --- | --- | --- | --- | --- |
| | Hennepin County | Hennepin County | Ramsey County | Ramsey County | Rice County | Rice County |
| | Coefficient (95% CI) | p-value | Coefficient (95% CI) | p-value | Coefficient (95% CI) | p-value |
| Change in systolic blood pressure (mmHg) | | | | | | |
| Unadjusted model | | | | | | |
| HealthRise-endline interaction | -5.6 (-11.7 to 0.4) | 0.068 | -10.7 (-21.7 to 0.2) | 0.054 | -6.9 (-12.9 to -0.9) | 0.024 |
| Adjusted model | | | | | | |
| HealthRise-endline interaction | -5.6 (-11.7 to 0.5) | 0.070 | -10.7 (-21.8 to 0.3) | 0.057 | -6.9 (-13.0 to -0.9) | 0.025 |
| Age | | | | | | |
| < 50 years | - | - | - | - | - | - |
| ≥ 50 years | 1.1 (-4.2 to 6.4) | 0.677 | 3.6 (-2.8 to 10.1) | 0.266 | 1.4 (-2.9 to 5.7) | 0.531 |
| Sex | | | | | | |
| Male | - | - | - | - | - | - |
| Female | -0.3 (-4.6 to 4.1) | 0.909 | 2.6 (-4.3 to 9.6) | 0.455 | 0.5 (-3.2 to 4.2) | 0.780 |
| Duration from baseline to endline | | | | | | |
| < 12 months | - | - | - | - | - | - |
| ≥ 12 months | -0.8 (-8.0 to 6.4) | 0.829 | -2.0 (-11.3 to 7.2) | 0.663 | -9.1 (-14.1 to -4.2) | < 0.001 |
| Comorbid | | | | | | |
| No (hypertension only) | - | - | - | - | - | - |
| Yes (hypertension and diabetes) | -5.6 (-11.7 to 0.5) | 0.266 | -7.4 (-15.1 to 0.3) | 0.061 | 8.5 (2.5 to 14.5) | 0.006 |
| Change in patients meeting treatment targets (% points) | | | | | | |
| Unadjusted model | | | | | | |
| HealthRise-endline interaction | 27.3 (9.8 to 44.9) | 0.002 | 22.3 (-0.0 to 44.8) | 0.051 | 17.1 (0.9 to 33.3) | 0.038 |
| Adjusted model | | | | | | |
| HealthRise-endline interaction | 27.3 (9.7 to 45.0) | 0.003 | 22.3 (-0.0 to 44.5) | 0.054 | 17.1 (0.9 to 33.4) | 0.039 |
| Age | | | | | | |
| < 50 years | - | - | - | - | - | - |
| ≥ 50 years | 2.9 (-12.7 to 18.6) | 0.714 | -2.4 (-18.8 to 14.1) | 0.777 | 7.8 (-1.6 to 17.2) | 0.102 |
| Sex | | | | | | |
| Male | - | - | - | - | - | - |
| Female | 0.1 (-9.0 to 11.0) | 0.844 | -6.5 (-21.4 to 8.4) | 0.389 | -1.6 (-9.8 to 6.5) | 0.697 |
| Duration from baseline to endline | | | | | | |
| < 12 months | - | - | - | - | - | - |
| ≥ 12 months | 6.0 (-12.1 to 24.1) | 0.511 | 5.2 (11.9 to 22.3) | 0.546 | 21.2 (11.1 to 31.2) | < 0.001 |
| Comorbid | | | | | | |
| No (hypertension only) | - | - | - | - | - | - |
| Yes (hypertension and diabetes) | 11.0 (-0.4 to 22.4) | 0.058 | 10.3 (-6.9 to 27.4) | 0.238 | -8.2 (-21.8 to 5.4) | 0.236 |
| B) Diabetes | | | | | | |
| | Hennepin County | Hennepin County | Ramsey County | Ramsey County | Rice County | Rice County |
| | Coefficient (95% CI) | p-value | Coefficient (95% CI) | p-value | Coefficient (95% CI) | p-value |
| Change in A1c (%) | | | | | | |
| Unadjusted model | | | | | | |
| HealthRise-endline interaction | -0.4 (-1.5 to 0.8) | 0.551 | -1.3 (-2.2 to -0.4) | 0.007 | -0.03 (-0.5 to 0.4) | 0.904 |
| Adjusted model | | | | | | |
| HealthRise-endline interaction | -0.4 (-1.5 to 0.8) | 0.556 | -1.3 (-2.3 to -0.4) | 0.007 | -0.03 (-0.5 to 0.4) | 0.905 |
| Age | | | | | | |
| < 50 years | - | - | - | - | - | - |
| ≥ 50 years | -1.8 (-3.0 to -0.6) | 0.006 | 0.1 (-0.8 to 1.0) | 0.828 | -0.1 (-0.6 to 0.4) | 0.369 |
| Sex | | | | | | |
| Male | - | - | - | - | - | - |
| Female | 0.1 (-0.7 to 0.9) | 0.833 | -0.1 (-0.9 to 0.7) | 0.791 | -0.1 (-0.5 to 0.3) | 0.622 |
| Duration from baseline to endline | | | | | | |
| < 12 months | - | - | - | - | - | - |
| ≥ 12 months | -1.5 (-2.8 to -0.1) | 0.031 | 0.2 (-0.8 to 1.1) | 0.807 | 0.2 (-0.2 to -0.6) | 0.369 |
| Comorbid | | | | | | |
| No (diabetes only) | - | - | - | - | - | - |
| Yes (hypertension and diabetes) | -0.5 (-1.7 to 0.7) | 0.417 | -0.5 (-1.3 to 0.4) | 0.260 | 0.1 (-0.3 to -0.6) | 0.510 |
| Change in patients meeting treatment targets (% points) | | | | | | |
| Unadjusted model | | | | | | |
| HealthRise-endline interaction | -5.1 (-30.2 to 19.9) | 0.684 | 18.2 (-0.2 to 36.5) | 0.052 | 3.0 (-8.3 to 14.1) | 0.612 |
| Adjusted model | | | | | | |
| HealthRise-endline interaction | -5.1 (-30.5 to 20.2) | 0.688 | 18.2 (-0.4 to 36.7) | 0.055 | 3.0 (-8.4 to 14.2) | 0.613 |
| Age | | | | | | |
| < 50 years | - | - | - | - | - | - |
| ≥ 50 years | 9.0 (-10.1 to 28.2) | 0.351 | 4.5 (-8.6 to 17.6) | 0.498 | -2.6 (-12.5 to 7.2) | 0.596 |
| Sex | | | | | | |
| Male | - | - | - | - | - | - |
| Female | -7.8 (-24.0 to 8.3) | 0.336 | -2.1 (-14.5 to 10.3) | 0.739 | 0.8 (-8.2 to 9.8) | 0.867 |
| Duration from baseline to endline | | | | | | |
| < 12 months | - | - | - | - | - | - |
| ≥ 12 months | 13.8 (-11.2 to 38.8) | 0.274 | -0.3 (-14.8 to 14.1) | 0.963 | -2.7 (-12.4 to 7.0) | 0.582 |
| Comorbid | | | | | | |
| No (diabetes only) | - | - | - | - | - | - |
| Yes (hypertension and diabetes) | 0.9 (-19.6 to 21.5) | 0.929 | 1.9 (-10.5 to 14.3) | 0.760 | 2.9 (-8.4 to 14.2) | 0.648 |
Among diabetes patients, HealthRise participation was associated with statistically significant reductions in A1c in Ramsey (1.3 decrease in A1c [0.4–3.2; $p \leq 0.01$) relative to comparison patients. While the percentage of HealthRise patients meeting treatment targets for diabetes did not statistically differ from that of comparison patients in Ramsey, this indicator trended toward improvement as well (a 18.2 percentage-point increase [-0.4–36.7; $$p \leq 0.054$$]).
Sensitivity analyses were conducted adjusting for baseline readings of SBP and A1c to test whether patients experiencing worse clinical profiles at baseline were more likely to experience improvements by endline (S2 Table). Models including these adjustments indicated otherwise, such that not meeting treatment targets at baseline (i.e., < 140 mmHg for SBP; < $8\%$ A1c) was associated with higher average SBP or A1c readings at endline measurement. The estimated effects of HealthRise participation did not change after adjusting for baseline readings, both in terms of continuous baseline measures and binary indicators of meeting treatment targets.
## Qualitative findings
Across HealthRise sites, six main themes emerged for the qualitative data synthesis (Table 5). First, respondents viewed home-based providers as critical to bridging barriers experienced by patients (e.g., linguistic and cultural divides), and clinic-based providers indicated high value in meeting patients beyond clinical settings. Coordination of care and a focus on social determinants of health, such as access to healthier food and nutrition, were highlighted as key program features. Second, program strengths involved learning from HealthRise sites in other countries and enabling many clinical staff to work with in-home providers for the first time. Clinical providers reported improved quality and efficiency in clinical appointments due to having additional details about patient needs from in-home providers. Home visits also enabled providers to connect patients with non-clinical resources (e.g., housing) to support improved outcomes. The theme of perceived program impacts extended program strengths, with providers reporting positive changes in the health and lives of patients and their families. Further, several interviewees emphasized the synergistic effects of pairing CHWs and community paramedics (CPs) within care teams, and reported efforts to adopt home-based provider models by other local organizations because of HealthRise experiences.
**Table 5**
| HealthRise thematic area components and contexts | Sample thematic quotes |
| --- | --- |
| Theme 1: Key program features | |
| • Home-based providers as the cornerstone to HealthRise model, bridging linguistic and cultural divides and gaining valuable new information on home visits | “…through care coordination, communication, and use of frontline health workers…in ways we’ve never been able to before, connect[ing] with families and follow[ing] up with specific patients to help them really understand and manage their chronic disease.”–Clinic-based provider |
| • Emphasis on care coordination and extending care outside the clinic to address social determinants of health | “There are so many hard things about managing diabetes. . .it takes so much time for any patient to fully understand how to put the different parts of diabetes treatment together. . .giving people the time they need to really understand all the components of diabetes control. . .that’s where our CHWs have really been massive assets.”–Clinic-based provider |
| | “If 80% of health happens outside the clinical setting, what are the ways we can foster healthy environments that allow individuals more capacity and agency to focus on these chronic diseases?”–Administrator |
| Theme 2: Program strengths | |
| • Global learning from other HealthRise sites to inform intervention design | “The global aspect is quite unique…utilizing similar strategies in different countries with very different health systems but with a similar population focus and similar workforce approaches…. I’m not aware of other projects that have attempted that across a set of different jurisdictions and landscapes.”–Policymaker |
| • Introduction to the value of home-based providers for many clinic staff | |
| • Strong relationships built between different types of providers over course of program | “The home visits contributed to more rational use of clinic time…and improved care on my end. From listening to CHWs, I have a better understanding of what’s going on in people’s lives.”–Clinic-based provider |
| • Patients’ receipt of extra support beyond what was typically possible in limited time of clinical appointment | |
| Theme 3: Program challenges | |
| • Clinical providers lacked familiarity with home-based providers | “We’ve learned that a lot of the hurdle we have to get past is educating other health care providers on what we do…what is a CP and how can we be part of their team and help to better serve their patients. . .the ones who do now understand our role. . .they are our champions, they get so excited. . .we definitely see resistance in the beginning.”–CP |
| • Facility administrators lacked experience managing home-based providers | |
| • High turnover of staff, both community health workers and management | |
| • Expectations to improve clinical outcomes in short period of time | “It was too short of a time. . .it took forever to get these communities up and running, get people hired. . .people quit, etc. . .need a longer lifespan than three years. . .to show enough impact to indicate policy change.”—Policymaker |
| • Times lags between patient recruitment/consent and actual program enrollment | |
| • Patients’ barriers to accessing healthcare (e.g., social determinants of health), as well as provider barriers to effectively accessing patients (e.g., language) | “One of the largest hurdles and barrier to successful implementation was the lack of cohesive patient data systems…you need layers of permission, use agreements, consent, you can’t compare across systems, you can’t look at anybody’s system but your own. . .if somebody could fix that, we could do a lot more good.”—Administrator |
| • Interoperability and data sharing between health systems, across platforms, and among care team (e.g., electronic medical record access for all team members) | |
| Theme 4: Perceived program impacts | |
| • Information gained during home visits improved the quality and efficiency of clinical appointments | “There’s just a synergy when you combine the two. . .the CP and the medical side and the CHW looking at the social issues. . .it was kind of exponential how much benefit we were able to provide versus just one or the other.”—CP |
| • Pairing of CP and CHW brought together complementary skill sets | |
| • Home visits helped connect patients with non-clinical resources to improve health | “We have nutritionists and a diabetic educator in the clinic, but patients for a variety of reasons are not always open to meeting with them….to have somebody go to their house and figure out what their specific food interests are and come up with recipes, that was great.”–Clinic-based provider |
| • Positive changes in health and lives of patients and their families | |
| • Efforts to institutionalize home-based providers as part of care model | |
| Theme 5: Recommendations for improvement | |
| • Tailor EMR software for care teams that include home-based providers | “EMRs are designed around providers and reimbursements, our model is around holistic care coordination across contexts. The tools we have been using are imperfect, we’re looking for other tools that might be able to plug in to our model better. . .We haven’t yet found the silver bullet.”–Administrator |
| • Determine ideal duration and/or frequency of home visits | |
| • Offer additional mental health resources to support patients with chronic disease | “We’re aggressively moving into providing more mental health care, plugging it into our model that we’ve perfected during HealthRise…Behavioral health, mental health, and chemical dependency are NCDs, similar to diabetes in that it’s all about what happens in the meantime.”–Administrator |
| • Provide more trainings for home-based providers (e.g. motivational interviewing) | |
| Theme 6: Future program directions | |
| • Expand use of home-based providers to other patient populations, particularly for other chronic conditions | “I think it should be implemented everywhere…that’s the response we’re getting from our partners, from others in the health care setting…everyone is saying ’CHW! I need 7 of you in my facility!’…everyone has a million questions about how to get it started, set up. . .I definitely think it’s a model to follow.”—CHW |
| • Generate additional evidence establishing the cost-effectiveness of home-based care for chronic conditions | |
| • Identify sustainable funding for home-based providers | |
Common challenges emerged across sites, often relating to new program establishment and incorporation of in-home providers within care teams. For example, some clinical providers showed initial skepticism about the added value of in-home providers, and most administrators did not have prior experience managing CHWs and CPs. Site-specific challenges also occurred; for instance, patient consent for enrollment took a long time during the initial phase of program implementation in Rice, while CHW turnover was an ongoing obstacle for both Hennepin and Ramsey. Data sharing was another pervasive challenge, mainly from poor interoperability and coordination between clinic data systems and electronic medical record (EMR) systems. Providers also expressed frustration with expectations for rapidly improving clinical outcomes, especially given the longstanding challenges in healthcare access and social determinants of health most patients faced.
The fifth theme pertained to recommendations for improvement, many of which stemmed from acknowledged challenges. Such suggestions included prioritizing better communication and coordination among care teams as well as EMR systems that could more seamlessly accommodate patient updates and provider notes from multiple care-team members. Interviewees reported having inadequate clarity on the ideal frequency and length of home visits, an area where efficiencies in resource deployment could be improved. The sixth theme, future program directions, involved many ideas about adapting and expanding HealthRise programming to new locations and conditions (e.g., mental health). Another common thread concerned program sustainability, namely longer-term financing and retaining home-based providers for the HealthRise model.
## Discussion
Increasingly more evidence shows that community-based programs can help underserved communities in the US better access health services, alleviate barriers to care, and improve at least some health behaviors and outcomes. The present study contributes to this evidence base through its prospective evaluation of HealthRise programs implemented within three different Minnesota communities. Relative to comparison patients in Rice County, HealthRise participation was significantly associated with SBP reductions, while the percentage of hypertension patients meeting treatment targets increased at the Hennepin and Rice HealthRise sites. For diabetes, HealthRise patients saw larger A1c declines in at the Ramsey site than comparison patients. Heterogeneous patterns in patient improvements since baseline highlight potential case management challenges among underserved individuals and communities, especially under short program implementation periods. As emphasized by HealthRise care teams, community-based programs show promise for improving NCD care and outcomes for underserved populations; nonetheless, more work is needed to better understand how such programs can be further brought to scale and sustained long-term.
HealthRise participation was related to SBP or A1c decreases and a higher percentage of patients meeting treatment targets at some sites relative to comparison patients. Variations found across sites may be related to the types of specific interventions and activities implemented, as HealthRise programs were meant to be tailored to local contexts and needs [14, 15]. For instance, one component of the Ramsey HealthRise program was to implement nutrition-focused interventions in both English and Spanish, as English language barriers were identified as a key challenge for patients served by the Ramsey site (Table 1) and $50\%$ of Ramsey HealthRise participants identified as Hispanic or Latinx at enrollment (S1 Table). Alternatively, such variation may be associated with the relative health status of each population at baseline and thus potential for future improvement. At the Ramsey site, HealthRise patients with diabetes averaged $11.4\%$ reductions in A1c and $9.4\%$ declines in SBP by endline; yet these patients also began HealthRise with highest risk profiles, averaging 150 mmHg SBP and $10\%$ A1c at baseline. Sensitivity analyses adjusting for baseline readings found no changes in the effects associated with HealthRise participation (S2 Table), whereas not meeting treatment targets at baseline measurement was associated with increases in SBP or A1c by endline. In combination, these results suggest that, all else being equal, patients with higher risk profiles may, on average, experience worsening indicators over time–and that community-based interventions such the HealthRise program could play a role in lessening or reversing such progression.
As found in past studies [5–7], several factors may have contributed to the observable effects of HealthRise. These included focusing on specific barriers patients faced in each community (e.g., home-based care provided by CHW and CP teams); maintaining small patient loads, enabling more individualized attention and tailoring of visit frequency to patient need; and explicitly providing non-medical support, such as health education and community resources like transportation. Further, the overall positive views of HealthRise by grantees and care teams alike may have contributed to the program’s effects. Despite challenges during earlier stages of program implementation (e.g., recruiting and retaining CHWs, ensuring adequate access to EMRs), providers voiced valuing home-based health workers and were eager to expand this model of care. In combination, these factors may have set the foundation for HealthRise’s impact for underserved patients with hypertension and diabetes in the US.
Amid such promising findings, however, important challenges remained for each site and for broader applications of the HealthRise model elsewhere. Across sites, some proportion of HealthRise patients failed to meet clinical targets for hypertension or diabetes at both time points–and concerningly, some percentage moved from being below biometric thresholds at baseline to exceeding them at endline. These patterns underscore the complexity of effectively managing chronic conditions like hypertension and diabetes, especially in environments where patients face compounding barriers to medical care and health-promoting behavior (e.g., limited options for nutritious food, minimal time for exercise amid job and family demands, inferior access to adequate transportation and housing). CHWs and integrated care teams may be able to mitigate some of these obstacles and better support patients’ medical needs, a critical step in addressing deep-seeded health disparities; however, in the absence of more macro-level socioeconomic policies and health system investments to support underserved patients in the US, many community-based programs will continue facing need and demand that far exceeds their limited capacities and resources. As laid bare by COVID-19, the health challenges underserved communities experience do not begin and end at clinics: rather, they stem from and are exacerbated by structural inequalities that require intervention and engagement well beyond the formal health system [20, 21] CHWs can provide a vital role in at least overcoming some of these access and sociocultural barriers, ranging from house-based visits made by CHWs fluent in patients’ native languages [3] to connecting patients with services that can facilitate better overall care. Nevertheless, the potential benefits of encouraging health programs can be easily blunted if actors–and actions–beyond the immediate health system are not also actively addressing fundamental drivers of health disparities.
## Limitations
This study is subject to several limitations. First, small program sizes and thus study samples at each HealthRise site likely affected the degree to which potential program impact could be detected and conclusively attributed to HealthRise participation. This is particularly true for site-condition combinations in which very few patients were prevalent cases at baseline and had at least two biometric readings within the program implementation period (e.g., 37 HealthRise patients with diabetes in at the Hennepin site, 32 HealthRise patients with hypertension in Ramsey). The inclusion or exclusion of even a few patients for several site-condition groupings could shift effect sizes and statistical significance estimated by the difference-in-difference models.
Second, comparison groups were constructed retrospectively based on available patient record information and were not selected by random assignment. While efforts were made to ensure that comparison patient data were chosen to generally represent individuals who would have been eligible for HealthRise enrollment, they may have differed from individuals who enrolled.
Third, only patients who remained enrolled at endline were included in the present study; by taking this ‘as treated’ analytic approach, which provides insights into program effects closer to full adherence, these patients may not represent all potential target populations for HealthRise interventions and results may be positively biased. For instance, relatively high rates of program withdrawal at some sites could have led to a bias toward healthier patients remaining in the program (i.e., sicker patients may not go into the clinic). However, due to the home-based care model espoused by HealthRise, it is equally possible that patients who did not withdraw were less healthy and stayed enrolled because HealthRise offered important access to services, like home visits, otherwise unavailable to them.
Fourth, single biometric readings comprised baseline and endline measures, as well as patients meeting clinical targets at each time period; subsequently, analyses could be sensitive to outliers in patient records, particularly given the relatively small sample sizes for each site. If more readings could have informed baseline and endline indicators, it is possible patient-level patterns could have differed from what observed on the basis of single readings.
Fifth, medication data were not available and thus we were unable to assess the full cascade of care for hypertension and diabetes case management. Medication adherence may have been important factor for patients who either failed to see improvements in clinical indicators or experienced worsening outcomes. Furthermore, additional information on risk factors or health behaviors (e.g., smoking status, alcohol consumption) that may have affected patient-level outcomes were not available and thus could not be accounted for.
Sixth, due to the small sample sizes for each site, we could not further analyze the potential effects of visit frequency (and thus approximate dose-response) or intensity of intervention exposure on patient outcomes. This is further complicated by issues related to endogeneity, such that patients with worse clinical profiles and thus greater need are likely to receive more frequent visits by CHWs or care teams.
Seventh, small sample size and inconsistent data availability limited our ability to conduct more disaggregated analyses by reported sex, race or ethnicity, income, and other potentially important factors (e.g., primary language) for understanding how community-based health interventions can promote greater equity for underserved communities. Descriptive statistics imply that there could be differential disease prevalence by reported race or ethnicity (S1 Table): for instance, at the Rice County site, a comparatively smaller percentage of HealthRise patients identifying as Black or African American had both hypertension and diabetes at enrollment ($7.0\%$, $$n = 4$$ of 57 total patients with both conditions) than the overall percentage of patients identifying as Black or African American had ($10.2\%$, $$n = 16$$ out 157 total patients). Yet based on such small samples, formal analyses comparing prevalence by reported race or ethnicity could yield inconclusive–or worse, potentially spurious or unrepresentative–results. Combining different reported race or ethnicity groups is one option often used to mitigate small-sample size issues, though to doing so risks minimizing the experiences of a given racial or ethnicity group [22]. For the present study, we ultimately determined that further disaggregation posed more potential analytic harms or opportunities for misinterpretation than possible overall benefits; in doing so, however, we recognize its substantive limitation for the present study.
## Conclusions
With its focus on community-based health programs and improving NCD care for underserved populations, HealthRise showed some positive effects for hypertension and diabetes patients in Hennepin, Ramsey, and Rice counties. Provider experiences indicated enthusiasm for expanding the home-based model of care for NCDs, though the resource requirements–as well feasibility–to sustain impact at a larger scale remain unknown. While community-based NCD interventions show promise for overcoming barriers to effective care for hypertension and diabetes among underserved populations, continued monitoring and robust evaluations of local impact are vital to ensuring maximum benefit for individuals with the greatest need.
## References
1. Mokdad AH, Ballestros K, Echko M, Glenn S, Olsen HE, Mullany E. **The State of US Health, 1990–2016: Burden of Diseases, Injuries, and Risk Factors Among US States**. *JAMA* (2018.0) **319** 1444-1472. DOI: 10.1001/jama.2018.0158
2. Baciu A, Negussie Y. (2017.0)
3. Witmer A, Seifer SD, Finocchio L, Leslie J, O’Neil EH. **Community health workers: integral members of the health care work force.**. *Am J Public Health* (1995.0) **85** 1055-1058. DOI: 10.2105/ajph.85.8_pt_1.1055
4. Rosenthal EL, Brownstein JN, Rush CH, Hirsch GR, Willaert AM, Scott JR. **Community Health Workers: Part Of The Solution.**. *Health Affairs.* (2010.0) **29** 1338-1342. DOI: 10.1377/hlthaff.2010.0081
5. Norris SL, Chowdhury FM, Le KV, Horsley T, Brownstein JN, Zhang X. **Effectiveness of community health workers in the care of persons with diabetes**. *Diabetic Medicine* (2006.0) **23** 544-556. DOI: 10.1111/j.1464-5491.2006.01845.x
6. Brownstein JN, Chowdhury FM, Norris SL, Horsley T, Jack L, Zhang X. **Effectiveness of Community Health Workers in the Care of People with Hypertension.**. *American Journal of Preventive Medicine.* (2007.0) **32** 435-447. DOI: 10.1016/j.amepre.2007.01.011
7. Kim K, Choi JS, Choi E, Nieman CL, Joo JH, Lin FR. **Effects of Community-Based Health Worker Interventions to Improve Chronic Disease Management and Care Among Vulnerable Populations: A Systematic Review.**. *Am J Public Health.* (2016.0) **106** e3-e28. DOI: 10.2105/AJPH.2015.302987
8. Kangovi S, Mitra N, Grande D, Huo H, Smith RA, Long JA. **Community Health Worker Support for Disadvantaged Patients With Multiple Chronic Diseases: A Randomized Clinical Trial.**. *Am J Public Health.* (2017.0) **107** 1660-1667. DOI: 10.2105/AJPH.2017.303985
9. Kangovi S, Mitra N, Norton L, Harte R, Zhao X, Carter T. **Effect of Community Health Worker Support on Clinical Outcomes of Low-Income Patients Across Primary Care Facilities: A Randomized Clinical Trial.**. *JAMA Intern Med.* (2018.0) **178** 1635-1643. DOI: 10.1001/jamainternmed.2018.4630
10. Gary TL, Bone LR, Hill MN, Levine DM, McGuire M, Saudek C. **Randomized controlled trial of the effects of nurse case manager and community health worker interventions on risk factors for diabetes-related complications in urban African Americans.**. *Prev Med* (2003.0) **37** 23-32. DOI: 10.1016/s0091-7435(03)00040-9
11. Brown SA, Garcia AA, Kouzekanani K, Hanis CL. **Culturally competent diabetes self-management education for Mexican Americans: the Starr County border health initiative**. *Diabetes Care* (2002.0) **25** 259-268. DOI: 10.2337/diacare.25.2.259
12. Flor LS, Wilson S, Bhatt P, Bryant M, Burnett A, Camarda JN. **Community-based interventions for detection and management of diabetes and hypertension in underserved communities: a mixed-methods evaluation in Brazil, India, South Africa and the USA.**. *BMJ Global Health.* (2020.0) **5** e001959. DOI: 10.1136/bmjgh-2019-001959
13. 13Institute for Health Metrics and Evaluation (IHME). HealthRise Evaluation: Final Report. Seattle, WA, USA: IHME; 2019. Available: https://www.healthdata.org/policy-report/healthrise-evaluation-final-report
14. Associates Abt. **HealthRise Final Report: Expanding Access to Chronic Disease Care through Community Approaches in Four Countries.**. *Abt Associates* (2019.0)
15. Roth G, Gabert R, Thomson B, Gakidou E. (2015.0)
16. 16MN Community Measurement.
2017 Health Equity of Care Report: Stratification of Health Care Performance Results in Minnesota by Race, Hispanic Ethnicity, Preferred Language, & Country of Origin.
Minneapolis, MN: MN Community Measurement; 2018. Available: https://mncm.org/wp-content/uploads/2018/01/2017-Health-Equity-of-Care-Report_unencrypted-1.pdf. (2018.0)
17. Gabert R, Thomson B, Gakidou E, Roth G. **Identifying High-Risk Neighborhoods Using Electronic Medical Records: A Population-Based Approach for Targeting Diabetes Prevention and Treatment Interventions.**. *PLOS ONE.* (2016.0) **11** e0159227. DOI: 10.1371/journal.pone.0159227
18. 18StataCorp.
Stata Statistical Software: Release 15.
College Station, TX: StataCorp LLC; 2017.. (2017.0)
19. 19R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2018. Available: https://www.R-project.org. (2018.0)
20. Norrisa Haynes, Cooper Lisa A, Albert Michelle A. **At the Heart of the Matter: Unmasking and Addressing COVID-19’s Toll on Diverse Populations.**. *Circulation*. DOI: 10.1161/CIRCULATIONAHA.120.048126
21. Dorn A van Cooney RE, Sabin ML. **COVID-19 exacerbating inequalities in the US**. *Lancet* (2020.0) **395** 1243-1244. DOI: 10.1016/S0140-6736(20)30893-X
22. Flanagin A, Frey T, Christiansen SL. **Updated Guidance on the Reporting of Race and Ethnicity in Medical and Science Journals**. *JAMA* (2021.0) **326** 621-627. DOI: 10.1001/jama.2021.13304
|
---
title: The metabolic profile of waist to hip ratio–A multi-cohort study
authors:
- Lars Lind
- Shafqat Ahmad
- Sölve Elmståhl
- Tove Fall
journal: PLOS ONE
year: 2023
pmcid: PMC9970070
doi: 10.1371/journal.pone.0282433
license: CC BY 4.0
---
# The metabolic profile of waist to hip ratio–A multi-cohort study
## Abstract
### Background
*The* genetic background of general obesity and fat distribution is different, pointing to separate underlying physiology. Here, we searched for metabolites and lipoprotein particles associated with fat distribution, measured as waist/hip ratio adjusted for fat mass (WHRadjfatmass), and general adiposity measured as percentage fat mass.
### Method
The sex-stratified association of 791 metabolites detected by liquid chromatography–mass spectrometry (LC-MS) and 91 lipoprotein particles measured by nuclear magnetic spectroscopy (NMR) with WHRadjfatmass and fat mass were assessed using three population-based cohorts: EpiHealth ($$n = 2350$$) as discovery cohort, with PIVUS ($$n = 603$$) and POEM ($$n = 502$$) as replication cohorts.
### Results
Of the 193 LC-MS-metabolites being associated with WHRadjfatmass in EpiHealth (false discovery rate (FDR) <$5\%$), 52 were replicated in a meta-analysis of PIVUS and POEM. Nine metabolites, including ceramides, sphingomyelins or glycerophosphatidylcholines, were inversely associated with WHRadjfatmass in both sexes. Two of the sphingomyelins (d18:$\frac{2}{24}$:1, d18:$\frac{1}{24}$:2 and d18:$\frac{2}{24}$:2) were not associated with fat mass ($p \leq 0.50$). Out of 91, 82 lipoprotein particles were associated with WHRadjfatmass in EpiHealth and 42 were replicated. Fourteen of those were associated in both sexes and belonged to very-large or large HDL particles, all being inversely associated with both WHRadjfatmass and fat mass.
### Conclusion
Two sphingomyelins were inversely linked to body fat distribution in both men and women without being associated with fat mass, while very-large and large HDL particles were inversely associated with both fat distribution and fat mass. If these metabolites represent a link between an impaired fat distribution and cardiometabolic diseases remains to be established.
## Introduction
Excess body adipose tissue could be analyzed in at least two dimensions, total fat mass and fat distribution. In epidemiology, increased fat mass is most often evaluated by use of body mass index (BMI), while fat distribution is assessed by either waist circumference (WC) or the waist/hip circumference ratio (WHR). Mendelian randomization studies have shown a positive causal effect of BMI [1, 2], as well as BMI-adjusted WHR [3], on cardiovascular risk factors and diseases, supporting the view that neither general obesity, nor a disadvantageous fat distribution are innocent phenomena.
Large-scale genetic studies are a way to search for pathophysiological pathways involved in obesity. Using data from several hundred thousand individuals have identified >250 genetic loci linked to BMI [4], and around 350 loci linked to WHR when adjusted for BMI [5]. In a review of the association between genetic loci and total fat mass and fat distribution, Fall et al. [ 6] pointed out that the biological pathways identified by these genes and further expression analyses are very different. For BMI, central nervous system pathways, with especially hippocampus, hypothalamus and the limbic system, plays a major role in terms of neurotransmission and energy balance. In contrast, genes related to WHR mainly reflects adipose tissue biology, insulin resistance and angiogenesis. Thus, genes regulating total fat mass and fat distribution points towards different mechanisms involved in these two dimensions of obesity. Another important finding is that for around one-third of the genetic loci linked to WHR a sex-interaction was seen, with generally stronger genetic effects in women compared to men.
Metabolomics is the study of small molecules (<1.5kD). A great number of studies have investigated the metabolomic profile of obesity, and a meta-analysis of 11 studies found high levels of branched-chain and aromatic amino acids, certain fatty acids and reduced levels of acylcarnitines and lysophosphatidylcholines to be the most common metabolic alterations in obese individuals [7]. There are also studies on the metabolomic profile of an altered fat distribution [8–13]. However, only a few studies have tried to disentangle if the metabolic profile of a disadvantageous fat distribution was different from that found in general obesity [14, 15].
A detailed description of the lipoprotein metabolic profile could be obtained by magnetic resonance spectroscopy (NMR) and a certain profile, with high levels of cholesterol in all VLDL and LDL subclasses and low levels in the larger classes of HDL together with elevated triglyceride levels in all lipoprotein classes except the largest classes of HDL, have been associated with cardiometabolic disease, such as myocardial infarction and stroke [16]. It is unclear if this lipoprotein profile is seen in subjects with a high WHR independently of general obesity.
The major aim of the present study is to investigate if the metabolic profile, assessed through LC-MS and lipoprotein profile measured through NMR, of a disadvantageous fat distribution is different from that found in general obesity, using a similar approach as in the genetic studies. Since we have measured fat mass by bioimpedance in the samples used in the present study, we adjusted WHR for fat mass percentage instead of BMI. Fat mass percentage is related to cardiovascular mortality [17], and all-cause mortality [18] independently of BMI, and is a more precise measure of the amount of adipose tissue than BMI. As the genetic studies of WHR adjusted for BMI points to sex-differences, we stratified the analyses of WHR by sex. In this study, we used one population-based study for discovery (EpiHealth) and a meta-analysis of another two studies as replication (PIVUS and POEM) in order to validate the findings in independent samples. The hypothesis tested was that we would find metabolites and lipoproteins being related to WHR independently of fat mass.
## EpiHealth
Starting 2011, a random sample of men and women in the age range 45 to 75 years were invited to a health screening survey, called EpiHealth, in the two Swedish cities Uppsala and Malmö [19]. In 2018, 25,000 individuals were included. Metabolomic data have been collected in a random subsample of 2,342 subjects attending the Uppsala site.
## POEM (Prospective investigation of Obesity, Energy and Metabolism)
The population-based POEM study is based on invitations to a random sample of 50-year old men and women living in Uppsala, Sweden [20]. Between Oct 2010 and Oct 2016, 502 individuals were included and metabolomics measurements have been performed in the total sample.
## PIVUS (Prospective Investigation of the Vasculature in Uppsala Seniors)
Between 2001 and 2004, 1,016 randomly selected men and women, all aged 70 years, were investigated [21]. They were all offered a new examination at the ages of 75 and 80. The present study use data from the 80-year examination in which metabolomics measurements have been performed in the total sample ($$n = 603$$).
All studies were approved by the Uppsala ethics committee (Application numbers $\frac{2010}{402}$, $\frac{2011}{045}$ and $\frac{2009}{057}$) and all participants provided their informed written consent.
## Physical measurements and blood sampling
Waist circumference was measured at the umbilical level, while hip circumference was measured at the level of trochanter major. WHR is the ratio of those two measurements (waist/hip). Fat mass and body weight were assessed through a weight scale that also calculates fat mass by mean of bioimpedance (Tanita BC-418MA, Tokyo, Japan). Fat mass percentage is the fat mass divided by body weight and is the measurement used in the present study. Using both total body potassium and total body water [22], as well as dual-energy x-ray absorptiometry (DXA) [23] as comparative methods, measurement of fat mass with bioimpedance has been proven to be valid in previous studies. In addition, to validate that fat mass measured by bioimpedance is an accurate measurement of body fat in our setting, we compared the bioimpedance measurement with measurements with dual X-ray absorbmetry (DEXA, Lunar Prodigy, GE Healthcare) performed in 486 individuals in the POEM study. The Pearson´s correlation coefficient between these two measurements were 0.93.
The physical measurements were performed in the same fashion in all of the three studies.
Blood was drawn after an overnight fast in the POEM and the PIVUS cohorts, while 6 hours of fasting was required in EpiHealth. The blood was collected in EDTA tubes that were centrifuged and plasma was frozen in -80°C for later analysis.
## Questionnaire
Life-style factors were evaluated using a questionnaire in all samples. In EpiHealth, leisure-time physical activity was assessed on a 5-level scale with 1 as sedentary and 5 as athlete training. Smoking was defined as years of smoking in life. Alcohol intake was assessed as drinks per week.
In the POEM and the PIVUS cohorts, leisure-time physical activity was assessed on a 4-level scale with 1 as sedentary and 4 as athlete training. Smoking variable was used as current smoking. Alcohol intake was not assessed in the PIVUS cohort at age 80 years neither in the POEM cohort. In all three cohort studies, education was defined on a three-level scale; <10, 10–12 and >12 years in school.
## Metabolomics
In all three study samples, metabolomics (Metabolon inc., USA) was performed on plasma samples being stored at -80°C. 100μl of human plasma was utilized for analysis. 500μl of methanol was added to each sample. Samples were prepared using the automated MicroLab STAR® system from Hamilton Company. Several internal standards were added prior to the first step in the extraction process for QC purposes. To remove protein, dissociate small molecules bound to protein or trapped in the precipitated protein matrix, and to recover chemically diverse metabolites, proteins were precipitated with methanol under vigorous shaking for 2 min (Glen Mills GenoGrinder 2000) followed by centrifugation. The resulting extract was divided into five fractions: two for analysis by two separate reverse phases (RP)/UPLC-MS/MS methods with positive ion mode electrospray ionization (ESI), one for analysis by RP/UPLC-MS/MS with negative ion mode ESI, one for analysis by hydrophilic interaction (HILIC)/UPLC-MS/MS with negative ion mode ESI, and one sample was reserved for backup. Metabolon’s untargeted metabolomics panel utilizes a Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive or Q-Exactive Plus high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-I). The columns utilized were Waters BEH C18 2.1x100mm 1.7μm and Waters BEH Amide 2.1 x150mm 1.7μm.
Only annotated, non-xenobiotic metabolites with a detection rate >$75\%$ in all samples were used in the analyses ($$n = 791$$). The values were normalized and given in arbitrary units. The relative concentration of identified peaks associated with each chemical in the Metabolon library (where present), are obtained by measuring the area of the peak relative to the surrounding baseline. All peak areas are integrated for each biochemical, based on the authentic standard for each biochemical, providing a consistent quantitation of relative abundance.
Compounds were identified by comparison to library entries of purified standards or recurrent unknown entities. Metabolon maintains a library based on >3,000 authenticated standards that contains the retention time/index (RI), mass to charge ratio (m/z), and chromatographic data (including MS/MS spectral data) on all molecules present in the library. Metabolon has built and maintains a proprietary chemical library based on authentic standards that contains the retention time (RT), retention index (RI), mass to charge ratio (m/z), and mass spectral data (including MS/MS spectral data) on all molecules present in the library per method. Biochemical identifications are therefore based on three criteria: retention index within a narrow retention window of the proposed identification, accurate mass match to the library, and the MS/MS forward and reverse scores between the experimental data and authentic standards. The MS/MS scores are based on a comparison of the ions present in the experimental spectrum to the ions present in the library spectrum. While there may be similarities between molecules based on one of these factors, the use of all three data points can be used to accurately identify biochemicals.
Several types of controls were analyzed in concert with the experimental samples: a pooled matrix sample generated by taking a small volume of each experimental sample (or alternatively, use of a pool of well-characterized human plasma) served as a technical replicate throughout the data set; extracted water samples served as process blanks; and a cocktail of QC standards that were carefully chosen not to interfere with the measurement of endogenous compounds were spiked into every analyzed sample, allowed instrument performance monitoring and aided chromatographic alignment.
Internal standards (IS) were used for alignment of data and for QC of instrument performance. The IS were selected to span the chromatogram and allow the creation of an RI ladder. Furthermore, they were selected to be representative of the type of endogenous compounds detected and therefore can be used to monitor consistency of chromatographic behavior and MS response.
Process standards are added during sample extraction to ensure consistent performance of the entire process from sample preparation through sample analysis.
## Instrument performance standards
d35-octadecanoic acid, fluorophenylglycine, d5-indole acetate, chlorophenylalanine, Br-phenylalanine, d5-tryptophan, d4-tyrosine, d3-serine, d3-aspartic acid, d7-ornithine, d4-lysine.
## Process assessment standards
Fluorophenylglycine, chlorophenylalanine.
Instrument variability was $5\%$ as determined by calculating the median relative standard deviation (RSD) for the internal standards that were added to each sample prior to injection into the mass spectrometers. Overall process variability was $7\%$ as determined by calculating the median RSD for all endogenous metabolites (i.e., non-instrument standards) present in $100\%$ of the Client Matrix samples, which are technical replicates of pooled client samples. In a published comparison between the 4 MS platforms used, the average laboratory coefficient of variation (CV) on the 4 platforms was between 9.3 and $11.5\%$, average inter-assay CV ranged from $9.9\%$ to $12.6\%$ and average intra-assay CV ranged from $5.7\%$ to $6.9\%$ [24].
## Lipoprotein measurements
Lipoproteins and their content were quantified using high-throughput NMR metabolomics (Nightingale Health Ltd, Helsinki, Finland) [25]. The 14 lipoprotein subclass sizes were defined as follows: extremely large VLDL with particle diameters from 75 nm upwards and a possible contribution of chylomicrons, five VLDL subclasses, IDL, three LDL subclasses and four HDL subclasses. The following components of the lipoprotein subclasses were quantified: phospholipids (PL), triglycerides (TG), cholesterol (C), free cholesterol (FC), and cholesteryl esters (CE). Very few of the measurements of the extremely large VLDL were above the level of detection, so this subclass was not used in the further analysis in the present study.
Two NMR spectra were recorded for each plasma sample using 500 MHz NMR spectrometers (Bruker AVANCE IIIHD). The first spectrum is a presaturated proton spectrum, which features resonances arising mainly from proteins and lipids within various lipoprotein particles. The other spectrum is a Carr-Purcell-Meiboom-Gill T2-relaxation-filtered spectrum where most of the broad macromolecule and lipoprotein lipid signals are suppressed, leading to enhanced detection of low-molecular-weight metabolites. The identification and quantification used a company proprietary software (version 2020). Two internal control samples provided by the company were included in each 96-well plate for tracking the consistency over time. When the coefficient of variation (CV) was calculated based on these internal controls and duplicate samples, the mean CV was below $4\%$, and only a few metabolites showed a CV>$10\%$.
## Statistics
All metabolites and lipoproteins were rank based inverse-normal transformed to obtain a normal distribution and the same mean level for each metabolite. Fat mass (percentage) and WHR were normally distributed.
Separate analyses were performed for the LC-MS metabolomics and NMR-based lipoprotein measurements.
For WHRadjfatmass, linear regression analyses were performed using metabolites as dependent variables and WHRadjfatmass as the independent variable. Potential confounders were used including fat mass percentage, age, education, smoking, alcohol, exercise habits and statin use (other antilipidemic agents are rarely used in Sweden). The same model was used in all three study samples (except that alcohol was not included in the PIVUS and the POEM cohorts). Sex stratified analysis were performed for WHRadjfatmass. An interaction term between WHRadjfatmass and sex was used in a set of separate models to test the significance of any sex-interactions regarding the relationships between WHRadjfatmass and metabolites.
To be able to identify metabolites associated with WHRadjfatmass only, we also assessed the association of fat mass percentage with metabolites using linear regression models where metabolites were used as dependent variables and fat mass percentage as the independent variable. Confounders in the model were age, sex, education, smoking, alcohol, exercise habits and statin. The same model was used in all three study cohorts (except that data on alcohol intake was not available in the PIVUS and POEM samples).
EpiHealth was used as the discovery sample and a false discovery rate (FDR, Benjamin-Hochenberg) <0.05 was used to qualify metabolites to be evaluated in the validation step. The validation step was performed using results from a meta-analysis (inverse-variance weighted (IVW) fixed effect meta-analysis) of the POEM and the PIVUS results. Also in the validation step, a FDR<0.05 was used to assess significance.
STATA 16.1 (Stata inc, College Station, TX, USA) was used for these analyses.
## Results
Basic characteristics of the three samples are provided in Table 1.
**Table 1**
| Unnamed: 0 | EpiHealth | POEM | PIVUS |
| --- | --- | --- | --- |
| n | 2342 | 502 | 604 |
| Age (years) | 61 (8.4) | 50 (0.1) | 80 (0.2) |
| Female sex (%) | 50% | 50% | 50% |
| BMI (kg/m 2 ) | 26.5 (3.8) | 26.4 (4.2) | 26.9 (4.6) |
| Weight (%) | | | |
| Normal-weight | 37 | 41 | 35 |
| Overweight | 47 | 41 | 44 |
| Obese | 16 | 18 | 21 |
| Waist/hip ratio (WHR) | | | |
| Total | 0.90 (0.08) | 0.90 (0.08) | 0.90 (0.07) |
| Men | 0.95 (0.06) | 0.94 (0.05) | 0.94 (0.06) |
| Women | 0.85 (0.07) | 0.87 (0.08) | 0.86 (0.06) |
| Fat mass (%) | | | |
| Total | 30 (8.0) | 28 (8.0) | 32 (8.0) |
| Men | 24 (5.2) | 22 (5.2) | 27 (6.2) |
| Women | 36 (6.5) | 33 (7.2) | 37 (7.5) |
| Alcohol intake | 2.43 (2.92) (drinks/week) | | |
| Exercise habits | 2.29 (.8) (On a 5-grade scale) | 2.8 (1.01) (On a 4-grade scale) | 1.21 (1.31) (On a 4-grade scale) |
| Education (%) | | | |
| <10 years | 21 | 8 | 56 |
| 10–12 years | 29 | 44 | 19 |
| >12 years | 50 | 48 | 25 |
| Smoking | 6.7 years of smoking | 9.8% | 3.2% |
| Statin treatment (%) | 10 | 3.4 | 30 |
| Total cholesterol (mmol/l) | 5.9 (1.1) | 5.3 (1.0) | 5.1 (1.0) |
| Total triglycerides (mmol/l) | 1.3 (0.8) | 1.2 (0.9) | 1.2 (0.6) |
| LDL-cholesterol (mmol/l) | 3.9 (1.0) | 3.4 (0.9) | 3.3 (0.9) |
| HDL-cholesterol (mmol/l) | 1.5 (0.4) | 1.4 (0.4) | 1.4 (0.4) |
## Metabolomics by liquid chromatography–mass spectrometry methods
In the discovery step, 193 out of the 791 evaluated metabolites were associated with WHRadjfatmass in either men or women at FDR<0.05. 154 of the metabolites were associated with fat mass.
In the replication sample of the PIVUS and the POEM cohorts, 52 of the 193 metabolites were associated with WHRadjfatmass in either men or women at FDR<0.05. Nine of those metabolites were associated with WHRadjfatmass in both men and women (Fig 1).
**Fig 1:** *Relationships between mass spectrometry-based metabolites and waist-hip ratio (WHR) in males and females and vs fat mass in both sexes combined.Only metabolites showing a false discovery rate (FDR) <0.05 in both the discovery and validation analyses in both sexes are shown. The estimates are from the validation meta-analysis of the PIVUS and POEM samples. The betas are for one SD change in WHR or fat mass. Also 95%CIs are given.*
The nine validated metabolites being significantly associated with WHRadjfatmass in both sexes include ceramides, sphingomyelins and glycerophosphatidylcholines (GPCs) (see Fig 1 and S1 Table for details) and were all inversely related to WHRadjfatmass. Of those nine metabolites, two sphingomyelins ((d18:$\frac{2}{24}$:1, d18:$\frac{1}{24}$:2) and (d18:$\frac{2}{24}$:2)) were very far from being significantly associated with fat mass ($$p \leq 0.53$$ and $$p \leq 0.67$$, respectively), while the other 7 showed significant negative association with fat mass. For sphingomyelin (d18:$\frac{1}{22}$:2, d18:$\frac{2}{22}$:1, d16:$\frac{1}{24}$:2), the association was in opposite direction for fat mass compared to WHRadjfatmass.
Nineteen of the 52 replicated metabolites were significantly associated with WHRadjfatmass in women only (Table 2). Of those, six showed an interaction with sex with $p \leq 0.05.$ Those 6 represents different chemical classes (GPCs, fatty acids, carotenes, and bile acids). All of these 6 metabolites were also associated with fat mass.
## Lipoprotein particles by NMR
In the discovery step, 82 of the 91 evaluated metabolites were associated with WHRadjfatmass in either men or women at FDR<0.05. All of these 82 metabolites except one, very small VLDL cholesterol, were also associated with fat mass.
In the replication step, 43 of these 82 metabolites were associated with WHRadjfatmass in either men or women at FDR<0.05. Fourteen of those metabolites were related to WHR in both men and women (see Fig 2 and S2 Table).
**Fig 2:** *Relationships between nuclear magnetic resonance spectrometry-based lipoprotein metabolites and waist-hip ratio (WHR) in males and females in both sexes combined.Only metabolites showing a false discovery rate (FDR) <0.05 in both the discovery and validation analyses in both sexes are shown.*
The estimates are from the validation meta-analysis of the PIVUS and POEM samples. The betas are for one SD change in WHR or fat mass. Also $95\%$CIs are given.
All of these metabolites were also related to fat mass. The betas and $95\%$CIs are given in S2 Table.
The 14 validated metabolites being associated with WHRadjfatmass in both sexes include different lipid fractions in very-large or large HDL particles, all being inversely related to WHRadjfatmass. In addition, very-large or large VLDL triglycerides were positively associated with WHRadjfatmass. All 14 validated metabolites were also associated with fat mass.
Only one validated metabolite was significantly associated with WHRadjfatmass in women only (Very-large HDL triglycerides, beta -2.4, SE 0.62, p-value = 8.5e10-5, with $$p \leq 0.36$$ in men, sex-interaction term $$p \leq 0.17$$).
Twenty-eight validated metabolites were associated with WHRadjfatmass in men only. They represent mainly large to small VLDL particles and medium HDL (inverse relationships vs WHRadjfatmass). None of those showed a sex-interaction at $p \leq 0.05$, but all 28 metabolites were associated with fat mass (Table 4).
**Table 4**
| Unnamed: 0 | WHR adjusted for fat mass | WHR adjusted for fat mass.1 | WHR adjusted for fat mass.2 | WHR adjusted for fat mass.3 | WHR adjusted for fat mass.4 | WHR adjusted for fat mass.5 | Fat mass | Fat mass.1 | Fat mass.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Women | Women | Women | Men | Men | Men | | | |
| Lipoprotein | Beta | SE | p-value | Beta | SE | p-value | Beta | SE | p-value |
| Medium HDL cholesterol esters | -1.18 | 0.61 | 0.054 | -3.19 | 0.85 | 0.0002 | -0.028 | 0.0046 | 1.2e-09 |
| Medium HDL cholesterol | -1.17 | 0.61 | 0.054 | -3.11 | 0.85 | 0.0002 | -0.028 | 0.0046 | 2.4e-09 |
| Small VLDL triglycerides | 1.52 | 0.67 | 0.024 | 2.74 | 0.85 | 0.0013 | 0.042 | 0.0048 | 4.1e-18 |
| Large VLDL phospholipids | 1.59 | 0.63 | 0.011 | 2.85 | 0.89 | 0.0014 | 0.040 | 0.0047 | 1.8e-17 |
| Small VLDL total concentration | 1.12 | 0.67 | 0.097 | 2.70 | 0.86 | 0.0017 | 0.041 | 0.0048 | 3.3e-17 |
| Small HDL triglycerides | 1.44 | 0.68 | 0.034 | 2.64 | 0.85 | 0.0019 | 0.043 | 0.0048 | 1.8e-19 |
| Medium HDL free cholesterol | -1.32 | 0.59 | 0.025 | -2.69 | 0.87 | 0.0021 | -0.023 | 0.0045 | 4.3e-07 |
| Medium VLDL triglycerides | 1.53 | 0.64 | 0.017 | 2.67 | 0.87 | 0.0022 | 0.041 | 0.0047 | 6.1e-18 |
| Small VLDL lipids | 0.91 | 0.66 | 0.170 | 2.60 | 0.85 | 0.0023 | 0.039 | 0.0048 | 2.4e-16 |
| Large VLDL total concentration | 1.54 | 0.62 | 0.013 | 2.62 | 0.88 | 0.0027 | 0.041 | 0.0047 | 3.2e-18 |
| Large VLDL lipids | 1.51 | 0.62 | 0.015 | 2.62 | 0.88 | 0.0028 | 0.041 | 0.0047 | 3.8e-18 |
| Very-large VLDL phospholipids | 1.31 | 0.59 | 0.026 | 2.61 | 0.88 | 0.0029 | 0.039 | 0.0046 | 1.3e-17 |
| Large VLDL free cholesterol | 1.50 | 0.63 | 0.018 | 2.55 | 0.86 | 0.0030 | 0.039 | 0.0047 | 2.4e-17 |
| Medium HDL total concentration | -1.16 | 0.60 | 0.052 | -2.65 | 0.89 | 0.0030 | -0.022 | 0.0046 | 2.8e-06 |
| Small VLDL phospholipids | 0.56 | 0.65 | 0.40 | 2.53 | 0.89 | 0.0043 | 0.035 | 0.0048 | 4.3e-13 |
| Large VLDL cholesterol | 1.19 | 0.63 | 0.061 | 2.47 | 0.87 | 0.0047 | 0.039 | 0.0047 | 3.0e-16 |
| Very-large VLDL lipids | 1.38 | 0.58 | 0.018 | 2.45 | 0.87 | 0.0048 | 0.040 | 0.0046 | 3.3e-18 |
| Very-large VLDL total concentration | 1.43 | 0.58 | 0.014 | 2.43 | 0.87 | 0.0052 | 0.039 | 0.0046 | 6.3e-18 |
| Large VLDL cholesterol esters | 1.01 | 0.63 | 0.110 | 2.42 | 0.88 | 0.0059 | 0.037 | 0.0048 | 5.2e-15 |
| Very-large VLDL free cholesterol | 1.34 | 0.59 | 0.024 | 2.36 | 0.86 | 0.0062 | 0.038 | 0.0046 | 2.8e-17 |
| Small VLDL cholesterol esters | 0.49 | 0.66 | 0.460 | 2.27 | 0.83 | 0.0064 | 0.037 | 0.0047 | 6.6e-15 |
| Very-large VLDL cholesterol | 1.12 | 0.60 | 0.060 | 2.33 | 0.86 | 0.0070 | 0.038 | 0.0046 | 1.4e-16 |
| Medium HDL lipids | -1.09 | 0.60 | 0.067 | -2.43 | 0.90 | 0.0071 | -0.016 | 0.0046 | 4.1e-04 |
| Very-large VLDL cholesterol esters | 0.99 | 0.60 | 0.10 | 2.30 | 0.86 | 0.0077 | 0.037 | 0.0046 | 1.6e-15 |
| Very small VLDL triglycerides | 1.06 | 0.69 | 0.13 | 2.38 | 0.93 | 0.010 | 0.037 | 0.0049 | 3.8e-14 |
| Small VLDL cholesterol | 0.42 | 0.65 | 0.52 | 2.13 | 0.84 | 0.011 | 0.035 | 0.0047 | 9.6e-14 |
| Medium VLDL total concentration | 0.67 | 0.64 | 0.30 | 2.08 | 0.89 | 0.019 | 0.034 | 0.0047 | 1.4e-12 |
| Medium VLDL phospholipids | 0.49 | 0.64 | 0.44 | 2.12 | 0.90 | 0.019 | 0.030 | 0.0048 | 4.2e-10 |
## Discussion
The present study using a discovery/validation approach in more than 3,000 individuals showed that several metabolites were related to fat distribution independent of fat mass. Some of these associations were observed among both sexes, while a significant interaction between WHRadjfatmass and sex were seen for many metabolites. Most of the WHRadjfatmass-associated metabolites were also related to fat mass. Of particular interest were two sphingomyelins being inversely related to WHRadjfatmass in both sexes, but not associated with fat mass as such. A certain lipoprotein profile was associated with a high WHRadjfatmass, but in this case, this profile was also associated with fat mass.
## Comparison with the literature
A great number of studies have investigated the metabolic profile of obesity, as reviewed in [7]. Some studies have also evaluated the metabolomics of fat distribution [8–15], but to the best of our knowledge, no other study have tried to link metabolites to WHR when taking fat mass into account. This approach aimed to identify metabolites linked to an altered fat distribution, independently of general obesity.
The study from the EPIC-Potsdam cohort [14], including Germans with almost exclusively European descent, studied the waist circumference (adjusted for hip circumference) and hip circumference (adjusted for waist circumference) and came to the conclusion that the metabolic profile for waist circumference was similar to that of BMI, but hip circumference showed a unique metabolomic profile. That study did however not evaluate WHR, but they showed isoleucine and several phosphatidylcholines (aa or ae C42:0, C34:3, C42:4, C42:5, C44:4 and C44:6) to show different directions in their relationships with waist and hip circumference. None of these metabolites were however related to WHRajdfatmass in our study.
In the other study based on three other Swedish samples using around 200 named metabolites [15], with almost exclusively European descent, one sphingomyelin (32:2) were amongst the metabolites found to be related to WHRadjBMI. In that study fat mass was not evaluated directly. In the present study, we found several sphingomyelin with a higher number of carbons to be linked to WHRadjfatmass. Neither could we find any other association being similar across the two studies. If that was due to the use of adjustment for fat mass in one study and BMI in the other is not known.
We identified two sphingomyelins of particular interest ((d18:$\frac{2}{24}$:1, d18:$\frac{1}{24}$:2) and (d18:$\frac{2}{24}$:2)) being inversely related to central adiposity in both sexes, but not being linked to fat mass. It is not likely that we would not be able to detect a relevant relationship between a metabolite and fat mass, since we have an $80\%$ power to detect a relationship with an R2 of 0.0071, and for those two sphingomyelins the relationships vs fat mass were far from being significant.
Sphingomyelin (SM) is a class of sphingolipids formed by adding phosphocholine to ceramide. As recently reviewed in [26], sphingolipids were initially thought to merely be structural components of the cell membrane, but recently a number of regulatory properties have been coupled to sphingolipids, such as cell growth and death, inflammation, angiogenesis and metabolism. Low levels of SMs have also recently been associated with cardiovascular diseases, such as stroke [27] and heart failure [28], as well as with the structure of the arterial wall [29]. Thus, certain sphingolipids might be a link between an altered fat distribution and cardiovascular disease. Unfortunately, no GWAS studies for these sphingomyelins with a 42-carbon chain length are available according to GWAS catalogue (https://www.ebi.ac.uk/gwas/) and the Helmholtz/KORA Metabolite GWAS server (http://metabolomics.helmholtz-muenchen.de/pgwas/), so Mendelian randomization studies cannot be carried out in order to investigate causality.
Why some of the SM were related to WHR and not others, and why two SMs were related to WHR and not fat mass, while other were related to both WHR and fat mass is not known. One explanation for this finding is that only slight changes in the SM molecule could have profound effects on the physical properties. This is exemplified in a paper on lipidomics and obesity conducted in Sweden with almost exclusively European descent, in which it was found that SM 34:1;2 has the greatest negative and SM 34:2;2 the greatest positive estimates vs fat mass [10]. Thus, only a change in a double bond could result in different signs of the association vs fat mass. In another study, high levels of serum SM species with distinct saturated acyl chains (C18:0, C20:0, C22:0 and C24:0) closely correlate with the parameters of obesity in young adult Japanese individuals [30], but this is not a universal finding, as could be seen in a study sample collected in Iran [31]. Thus, it is likely that all SM should not be regarded as equal, although we do not have the knowledge today to understand how the chemical properties of different SMs translates to relationships vs fat mass and fat distribution.
Another question arises if SMs are causally related to fat mass and fat distribution. An experimental study using knock-out of the enzyme involved in SM synthesis (SMS2) in mice supports a role of SMs in obesity [32]. Mice with SMS2 deficiency developed obesity when challenged with a high-fat diet. However, also the other direction of the relationship between SMs and fat mass might exist. In a study of weight loss induced by a low-caloric diet conducted in Spain and Denmark, the change in fat mass over 8 weeks was inversely related to the change in certain SMs (33:1, 35:1, 36:0, and 36:1), while the change in fat mass was directly related to the change in other SMs (32:1, 32:2, 38:1, 40:1 and 41:1) [33].
Apart from the 9 metabolites linked to WHRadjfatmass in both sexes, we identified a number of metabolites being related to WHRadjfatmass in one sex only. This is not surprising since WHR is very different in men and women and that most genetic correlates to WHR also are sex-specific. Both the male-specific and the female-specific metabolites being related to WHR comes from several chemical classes and it is hard to see any clear pattern in these two metabolomic profiles. Again, genetic studies might be a way forward to disentangle these sex-specific metabolomic profiles in the future.
As with the MS-based metabolomics analysis, the analysis of the lipoprotein profile showed several validated metabolites to be linked to a high WHRadjfatmass. It was mainly very-large and large HDL-cholesterol (inverse) and large VLDL-triglycerides that were related to WHRadjfatmass in both sexes. In the sex-stratified analyses, associations with WHRadjfatmass were much more common in males than in females, with predominantly small VLDL being sex-specific. Very-large and large HDL-cholesterol (inverse) and large VLDL-triglycerides have been linked to cardiovascular disease [16], but these lipoprotein fractions were also related to fat mass, not only to fat distribution.
In the relationship between WHRadjfatmass and metabolites, it is likely that most relationships have a causal direction from an unfavorable fat distribution to a change in metabolites rather than the opposite, since a major weight change induces profound metabolic effects [33, 34]. Since we also know that WHR is a major risk factor for CVD both in observational [35] and genetic studies [3], it would be of interest to find metabolites being mediators in the causal WHR->CVD relationship. Regarding the 14 lipoprotein measurements identified (Fig 2) to be linked to WHRadjfatmass in both men and women, a mediating role for those lipoprotein-based metabolites are plausible, since they all have been linked to atherosclerosis and incident CVD [16] with the same direction of associations as found vs WHRadjfatmass in the present study. As discussed above, the role of the 9 MS-based metabolites (Table 1) are less clear, since the knowledge on sphingomyelins and GPCs in CVD are much less advanced than the knowledge on lipoproteins in CVD. We are still awaiting to obtain robust, powerful genetic instruments for different sphingomyelins and GPCs that are not pleotropic to be used in Mendelian randomization studies. Also mice models with knock-out of different sphingomyelins and GPCs on a genetic atherosclerotic background would be a way forward to disentangle if sphingomyelins and GPCs are important players in atherosclerosis formation.
It is also be emphasized that the MS-based part of the study should be regarded as an untargeted approach, since a large number of metabolites from a great number of chemical classes were analyzed, and therefore the evaluation of the WHRadjfatmass vs metabolite associations were hypothesis-free, like in a GWAS study. It is therefore not surprising that many of the findings were not expected and given the unperfect knowledge of many of the metabolites, the findings are hard to understand in detail at this stage.
The strength of the present study is the use of the same MS-based metabolomic platform and lipoprotein NMR analysis with a large number of metabolites and lipoproteins subfractions in three different studies, so we could obtain validated relationships.
One limitation is the cross-sectional nature of the studies, which hinders any conclusion of causal directions. Another limitation is the fact that we do not have genetic instruments for the metabolites of interest to be used in Mendelian randomization studies. Metabolomic measurements were performed by a commercial company. Some of the details regarding standards and QC determinations were considered proprietary to their platform and therefore not shared. How this translates into the findings in the present study is unknown, but as a general rule any poor performance of a technique would only increase the probability of the null hypothesis and would not produce any false positive findings.
We used bioimpedance to evaluate fat mass in the present study, a technique that has been present for decades. However, DEXA is the gold standard in this respect, but was only used in one of the samples, the POEM study, being the smallest of the three cohorts used. Therefore, we did not use DEXA due to the limited power and lack of replication sample. However, when we related bioimpedance to DEXA measurements for fat mass, the correlation was very good (correlation coefficient 0.93), so we regard the bioimpedance measurement of fat mass to be valid, in accordance with previous evaluations of the bioimpedance technique [22, 23].
DEXA has the advantage over bioimpedance that regional fat distribution could be assessed in detail. In the absence of DEXA, we used an indirect measurement of regional fat distribution, the WHR, a measure that is a better predictor of myocardial infarction than measurements of general obesity, such as BMI [35]. WHR has the advantage over DEXA of being a cheap and easy measurement and is widely used in the clinic, but it would be of great interest to validate our present findings in a sample with DEXA measurements.
The samples used in the present studies are almost exclusively including subjects with a European descent. In order to generalize the findings to other populations, replication studies in subjects from other parts of the world has to be undertaken, especially since most other studies in this field also have been conducted in subjects with a European descent. It is also emphasized that the MS-based part of the study should be regarded as an untargeted approach, since a large number of metabolites from a great number of chemical classes were analyzed, and therefore the evaluation of the WHRadjfatmass vs metabolite associations were hypothesis-free, like in a GWAS study. It is therefore not surprising that many of the findings were not expected and given the imperfect knowledge of many of the metabolites, the findings are hard to understand in detail at this stage. Regarding the lipoprotein-based metabolites, they were preselected to cover the lipoprotein spectra and therefore this analysis should be considered as targeted and due to the greater knowledge on lipoproteins, these results could be interpreted more in detail.
Thyroid function might have been a confounder in the present evaluation of the metabolomic profile of WHR, but unfortunately we do not have valid measurements of thyroid function in the samples.
In conclusion, two sphingomyelins were inversely linked to WHR (fat mass adjusted) in both men and women without being related to fat mass, while very-large or large HDL particles were inversely related to WHR as well as to fat mass. If these sphingomyelins represent a link between WHR and cardiometabolic diseases remains to be established when genetic instruments might become available in the future.
## References
1. Fall T, Hägg S, Mägi R, Ploner A, Fischer K, Horikoshi M. **The role of adiposity in cardiometabolic traits: a Mendelian randomization analysis**. *PLoS Med* (2013) **10** e1001474. DOI: 10.1371/journal.pmed.1001474
2. Hägg S, Fall T, Ploner A, Mägi R, Fischer K, Draisma HH. **Adiposity as a cause of cardiovascular disease: a Mendelian randomization study**. *Int J Epidemiol* (2015) **44** 578-86. DOI: 10.1093/ije/dyv094
3. Emdin CA, Khera AV, Natarajan P, Klarin D, Zekavat SM, Hsiao AJ. **Genetic Association of Waist-to-Hip Ratio With Cardiometabolic Traits, Type 2 Diabetes, and Coronary Heart Disease**. *JAMA* (2017) **317** 626-34. DOI: 10.1001/jama.2016.21042
4. Turcot V, Lu Y, Highland HM, Schurmann C, Justice AE, Fine RS. **Protein-altering variants associated with body mass index implicate pathways that control energy intake and expenditure in obesity**. *Nat Genet* (2018) **50** 26-41. DOI: 10.1038/s41588-017-0011-x
5. Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J. **Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry**. *Hum Mol Genet* (2019) **28** 166-74. PMID: 30239722
6. Fall T, Mendelson M, Speliotes EK. **Recent Advances in Human Genetics and Epigenetics of Adiposity: Pathway to Precision Medicine?**. *Gastroenterology* (2017) **152** 1695-706. DOI: 10.1053/j.gastro.2017.01.054
7. Rangel-Huerta OD, Pastor-Villaescusa B, Gil A. **Are we close to defining a metabolomic signature of human obesity? A systematic review of metabolomics studies**. *Metabolomics* (2019) **15** 93. DOI: 10.1007/s11306-019-1553-y
8. Beyene HB, Olshansky G, Giles C, Huynh K, Cinel M, Mellett NA. **Lipidomic Signatures of Changes in Adiposity: A Large Prospective Study of 5849 Adults from the Australian Diabetes, Obesity and Lifestyle Study**. *Metabolites* (2021) **11** 646. DOI: 10.3390/metabo11090646
9. Bogl LH, Kaye SM, Rämö JT, Kangas AJ, Soininen P, Hakkarainen A. **Abdominal obesity and circulating metabolites: A twin study approach**. *Metabolism* (2016) **65** 111-21. DOI: 10.1016/j.metabol.2015.10.027
10. Gerl MJ, Klose C, Surma MA, Fernandez C, Melander O, Mannisto S. **Machine learning of human plasma lipidomes for obesity estimation in a large population cohort**. *PLoS Biol* (2019) **17** e3000443. DOI: 10.1371/journal.pbio.3000443
11. Stevens VL, Carter BD, McCullough ML, Campbell PT, Wang Y. **Metabolomic Profiles Associated with BMI, Waist Circumference, and Diabetes and Inflammation Biomarkers in Women**. *Obesity (Silver Spring)* (2020) **28** 187-96. DOI: 10.1002/oby.22670
12. Wulaningsih W, Proitsi P, Wong A, Kuh D, Hardy R. **Metabolomic correlates of central adiposity and earlier-life body mass index**. *J Lipid Res* (2019) **60** 1136-43. DOI: 10.1194/jlr.P085944
13. Xu K, Shi L, Zhang B, Mi B, Yang J, Sun X. **Distinct metabolite profiles of adiposity indices and their relationships with habitual diet in young adults**. *Nutr Metab Cardiovasc Dis* (2021) **31** 2122-30. DOI: 10.1016/j.numecd.2021.03.025
14. Bachlechner U, Floegel A, Steffen A, Prehn C, Adamski J, Pischon T. **Associations of anthropometric markers with serum metabolites using a targeted metabolomics approach: results of the EPIC-potsdam study**. *Nutr Diabetes* (2016) **6** e215. DOI: 10.1038/nutd.2016.23
15. Ahmad S, Hammar U, Kennedy B, Salihovic S, Ganna A, Lind L. **Effect of General Adiposity and Central Body Fat Distribution on the Circulating Metabolome: A Multicohort Nontargeted Metabolomics Observational and Mendelian Randomization Study**. *Diabetes* (2022) **71** 329-39. DOI: 10.2337/db20-1120
16. Holmes MV, Millwood IY, Kartsonaki C, Hill MR, Bennett DA, Boxall R. **Lipids, Lipoproteins, and Metabolites and Risk of Myocardial Infarction and Stroke**. *J Am Coll Cardiol* (2018) **71** 620-32. PMID: 29420958
17. Romero-Corral A, Somers VK, Sierra-Johnson J, Korenfeld Y, Boarin S, Korinek J. **Normal weight obesity: a risk factor for cardiometabolic dysregulation and cardiovascular mortality**. *Eur Heart J* (2010) **31** 737-46. DOI: 10.1093/eurheartj/ehp487
18. Padwal R, Leslie WD, Lix LM, Majumdar SR. **Relationship Among Body Fat Percentage, Body Mass Index, and All-Cause Mortality: A Cohort Study**. *Ann Intern Med* (2016) **164** 532-41. DOI: 10.7326/M15-1181
19. Lind L, Elmståhl S, Bergman E, Englund M, Lindberg E, Michaelsson K. **EpiHealth: a large population-based cohort study for investigation of gene-lifestyle interactions in the pathogenesis of common diseases**. *Eur J Epidemiol* (2013) **28** 189-97. DOI: 10.1007/s10654-013-9787-x
20. Lind L.. **Impaired glucose control is associated with multiple cardiovascular impairments**. *Clin Physiol Funct Imaging* (2020) **40** 257-68. DOI: 10.1111/cpf.12634
21. Lind L, Fors N, Hall J, Marttala K, Stenborg A. **A comparison of three different methods to evaluate endothelium-dependent vasodilation in the elderly: the Prospective Investigation of the Vasculature in Uppsala Seniors (PIVUS) study**. *Arterioscler Thromb Vasc Biol* (2005) **25** 2368-75. DOI: 10.1161/01.ATV.0000184769.22061.da
22. Steen B, Bosaeus I, Elmståhl S, Galvard H, Isaksson B, Robertsson E. **Body composition in the elderly estimated with an electrical impedance method**. *Compr Gerontol A* (1987) **1** 102-5. PMID: 3453289
23. Ling CH, de Craen AJ, Slagboom PE, Gunn DA, Stokkel MP, Westendorp RG. **Accuracy of direct segmental multi-frequency bioimpedance analysis in the assessment of total body and segmental body composition in middle-aged adult population**. *Clin Nutr* (2011) **30** 610-5. DOI: 10.1016/j.clnu.2011.04.001
24. Ford L, Kennedy AD, Goodman KD, Pappan KL, Evans AM, Miller LAD. **Precision of a Clinical Metabolomics Profiling Platform for Use in the Identification of Inborn Errors of Metabolism**. *J Appl Lab Med* (2020) **5** 342-56. DOI: 10.1093/jalm/jfz026
25. Wurtz P, Kangas AJ, Soininen P, Lawlor DA, Davey Smith G, Ala-Korpela M. **Quantitative Serum Nuclear Magnetic Resonance Metabolomics in Large-Scale Epidemiology: A Primer on -Omic Technologies**. *Am J Epidemiol* (2017) **186** 1084-96. DOI: 10.1093/aje/kwx016
26. Insausti-Urkia N, Solsona-Vilarrasa E, Garcia-Ruiz C, Fernandez-Checa JC. **Sphingomyelinases and Liver Diseases**. *Biomolecules* (2020) **10** 1497. DOI: 10.3390/biom10111497
27. Lind L, Salihovic S, Ganna A, Sundström J, Broeckling CD, Magnusson PK. **A Multi-Cohort Metabolomics Analysis Discloses Sphingomyelin (32:1) Levels to be Inversely Related to Incident Ischemic Stroke**. *J Stroke Cerebrovasc Dis* (2020) **29** 104476. DOI: 10.1016/j.jstrokecerebrovasdis.2019.104476
28. Stenemo M, Ganna A, Salihovic S, Nowak C, Sundstrom J, Giedraitis V. **The metabolites urobilin and sphingomyelin (30:1) are associated with incident heart failure in the general population**. *ESC Heart Fail* (2019) **6** 764-73. DOI: 10.1002/ehf2.12453
29. Lind L.. **The metabolomic profile of carotid artery intima-media thickness and echogenicity**. *Atherosclerosis* (2021) **335** 142-7. DOI: 10.1016/j.atherosclerosis.2021.09.011
30. Hanamatsu H, Ohnishi S, Sakai S, Yuyama K, Mitsutake S, Takeda H. **Altered levels of serum sphingomyelin and ceramide containing distinct acyl chains in young obese adults**. *Nutr Diabetes* (2014) **4** e141. DOI: 10.1038/nutd.2014.38
31. Bagheri M, Djazayery A, Farzadfar F, Qi L, Yekaninejad MS, Aslibekyan S. **Plasma metabolomic profiling of amino acids and polar lipids in Iranian obese adults**. *Lipids Health Dis* (2019) **18** 94. DOI: 10.1186/s12944-019-1037-0
32. Mitsutake S, Zama K, Yokota H, Yoshida T, Tanaka M, Mitsui M. **Dynamic modification of sphingomyelin in lipid microdomains controls development of obesity, fatty liver, and type 2 diabetes**. *J Biol Chem* (2011) **286** 28544-55. DOI: 10.1074/jbc.M111.255646
33. Papandreou C, Garcia-Gavilan J, Camacho-Barcia L, Toft Hansen T, Harrold JA, Sjodin A. **Changes in Circulating Metabolites During Weight Loss are Associated with Adiposity Improvement, and Body Weight and Adiposity Regain During Weight Loss Maintenance**. *The SATIN Study. Mol Nutr Food Res* (2021) **65** e2001154. PMID: 34184401
34. Palau-Rodriguez M, Tulipani S, Marco-Ramell A, Miñarro A, Jauregui O, Gonzalez-Dominguez R. **Characterization of Metabolomic Profile Associated with Metabolic Improvement after Bariatric Surgery in Subjects with Morbid Obesity**. *J Proteome Res* (2018) **17** 2704-14. DOI: 10.1021/acs.jproteome.8b00144
35. Yusuf S, Hawken S, Ounpuu S, Dans T, Avezum A, Lanas F. **Effect of potentially modifiable risk factors associated with myocardial infarction in 52 countries (the INTERHEART study): case-control study**. *Lancet* (2004) **364** 937-52. DOI: 10.1016/S0140-6736(04)17018-9
|
---
title: Application of the skills network approach to measure physician competence
in shared decision making based on self-assessment
authors:
- Levente Kriston
- Lea Schumacher
- Pola Hahlweg
- Martin Härter
- Isabelle Scholl
journal: PLOS ONE
year: 2023
pmcid: PMC9970074
doi: 10.1371/journal.pone.0282283
license: CC BY 4.0
---
# Application of the skills network approach to measure physician competence in shared decision making based on self-assessment
## Abstract
Several approaches to and definitions of ‘shared decision making’ (SDM) exist, which makes measurement challenging. Recently, a skills network approach was proposed, which conceptualizes SDM competence as an organized network of interacting SDM skills. With this approach, it was possible to accurately predict observer-rated SDM competence of physicians from the patients’ assessments of the physician’s SDM skills. The aim of this study was to assess whether using the skills network approach allows to predict observer-rated SDM competence of physicians from their self-reported SDM skills. We conducted a secondary data analysis of an observational study, in which outpatient care physicians rated their use of SDM skills with the physician version of the 9-item Shared Decision Making Questionnaire (SDM-Q-Doc) during consultations with chronically ill adult patients. Based on the estimated association of each skill with all other skills, an SDM skills network for each physician was constructed. Network parameters were used to predict observer-rated SDM competence, which was determined from audio-recorded consultations using three widely used measures (OPTION-12, OPTION-5, Four Habits Coding Scheme). In our study, 28 physicians rated consultations with 308 patients. The skill ‘deliberating the decision’ was central in the population skills network averaged across physicians. The correlation between parameters of the skills networks and observer-rated competence ranged from 0.65 to 0.82 across analyses. The use and connectedness of the skill ‘eliciting treatment preference of the patient’ showed the strongest unique association with observer-rated competence. Thus, we found evidence that processing SDM skill ratings from the physicians’ perspective according to the skills network approach offers new theoretically and empirically grounded opportunities for the assessment of SDM competence. A feasible and robust measurement of SDM competence is essential for research on SDM and can be applied for evaluating SDM competence during medical education, for training evaluation, and for quality management purposes. [ A plain language summary of the study is available at https://osf.io/3wy4v.]
## Introduction
An essential patient-centered communication competence in health care delivery is the ability to support shared decision making (SDM) in medical consultations. SDM is frequently described as an interpersonal decision making process with a strong emphasis on a balanced flow and exchange of information, values, preferences, power, and responsibility between the patient and the health care professional during medical consultations [1, 2]. SDM has been considered ethical in medical consultations because it ensures that patients are informed about various treatment options and that the patients’ preferences are valued in medical decision-making [3]. This seems particularly important considering that physicians´ assumptions on their patients’ preferences often mismatch the patients’ actual preferences, and patients tend to choose different treatment options when they are better informed [4]. Further, SDM may help to reduce the use of inappropriate tests and interventions when benefits and drawbacks of these are clearly discussed [5] and could lead to better medication adherence [6]. Finally, as patients tend to choose more conservative options when asked, SDM might even reduce health care costs [7]. Thus, it is not surprising that major health care organizations have adopted the principles of SDM [8–10].
Although several definitions of SDM exist, it is rarely acknowledged explicitly that the same term can refer to ontologically very different concepts [11]. It is frequently unclear, whether ‘SDM’ is used to denote observable attributes of the communication process in a medical encounter, the perception of these attributes by the patient or the physician, attitudes of the participating individuals, a specific method or technique which physicians can utilize, a general philosophy of shaping health care, or a scientific model of medical communication. Conceptual clarity is indispensable for the measurement of latent constructs [12]. From a competence-focused perspective, SDM competence can be defined as the physician’s ability to use specific behavioral skills in a way which supports building a consensus with the patient regarding the favored treatment among multiple viable options in accordance with the patient’s preferences and values [13]. According to this approach, SDM competence requires physicians to organize a defined set of behavioral skills into a certain pattern or network to make a patient-centered decision in the medical consultation more likely.
In a recent study, we found that modelling SDM competence as a network of skills can be used to predict physicians’ observer-rated SDM competence [13]. In that study, patients rated the degree to which certain SDM-related skills were shown by the physicians in their routine medical consultations. These ratings were used to create an SDM skills network for each physician, which models how individual SDM skills are related to each other. Attributes of these networks, e.g., how strongly a skill was related to other skills, predicted observer-rated competence with high accuracy. Using this approach with skill rating input from other sources than patients would substantiate the validity of conceptualizing SDM competence as an organized network of behavioral skills. In the present study, we investigated whether processing physician-reported data on their SDM skills according to the skills network approach can be used to predict observer-rated SDM competence.
## Design and procedures
The design of the present study was based on a previous investigation [13]. We re-analyzed data from a study on measuring SDM, collected between August 2009 and September 2010 in Hamburg, Germany [14]. In that study, consultations between adult patients with chronic conditions who faced a treatment decision and physicians providing primary and specialty outpatient care were examined using ratings from patients, physicians, and external observers. The investigators aimed to include thirty physicians with written documentation of ten consultations and audio-recordings of three consultations each. The ethics committee of the state chamber of physicians in Hamburg approved the study protocol (record no. PV3180). All participants provided written informed consent. In the present analysis, we used data from the physicians and the external observers.
## Measures
Basic demographic and clinical data on the participating patients and physicians were collected by administering written questionnaires.
Physician-reported data on SDM skills were collected with the physician version of the 9-item Shared Decision Making Questionnaire (SDM-Q-Doc), which was filled out after the respective consultations [15]. This measure requires physicians to rate the degree to which they showed nine behaviors in the consultation using a six-step Likert-type scale ranging from zero to five. The behaviors captured by the SDM-Q-Doc correspond to key SDM skills: focusing the decision, sharing the decision, presenting options, informing on options, supporting comprehension, eliciting preferences, deliberating the decision, selecting an option, and planning actions [13, 15].
Observer-rated SDM competence of the physicians was measured by three widely used validated measures, the OPTION-12 [16, 17], the OPTION-5 [18, 19], and the Invest in the End subscale of the Four Habits Coding Scheme (4HCS) [20, 21], based on the audio-recorded consultations. We decided to include all three measures in the present analysis, because they capture SDM competence from different perspectives. The OPTION measures focus on decision making, while the 4HCS assesses primarily communication. We decided to include the OPTION-5 in addition to the OPTION-12, because it has a stronger focus on patient preferences and is based on a revised model of SDM [18]. As shown also by empirical analysis, the OPTION-12, the OPTION-5, and the Invest in the End subscale of the 4HCS capture overlapping but notably distinct constructs [13].
Two independent raters assessed each consultation using pilot sessions and manuals to achieve sufficient agreement. Inter-rater reliability varied between 0.69 to 0.76 across instruments for averaged ratings of the physicians’ SDM competence, showing substantial agreement between raters [13]. Raters were blinded to the results of the assessment with other measures. For analysis, we transformed all scores to range from 0 to 100, with higher values indicating a higher level of SDM competence. Each of the measures was averaged across consultations in order to obtain three observer-rated SDM competence scores for each physician. Validity of this method of estimating competence was supported by substantial physician-level variance of the three scores and moderate to high physician-level correlation between them [13]. This means that while SDM competence considerably varied between different physicians, the three observer-rated measures indicated similar SDM competence for each individual physician. An overview on the design, measures and analysis is displayed in Fig 1.
**Fig 1:** *Overview of the research design, measures and analysis.SDM, Shared Decision Making; SDM-Q-Doc, Shared Decision Making Questionnaire—physician version; 4HCS, Invest in the End subscale of the Four Habits Coding Scheme.*
## Statistical analysis
The physicians’ self-rated data on their SDM skills were analyzed according to the skills network model of competence [13]. We assessed the associations between the nine SDM skills and constructed a skills network for each physician. These networks display individual SDM skills as nodes. The connections between nodes are called edges, which indicate how strongly individual SDM skills are related to each other.
For each SDM skill, a Bayesian multilevel linear regression was estimated with the skill as the outcome variable and all other skills as predictors, based on the physician-rated data from all consultations of all physicians. The intercept and slopes were allowed to vary between physicians yielding estimates for each individual physician. Thus, the strength of the associations between individual SDM skills was expected to vary across physicians. Bayesian analysis requires the definition of a prior distribution for each estimated parameter. This prior distribution is updated during the analysis by combining it with the observed data to obtain a posterior distribution, which informs on how probable certain values of the estimated parameter are. We used weakly informative priors, reflecting that we had an approximate but not exact idea of the expected size of the statistical parameters before calculation (see S1 File). If more than two of the nine skills were missing for a consultation, data points from that consultation were excluded. One or two missing ratings per consultations were imputed using the expectation-maximization algorithm.
Based on the estimated coefficients from the multilevel regression models described above, a skills network was constructed for each physician. The regression estimates describing the direction and strength of the association between the different skills for each physician were used as edge weights. When the $95\%$ credible interval of a regression estimate included a zero, this association was excluded to avoid spurious associations. Nodes in the skills networks were placed using the Fruchterman-Reingold algorithm, thus, as far as possible in two dimensions, their distance is relative to the strength of their association [22]. Consequently, skills that were strongly related were placed closer to each other in the networks. Three network parameters, namely activation, outstrength and instrength, of each skill for each physician were calculated. Activation of a skill was defined as the mean of that skill across consultations, i.e., how strongly each physician indicated to have used the skill across their consultations. Outstrength of a skill was calculated by summing the weights of the outgoing edges of that skill and indicates how strongly a skill influences other skills. Instrength was calculated by summing the weights of the ingoing edges of that skill, showing how strongly that skill is influenced by other skills. In addition to the physician-specific networks, we created a population network through averaging the network parameters across physicians. Thus, in addition to constructing a network for each physician, a population network showing how skills are related on average across all physicians was also created. A more detailed description including a step-by-step instruction for calculations can be found elsewhere [13].
Finally, we performed Bayesian linear regression analyses to test whether the network parameters of each physician can predict observer-rated SDM competence as measured with the OPTION-5, OPTION-12 and the Invest in the End subscale of the 4HCS. By doing so, we tested whether characteristics of the skills networks predicted the SDM competence of individual physicians as rated by external observers. First, a confirmatory model with the activation, outstrength and instrength of the skills ‘focusing the decision’, ‘eliciting preferences’ and ‘deliberating the decision’ as predictors was tested, since these skills were relevant in the previous analysis with patient-rated data [13]. We used informative priors with means and standard deviations estimated from the posterior distribution of the estimates observed in the analysis of the patient-reported data (see S1 File) [13]. Subsequently, we created an exploratory model to investigate whether ignoring previous results changes the conclusions substantively. For this, three Bayesian linear regression models were fitted for predicting each observer-rated measure of SDM competence with the activation, the instrength, and the outstrength of all skills as predictors, respectively. The network parameters of the skills, which were significant predictors in this first step for at least one of the observer-rated measures, were regressed onto the three observer-rated measures in the final exploratory model. Weakly informative priors were chosen for all exploratory analyses (see S1 File).
All analyses were conducted in R version 4.0.4 [23]. Bayesian (multilevel) regression analyses were conducted with the package brms utilizing Markov chain Monte Carlo sampling methods [24]. Networks were plotted using qgraph [25]. All regression models were run with four chains, a total of 20,000 iterations, a thinning rate of 10, and 12,000 burn-in simulations, resulting in a posterior sample of 2,000. For each model, the Gelman-Rubin potential scale reduction statistic [26] and traceplots were checked for convergence. We labeled a regression coefficient as statically significant when its $95\%$ credible interval did not include zero. The R code of all analyses is available at https://osf.io/z$\frac{7368}{.}$
## Sample
In the original study, 33 physicians agreed to participate [14], of which 28 provided self-assessment of their SDM skills in 326 consultations. Ratings of 18 consultations were excluded as they had more than two missing data points, resulting in data from 308 consultations included in the analyses (on average 11 consultations per physician). Audio recordings were available from 24 physicians and 80 consultations (on average 3.3 consultations per physician).
Over 70 percent of the participating physicians (42.9 percent female, mean age 50.4 years) were specialized in family or internal medicine and less than one in four had 20 years or more experience (Table 1). The majority of the patients in the investigated consultations (60.3 percent female, mean age 54.2 years) were married, had a low to medium formal education, and were employed or retired (Table 2). About one third of the patients were diagnosed with type 2 diabetes, chronic back pain, and depressive disorder, respectively. The subsample of the physicians and patients contributing audio-recorded consultations were comparable to the total sample.
## Population network of SDM skills
The average skills network (Fig 2) showed that the skills ‘focusing the decision’ and ‘sharing the decision’ were, despite their strong reciprocal association, disconnected from the remaining network, suggesting that these skills were only related to each other. ‘ Presenting options’, ‘informing on options’, ‘eliciting preferences’ and ‘selecting an option’ were strongly connected, with ‘deliberating the decision’ being in the center of this skill cluster, showing a high level of interrelatedness between these skills. The skills ‘supporting comprehension’ and ‘planning actions’ were more peripheral in the skills network, as they were only related to ‘informing on options’ and ‘selecting an option’, respectively.
**Fig 2:** *Average skills network across physicians.The width of the arrows represents the strength of the skills associations. The pie around each node indicates the extent of activation of each item. The labels refer to the following skills: 1. focusing the decision; 2. sharing the decision; 3. presenting options; 4. informing on options; 5. supporting comprehension; 6. eliciting preferences; 7. deliberating the decision; 8. selecting an option; 9. planning actions.*
On average, the skill ‘planning actions’ were most frequently shown (Fig 3, panel A). ‘ Presenting options’ had the strongest influence on other skills (Fig 3, panel B), and ´informing on options’ was most strongly influenced by other skills (Fig 4, panel C). There was considerable variation between the physicians in their network structure and network parameters (Fig 3; skills networks of individual physicians can be seen in S1 Fig). Thus, how skills were related to each other differed between physicians.
**Fig 3:** *Network parameters of the investigated skills.Black dots represent the average score, and grey dots indicate estimates from each physician network. The labels refer to the following skills: 1. focusing the decision; 2. sharing the decision; 3. presenting options; 4. informing on options; 5. supporting comprehension; 6. eliciting preferences; 7. deliberating the decision; 8. selecting an option; 9. planning actions.* **Fig 4:** *Calibration plots for the confirmatory and exploratory prediction of observer-rated SDM competence.Panels A, B and C show predicted and observed scores for the confirmatory model, panels D, E and F for the exploratory model. Black dots represent the physicians’ scores; smoothing (loess) curves are displayed for each outcome by the grey line. 4HCS, Four Habits Coding Scheme.*
## Confirmatory prediction of observed SDM competence from skills networks
The skill ‘eliciting preference’ played an important role in the prediction of observer-rated SDM competence in the confirmatory model, as its activation was significantly positively related to SDM competence ratings with the OPTION-12 and the OPTION-5 and its outstrength was significantly positively related to the SDM competence rating with the 4HCS (Table 3). This means that how often this skill was used and how strongly it was associated with other skills could predict observer-rated SDM competence. Further, the outstrength of ‘deliberating the decision’ was significantly negatively associated with SDM competence as measured with the OPTION-12. This indicated that when a physician’s network showed that ‘deliberating the decision’ influenced many other skills, the SDM competence of that physician was rated lower. The confirmatory model explained about half of the variance of the observer-rated SDM competence with correlations between predicted and observed values ranging from 0.65 to 0.75. Thus, skills network characteristics explained a considerable amount of variation in the observer-rated SDM competence of physicians. Predicted and observed values of the confirmatory models are depicted in Fig 4, panels A-C.
**Table 3**
| Unnamed: 0 | OPTION-12(n = 22) | OPTION-5(n = 24) | 4HCS(n = 22) |
| --- | --- | --- | --- |
| | Estimate [95% CI] | Estimate [95% CI] | Estimate [95% CI] |
| Intercept | 15.96 [14.67 to17.23] | 11.96 [10.31 to 13.51] | 33.07 [32.02 to 34.10] |
| Activation | | | |
| Skill 1 | 4.18 [-1.84 to 9.81] | 2.16 [-4.67 to 9.05] | 1.51 [-3.72 to 6.61] |
| Skill 6 | 43.44 [0.32 to 86.45]* | 65.62 [9.98 to 119.88]* | 18.84 [-14.33 to 51.00] |
| Skill 7 | 2.51 [-14.64 to 20.22] | 7.90 [-11.54 to 27.26] | 7.96 [-6.18 to 23.66] |
| Instrength | | | |
| Skill 1 | 2.91 [-5.92 to 11.71] | -3.47 [-13.35 to 6.75] | 4.46 [-2.52 to 11.04] |
| Skill 6 | 2.32 [-4.47 to 8.83] | 7.42 [-0.27 to 15.75] | 1.38 [-3.57 to 6.44] |
| Skill 7 | 1.43 [-4.37 to 7.10] | 1.89 [-4.98 to 9.07] | -1.69 [-6.73 to 2.77] |
| Outstrength | | | |
| Skill 1 | -7.15 [-16.63 to 2.64] | -5.78 [-19.25 to 7.83] | -5.38 [-12.18 to 1.20] |
| Skill 6 | 3.16 [-2.85 to 9.05] | 0.32 [-6.97 to 7.43] | 4.70 [0.12 to 9.70]* |
| Skill 7 | -6.30 [-11.57 to -1.00]* | -4.77 [-11.20 to 1.98] | -1.40 [-5.62 to 2.77] |
| R | 0.750 [0.620 to 0.812] | 0.750 [0.619 to 0.819] | 0.653 [0.504 to 0.732] |
| R2 | 0.563 [0.385 to 0.660] | 0.562 [0.383 to 0.670] | 0.427 [0.254 to 0.536] |
## Exploratory prediction of observed SDM competence from skills networks
When the activation, instrength and outstrength of all skills were regressed on the observer-rated SDM competence, the skills ´focusing on the decision´, ‘presenting option’, ‘informing on options’ and ‘eliciting preferences’ were significantly related to at least one of the three observer measures (S1–S3 Tables). Results from the subsequent analysis, which included the activation, instrength and outstrength of these four skills, are reported in Table 4. Only the activation of ‘eliciting preference’ was significantly related to SDM competence as measured by the OPTION-5. Still, the model explained about half of the variance for each of the observer measures, with multiple correlation coefficients ranging from 0.69 to 0.82 (Table 4). Predicted and observed values of the exploratory models are displayed in Fig 4, Panels D-F.
**Table 4**
| Unnamed: 0 | OPTION-12(n = 22) | OPTION-5(n = 24) | 4HCS(n = 22) |
| --- | --- | --- | --- |
| | Estimate [95% CI] | Estimate [95% CI] | Estimate [95% CI] |
| Intercept | 15.84 [13.50 to 17.86] | 11.77 [9.84 to 13.68] | 33.07 [31.04–34.98] |
| Activation | | | |
| Skill 1 | 2.10 [-7.02 to 10.96] | 1.11 [-7.25 to 8.79] | 0.50 [-7.31–8.09] |
| Skill 3 | 12.48 [-27.76 to 52.85] | 22.21 [-11.11 to 52.92] | 1.42 [-33.00–34.40] |
| Skill 4 | 2.34 [-33.13 to 35.67] | 15.47 [-15.52 to 42.52] | -4.13 [-33.19–26.41] |
| Skill 6 | 18.76 [-89.19 to 121.36] | 100.09 [2.61 to 189.00]* | -17.78 [-108.27–72.45] |
| Instrength | | | |
| Skill 1 | 4.93 [-18.16 to 28.10] | -3.78 [-24.06 to 16.69] | 13.98 [-6.47–33.00] |
| Skill 3 | 5.39 [-8.39 to 19.41] | -2.69 [-12.15 to 7.10] | 4.52 [-6.90–16.28] |
| Skill 4 | -3.57 [-16.50 to 9.43] | -4.01 [-13.10 to 5.69] | -1.17 [-12.84–9.89] |
| Skill 6 | 1.59 [-8.31 to 11.51] | -1.65 [-9.80 to 7.12] | 3.75 [-4.71–12.66] |
| Outstrength | | | |
| Skill 1 | -19.27 [-61.96 to 23.77] | -8.06 [-47.55 to 30.42] | -28.64 [-65.74–11.31] |
| Skill 3 | -1.73 [-15.10 to 10.86] | 4.40 [-4.85 to 13.06] | -3.95 [-15.36–7.29] |
| Skill 4 | -3.09 [-15.39 to 9.14] | -4.37 [-13.61 to 4.85] | -2.38 [-13.53–8.33] |
| Skill 6 | 4.72 [-3.42 to 12.63] | 5.96 [-1.17 to 13.20] | 0.86 [-5.92–7.79] |
| R | 0.755 [0.620 to 0.829] | 0.821 [0.700 to 0.871] | 0.693 [0.559 to 0.775] |
| R 2 | 0.570 [0.384 to 0.686] | 0.674 [0.490 to 0.785] | 0.480 [0.312 to 0.600] |
## Discussion
A wide range of empirical results suggest that physicians have a limited ability to assess their professional competences accurately [27]. This includes communication competences, where studies frequently show a lack of association between physicians’ self-assessment and external rating by trained observers [28–30]. Here, we found encouraging evidence that it is possible to use physicians’ self-assessment of behavioral skills for measuring competence, even though the measurement is computationally more complex than using simple (averaged) global ratings as a direct measure of competence.
In the population network, the most central SDM skills were presenting options, informing on options, eliciting preferences, deliberating the decision, and selecting an option. Supporting comprehension and planning actions seem to be somewhat more peripheral skills, while focusing the decision and sharing the decision are (albeit strongly associated with each other) completely disconnected from the rest of the network. This architecture is strikingly similar to the structure of the population network of SDM skills based on patient-reported data [13], even though patient and physician assessments of the specific skills from the same consultation considerably disagreed in previous investigations [31, 32]. It should also be noted that, although we did not attempt to cluster skills in the present study explicitly, the identified structure of the SDM skills shows similarities with the categorization of the skills postulated by the three-talk model of SDM by Elwyn and colleagues [33]. These findings suggest that skills networks are able to capture a robust and replicable physician-level construct, which we hypothesize to be SDM competence.
Validity of interpreting the information contained in the network structure as an indicator of SDM competence was supported by its association with observer-rated data. In a confirmatory approach, we found that the combination of data-based inference with findings from the analysis of patient-reported data [13] (in the form of informative priors for Bayesian analysis) produced strong predictions of observer-rated competence. In the spirit of a continuous Bayesian accumulation of evidence, the results of the confirmatory analysis can be considered to synthesize the findings of the previously reported investigation using patient-reported data and the current study based on physicians’ self-assessment quantitatively. Results of the exploratory analysis led to models with even stronger predictive accuracy. This indicates that skills networks based on physicians`self-assessment of their SDM skills were highly predictive of their SDM competence as rated by external observers. *In* general, the findings support the hypothesis that patient and physician rated data may be used interchangeably for competence assessment if handled in the context of the network approach.
Both patients`and physicians`ratings of SDM processed according to the skills network approach seem to yield an objective assessment of SDM competence, which highly relates to external assessments of this competence. This finding has various implications. From a theoretical perspective, it suggests a new definition of professional competence, which can be contrasted to and integrated with existing ones [34]. For the network science of psychological phenomena [35], it means a methodological extension and a new field of application. Lastly, for assessing professional SDM competence [36], it offers a new way of measurement based on self-rating of physicians. By applying the skills network model of SDM competence to physician-rated data, we provided a promising opportunity for a feasible assessment of SDM competence. Self-ratings are, in contrast to observer ratings, more easily applicable and less time-intensive, offering a genuine opportunity for their application in routine practice.
Since measuring SDM competence with skills networks seem to offer a replicable and robust assessment of this professional skill (high agreement between patient, physician, and observer assessment), our proposed method is of relevance and could be applied to areas in which a feasible and robust assessment of SDM competence is highly needed. First, research on SDM depends largely on a valid measurement of SDM competence, for example to assess predictors and treatment outcomes for different levels of SDM competence. Considering that assessing competence by observation is very resource intensive, utilizing brief self-assessment increases the range of options for research projects. Second, to evaluate the effectiveness of a trainings for SDM, including education of health care professionals, the assessment of this competence is of central importance. Novel measures without the need for external judgment by qualified experts could contribute to a more comprehensive evaluation of interventions aiming to implement SDM. Finally, the network approach to SDM competence could be applied when assessing SDM as a part of quality management in clinical routine care. Here, a robust assessment can be gained from quite easily attainable patient or physician ratings of SDM. In this context, analysis of a continuous data stream from SDM surveys may enable monitoring of the SDM competence of individuals, teams, departments, or hospitals. Furthermore, a detailed analysis of the obtained skills networks could reveal specific and actionable targets (i.e., skills or skill connections) for improvement. Being able to create individual skills networks and to precisely pinpoint skills and skill connections that need to be improved could open the way to a data-driven and individualized measurement, education, training, and monitoring of complex competences.
Current findings are limited by the restricted sample size and the considerable complexity of the statistical models relative to the sample size. These factors are likely to be partly responsible for the wide credible intervals of the estimated parameters. Due to this imprecision and to collinearity between network parameters, the influence of specific network parameters of individual skills could be investigated only to a limited extend. As network parameters were correlated to each other, it remains unclear how each individual network parameter relates to observer-rated SDM competence and which network parameters are most important for indicating SDM competence. Jointly, the network parameters showed a high predictive accuracy for the observer-rated SDM competence, and future studies need to assess which specific network parameters are most important for this. Furthermore, since the approach has been only applied to data from a self-selected sample from outpatient care in Germany, generalizability to other contexts needs to be investigated in future studies. This should also include comparing results between various contexts and subgroups, for example, defined by the primary specialty of the physician or the disease of the consulted patients, which was unfortunately not possible in the present study due to the limited sample size. Finally, results from the exploratory analyses need to be interpreted with due caution, as different model building procedures could have led to different results and current results could not be cross-validated. Still, especially through the confirmatory testing and the replication of findings from previous analyses with patient data, the current study offered considerable support for the skills network approach to SDM competence. By applying a Bayesian framework, some previously mentioned weaknesses could be extenuated and problems such as the multiple testing problem avoided. Future studies need to test this new approach with larger datasets to assess the relative importance of individual network parameters and skills.
Structuring clinical competences into a hierarchically organized categorical system is challenging, particularly in the interpersonal and communication domains [37]. “ Choosing the right boundaries for a unit of analysis is a central problem in every science” [38], and this is particularly true for clinical skills and competences, which are strongly interrelated and frequently overlapping. Thus, it is not always clear how to narrow down the densely connected network of clinical skills into well definable and analyzable competences. Whether SDM is a sufficiently distinct concept from this perspective, i.e., whether it is operationally sufficiently closed in the environment of other skills and competences, should be empirically investigated in further studies by collecting data on a broader range of skills and competences for network analysis.
## Conclusions
Our findings provide further support for conceptualizing and modeling physicians’ SDM competence as a network of SDM skills. This conceptualization suggests a new definition of professional competence, offers a methodological extension and a new field of application for network science and, most importantly, provides a new way of measuring professional competence based on self-rating of physicians. A robust measurement of SDM competence offers new opportunities for research, for evaluating learning success in education and training, and for monitoring SDM competence for quality management purposes in clinical routine care. In combination, these consistent theoretical, empirical, and practical implications have the potential to open up a new approach to professional competence in health care.
## References
1. Spatz ES, Krumholz HM, Moulton BW. **Prime time for shared decision making**. *JAMA* (2017) **317** 1309-1310. DOI: 10.1001/jama.2017.0616
2. Barry MJ, Edgman-Levitan S. **Shared decision making—pinnacle of patient-centered care**. *N Engl J Med* (2012) **366** 780-781. DOI: 10.1056/NEJMp1109283
3. **Salzburg statement on shared decision making**. *BMJ* (2011) **342** d1745. DOI: 10.1136/bmj.d1745
4. Mulley AG, Trimble C, Elwyn G. **Stop the silent misdiagnosis: patients’ preferences matter**. *BMJ* (2012) **345** e6572. DOI: 10.1136/bmj.e6572
5. Hoffmann TC, Légaré F, Simmons MB, McNamara K, McCaffery K, Trevena LJ. **Shared decision making: what do clinicians need to know and why should they bother?**. *Med J Aust* (2014) **201** 35-39. DOI: 10.5694/mja14.00002
6. Ratanawongsa N, Karter AJ, Parker MM, Lyles CR, Heisler M, Moffet HH. **Communication and medication refill adherence: the Diabetes Study of Northern California**. *JAMA Intern Med* (2013) **173** 210-218. DOI: 10.1001/jamainternmed.2013.1216
7. Oshima Lee E, Emanuel EJ. **Shared decision making to improve care and reduce costs**. *N Engl J Med* (2013) **368** 6-8. DOI: 10.1056/NEJMp1209500
8. Nickel WK, Weinberger SE, Guze PA. **Patient Partnership in Healthcare Committee of the American College of Physicians. Principles for patient and family partnership in care: an American College of Physicians position paper**. *Ann Intern Med* (2018) **169** 796-799. DOI: 10.7326/M18-0018
9. Carmona C, Crutwell J, Burnham M, Polak L. **Shared decision-making: summary of NICE guidance**. *BMJ* (2021) **373** n1430. DOI: 10.1136/bmj.n1430
10. **Collaboration and shared decision-making between patients and clinicians in preventive health care decisions and US Preventive Services Task Force recommendations**. *JAMA* (2022) **327** 1171-1176. DOI: 10.1001/jama.2022.3267
11. Wollschläger D.. **Where is SDM at home? Putting theoretical constraints on the way shared decision making is measured.**. *Z Evid Fortbild Qual Gesundhwes* (2012) **106** 272-274. DOI: 10.1016/j.zefq.2012.04.004
12. Borsboom D, Mellenbergh GJ, van Heerden J. **The theoretical status of latent variables.**. *Psychol Rev* (2003) **110** 203-219. DOI: 10.1037/0033-295X.110.2.203
13. Kriston L, Hahlweg P, Härter M, Scholl I. **A skills network approach to physicians’ competence in shared decision making**. *Health Expect* (2020) **23** 1466-1476. DOI: 10.1111/hex.13130
14. Scholl I, Kriston L, Dirmaier J, Härter M. **Comparing the nine-item Shared Decision-Making Questionnaire to the OPTION Scale—an attempt to establish convergent validity**. *Health Expect* (2015) **18** 137-150. DOI: 10.1111/hex.12022
15. Scholl I, Kriston L, Dirmaier J, Buchholz A, Härter M. **Development and psychometric properties of the Shared Decision Making Questionnaire–physician version (SDM-Q-Doc).**. *Patient Educ Couns* (2012) **88** 284-290. DOI: 10.1016/j.pec.2012.03.005
16. Elwyn G, Hutchings H, Edwards A, Rapport F, Wensing M, Cheung W-Y. **The OPTION scale: measuring the extent that clinicians involve patients in decision-making tasks**. *Health Expect* (2005) **8** 34-42. DOI: 10.1111/j.1369-7625.2004.00311.x
17. Hirsch O, Keller H, Müller‐Engelmann M, Gutenbrunner MH, Krones T, Donner‐Banzhoff N. **Reliability and validity of the German version of the OPTION scale**. *Health Expect* (2012) **15** 379-388. DOI: 10.1111/j.1369-7625.2011.00689.x
18. Barr PJ, O’Malley AJ, Tsulukidze M, Gionfriddo MR, Montori V, Elwyn G. **The psychometric properties of Observer OPTION(5), an observer measure of shared decision making**. *Patient Educ Couns* (2015) **98** 970-976. DOI: 10.1016/j.pec.2015.04.010
19. Kölker M, Topp J, Elwyn G, Härter M, Scholl I. **Psychometric properties of the German version of Observer OPTION5**. *BMC Health Serv Res* (2018) 18. DOI: 10.1186/s12913-018-2891-6
20. Krupat E, Frankel R, Stein T, Irish J. **The Four Habits Coding Scheme: validation of an instrument to assess clinicians’ communication behavior**. *Patient Educ Couns* (2006) **62** 38-45. DOI: 10.1016/j.pec.2005.04.015
21. Scholl I, Nicolai J, Pahlke S, Kriston L, Krupat E, Härter M. **The German version of the Four Habits Coding Scheme—association between physicians’ communication and shared decision making skills in the medical encounter**. *Patient Educ Couns* (2014) **94** 224-229. DOI: 10.1016/j.pec.2013.10.006
22. Fruchterman TMJ, Reingold EM. **Graph drawing by force-directed placement**. *Softw Pract Exp* (1991) **21** 1129-1164. DOI: 10.1002/spe.4380211102
23. 23R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2014.. (2014)
24. Bürkner PC. **Advanced Bayesian multilevel modeling with the R package brms**. *R J.* (2018) **10** 395-411. DOI: 10.32614/RJ-2018-017
25. Epskamp S, Cramer AOJ, Waldorp LJ, Schmittmann VD, Borsboom D. **qgraph: Network visualizations of relationships in psychometric data**. *J Stat Softw* (2012) **48** 1-18. DOI: 10.18637/jss.v048.i04
26. Gelman A, Rubin DB. **Inference from iterative simulation using multiple sequences**. *Stat Sci* (1992) **7** 457-472
27. Davis DA, Mazmanian PE, Fordis M, Van Harrison R, Thorpe KE, Perrier L. **Accuracy of physician self-assessment compared with observed measures of competence: a systematic review**. *JAMA* (2006) **296** 1094-1102. DOI: 10.1001/jama.296.9.1094
28. Burt J, Abel G, Elliott MN, Elmore N, Newbould J, Davey A. **The evaluation of physicians’ communication skills from multiple perspectives.**. *Ann Fam Med* (2018) **16** 330-337. DOI: 10.1370/afm.2241
29. Pollak KI, Arnold RM, Jeffreys AS, Alexander SC, Olsen MK, Abernethy AP. **Oncologist communication about emotion during visits with patients with advanced cancer**. *J Clin Oncol* (2007) **25** 5748-5752. DOI: 10.1200/JCO.2007.12.4180
30. Gude T, Finset A, Anvik T, Bærheim A, Fasmer OB, Grimstad H. **Do medical students and young physicians assess reliably their self-efficacy regarding communication skills? A prospective study from end of medical school until end of internship.**. *BMC Med Educ* (2017) **17** 107. DOI: 10.1186/s12909-017-0943-y
31. Kriston L, Härter M, Scholl I. **A latent variable framework for modeling dyadic measures in research on shared decision-making**. *Z Evid Fortbild Qual Gesundhwes* (2012) **106** 253-263. DOI: 10.1016/j.zefq.2012.03.021
32. Röttele N, Schöpf-Lazzarino AC, Becker S, Körner M, Boeker M, Wirtz MA. **Agreement of physician and patient ratings of communication in medical encounters: A systematic review and meta-analysis of interrater agreement**. *Patient Educ Couns* (2020) **103** 1873-1882. DOI: 10.1016/j.pec.2020.04.002
33. Elwyn G, Durand MA, Song J, Aarts J, Barr PJ, Berger Z. **A three-talk model for shared decision making: multistage consultation process**. *BMJ* (2017) **359** j4891. DOI: 10.1136/bmj.j4891
34. Epstein RM, Hundert EM. **Defining and assessing professional competence**. *JAMA* (2002) **287** 226-235. DOI: 10.1001/jama.287.2.226
35. Schmittmann VD, Cramer AOJ, Waldorp LJ, Epskamp S, Kievit RA, Borsboom D. **Deconstructing the construct: a network perspective on psychological phenomena**. *New Ideas Psychol* (2013) **31** 43-53. DOI: 10.1016/j.newideapsych.2011.02.007
36. Miller GE. **The assessment of clinical skills/competence/performance**. *Acad Med* (1990) **65** S63-67. DOI: 10.1097/00001888-199009000-00045
37. Makoul G.. **Essential elements of communication in medical encounters: the Kalamazoo consensus statement**. *Acad Med* (2001) **76** 390-393. DOI: 10.1097/00001888-200104000-00021
38. Hutchins E.. **Cognitive ecology**. *Topics Cogn Sci* (2010) **2** 705-715. DOI: 10.1111/j.1756-8765.2010.01089.x
|
---
title: Novel in-frame duplication variant characterization in late infantile metachromatic
leukodystrophy using whole-exome sequencing and molecular dynamics simulation
authors:
- Zahra Ataei
- Zahra Nouri
- Farial Tavakoli
- Mohammad Reza Pourreza
- Sina Narrei
- Mohammad Amin Tabatabaiefar
journal: PLOS ONE
year: 2023
pmcid: PMC9970088
doi: 10.1371/journal.pone.0282304
license: CC BY 4.0
---
# Novel in-frame duplication variant characterization in late infantile metachromatic leukodystrophy using whole-exome sequencing and molecular dynamics simulation
## Abstract
Metachromatic leukodystrophy (MLD) is a neurodegenerative lysosomal storage disease caused by a deficiency in the arylsulfatase A (ARSA). ARSA deficiency leads to sulfatide accumulation, which involves progressive demyelination. The profound impact of early diagnosis on MLD treatment options necessitates the development of new or updated analysis tools and approaches. In this study, to identify the genetic etiology in a proband from a consanguineous family with MLD presentation and low ARSA activity, we employed Whole-Exome Sequencing (WES) followed by co-segregation analysis using Sanger sequencing. Also, MD simulation was utilized to study how the variant alters the structural behavior and function of the ARSA protein. GROMACS was applied and the data was analyzed by RMSD, RMSF, Rg, SASA, HB, atomic distance, PCA, and FEL. Variant interpretation was done based on the American College of Medical Genetics and Genomics (ACMG) guidelines. WES results showed a novel homozygous insertion mutation, c.109_126dup (p.Asp37_Gly42dup), in the ARSA gene. This variant is located in the first exon of ARSA, fulfilling the criteria of being categorized as likely pathogenic, according to the ACMG guidelines and it was also found to be co-segregating in the family. The MD simulation analysis revealed this mutation influenced the structure and the stabilization of ARSA and led to the protein function impairment. Here, we report a useful application of WES and MD to identify the causes of a neurometabolic disorder.
## Introduction
Leukodystrophies are a group of usually inherited disorders that affect the white matter of the central nervous system [1]. Metachromatic leukodystrophy (MLD) (OMIM 250100), as an autosomal recessive neurodegenerative disorder, is one of the most common leukodystrophies. MLD is one of the lysosomal storage diseases (LSDs) with a prevalence of 1.45 per 100,000 births worldwide [2]. MLD is a sphingolipidosis caused by lysosomal enzyme arylsulfatase A (ARSA) deficiency or its sphingolipid activator protein B (SapB) [3]. The ARSA gene is located on chromosome 22q13 with eight exons spanning a genomic region of 3kb and encoding a 509 amino acids [4, 5] ARSA is responsible for the hydrolysis of the 3-O ester bond of sphingolipid 3′-O-sulfogalactosylceramide, known as sulfatide. ARSA deficiency leads to an increase in the sulfatide within oligodendrocytes, macrophages, and some subtypes of neurons in the CNS, in Schwann cells, and in the peripheral nervous system (PNS) macrophages, which exhibit metachromatic staining characteristics [6]. Instability of the myelin sheath, change in calcium homeostasis, cell stress, apoptosis, and inflammatory response are consequences of sulfatide deposition in the cell cytoplasm [1, 7].
Depending upon the age which symptoms present, MLD is classified into three different clinical forms: late infantile, juvenile, and adult forms [8]. Late-infantile MLD, the most severe type with poor prognosis, is characterized by gait abnormalities, seizures, ataxia, hypotonia, extensor planters, optic atrophy, and regression of motor skills that lead to complete gross motor deterioration before the age of 40 months [9, 10].
There is no newborn screening test available for MLD diagnosis, and MLD is diagnosed after birth [11]. Progressive demyelination and subsequent neurological symptoms, biochemical procedures, genetic analysis, and imaging results should be applied for a specific diagnosis of MLD. Biochemical assays, including the quantification of Sulfatide accumulation in urine, and ARSA activity in peripheral blood leukocytes, are used for accurate diagnosis of MLD [12].
*Next* generation sequencing techniques provide an opportunity for diagnosis of hereditary genetic diseases in research and clinical settings. Whole exome sequencing (WES) has been applied in identifying genetic variants associated with a variety of diseases [13, 14]. Moreover, the biocomputational techniques have been recently applied broadly as a potential tool to investigate the variant effect on a mutant protein structure, in a rapid and cost-effective manner [15, 16] In the current study, for accurate diagnosis of a 34- month patient, the first child of consanguineous parents with a progressive decline of motor and cognitive abilities, hypotonia, and spasticity, we used WES in the diagnosis of MLD, followed by imaging analysis and enzymatic tests aimed at confirming the molecular results, which led to a novel in-frame mutation identification. Furthermore, Molecular dynamics (MD) simulation was applied to investigate the ARSA in wild type form (WT-ARSA) and structural and conformational changes of mutant form of ARSA (mutant-ARSA) to understand the pathogenic mechanism of MLD disease at atomic level.
## Subject and clinical evaluations and imaging results
The study was approved by the Research Ethics Committee of “Alzahra Research Center” (grant no:2400173, IR.ARI.MUI.REC.1400.011), and the patient and his parents were recruited to the study after obtaining informed consent. The proband was a 34-month boy and the first child of consanguineous Iranian parents who were first cousins without a family history of neurological disease. A detailed clinical examination and comprehensive family history were done by a clinical geneticist.
## ARSA enzymatic assay
ARSA activity was estimated as mu/mg protein in leukocytes, using 4-nitrocatechol sulfate. In brief: 0.25 mM sodium pyrophosphate was utilized to inactivate arylsulfatase B(ARSB). Then, the amount of sulfate released was measured by the absorbance of free 4-nitrocatechol at 515 nm on a spectrophotometer (Beckmann Coulter, Brea, CA, USA), which is associated with sulfatase activity [17].
## Metabolic panel test
Amino Acid Profile and AcylCarnitine Profile were investigated in plasma using LC-MS/MS and MS/MS, respectively. Also, Organic Acids and Acylglycines measurements in urine were performed through GC-MS and LC-MS/MS.
## Whole exome sequencing
Blood samples were taken from the patient and his parents. Using a Qiagen DNA extraction kit, genomic DNA was extracted from peripheral blood lymphocytes, and assessment of its purity, was done on a Nanospec Cube Biophotometer (Nanolytik®, Dusseldorf, Germany). The sample was sent to Macrogen (South Korea) (https://www.macrogen.com/) for WES analysis using the Novaseq 4000 platform (Illumina, San Diego, CA, USA) with the mean depth of coverage 100X. These samples were sheared into 151-bp fragments by a hydrodynamic shearing system (Covaris, Massachusetts, USA), and whole exome was captured through in-solution targeted genomic enrichment using Agilent SureSelect Human All Exon kit v6 (Agilent Technologies, CA, USA).
## Bioinformatics analysis
Following sequencing, image analysis and base-calling were performed using the standard *Illumina data* analysis pipeline Real-Time Analysis (RTA) version (RTA) v1.12.4. CASAVA v1.8.2. The raw reads quality was assessed using the FastQC [18]. The low quality reads have been removed with TRIMMOMATIC [19] and quality control was done after trimming. Reads were mapped to the human reference genome build UCSC hg19 (http://genome.ucsc.edu/) using BWA (Burrows-Wheeler Aligner) (http://bio-bwa.sourceforge.net/). SAMtools was used to convert sequence alignment map (SAM) format to sorted, indexed binary alignment map files [20]. GATK software tools (https://gatk.broadinstitute.org/hc/en-us) were used to improve alignments and genotype calling. annotation was performed using ANNOVAR [21].
Missense, nonsense, start codon change, stop loss, indel variants, and splice site with Minor Allele Frequency <$1\%$ in dbSNP version 147, 1000 genomes project phase 3 database(https://www.internationalgenome.org/), NHLBI GO exome sequencing project (ESP) (https://evs.gs.washington.edu/EVS/), exome aggregation consortium (ExAC) (https://exac.broadinstitute.org/), Iranome database (http://www.iranome.ir/), and our locally developed database (GTAC) are considered for further analysis. The novelty of the variant was investigated in the Human Gene Mutation Database (HGMD) (http://www.hgmd.cf.ac.uk/ac/index.php) and the literature. Finally, the identified variants’ pathogenicity can be interpreted according to ACMG 2015 standards and guidelines for interpreting sequence variants [22]. The MEGA6 software was utilized to investigate the conservation of the mutated region among several species [23].
## Variant confirmation
The candidate variant (p.Asp37_Gly42dup) that was identified by WES was subsequently confirmed using polymerase chain reaction (PCR) and bidirectional Sanger sequencing in the proband Using the SeqStudio Genetic Analyzer (Applied Biosystem Inc., Foster City, CA, USA). Then, to examine the segregation of genotype among the family members, co-segregation analysis was performed on his parents. Primers (forward primer: 5′- GTATTTGGGTCCGGGGTCTC-3′ and the reverse primer: 5′-TGTGGCCTTCCCTAGAGAGA-3′ (designed by Primer 3 software Input version 0.4.0 (https://primer3.ut.ee/) and NCBI primer BLAST software)) encompassed exon 1 of the ARSA gene. Chromatogram sequences files were compared with the reference sequence (NM_000487), via SeqMan software version 5.00© (DNASTAR, Madison, WI, USA).
## Molecular modeling
Crystal structure of WT-ARSA protein in complex with NAG1 and NAG2 was obtained from Protein Data Bank (PDB ID: 1AUK with resolution 2.10 Å). Only protein molecule was retained and all excess molecules were removed from the complex using Discovery Studio 2016 Visualizer software (DS 2016) (DS 4.0, Accelrys Software Inc., San Diego, CA). Swiss Model webserver [24] was applied in order to repair the chain breaks in the WT-ARSA and construct a model for the mutant-ARSA. The modeled WT-ARSA is composed of 485 residues starting with residue Arg1 and ending with residue Pro485 while the modeled mutant-ARSA is composed of 491 residues containing the six-amino acid insertion mutation in 23–28 position (Asp23, Leu24, Gly25, Cys26, Tyr27, and Gly28) starting with residue Arg1 and ending with residue Pro491. The amino acid compositions of the active site of WT-ARSA are Ala10, Asp11, Asp12, FGly51, Arg55, Lys105, His107, His211, Asp263, Asn264, Lys284 and those of mutant-ARSA are Ala10, Asp11, Asp12, FGly57, Arg61, Lys111, His113, His217, Asp269, Asn270, and Lys290.
## Molecular dynamics simulation
In order to investigate the structural and conformational changes in ARSA proteins, GROMACS 2018 package was applied to perform molecular dynamics simulation [25]. PDB2PQR web server [26] was retrieved to determine the protonation state (PH = 5) of His residues in both proteins. All the required files, topology and coordinate files, for both molecules were constructed using GROMACS [25] through CHARMM27 all-atom force field [27].
A dodecahedron TIP3P water box with a direction of 9 Å as an unit cell was constructed to solvate all models [28]. In order to calculate each system in water, the neutralization of the negative charges of the system was carried out through adding Na+ ions. Energy-minimization of the solvated system was performed applying 50000 steps of steepest-descent method to remove steric clashes. Equilibration of the minimized systems with position restrain on the proteins were performed by NVT and NPT ensembles for 400 ps at a temperature of 300 K and pressure of 1 bar, respectively.
V-rescale temperature [29] and Parrinello-Rahman pressure [30] coupling methods were applied to stabilize the temperature and pressure at 300 K and 1 bar for the system. 100-ns MD simulations were carried out for ARSA proteins under periodic boundary condition with the time step of 2 fs applying LINCS [31] and Partial Mesh Ewald (PME) [32] algorithms. The coulomb and van der Waals interactions were calculated by cut-off value of 1.2 nm.
Molecular dynamics simulation method was employed to investigate the effect of the variant on the ARSA structure and function. To do this, 100-ns MD simulations were performed for WT-ARSA and mutant-ARSA to calculate all the required data. The obtained final structures of both ARSA proteins from 100-ns MD simulation were superimposed and are displayed in S1 Fig. No significant structural changes were found in the mutant-ARSA protein except in the loop shape region is composed of mutant residues Asp23, Leu24, Gly25, Cyc26, Tyr27, Gly28.
RMSD and RMSF were calculated for each production simulation. The average all-atom RMSD values for both ARSA proteins relative to the initial structures were calculated as 0.16 and 0.17 nm, respectively (Table 1 and Fig 3A).
**Fig 3:** *Conformational changes in ARSA proteins.(A) all-atom RMSD, (B) RMSF and (C) Rg of WT-ARSA and mutant-ARSA.* TABLE_PLACEHOLDER:Table 1 Residual flexibility of both proteins was obtained and is displayed in Fig 3B. The RMSF graphs reveal that this variant significantly increased the fluctuation of residues in the loop shape regions, particularly the regions are composed of mutant residues in 23–28 position and residues in 480–485 position, and the regions with residues 160–163. In addition, the RMSF graphs show that the fluctuation of active site entrance residues of mutant-ARSA including Pro77, Gly78, Val79, His139, Gly158, Asp161, Gln162, Gly163, Tyr218, Thr274, Arg276, His393 are also increased, while the fluctuation of active site residues of mutant-ARSA was small.
The radius of gyration, which is defined as Rg, shows the compression and density of protein structure and stability, and the more compact a protein is, the more stable it will be [36] The average Rg values of WT-ARSA and mutant-ARSA were calculated as 2.18 and 2.21 nm, respectively (Table 1). The graphs of *Rg data* between both proteins are displayed in Fig 3C. The mutant-ARSA demonstrates a significant increment in Rg value.
The hydrophobic core region of both ARSA proteins was investigated by calculating the total and residue solvent accessible surface area (SASA). Remarkable increment in the average of total SASA and relatively small increase in the average of active site SASA were obtained in the mutant-ARSA compared to WT-ARSA (Table 1 and Fig 4A). Among the active site residues of mutant-ARSA, residues Ala10, FGly57, Arg61, Lys111 and Asp269 with SASA of 0.02, 0.38, 0.061, 0.105 and 0.072 nm2 respectively indicate the most accessibility to solvent compared to those of in WT-ARSA with SASA of 0.002, 0.125, 0.011, 0.04, and 0.017nm2 respectively (Fig 4B).
**Fig 4:** *Graphical representation of solvent accessible surface area (SASA) of ARSA proteins.(A) Total SASA and (B) active site SASA of WT-ARSA and mutant-ARSA.*
Hydrogen bond (HB) analysis during MD simulation was investigated internally and the same analysis was conducted between the proteins and the solvent (protein-solvent) to understand the stability and the solubility of the ARSA proteins. The average number of intramolecular and protein-solvent HB networks and the patterns of them were obtained (Table 1 and Fig 5A and 5B).
**Fig 5:** *The pattern of hydrogen bonds during 100 ns MD simulation for ARSA proteins.(A) Intramolecular and (B) protein-solvent hydrogen bonds of WT-ARSA and mutant-ARSA.*
As can be seen in Table 1 and Fig 5A, there is no substantial difference between the number of intramolecular HB of ARSA proteins, while a significant increment in the number of protein-solvent HB is observed in mutant-ARSA with an average of 833 HB compared to WT-ARSA with an average of 799 HB (Table 1 and Fig 5B).
The assignment of secondary structure elements is an essential component to investigate the structural behavior of protein. The variations in the secondary structure in WT-ARSA and mutant-ARSA were investigated using do-dssp function. The secondary structures of both proteins are almost similar and the six-amino acid insertion mutation is not able to induce any remarkable change in the secondary structure content because the mutation was in the loop region but not in a critical position for the secondary structure formation (S2A and S2B Fig) as illustrated in S1 Fig.
In order to calculate distances between all possible amino acid residues pairs of active site in both ARSA proteins, distances between Cα atoms of active site residues as a function of time were measured using gmx distance module. Eleven groups consist of Cα atoms of amino acid residues pairs of active site in WT-ARSA and mutant-ARSA were created and then were chosen to calculate the distances. These groups include Ala10—Asp11, Asp11—Asp12, Asp12—FGly51, FGly51—Arg55, Arg55—Lys105, Lys105—His107, His107—His211, His211—Asp263, Asp263—Asn264, Asn264—Lys284, Lys284—Ala10 in WT-ARSA and consist of Ala10—Asp11, Asp11—Asp12, Asp12—FGly57, FGly57—Arg61, Arg61—Lys111, Lys111—His113, His113—His217, His217—Asp269, Asp269—Asn270, Asn270—Lys290, Lys290—Ala10 in mutant-ARSA. As can be seen in distance plots in S3 Fig, the distances between almost all amino acid residues pairs in WT-ARSA are stable during 100-ns simulation while the groups of Asp12—FGly57, Arg61—Lys111, His217—Asp269 and Lys290—Ala10 in mutant-ARSA did not show stability in the course of 100-ns simulation.
Principal Component Analysis (PCA) was performed to understand the dynamics of both ARSA proteins into a few principal motions, defined by eigenvalues and eigenvectors. The sampled conformations of WT-ARSA and mutant-ARSA in the essential subspace were calculated by projecting the protein backbone of the MD trajectory on eigenvectors 1 and 2 (Fig 6A). The projection of the PC1 and PC2 of a trajectory can represent a cluster of stable states in protein. As can be seen in Fig 6A, the mutant-ARSA displays a transition towards a more distant region of the phase spaces compared to WT-ARSA.
**Fig 6:** *Conformational sampling and Free energy landscape (FEL) analysis of ARSA proteins.(A) Conformational sampling of WT-ARSA and mutant-ARSA proteins by 2D projection of the MD trajectory on PC1 and PC2. (B) FEL of WT-ARSA and (C) FEL of mutant-ARSA.*
Free energy landscape (FEL) was calculated for two systems applying the first two PCs as reaction coordinates to find the conformational states of ARSA proteins. The FEL maps provide noticeable data on the diverse conformational states to the ARSAs in the 100 ns-MD simulation (Fig 6B and 6C). It is observed from Fig 6 that the PC1 and PC2 motion modes of the mutant-ARSA occupied larger spaces than those of the WT-ARSA, illustrating the conformational rearrangements which lie at the root of the mutation.
## Analysis of MD simulations
Production simulations were analyzed applying trajectory analysis modules in the GROMACS simulation package and visualized using VMD [33], Root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), solvent accessible surface area (SASA), hydrogen bond (HB), secondary structure, atomic distance, principal component analysis (PCA) and gibbs free energy landscape (FEL) were obtained using GROMACS and Grace-5.1.22/QTGrace v0.2.6 program was retrieved to visualize all the graphs.
## Principal component analysis
Principal component analysis (PCA) was calculated to investigate the biomolecules motion [34] applying calculating the covariance matrix C: Cij=<(xi−<xi>)(xj−<xj>)> [1] Where xi and xj are the coordinate of the ith and jth atoms of the systems and <xi> and <xj> represent the average coordinate of the ith and jth atoms over the ensemble. Then, the principal components (PCs) are calculated by diagonalization of covariance matrix.
## Free energy landscape
Free energy landscape (FEL) was calculated to investigate the conformational changes of biomolecules to identify the stable state and transient state of proteins and to assign their stability and their function [35]. The FEL can be calculated as: ΔG(X)=−KBTlnP(X) [2] Where ΔG (X) is the free energy, KB and T, and P(X) represent the Boltzman constant, absolute temperature, and the probability distribution of the conformation ensemble along the PCs.
## Clinical evaluations
The proband was a 34-month boy and the first child of consanguineous Iranian parents without a family history of neurological disease. The function of kidneys, liver, thyroid, and heart was normal. He was born at full term via normal vaginal delivery after an uncomplicated pregnancy.
Prenatal screening tests did not show any abnormalities in the fetus. At the age of 15 months, he presented with spastic gait, and he could not walk independently. However, MRI results suggested a normal brain. Clonazepam was administered orally until two years, and manifestations of spasticity were controlled. The patient had no swallowing difficulties. At 32 months, the child had language disorder, weakness started in the hands, muscle spasms progressed, and muscular hypotonia was evident. Electroencephalogram (EEG) showed mildly abnormal, and MRI results show evidence of bilateral symmetrical abnormal signal intensity in deep white matter, centrum semiovale, paraventricular regions, and subcortical region as low signal intensity in T1, high signal intensity in T2W and flair images.
## ARSA enzyme activity
The enzyme activity using the colorimetric method was 0.066 mu/mg protein. The reference was 0.37–1.815.
## Metabolic panel
There was no abnormality in the metabolic profile of the patient in the mentioned factors.
## Molecular findings
Upon exome analysis, a total number of 28135 variants was obtained. The number of variants was limited by applying the following criteria for variant filtering, and a homozygous insertion of 18 nucleotides (c.109_126dup) in the ARSA gene was found. It causes in-frame insertion in exon 1 of this gene, resulting in a protein with 515 residues (versus 509 residues in the intact protein). The in-frame variant was absent from dbSNP version 147, 1000 genomes project phase 3, NHLBI GO ESP, ExAC, Iranome, and our locally developed (GTAC) databases. It was not found in the literature, either. The six residues are inserted into a highly conserved region of ARSA, as evident through multiple-species alignment (Fig 1). The novel in-frame variant co-segregated with the disease in the family and was heterozygous in parents but homozygous in the patient (Fig 2). Given the PM2, PM4, PP3, PP4 ACMG/AMP criteria being fulfilled and the consistent phenotype, the c.109_126dup in ARSA was considered likely pathogenic in the patient.
**Fig 1:** *Phylogenetic alignment performed with MEGA6.The modified region is located in a highly conserved region among species.* **Fig 2:** *(A) Pedigree chart of the family. (B) Chromatogram of the homozygous sequence variant c.109_126dup detected in the ARSA gene in the proband compared with the heterozygous sequence in the parents. The comparison of three sequences with the reference sequence is given at the bottom of the electropherograms. The red and yellow boxes indicate the duplicated and the previous sequences, respectively.*
## Discussion
Owing to the rarity of MLD and the heterogeneity in its presentation, early diagnosis of MLD is challenging. To treat these individuals at an early stage of disease, accurate methods will be required to obtain an early diagnosis. Patients with MLD who get Haematopoietic Stem Cell Transplantation (HSCT) or gene therapy earlier in the disease have a better prognosis [11, 37, 38].
As a common molecular diagnostic tool, WES had a significant impact on identifying causative variations and determining the appropriate illness management strategy [39]. In the case of MLD, NGS can assist in illness identification, differentiating *Pseudodeficiency diagnosis* to avoid erroneous diagnoses based on ARSA activity, and identifying juvenile and adult variants of MLD to consider early treatment.
According to the Human Gene Mutation Database (HGMD) (http://www.hgmd.cf.ac.uk/ac/index.php), 303 ARSA mutations have been reported. Amongst Iranian MLD patients, c.931G>A(p. Gly311Ser) and c.465+1G>A variants are the most frequent alleles [40].
The structure of the DNA sequence around the 18-base-pair tandem duplication allele was investigated. In the normal sequence, we found a region with two short repeated sequences (Fig 7), which is indicative of a replication slippage. Polymerase slippage, that is expected to be as the possible cause of up to $75\%$ of all indels, might explain the mechanism through which the mutation was generated [41]. This repeat, in our case is imperfect (GACCTCGGC and GACCTGGGC) (Fig 7), and the duplicated region is (GACCTGGGCTGCTATGGG). As a hypothesis, it appears polymerase slippage can cause this duplication at the genomic level, and subsequently leading to protein instability, and the resultant clinical manifestation in patients. We suggest further functional studies for interpreting the pathogenic mechanism of MLD.
**Fig 7:** *DNA sequences of Wilde Type ARSA and the mutated allele.The 5′ one is shown in red (GACCTCGGC) and the 3′ one is shown in blue (GACCTGGGC). The underlined sequence is duplicated in the mutated allele.*
In computational part of our study, the protein structure of the mutant-ARSA was constructed using Swiss Model server, then structural and functional effects of the six-amino acid insertion mutation in the ARSA protein were studied using 100 ns MD simulation to investigate the stability of mutant protein and the molecular mechanism of disease.
The RMSD plots of both proteins showed that they are constant during the simulation, with RMSD value of approximately 0.15 nm. The RMSF plots displayed that the most mutation effect on the fluctuation of residues was related to the loop shape regions and the active site entrance residues. It revealed that the regions with high fluctuation have a low contribution to creating contact with other amino acids [42], particularly the loop regions in positions 160–163, 480–485 and 23–28 with mutant residues. A comparison of *Rg data* of both proteins revealed that the Rg value of mutant-ARSA increased significantly, indicating a loss in compactness of the protein.
The investigation of the hydrophobic core region, a key parameter to assess protein folding and stability [43] in both proteins, displayed a noticeable increment in average total SASA and a small increase in average active site SASA in mutant-ARSA, demonstrating a large surface exposed to the solvent and this could be due to the solvent exposure of hydrophobic amino acids and subsequently influencing the protein folding [35, 43]. Furthermore, the HB analysis revealed that although there was no significant change in intramolecular HB of both proteins, the average number of protein-solvent HB in mutant-ARSA increased considerably, which indicates higher solubility than that of the WT-ARSA [44]. The insignificant reduction in the intramolecular HB formation in mutant- ARSA might be due to the fact that the variant occurs in a highly flexible loop at the protein surface and, subsequently it has minimal effect on the existing interactions [42, 45, 46], as clarified from the RMSF data. As elucidated from our *Rg data* as a measure of the protein stability, change in compression and size of mutant-ARSA could have been a plausible reason behind the changing in protein folding, increasing its solubility and following instability of mutant-ARSA [15]. Above all, the examination of the atomic distance between amino acid residues pairs of the active site in both proteins as a function of time demonstrated clearly that the mutation causes instability in mutant-ARSA active site, whereas the active site structure of WT-ARSA was stable in the course of simulation.
The most important parameters, principal component, and free energy landscape analysis were retrieved to investigate the global motion, folding, function, and, eventually stability of the protein [34, 47]. PCA demonstrated that mutant-ARSA covered a wide range of phase spaces compared to WT-ARSA. Given that the proteins perform their function via collective atomic motions and the stability of a protein is related to its collective atomic motion, PCA results indicate that the underlying reason for the impairment of the protein function might be given rise to an increase in the overall motion in mutant-ARSA. FEL analysis revealed that, compared to the WT-ARSA which displayed a single global energy minimum basin, mutant-ARSA illustrated wide global energy minima via a transition state which demonstrates the mutation caused the conformational rearrangement in protein. Indeed, the existence of several energy minima in the conformational space of mutant-ARSA proves the significant destabilization of the protein [47].
According to the in silico study, although no remarkable changes were observed in the overall structure, the secondary structure content and internal HB of both proteins, due to the position of the mutation in the flexible loop region far from the active site [45, 46, 48], changes in compactness, solvent accessibility, protein-solvent hydrogen bond, atomic distance measurement, protein motion, and FEL analysis obviously demonstrate the instability and dysfunction of the mutant-ARSA, which is in agreement with a remarkable reduction in ARSA Activity (0.066 mu/mg pro).
## Conclusion
In this paper, we report the successful application of WES for diagnosis and proper genetic counseling of MLD. The in silico data analysis of the ARSA protein provided an in-depth comprehension of destabilization and loss of conformational dynamics of the mutant-ARSA and elucidated clearly that the structure and function of mutant-ARSA are negatively influenced due to the mutation. The approach would be highly useful in pre-clinical studies for the discovery of novel therapeutic and management strategies for MLD.
## References
1. van der Knaap MS, Bugiani M. **Leukodystrophies: a proposed classification system based on pathological changes and pathogenetic mechanisms**. *Acta neuropathologica* (2017) **134** 351-82. DOI: 10.1007/s00401-017-1739-1
2. Giugliani R.. **Newborn screening for lysosomal diseases: current status and potential interface with population medical genetics in Latin America**. *Journal of inherited metabolic disease* (2012) **35** 871-7. DOI: 10.1007/s10545-011-9436-z
3. Barboura I, Maire I, Ferchichi S, Miled A. **Metachromatic Leukodystrophy Clinical, Biological and Therapeutic Aspects**. *Latest Findings in Intellectual and Developmental Disabilities Research: IntechOpen* (2012)
4. Cesani M, Lorioli L, Grossi S, Amico G, Fumagalli F, Spiga I. **Mutation update of ARSA and PSAP genes causing metachromatic leukodystrophy**. *Human mutation* (2016) **37** 16-27. DOI: 10.1002/humu.22919
5. Stein C, Gieselmann V, Kreysing J, Schmidt B, Pohlmann R, Waheed A. **Cloning and expression of human arylsulfatase A**. *Journal of Biological Chemistry* (1989) **264** 1252-9. PMID: 2562955
6. Biffi A, Lucchini G, Rovelli A, Sessa M. **Metachromatic leukodystrophy: an overview of current and prospective treatments**. *Bone Marrow Transplantation* (2008) **42** S2-S6. DOI: 10.1038/bmt.2008.275
7. Barkovich AJ. **Concepts of myelin and myelination in neuroradiology**. *American Journal of Neuroradiology* (2000) **21** 1099-109. PMID: 10871022
8. Van Rappard DF, Boelens JJ, Wolf NI. **Metachromatic leukodystrophy: disease spectrum and approaches for treatment.**. *Best Practice & Research Clinical Endocrinology & Metabolism* (2015) **29** 261-73. DOI: 10.1016/j.beem.2014.10.001
9. Lynch DS, Wade C, de Paiva ARB, John N, Kinsella JA, Merwick Á. **Practical approach to the diagnosis of adult-onset leukodystrophies: an updated guide in the genomic era**. *Journal of Neurology, Neurosurgery & Psychiatry* (2019) **90** 543-54. PMID: 30467211
10. Kehrer C, Blumenstock G, Gieselmann V, KRÄGELOH‐MANN I, LEUKONET G. **The natural course of gross motor deterioration in metachromatic leukodystrophy.**. *Developmental Medicine & Child Neurology* (2011) **53** 850-5. DOI: 10.1111/j.1469-8749.2011.04028.x
11. Sessa M, Lorioli L, Fumagalli F, Acquati S, Redaelli D, Baldoli C. **Lentiviral haemopoietic stem-cell gene therapy in early-onset metachromatic leukodystrophy: an ad-hoc analysis of a non-randomised, open-label, phase 1/2 trial**. *The Lancet* (2016) **388** 476-87. DOI: 10.1016/S0140-6736(16)30374-9
12. Shaimardanova AA, Chulpanova DS, Solovyeva VV, Mullagulova AI, Kitaeva KV, Allegrucci C. **Metachromatic leukodystrophy: diagnosis, modeling, and treatment approaches**. *Frontiers in medicine* (2020) **7** 576221. DOI: 10.3389/fmed.2020.576221
13. Wang Z, Lin Y, Zheng D, Yan A, Tu X, Lin J. **Whole-exome sequencing identifies compound heterozygous mutations in ARSA of two siblings presented with atypical onset of metachromatic leukodystrophy from a Chinese pedigree**. *Clinica Chimica Acta* (2016) **460** 135-7. DOI: 10.1016/j.cca.2016.06.038
14. Sarmadi A, Mohammadi A, Tabatabaei F, Nouri Z, Chaleshtori MH, Tabatabaiefar MA. **Molecular genetic study in a cohort of Iranian families suspected to maturity-onset diabetes of the young, reveals a recurrent mutation and a high-risk variant in the CEL gene**. *Advanced biomedical research* (2020) 9. PMID: 32318359
15. Sadr AS, Eslahchi C, Ghassempour A, Kiaei M. **In silico studies reveal structural deviations of mutant profilin-1 and interaction with riluzole and edaravone in amyotrophic lateral sclerosis**. *Scientific reports* (2021) **11** 1-14. PMID: 33414495
16. Tavakoli F, Ganjalikhany MR. **Structure-based inhibitory peptide design targeting peptide-substrate binding site in EGFR tyrosine kinase**. *PloS one* (2019) **14** e0217031. DOI: 10.1371/journal.pone.0217031
17. Wenger D, Williams C. *Techniques on Diagnostic Human Biochemical Genetics A Laboratory Manual* (1991) **10003** 587-617
18. Andrews S.. *FastQC: a quality control tool for high throughput sequence data. Babraham Bioinformatics* (2010)
19. Bolger AM, Lohse M, Usadel B. **Trimmomatic: a flexible trimmer for Illumina sequence data**. *Bioinformatics* (2014) **30** 2114-20. DOI: 10.1093/bioinformatics/btu170
20. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N. **The sequence alignment/map format and SAMtools**. *Bioinformatics* (2009) **25** 2078-9. DOI: 10.1093/bioinformatics/btp352
21. Wang K, Li M, Hakonarson H. **ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data**. *Nucleic acids research* (2010) **38** e164. DOI: 10.1093/nar/gkq603
22. Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J. **Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology**. *Genetics in medicine* (2015) **17** 405-23. DOI: 10.1038/gim.2015.30
23. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. **MEGA6: molecular evolutionary genetics analysis version 6.0**. *Molecular biology and evolution* (2013) **30** 2725-9. DOI: 10.1093/molbev/mst197
24. Biasini M, Bienert S, Waterhouse A, Arnold K, Studer G, Schmidt T. **SWISS-MODEL: modelling protein tertiary and quaternary structure using evolutionary information**. *Nucleic acids research* (2014) **42** W252-W8. DOI: 10.1093/nar/gku340
25. Abraham MJ, Murtola T, Schulz R, Páll S, Smith JC, Hess B. **GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers**. *SoftwareX* (2015) **1** 19-25
26. Dolinsky TJ, Nielsen JE, McCammon JA, Baker NA. **PDB2PQR: an automated pipeline for the setup of Poisson–Boltzmann electrostatics calculations**. *Nucleic acids research* (2004) **32** W665-W7. DOI: 10.1093/nar/gkh381
27. Bjelkmar P, Larsson P, Cuendet MA, Hess B, Lindahl E. **Implementation of the CHARMM force field in GROMACS: analysis of protein stability effects from correction maps, virtual interaction sites, and water models**. *Journal of chemical theory and computation* (2010) **6** 459-66. DOI: 10.1021/ct900549r
28. Mark P, Nilsson L. **Structure and dynamics of the TIP3P, SPC, and SPC/E water models at 298 K**. *The Journal of Physical Chemistry A* (2001) **105** 9954-60
29. Bussi G, Donadio D, Parrinello M. **Canonical sampling through velocity rescaling**. *The Journal of chemical physics* (2007) **126** 014101. DOI: 10.1063/1.2408420
30. Nosé S, Klein M. **Constant pressure molecular dynamics for molecular systems**. *Molecular Physics* (1983) **50** 1055-76
31. Hess B.. **P-LINCS: A parallel linear constraint solver for molecular simulation**. *Journal of Chemical Theory and Computation* (2008) **4** 116-22. PMID: 26619985
32. Sagui C, Darden TA. **Molecular dynamics simulations of biomolecules: long-range electrostatic effects**. *Annual review of biophysics and biomolecular structure* (1999) **28** 155-79. DOI: 10.1146/annurev.biophys.28.1.155
33. Humphrey W, Dalke A, Schulten K. **VMD: visual molecular dynamics**. *Journal of molecular graphics* (1996) **14** 33-8. DOI: 10.1016/0263-7855(96)00018-5
34. David CC, Jacobs DJ. *Protein dynamics* (2014) 193-226
35. Tanwar H, George Priya Doss C. **An Integrated Computational Framework to Assess the Mutational Landscape of α‐L‐Iduronidase IDUA Gene**. *Journal of Cellular Biochemistry* (2018) **119** 555-65. PMID: 28608934
36. MIu L, Bogatyreva N, Galzitskaia O. **Radius of gyration is indicator of compactness of protein structure**. *Molekuliarnaia biologiia* (2008) **42** 701-6. PMID: 18856071
37. van Rappard DF, Boelens JJ, van Egmond ME, Kuball J, van Hasselt PM, Oostrom KJ. **Efficacy of hematopoietic cell transplantation in metachromatic leukodystrophy: the Dutch experience**. *Blood, The Journal of the American Society of Hematology* (2016) **127** 3098-101. DOI: 10.1182/blood-2016-03-708479
38. van der Knaap MS, Schiffmann R, Mochel F, Wolf NI. **Diagnosis, prognosis, and treatment of leukodystrophies**. *The Lancet Neurology* (2019) **18** 962-72. DOI: 10.1016/S1474-4422(19)30143-7
39. Di Resta C, Galbiati S, Carrera P, Ferrari M. **Next-generation sequencing approach for the diagnosis of human diseases: open challenges and new opportunities**. *Ejifcc* (2018) **29** 4. PMID: 29765282
40. Mahdieh N, Sharifi A, Rabbani A, Ashrafi M, Tavasoli AR, Badv RS. **Novel disease-causing variants in a cohort of Iranian patients with metachromatic leukodystrophy and in silico analysis of their pathogenicity**. *Clinical Neurology and Neurosurgery* (2021) **201** 106448. DOI: 10.1016/j.clineuro.2020.106448
41. Montgomery SB, Goode DL, Kvikstad E, Albers CA, Zhang ZD, Mu XJ. **The origin, evolution, and functional impact of short insertion–deletion variants identified in 179 human genomes**. *Genome research* (2013) **23** 749-61. DOI: 10.1101/gr.148718.112
42. Yu H, Yan Y, Zhang C, Dalby PA. **Two strategies to engineer flexible loops for improved enzyme thermostability**. *Scientific reports* (2017) **7** 1-15. PMID: 28127051
43. Savojardo C, Manfredi M, Martelli PL, Casadio R. **Solvent accessibility of residues undergoing pathogenic variations in humans: from protein structures to protein sequences**. *Frontiers in molecular biosciences* (2021) **7** 460. DOI: 10.3389/fmolb.2020.626363
44. Mohammadian H, Mahnam K, Sadeghi HM, Ganjalikhany MR, Akbari V. **Rational design of a new mutant of tobacco etch virus protease in order to increase the in vitro solubility**. *Research in Pharmaceutical Sciences* (2020) **15** 164. DOI: 10.4103/1735-5362.283816
45. Batori V, Koide A, Koide S. **Exploring the potential of the monobody scaffold: effects of loop elongation on the stability of a fibronectin type III domain**. *Protein engineering* (2002) **15** 1015-20. DOI: 10.1093/protein/15.12.1015
46. Khan MT, Ali S, Zeb MT, Kaushik AC, Malik SI, Wei D-Q. **Gibbs free energy calculation of mutation in pnca and rpsa associated with pyrazinamide resistance**. *Frontiers in molecular biosciences* (2020) **7** 52. DOI: 10.3389/fmolb.2020.00052
47. Amir M, Mohammad T, Kumar V, Alajmi MF, Rehman M, Hussain A. **Structural analysis and conformational dynamics of STN1 gene mutations involved in coat plus syndrome**. *Frontiers in molecular biosciences* (2019) **6** 41. DOI: 10.3389/fmolb.2019.00041
48. Studer RA, Dessailly BH, Orengo CA. **Residue mutations and their impact on protein structure and function: detecting beneficial and pathogenic changes**. *Biochemical journal* (2013) **449** 581-94. DOI: 10.1042/BJ20121221
|
---
title: Identification of shared gene signatures and molecular mechanisms between chronic
kidney disease and ulcerative colitis
authors:
- Zhou Liang
- Xinrong Hu
- Ruoni Lin
- Ziwen Tang
- Ziyin Ye
- Ren Mao
- Wei Chen
- Yi Zhou
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9970095
doi: 10.3389/fimmu.2023.1078310
license: CC BY 4.0
---
# Identification of shared gene signatures and molecular mechanisms between chronic kidney disease and ulcerative colitis
## Abstract
### Background
There is a complex interaction between chronic kidney disease (CKD) and ulcerative colitis (UC), but the pathophysiological mechanisms underlying the coexistence of CKD and UC are unclear. This study aimed to investigate the key molecules and pathways that may mediate the co-occurrence of CKD and UC through quantitative bioinformatics analysis based on a public RNA-sequencing database.
### Methods
The discovery datasets of CKD (GSE66494) and UC (GSE4183), as well as validation datasets of CKD (GSE115857) and UC (GSE10616), were downloaded from the Gene Expression Omnibus (GEO) database. After identifying differentially expressed genes (DEGs) with GEO2R online tool, the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses for the DEGs were performed. Next, protein-protein interaction network was constructed with Search Tool for the Retrieval of Interacting Genes (STRING) and visualized by Cytoscape. Gene modules were identified by the plug-in MCODE and hub genes were screened using the plug-in CytoHubba. Then, correlation between immune cell infiltration and hub genes was analyzed, and the receiver operating characteristic curves were used to assess the predictive value of hub genes. Finally, immunostaining of human specimens was used to validate the relevant findings.
### Results
A total of 462 common DEGs were identified and selected for further analyses. GO and KEGG enrichment analyses indicated that these DEGs were primarily enriched in immune- and inflammation-related pathways. Among them, the PI3K-Akt signaling pathway ranked top in both discovery and validation cohorts, and the key signal molecule phosphorylated Akt (p-Akt) was shown to be significantly overexpressed in human CKD kidneys and UC colons, and further elevated in CKD-UC comorbidity specimens. Moreover, nine candidate hub genes, including CXCL8, CCL2, CD44, ICAM1, IL1A, CXCR2, PTPRC, ITGAX, and CSF3, were identified, of which ICAM1 was validated as a common hub gene. Besides, immune infiltration analysis revealed that neutrophils, macrophages, and CD4+ T memory cells significantly accumulated in both diseases, and ICAM1 was remarkably associated with neutrophil infiltration. Furthermore, intercellular adhesion molecule1 (ICAM1)-mediated neutrophil infiltration was validated to be upregulated in kidney and colon biopsies of CKD and UC patients, and further increased in patients diagnosed with both CKD and UC. Finally, ICAM1 had shown critical value as a diagnostic marker for the co-occurrence of CKD and UC.
### Conclusions
Our study elucidated that immune response, PI3K-Akt signaling pathway, and ICAM1-mediated neutrophil infiltration might be the common pathogenesis of CKD and UC, and identified ICAM1 as a key potential biomarker and therapeutic target for the comorbidity of these two diseases.
## Introduction
Chronic kidney disease (CKD) refers to a progressive incurable disease characterized by immune dysregulation and multisystem involvement [1]. Growing evidence indicates a complex interplay between CKD and gut dysfunction (2–14). Studies revealed that ulcerative colitis (UC), a chronic inflammatory bowel disease affecting the colon and rectum, has a prevalence ranging from 0 to $4.4\%$ in IgA Nephropathy (IgAN, one of the leading causes of CKD), far higher than the highest morbidity of $0.505\%$ in the general population [3, 15]. Moreover, IgAN patients with UC displayed more severe renal injury with more proliferation of mesangial cells in renal biopsies [4]. Meanwhile, approximately $12\%$ of UC patients suffered from CKD, which is 2.46-fold higher than that of healthy individuals, making CKD one of the most common extraintestinal manifestations (EIMs) of UC [5, 7]. Evidently, there exists a strong connection between the occurrence of CKD and UC. However, the mechanisms underlying this phenomenon remain to be explored.
Shared pathological mechanisms, especially immune-mediated inflammation, might explain this clinical observation in CKD and UC. For UC, structural or functional impairment of the intestinal epithelial barrier drives an antigen-activated inflammatory cascade, triggering antigen-presenting cells such as dendritic cells initially [16]. As a result, neutrophils are attracted to the inflammatory sites by chemokines and exert pro-inflammatory effects via producing IL-23 and neutrophil extracellular traps (NETs) [17]. Additionally, immune responses mediated by lymphocytes play crucial roles in UC as well [17, 18]. IL-5, IL-6, IL-13, and tumor necrosis factor (TNF) produced by type 2 T helper cells (Th2s), and IL-17 derived from type 17 T helper cells (Th17s) as well as group 3 innate lymphoid cells (ILC3s) can exacerbate colonic barrier dysfunction. Likewise, in CKD, injured tubular epithelial cells secrete inflammatory mediators to recruit immune cells, including macrophages, neutrophils and lymphocytes, which amplify the inflammation and promote renal function failure [19, 20]. However, further researches are needed to illustrate the core signaling and associated immunological features in the co-occurrence of these two diseases.
Therefore, our study aimed to investigate the common pathogenic mechanism of CKD and UC, providing insights into both renal EIMs in UC and intestinal immune disorders in CKD patients. *Public* gene expression database Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) was employed to identify shared DEGs between CKD and UC for further analyses. We revealed that inflammatory responses and PI3K-Akt signaling pathway were of great importance in both CKD and UC. Besides, we demonstrated that ICAM1 was a hub gene of DEGs and analyzed its association with immune infiltration. All of the above results were confirmed in validation cohorts and verified in human specimens. Notably, this is the first study to elucidate shared genetic signatures and molecular mechanisms between CKD and UC, which brings a novel angle to the intrinsic link between these two diseases.
## Dataset collection
GEO (www.ncbi.nlm.nih.gov/geo/) is a huge public genomics data repository containing gene expression profiles of various diseases. We collected the datasets via using ulcerative colitis (UC) and chronic kidney disease (CKD) as keywords for search in the GEO database. The inclusion criteria were as follows: 1) The dataset contained at least fifteen samples in total; 2) *The* gene expression profiles were from adult human; 3) The included test samples should be from both patients and healthy individuals; 4) *Raw data* must be provided for further exploration. Finally, GSE66494 (CKD), GSE4183 (UC), GSE115857 (CKD), and GSE10616 (UC) were selected for further analysis (Table 1). For datasets GSE4183 and GSE10616, only samples from UC patients were included in the analysis. The GSM numbers of samples used in the analysis were provided in Supplementary Table 1.
**Table 1**
| No. | GSE number | Platform | Samples | Source type | Disease | Group |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | GSE66494 | GPL6480 | 53 patients and 8 controls | Renal biopsy | CKD | Discovery |
| 2 | GSE4183 | GPL570 | 9 patients and 8 controls | Colonic biopsy | UC | Discovery |
| 3 | GSE115857 | GPL14951 | 24 patients and 7 controls | Renal biopsy | CKD | Validation |
| 4 | GSE10616 | GPL5760 | 10 patients and 11 controls | Colonic biopsy | UC | Validation |
| 5 | GSE108112 | GPL19983 | 107 patients and 5 controls | Renal tubulointerstitial compartment biopsy | CKD | Test |
| 6 | GSE200818 | GPL19983 | 188 patients and 5 controls | Renal tubulointerstitial compartment biopsy | CKD | Test |
| 7 | GSE87466 | GPL13158 | 87 patients and 21 controls | Colon mucosal biopsy | UC | Test |
| 8 | GSE47908 | GPL570 | 39 patients and 15 controls | Colon mucosal biopsy | UC | Test |
## Identification of common DEGs in both CKD and UC
Differentially expressed genes (DEGs) between diseased group and control group were identified using GEO2R (www.ncbi.nlm.nih.gov/geo/geo2r/) online analysis tool based on R packages (GEOquery and Limma). Genes with p-value < 0.05 and |log2 fold change (log2FC)|>1 were defined as DEGs. The R package “ggplot” was employed to visualize DEGs from datasets using volcano maps. The online Venn diagram tool was used to extract common DEGs both up-regulated or down-regulated between GSE66494 and GSE4183, or GSE115857 and GSE10616 respectively. Additionally, we chose GSE66494 and GSE4183 as our discovery datasets, and GSE115857 and GSE10616 as our validation datasets.
## Functional enrichment analysis
The above shared genes were submitted to the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (https://david.ncifcrf.gov/) online tool for further Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. The enriched GO and KEGG pathways with p-value < 0.05 were selected and visualized using bar plots and bubble plots respectively.
## Establishment of protein-protein interaction network
Based on the overlapping DEGs, the Search Tool for the Retrieval of Interacting Genes (STRING) (http://string-db.org/) was employed for protein-protein interaction (PPI). The interaction sources of Textmining, Experiments, Databases, Co−expression, Neighborhood, Gene Fusion and Co−occurrence were adopted. The PPI pairs were extracted with an interaction score > 0.7 and then visualized by Cytoscape 3.9.0. The Cytoscape’s plug-in molecular complex detection (MCODE) was applied to the PPI network to explore key functional modules with selection criteria as follows: K-core = 2, degree cutoff = 2, max depth = 100, and node score cutoff = 0.2. And then, KEGG and GO analyses of the identified modular genes were performed using DAVID online tool and visualized by bubble plots and bar plots respectively.
## Selection and analysis of hub genes
The CytoHubba plug-in of Cytoscape identified hub genes from the PPI network, and then five algorithms (Radiality, MNC {Maximum Neighborhood Component}, MCC {Maximal Clique Centrality}, EPC {Edge Percolated Component}, and Degree) were used to confirm the final hub genes, which were illustrated in Venn diagram. Then, the hub genes were validated in GSE115857 and GSE10616. Ultimately, the above hub genes were submitted to GeneMANIA (http://genemania.org) to construct a co-expression network.
## Receiver operating characteristic curves of hub gene
Receiver Operating Characteristic (ROC) curves were constructed using GraphPad Prism 9, and the area under the ROC curve (AUC) was calculated to assess the predictive value of the hub gene in four testing datasets, including GSE108112 (CKD), GSE200818 (CKD), GSE87466 (UC) and GSE47908 (UC).
## Immune infiltration analysis
To further reveal the immune cell landscape in renal and colonic biopsy, we employed the analytical platform CIBERSORT (https://cibersort.stanford.edu/) and LM22 signature to decode immune cell infiltration profiles of discovery and validation datasets. Pearson correlation analysis was used to identify the relationship between different immune cell phenotypes and hub genes, which was illustrated in lollipop plots.
## Human biopsy specimens
Kidney biopsies in this study were obtained from patients with CKD-UC comorbidity ($$n = 8$$), patients with CKD without reported comorbid bowel disease (including UC) ($$n = 8$$) or living donors ($$n = 8$$). Colon biopsies were derived from patients with CKD-UC ($$n = 8$$), UC patients without kidney disease ($$n = 8$$) or normal tissues from colonoscopy ($$n = 8$$). Samples were obtained from the Department of Nephrology or Department of Pathology, The First Affiliated Hospital of Sun Yat-sen University. All participants provided informed consent and this study was approved by the Ethical Committee of the First Affiliated Hospital of Sun Yat-Sen University.
## Immunofluorescence staining and confocal microscopy
To detect p-Akt, ICAM1 and myeloperoxidase (MPO) in human specimens, slides were processed to remove paraffin and then hydrated in alcohol and phosphate-buffered saline. Then the slides were blocked with $10\%$ normal donkey serum (Sigma, D9663) and incubated with primary antibodies, respectively, at 4°C overnight, including ICAM1 (Santa Cruz, sc-8439, 1:50), p-Akt (CST, 4060S, 1:50), and MPO (abcam, ab208670, 1:150). Later, slides were stained with corresponding fluorescence-labeled secondary antibodies for 1h at room temperature, including anti-rabbit Alexa Fluor 488 (Thermo Fisher, A21206, 1:1000) and anti-mouse Alexa Fluor 546 (Thermo Fisher, A11030, 1:1000). Nuclei were stained with DAPI. Slides were then mounted with Prolong Gold Antifade reagent (Invitrogen, P36934), and examined on a ZEISS LSM880 confocal microscope.
## Statistical analysis
Unpaired t-test was used to detect the difference between two groups. And comparison among more than two groups of immunofluorescence results was assessed by the analysis of variance (ANOVA) using the statistical software GraphPad Prism and P values < 0.05 were considered statistically significant. Statistical significance is defined as * $p \leq 0.05$, ** $p \leq 0.01$, *** $p \leq 0.001$, **** $p \leq 0.0001$, or ns (not significant).
## Identification of common DEGs between CKD and UC
As shown in the workflow chart summarized in Figure 1, CKD microarray dataset GSE66494 and UC microarray dataset GSE4183 were downloaded from the GEO database. Details of selected datasets are provided in Table 1. Subsequently, DEGs were identified (6604 in GSE66494 and 2473 in GSE4183) after screening with the criteria of p-value < 0.05 and |log2FC| > 1, which were illustrated by the volcano plots in Figures 2A, B, respectively. In addition, a total of 784 shared differentially expressed genes were found after taking the intersection of CKD DEGs and UC DEGs based on Venn diagram analysis (Figure 2C). Finally, 462 co-up- or down-regulated DEGs in GSE66494 and GSE4183 were obtained after excluding genes showing opposite expression trends.
**Figure 1:** *Research design flow chart.* **Figure 2:** *Identification of common DEGs. (A) Volcano plot revealed 6604 DEGs between CKD patients and healthy controls. (B) Volcano plot revealed 2473 DEGs between UC patients and healthy controls. (C) A total of 784 common DEGs were identified after taking the intersection of DEGs in CKD and UC. DEG, differentially expressed gene; CKD, chronic kidney disease; UC, ulcerative colitis.*
## Functional enrichment of common DEGs
To decipher the biological functions of the common DEGs, GO and KEGG pathway enrichment analyses were performed on 462 shared genes between CKD and UC using the DAVID online tool. For the biological processes of GO enrichment analysis, these genes were primarily enriched in immune-related biological processes, such as inflammatory response, immune response, neutrophil chemotaxis, and chemokine-mediated signaling pathway (Figure 3A). For cellular components and molecular functions, these DEGs were enriched in extracellular space, extracellular region, chemokine activity, cytokine activity, heparin binding, and receptor binding. In terms of KEGG pathway enrichment analysis, significantly enriched pathways included cytokine-cytokine receptor interaction, pathways in cancer, PI3K-Akt signaling pathway, chemokine signaling pathway, TNF signaling pathway, IL-17 signaling pathway and NF-kappa B (NFκB) signaling pathway (Figure 3B). Taken together, these results indicate that inflammatory response plays a vital role in both CKD and UC, and that extracellular immune mediators such as chemokines and cytokines and PI3K-Akt signaling pathway largely participate in the process.
**Figure 3:** *Functional enrichment analysis of common DEGs. (A) GO terms in biological process, cellular component, and molecular function were used for functional enrichment clustering analysis on common DEGs. (B) KEGG pathway analysis was performed on common DEGs. (C, D) Representative immunofluorescence staining images of p-Akt (green) of kidney (C) and colon (D) biopsies from HC, CKD, UC and CKD-UC patients. Cell nuclei were counterstained with DAPI (blue). Scale bars, 20 μm. The MFI of p-Akt was measured by Image (J) DEG, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; HC, healthy control; CKD, chronic kidney disease; UC, ulcerative colitis; MPO, myeloperoxidase; MFI, mean fluorescence intensity; p-Akt, phospho-Akt. Data in C and D were presented as mean ± SEM, ****P < 0.0001; ***P < 0.0005; **P < 0.005 (ANOVA).*
To confirm our findings, we chose GSE115857 for CKD and GSE10616 for UC as validation datasets, the details of which are presented in Table 1. Initially, DEGs with p-value < 0.05 and |log2FC| > 1 were extracted from these two datasets (2001 in GSE115857 and 1450 in GSE10616). Then, Venn diagram analysis was performed on DEGs of the two groups, and a total of 155 common DEGs were spotted (Supplementary Figure 1A). Afterward, genes with contrary expression trends were removed, and the remaining 117 DEGs were introduced to GO and KEGG enrichment analyses. For GO enrichment analysis, inflammatory response was enriched, and DEGs were mostly presented in extracellular space as well (Supplementary Figure 1B). In terms of KEGG pathway enrichment analysis, PI3K-Akt signaling also ranked top as displayed in Supplementary Figure 1C.
Research on PI3K-Akt signaling pathway mainly focused on Akt, which is the most important downstream effector in the PI3K-Akt pathway, and phosphorylation is indispensable for Akt activation [21, 22]. Therefore, to validate the PI3K-Akt signaling pathway across disease states, we performed immunofluorescence staining to detect phosphorylated Akt (p-Akt) levels in colon and kidney biopsies from healthy individuals, CKD patients, UC patients, and those suffering from both diseases (CKD-UC). As shown in Figure 3C, p-Akt was sparsely expressed in healthy kidney tissues and significantly increased in tubular epithelial cells (TECs) of CKD renal biopsies. Moreover, the expression level of p-Akt was further elevated in CKD patients who also developed UC. An interesting parallel occurred in the colon, where p-Akt, nearly unexpressed in normal human colon, was highly upregulated in colon biopsies from UC patients. A more abundant expression of p-Akt was detected in colon tissues of patients diagnosed with both UC and CKD (Figure 3D). To sum all, these findings vigorously indicate that immune regulation and PI3K-Akt signaling pathway might play crucial roles in the occurrence and progression of these two diseases.
## Protein-protein interaction network analysis and submodule analysis
In order to further identify the potential relationships between proteins encoded by the shared DEGs of CKD and UC, the PPI network with an interaction score > 0.7 was performed by STRING and visualized with Cytoscape, including 438 nodes and 317 edges (Figure 4A). Next, the MCODE plug-in was used to detect significant gene clustering modules. Eight modules consisting of 52 common DEGs and 118 interaction pairs were extracted, and the top three significant modules were shown in Figures 4B-D. GO analysis of these modular genes revealed that these modules were mainly enriched in inflammatory response, chemokine-mediated signaling pathway and chemotaxis (Figure 4E). KEGG enrichment analysis showed that these modules were highly associated with cytokine-cytokine receptor interaction, chemokine signaling pathway, TNF signaling pathway, and IL-17 signaling pathway (Figure 4F). Altogether, these data indicate that inflammation is important in both CKD and UC.
**Figure 4:** *PPI and modular analysis of common DEGs. (A) The PPI network of common DEGs. Orange indicates up-regulated genes, and blue-green indicates down-regulated genes. (B-D) Top three gene clustering modules in MCODE analysis. (E, F) GO and KEGG enrichment analysis of total modular genes. PPI, protein-protein interaction; DEG, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.*
## Identification and co-expression network analysis of hub gene
To investigate the potential genes playing critical roles both in the occurrence of CKD and UC, hub genes of common DEGs were determined by the CytoHubba plug-in. We calculated the scores of common DEGs by five algorithms (Radiality, MNC, MCC, EPC, and Degree) and then took the intersection of the top 20 hub genes obtained from different algorithms by Venn diagram analysis (Table 2). Finally, nine candidate hub genes were identified, including CXCL8, CCL2, CD44, ICAM1, IL1A, CXCR2, PTPRC, ITGAX, and CSF3 (Figure 5A). The full names and related functions of these candidate hub genes are shown in Table 3. Interestingly, the majority of these genes were involved in immune cell migration and adhesion, suggesting that the infiltration of inflammatory cells is crucial in the progression of both CKD and UC.
We then validated our findings in GSE115857 for CKD and GSE10616 for UC (Supplementary Figure 2A). After intersecting the candidate hub genes from discovery cohorts and validation cohorts, we obtained only one shared hub gene, ICAM1 (Figure 5B), which was shown to be significantly upregulated in both CKD and UC of validation cohorts (Supplementary Figures 2B, C), suggesting that ICAM1 may be of great importance in both diseases.
To further verify the reliability and clinical significance of the hub gene discovered by bioinformatics analysis, we obtained the colon and kidney biopsies from healthy individuals, CKD patients, UC patients, and CKD-UC comorbidity patients with the same grouping settings as above, and detected ICAM1 by immunofluorescence. As shown in Figures 5D, E, ICAM1 was rarely expressed in healthy kidney and colon tissues, while significantly upregulated in renal TECs and vascular endothelium of CKD patients and colonic mucosae of UC patients, consistent with previous reports [23, 24]. Strikingly, the expression levels of ICAM1 were further dramatically elevated in both kidney and colon biopsies from CKD-UC patients, thus confirming the important role of ICAM1 in comorbid features and pathogenic mechanisms. On this basis, we analyzed the co-expression network and related functions of ICAM1 based on the GeneMANIA database. ICAM1 possessed a complex interaction network with physical interaction of $77.64\%$, co-expression of $8.01\%$, predicted of $5.37\%$, co-localization of $3.63\%$, genetic interaction of $2.87\%$, pathway of $1.83\%$, shared protein domains of $0.60\%$ (Figure 5C). *Twenty* genes associated with ICAM1 were identified, mainly related to leukocyte migration, regulation of leukocyte migration, T cell activation involved in immune response, positive regulation of leukocyte migration, lymphocyte activation involved in immune response, and cellular extravasation. These results re-emphasize the importance of immune cell-related inflammation in these two diseases, and ICAM1 might be the key regulatory gene.
## Association between the hub gene and immune infiltration
To further determine which types of immune cells in the kidney and colon are related to both CKD and UC, the proportions of 22 types of immune cells within kidney and colon samples were quantified by CIBERSORT (Figure 6A). Compared to healthy controls, both CKD and UC were commonly characterized by a remarkable enhancement of neutrophils, M0 macrophages (Mφ0), M1 macrophages (Mφ1), and CD4+ T memory cells (Figures 6A-D).
**Figure 6:** *Composition of infiltrating immune cells in CKD and UC. (A) The proportion of immune cell populations in kidney was determined by CIBERSORT. (B) Comparison of renal immune infiltration between healthy control and CKD patients. (C) The proportion of immune cell populations in colon was determined by CIBERSORT. (D) Comparison of colonic infiltrating immune cells between healthy control and UC patients. Stacked bar plots show the relative composition of immune cell subsets in CKD (A) and UC (B). CKD, chronic kidney disease; UC, ulcerative colitis; ****P < 0.0001; **P < 0.005; *P < 0.05; (unpaired t-test).*
Furthermore, the relationships between these immune cells and ICAM1 expression levels were analyzed by Pearson correlation analysis. Intriguingly, the expression level of ICAM1 was positively correlated with neutrophils, Mφ0, Mφ1, CD4+ T memory cells, activated mast cells, gamma delta (γδ) T cells, and eosinophils, while negatively linked with regulatory T cells, naive B cells, plasma cells, resting NK cells and resting mast cells in both CKD and UC (Figures 7A, B). These matching trends of immune infiltration indicate that the upregulation of ICAM1 resulted in a similar immune cell landscape in the two diseases. In particular, neutrophils, Mφ0, Mφ1, and activated CD4+ T memory cells were the top four significant immune cells positively associated with ICAM1, suggesting that ICAM1 may specifically regulate these four types of immune cells in CKD and UC. The correlation patterns were highly consistent in validated cohorts (Supplementary Figure 3).
**Figure 7:** *Correlation of hub gene and immune cell infiltration in CKD and UC. (A, B) Correlation of ICAM1 expression level and immune cell subtypes in GSE 66494 of CKD (A) and GSE4183 of UC (B). (C, D) Representative immunofluorescence images of ICAM1 (green) and neutrophil marker MPO (red) of renal (C) and colon (D) biopsies from HC, CKD, UC, CKD-UC patients. Cell nuclei were counterstained with DAPI (blue). Scale bars, 20 μm. The MFI of ICAM1 was measured by Image J. The quantitative analysis on the percentage of neutrophils per visual field was performed using Image J. (E, F) The MFI of ICAM1 was measured by Image J. The relationship between MFI of ICAM1 and neutrophils in renal (E) and colon (F) biopsies was analyzed with Pearson’s correlation analysis. HC, healthy control; CKD, chronic kidney disease; UC, ulcerative colitis; MPO, myeloperoxidase; MFI, mean fluorescence intensity. Data in C and D were presented as mean ± SEM, ****P < 0.0001; ***P < 0.0005; **P < 0.005; *P < 0.05; (ANONA).*
ICAM1 is a well-known adhesion receptor regulating leukocyte recruitment. In particular, existing evidence suggested that ICAM1 primarily mediates the infiltration of neutrophils in CKD and UC [23, 25]. Therefore, we validated the relationship between ICAM1 and myeloperoxidase (MPO) positive neutrophil infiltration in colon and kidney biopsies from healthy controls, CKD patients, UC patients and CKD-UC patients. As expected, ICAM1 expression levels and neutrophil infiltration were both markedly increased in the colon and kidney tissues of UC patients and CKD patients compared with healthy controls, and were further upregulated in CKD-UC patients (Figures 7C, D). Most importantly, there was a significant positive correlation between the expression levels of ICAM1 and neutrophil infiltration, which was more pronounced in comorbid patients (Figure 7E, F). These results suggest that ICAM1 and subsequent neutrophil infiltration mediated by it might play an important role in the co-occurrence of CKD and UC.
## ROC curves of hub gene
Next, we employed ROC curve analysis to explore whether hub gene ICAM1 possesses diagnostic efficiency in the four testing datasets GSE108112 (CKD), GSE200818 (CKD), GSE87466 (UC) and GSE47908 (UC). Details of these datasets are provided in Table 1. The AUC values in these four datasets were greater than 0.7, demonstrating that the predictive capabilities of ICAM1 are excellent. Specifically, in GSE108112 (AUC: 0.8106) and GSE200818 (AUC: 0.7754), ICAM1 displayed an excellent performance in distinguishing CKD patients from healthy controls (Figures 8A, B). Likewise, ICAM1 worked well to separate UC patients from healthy individuals in GSE87466 (AUC:0.9573) and GSE47908 (AUC:0.8274) (Figures 8C, D).
**Figure 8:** *ROC curves of the hub gene in CKD and UC. ROC curves were drawn to evaluate the accuracy of ICAM1 in diagnosing CKD (A, B) or UC (C, D). ROC: receiver operating characteristic; CKD, chronic kidney disease; UC, ulcerative colitis.*
## Discussion
A co-occurrence of CKD and inflammatory bowel disease (IBD), including UC and Crohn’s disease (CD), has long been observed. CKD patients have a higher prevalence of IBD, while renal dysfunction is one of the most common extraintestinal manifestations of IBD [26, 27]. Interestingly, it was observed that patients with UC had a significantly higher risk of renal disease than CD patients, suggesting a stronger relationship between UC and CKD. For example, a statistically increased risk for kidney disease was observed in UC but not in CD patients [5, 28, 29]. Moreover, CKD patients were more prone to develop UC (prevalence ratio {PR} of 2.46, $95\%$ confidence interval {CI} of 1.40-4.35) than CD (PR, 1.30; $95\%$ CI, 0.75-2.27) [5]. In patients with specific etiologies, such as IgAN, a leading cause of CKD, comorbidity of UC was also more common than CD [28]. Notably, although multiple etiologies can lead to CKD, including IgAN, diabetic nephropathy, glomerulonephritis, etc., the jury is still out as to which type of CKD is more frequently comorbid with IBD. In IBD patients, the most common renal pathological diagnosis varies across studies [30, 31], with amyloidosis, IgAN and tubulointerstitial nephritis being the most reported. For these reasons, we focused specifically on the link between UC and the onset of CKD, rather than a certain etiological type of CKD. To date, the molecular mechanisms of the connection between CKD and UC remain obscure. In the present study, we sought to investigate the shared gene features and molecular pathways between CKD and UC from a bioinformatics perspective based on the existing sequencing database.
A total of 462 common DEGs with identical expression trends in CKD and UC were identified and subjected to functional enrichment analysis. GO and KEGG analyses revealed that these DEGs were primarily enriched in inflammatory-related pathways, such as inflammatory response, chemokine activity, neutrophil chemotaxis, cytokine-cytokine receptor interaction, TNF and IL-17 signaling pathways. These discoveries underlined the importance of inflammation in the onset and development of CKD and UC. Notably, PI3K-Akt signaling pathway ranked top in both discovery and validation cohorts, suggesting its unique position in the pathogenesis of the two diseases. PI3K-Akt pathway is an intracellular signal transduction pathway involved in multiple biological processes, including metabolism, proliferation, cell survival, growth and angiogenesis [32]. The role of PI3K-Akt pathway is well established in cancer, while its importance has also been recognized in various non-cancer diseases (33–35). In CKD, PI3K-Akt signaling pathway mediates oxidized low-density lipoprotein (ox-LDL)-induced nephron loss in podocytes, resulting in renal injury [36]. While in UC, inhibition of PI3K-Akt alleviated disruptive epithelial barrier integrity in a mouse model of Dextran Sulfate Sodium (DSS)-induced colitis [33]. Additionally, PI3K-Akt pathway also contributes to immune regulation, and is involved in neutrophil chemotaxis and phagocytosis, B cell receptor signaling activation, and dendritic cells maturation (37–39). Here, we observed the expression levels of p-Akt, which represents the activation of PI3K-Akt pathway [22], and found that it was significantly upregulated in kidney tissues of CKD patients and colon tissues of UC patients. Even more impressively, the PI3K-Akt pathway was further activated in kidney and colon biopsies from patients with CKD-UC comorbidity. It is worth emphasizing here that our study obtained both colon and kidney tissue from the same patients with diagnostic comorbidity, and observed identical trends in validated signals within both tissues, which is of particular importance in supporting our proposed pathogenic mechanism for the comorbidity. Taken together, upregulation of the PI3K-Akt pathway might facilitate the progression of immune responses in these two diseases, promoting the co-occurrence of CKD and UC.
Subsequently, according to the CytoHubba plug-in of Cytoscape, we screened nine candidate hub genes, including CXCL8, CCL2, CD44, ICAM1, IL1A, CXCR2, PTPRC, ITGAX, and CSF3, which were all elevated in both CKD and UC patients. Their respective functions are described below. CXCR2, CXCL8 and CCL2 are chemokine receptor and ligands involved in immunoregulatory and inflammatory responses [40]. CXCR2 is the receptor for CXCL8, acting as a robust chemotactic factor for neutrophil recruitment and activation [41]. CCL2 displays chemotactic activity for monocytes and basophils [42]. CD44, ICAM1, and integrin alpha X (ITAX) are adhesion molecules that participate in cell-cell binding and interaction during inflammation or tumor (43–45). IL-1α is a pleiotropic cytokine produced primarily by monocytes and macrophages and is involved in various immune responses [46]. Colony stimulating factor 3 (CSF-3) plays major roles in the proliferation, differentiation, and activation of neutrophil cell line hematopoietic cells [47]. The identified functions of the above molecules highlighted the role of neutrophils and macrophages in the pathogenesis of CKD and UC. Congruously, taking immune infiltration analysis and histological staining, we revealed that neutrophils and macrophages were significantly accumulated in both CKD and UC biopsies. Collectively, our data indicated that these candidate hub genes regulated the infiltration of neutrophils and macrophages, subsequently resulting in an imbalanced immune response, which played conspicuous roles in both CKD and UC progression.
To further identify the common hub gene, we intersected the candidate hub genes from discovery cohorts and validation cohorts. Ultimately, we found ICAM1 might play an extremely important role in the progression of both CKD and UC. ICAM1 is a cell surface glycoprotein and adhesion receptor primarily expressed by endothelial cells (ECs), epithelial cells and some immune cells [44, 48, 49]. The best-characterized function of ICAM1 is to regulate leukocyte recruitment from the circulation to sites of inflammation and mediate their migration across the endothelium [50, 51]. Previous studies have demonstrated a pathogenic role of ICAM1 in UC [52, 53]. Colonic endothelial cells are by and large ICAM1-negative under the homeostatic state, while are remarkably up-regulated with ICAM1 expression during colitis, consistent with what we discovered here. Targeting ICAM1 was shown to effectively alleviate inflammation, reduce bloody stools and anemia in murine model of colitis. The underlying mechanism was that ICAM1 inhibition prevented the infiltration of pathogenic neutrophils [24]. Moreover, clinical trials showed that the antisense oligonucleotide inhibitor of ICAM1, Alicaforsen, was effective and durable in treating UC patients with an excellent safety profile [54]. For CKD, ICAM1 overexpression was also detected in human and murine injured kidneys and was found to promote renal dysfunction via potentiation of neutrophil-endothelial interactions (23, 55–57). In line with these reports, here we also discovered that ICAM1 showed a remarkably positive correlation with neutrophils, Mφ0, Mφ1, and activated CD4+ T memory cells in both CKD and UC. Given that ICAM1 contribute to the pathogenesis of CKD and UC via recruiting neutrophil as discussed above, we further validated the positive correlation between ICAM1 expression levels and neutrophil infiltration in human colon and kidney biopsies. Excitingly, the levels of ICAM1 and neutrophils in the colon and kidney biopsies from patients diagnosed with both CKD and UC were significantly higher than in biopsies from patients with CKD alone or UC alone. Therefore, ICAM1-mediated neutrophil infiltration might play a pivotal role in the pathophysiology of CKD and UC, and patients with a higher level of ICAM1 might be at risk for the two diseases.
It is worth noting that proinflammatory cytokines such as TNFα, IL-1β, and IL-17 were demonstrated to enhance ICAM1 expression via NFκB in ECs ex vivo [14, 58]. While, TNF signaling pathway, IL-17 signaling pathway, and NFκB signaling pathway were significantly enriched in our KEGG analysis of common DEGs, indicating the upregulation of ICAM1 in CKD and UC might be induced by TNFα- and IL-17-mediated NFκB activation in vivo. Consequently, the TNFα/IL-17-NFκB-ICAM1-neutrophil pathological pathway might be shared in these two diseases.
However, the present study also has some limitations. First of all, although we have observed the key pathways and hub molecules and immune cells identified in this study on human specimens, further laboratory studies, especially animal experiments, are still needed to verify their pathogenic importance. Secondly, the potential shared mechanisms of CKD and UC were based on common DEGs in patients with CKD or UC, rather than patients with both diseases. As such datasets are not available, validation with datasets including patients with both CKD and UC is not currently possible and should be carried out in the future. At last, we did not explore in the current analysis whether there are shared genetic signatures between CKD and CD, the other major form of IBD. Thus, in order to clarify the specificity between different etiologies, the comorbidity relationship between them will be further investigated in the future.
In conclusion, this bioinformatics study identified shared gene signatures between CKD and UC, illustrating the potential molecular mechanisms of these two diseases. We revealed that an unbalanced immune response, PI3K-Akt signaling pathway, and ICAM1-mediated neutrophil infiltration might be the common pathogenesis of CKD and UC.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethical Committee of the First Affiliated Hospital of Sun Yat-Sen University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
ZL and XH designed and conducted the entire study and contributed to the bioinformatics analysis. ZL analyzed GEO dataset of UC and XH was in charge of GEO dataset of CKD. RL and ZT performed the laboratory experiments and data analysis. ZY. RM collected human colon biopsies, and WC collected human kidney biopsies. ZL. XH wrote the original manuscript, and WC. YZ revised and finalized the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1078310/full#supplementary-material
## References
1. Kalantar-Zadeh K, Jafar TH, Nitsch D, Neuen BL, Perkovic V. **Chronic kidney disease**. *Lancet* (2021) **398** 786-802. DOI: 10.1016/s0140-6736(21)00519-5
2. Moon JM, Kang EA, Han K, Hong SW, Soh H, Park S. **Trends and risk factors of elderly-onset crohn’s disease: A nationwide cohort study**. *World J Gastroenterol* (2020) **26**. DOI: 10.3748/wjg.v26.i4.404
3. Nurmi R, Pohjonen J, Metso M, Porsti I, Niemela O, Huhtala H. **Prevalence of inflammatory bowel disease and celiac disease in patients with iga nephropathy over time**. *Nephron* (2021) **145** 78-84. DOI: 10.1159/000511555
4. Joher N, Gosset C, Guerrot D, Pillebout E, Hummel A, Boffa JJ. **Immunoglobulin a nephropathy in association with inflammatory bowel diseases: Results from a national study and systematic literature review**. *Nephrol Dial Transplant* (2022) **37**. DOI: 10.1093/ndt/gfaa378
5. Bernstein CN, Wajda A, Blanchard JF. **The clustering of other chronic inflammatory diseases in inflammatory bowel disease: A population-based study**. *Gastroenterology* (2005) **129**. DOI: 10.1053/j.gastro.2005.06.021
6. Oikonomou K, Kapsoritakis A, Eleftheriadis T, Stefanidis I, Potamianos S. **Renal manifestations and complications of inflammatory bowel disease**. *Inflammation Bowel Dis* (2011) **17**. DOI: 10.1002/ibd.21468
7. Lewis B, Mukewar S, Lopez R, Brzezinski A, Hall P, Shen B. **Frequency and risk factors of renal insufficiency in inflammatory bowel disease inpatients**. *Inflammation Bowel Dis* (2013) **19**. DOI: 10.1097/MIB.0b013e31828a661e
8. Primas C, Novacek G, Schweiger K, Mayer A, Eser A, Papay P. **Renal insufficiency in ibd–prevalence and possible pathogenetic aspects**. *J Crohns Colitis* (2013) **7**. DOI: 10.1016/j.crohns.2013.05.001
9. Ambruzs JM, Walker PD, Larsen CP. **The histopathologic spectrum of kidney biopsies in patients with inflammatory bowel disease**. *Clin J Am Soc Nephrol* (2014) **9**. DOI: 10.2215/CJN.04660513
10. Lauritzen D, Andreassen BU, Heegaard NHH, Klinge LG, Walsted AM, Neland M. **Pediatric inflammatory bowel diseases: Should we be looking for kidney abnormalities**. *Inflammation Bowel Dis* (2018) **24**. DOI: 10.1093/ibd/izy166
11. Vajravelu RK, Copelovitch L, Osterman MT, Scott FI, Mamtani R, Lewis JD. **Inflammatory bowel diseases are associated with an increased risk for chronic kidney disease, which decreases with age**. *Clin Gastroenterol Hepatol* (2020) **18**. DOI: 10.1016/j.cgh.2019.10.043
12. Mutalib M. **Renal involvement in paediatric inflammatory bowel disease**. *Pediatr Nephrol* (2021) **36**. DOI: 10.1007/s00467-019-04413-5
13. Dincer MT, Dincer ZT, Bakkaloglu OK, Yalin SF, Trabulus S, Celik AF. **Renal manifestations in inflammatory bowel disease: A cohort study during the biologic era**. *Med Sci Monit* (2022) **28**. DOI: 10.12659/MSM.936497
14. van Hoeve K, Hoffman I. **Renal manifestations in inflammatory bowel disease: A systematic review**. *J Gastroenterol* (2022) **57**. DOI: 10.1007/s00535-022-01903-6
15. Ng SC, Shi HY, Hamidi N, Underwood FE, Tang W, Benchimol EI. **Worldwide incidence and prevalence of inflammatory bowel disease in the 21st century: A systematic review of population-based studies**. *Lancet* (2017) **390**. DOI: 10.1016/s0140-6736(17)32448-0
16. Kobayashi T, Siegmund B, Le Berre C, Wei SC, Ferrante M, Shen B. **Ulcerative colitis**. *Nat Rev Dis Primers* (2020) **6** 74. DOI: 10.1038/s41572-020-0205-x
17. Neurath MF. **Targeting immune cell circuits and trafficking in inflammatory bowel disease**. *Nat Immunol* (2019) **20**. DOI: 10.1038/s41590-019-0415-0
18. Castro-Dopico T, Dennison TW, Ferdinand JR, Mathews RJ, Fleming A, Clift D. **Anti-commensal igg drives intestinal inflammation and type 17 immunity in ulcerative colitis**. *Immunity* (2019) **50** 1099-114 e10. DOI: 10.1016/j.immuni.2019.02.006
19. Ruiz-Ortega M, Rayego-Mateos S, Lamas S, Ortiz A, Rodrigues-Diez RR. **Targeting the progression of chronic kidney disease**. *Nat Rev Nephrol* (2020) **16**. DOI: 10.1038/s41581-019-0248-y
20. Kitching AR, Hickey MJ. **Immune cell behaviour and dynamics in the kidney - insights from in vivo imaging**. *Nat Rev Nephrol* (2022) **18** 22-37. DOI: 10.1038/s41581-021-00481-9
21. Alzahrani AS. **Pi3k/Akt/Mtor inhibitors in cancer: At the bench and bedside**. *Semin Cancer Biol* (2019) **59**. DOI: 10.1016/j.semcancer.2019.07.009
22. Chua BT, Gallego-Ortega D, Ramirez de Molina A, Ullrich A, Lacal JC, Downward J. **Regulation of Akt(Ser473) phosphorylation by choline kinase in breast carcinoma cells**. *Mol Cancer* (2009) **8**. DOI: 10.1186/1476-4598-8-131
23. Kelly KJ, Williams WW, Colvin RB, Meehan SM, Springer TA, Gutierrez-Ramos JC. **Intercellular adhesion molecule-1-Deficient mice are protected against ischemic renal injury**. *J Clin Invest* (1996) **97**. DOI: 10.1172/JCI118498
24. Sumagin R, Robin AZ, Nusrat A, Parkos CA. **Transmigrated neutrophils in the intestinal lumen engage icam-1 to regulate the epithelial barrier and neutrophil recruitment**. *Mucosal Immunol* (2014) **7**. DOI: 10.1038/mi.2013.106
25. Thomas PD FA, Price AB, Nicholls RJ, Ciclitra PJ. **Differential expression of cell adhesion molecules within inflamed ileal pouch mucosa: Relationship to recruited cell subtypes**. *Eur J Gastroenterol Hepatol* (2002) **14**. DOI: 10.1097/00042737-200202000-00007
26. Hubert D Fau - Beaufils M, Beaufils M Fau - Meyrier A, Meyrier A. **Immunoglobulin a glomerular nephropathy associated with inflammatory colitis**. *Apropos 2 Cases Presse Med* (1984) **13**
27. Wilcox GM AH, Roy MA, Roche JK. **Glomerulonephritis associated with inflammatory bowel disease. report of a patient with chronic ulcerative colitis, sclerosing cholangitis, and acute glomerulonephritis**. *Gastroenterology* (1990) **98)**. DOI: 10.1016/0016-5085(90)90304-J
28. Rehnberg J, Symreng A, Ludvigsson JF, Emilsson L. **Inflammatory bowel disease is more common in patients with iga nephropathy and predicts progression of eskd: A Swedish population-based cohort study**. *J Am Soc Nephrol* (2021) **32**. DOI: 10.1681/ASN.2020060848
29. Herrlinger KR NM, Fellermann K, Schmidt K, Steinhoff J, Stange EF. **Minimal renal dysfunction in inflammatory bowel disease is related to disease activity but not to 5-asa use**. *Aliment Pharmacol Ther* (2001) **15**. DOI: 10.1046/j.1365-2036.2001.00940.x
30. Elaziz MMA FA. **Patterns of renal involvement in a cohort of patients with inflammatory bowel disease in Egypt**. *Acta Gastroenterol Belg* (2018) **81**
31. Pohjonen J, Nurmi R, Metso M, Oksanen P, Huhtala H, Porsti I. **Inflammatory bowel disease in patients undergoing renal biopsies**. *Clin Kidney J* (2019) **12**. DOI: 10.1093/ckj/sfz004
32. Hoxhaj G, Manning BD. **The Pi3k-akt network at the interface of oncogenic signalling and cancer metabolism**. *Nat Rev Cancer* (2020) **20** 74-88. DOI: 10.1038/s41568-019-0216-7
33. Burger C, Shirsath N, Lang V, Diehl S, Kaufmann R, Weigert A. **Blocking mtor signalling with rapamycin ameliorates imiquimod-induced psoriasis in mice**. *Acta Derm Venereol* (2017) **97**. DOI: 10.2340/00015555-2724
34. Qin W, Cao L, Massey IY. **Role of Pi3k/Akt signaling pathway in cardiac fibrosis**. *Mol Cell Biochem* (2021) **476**. DOI: 10.1007/s11010-021-04219-w
35. Song T, Chen WD. **Berberine inhibited carotid atherosclerosis through Pi3k/Aktmtor signaling pathway**. *Bioengineered* (2021) **12**. DOI: 10.1080/21655979.2021.1987130
36. Bussolati B, Deregibus MC, Fonsato V, Doublier S, Spatola T, Procida S. **Statins prevent oxidized ldl-induced injury of glomerular podocytes by activating the phosphatidylinositol 3-Kinase/Akt-Signaling pathway**. *J Am Soc Nephrol* (2005) **16**. DOI: 10.1681/ASN.2004080629
37. Jing Y DX, Yang L, Kang D, Jiang P, Li N, Cheng J. **Sting couples with Pi3k to regulate actin reorganization during bcr activation**. *Sci Adv* (2020) **6**. DOI: 10.1126/sciadv.aax9455
38. Pan T, Sun S, Chen Y, Tian R, Chen E, Tan R. **Immune effects of Pi3k/Akt/Hif-1α-Regulated glycolysis in polymorphonuclear neutrophils during sepsis**. *Crit Care* (2022) **26** 29. DOI: 10.1186/s13054-022-03893-6
39. Xiang M, Liu T, Tian C, Ma K, Gou J, Huang R. **Kinsenoside attenuates liver fibro-inflammation by suppressing dendritic cells**. *Pharmacol Res* (2022) **177**. DOI: 10.1016/j.phrs.2022.106092
40. Griffith JW, Sokol CL, Luster AD. **Chemokines and chemokine receptors: Positioning cells for host defense and immunity**. *Annu Rev Immunol* (2014) **32** 659-702. DOI: 10.1146/annurev-immunol-032713-120145
41. SenGupta S, Hein LE, Parent CA. **The recruitment of neutrophils to the tumor microenvironment is regulated by multiple mediators**. *Front Immunol* (2021) **12**. DOI: 10.3389/fimmu.2021.734188
42. Gschwandtner M, Derler R, Midwood KS. **More than just attractive: How Ccl2 influences myeloid cell behavior beyond chemotaxis**. *Front Immunol* (2019) **10**. DOI: 10.3389/fimmu.2019.02759
43. Chen C, Zhao S, Karnad A, Freeman JW. **The biology and role of Cd44 in cancer progression: Therapeutic implications**. *J Hematol Oncol* (2018) **11** 64. DOI: 10.1186/s13045-018-0605-5
44. Bui TM, Wiesolek HL, Sumagin R. **Icam-1: A master regulator of cellular responses in inflammation, injury resolution, and tumorigenesis**. *J Leukoc Biol* (2020) **108**. DOI: 10.1002/JLB.2MR0220-549R
45. Moreno-Layseca P, Icha J, Hamidi H, Ivaska J. **Integrin trafficking in cells and tissues**. *Nat Cell Biol* (2019) **21**. DOI: 10.1038/s41556-018-0223-z
46. Dinarello CA. **Overview of the il-1 family in innate inflammation and acquired immunity**. *Immunol Rev* (2018) **281** 8-27. DOI: 10.1111/imr.12621
47. Hamilton JA, Cook AD, Tak PP. **Anti-Colony-Stimulating factor therapies for inflammatory and autoimmune diseases**. *Nat Rev Drug Discovery* (2016) **16** 53-70. DOI: 10.1038/nrd.2016.231
48. Kumar S, Pollok R, Goldsmith D. **Renal and urological disorders associated with inflammatory bowel disease**. *Inflammation Bowel Dis* (2022). DOI: 10.1093/ibd/izac140
49. Lawson C WS. **Icam-1 signaling in endothelial cells**. *Pharmacol Rep* (2009) **61** 22-32. DOI: 10.1016/s1734-1140(09)70004-0
50. Lyck R, Enzmann G. **The physiological roles of icam-1 and icam-2 in neutrophil migration into tissues**. *Curr Opin Hematol* (2015) **22**. DOI: 10.1097/MOH.0000000000000103
51. Harjunpaa H, Llort Asens M, Guenther C, Fagerholm SC. **Cell adhesion molecules and their roles and regulation in the immune and tumor microenvironment**. *Front Immunol* (2019) **10**. DOI: 10.3389/fimmu.2019.01078
52. Reinisch W, Hung K, Hassan-Zahraee M, Cataldi F. **Targeting endothelial ligands: Icam-1/Alicaforsen, madcam-1**. *J Crohns Colitis* (2018) **12**. DOI: 10.1093/ecco-jcc/jjy059
53. Vainer B. **Intercellular adhesion molecule-1 (Icam-1) in ulcerative colitis: Presence, visualization, and significance**. *Inflammation Res* (2005) **54**. DOI: 10.1007/s00011-005-1363-8
54. Jairath V, Khanna R, Feagan BG. **Alicaforsen for the treatment of inflammatory bowel disease**. *Expert Opin Investig Drugs* (2017) **26**. DOI: 10.1080/13543784.2017.1349753
55. Sun W, Choi HS, Kim CS, Bae EH, Ma SK, Kim SW. **Maslinic acid attenuates Ischemia/Reperfusion-induced acute kidney injury by suppressing inflammation and apoptosis through inhibiting nf-kappab and mapk signaling pathway**. *Front Pharmacol* (2022) **13**. DOI: 10.3389/fphar.2022.807452
56. Tang G, Li S, Zhang C, Chen H, Wang N, Feng Y. **Clinical efficacies, underlying mechanisms and molecular targets of Chinese medicines for diabetic nephropathy treatment and management**. *Acta Pharm Sin B* (2021) **11**. DOI: 10.1016/j.apsb.2020.12.020
57. Zhang X, Seman NA, Falhammar H, Brismar K, Gu HF. **Genetic and biological effects of icam-1 E469k polymorphism in diabetic kidney disease**. *J Diabetes Res* (2020) **2020**. DOI: 10.1155/2020/8305460
58. Guillo L, Delanaye P, Flamant M, Figueres L, Karam S, Lemoine S. **Kidney function monitoring in inflammatory bowel disease: The monitored consensus**. *Dig Liver Dis* (2022) **54**. DOI: 10.1016/j.dld.2021.11.008
|
---
title: 'Patients’ preferences in dental care: A discrete-choice experiment and an
analysis of willingness-to-pay'
authors:
- Susanne Felgner
- Cornelia Henschke
journal: PLOS ONE
year: 2023
pmcid: PMC9970100
doi: 10.1371/journal.pone.0280441
license: CC BY 4.0
---
# Patients’ preferences in dental care: A discrete-choice experiment and an analysis of willingness-to-pay
## Abstract
### Introduction
Dental diseases are a major problem worldwide. Costs are a burden on healthcare systems and patients. Missed treatments can have health and financial consequences. Compared to other health services, dental treatments are only covered in parts by statutory health insurance (SHI). Using the example of dental crowns for a cost-intensive treatment, our study aims to investigate whether [1] certain treatment attributes determine patients’ treatment choice, and [2] out-of-pocket payments represent a barrier to access dental care.
### Methods
We conducted a discrete-choice-experiment by mailing questionnaires to 10,752 people in Germany. In presented scenarios the participants could choose between treatment options (A, B, or none) composed of treatment attribute levels (e.g., color of teeth) for posterior (PT) and anterior teeth (AT). Considering interaction effects, we used a D-efficient fractional factorial design. Choice analysis was performed using different models. Furthermore, we analyzed willingness-to-pay (WTP), preference of choosing no and SHI standard care treatment, and influence of socioeconomic characteristics on individual WTP.
### Results
Out of $$n = 762$$ returned questionnaires (response rate of $r = 7.1$), $$n = 380$$ were included in the analysis. Most of the participants are in age group "50 to 59 years" ($$n = 103$$, $27.1\%$) and female ($$n = 249$$, $65.5\%$). The participants’ benefit allocations varied across treatment attributes. Aesthetics and durability of dental crowns play most important roles in decision-making. WTP regarding natural color teeth is higher than standard SHI out-of-pocket payment. Estimations for AT dominate. For both tooth areas, "no treatment" was a frequent choice (PT: $25.7\%$, AT: $37.2\%$). Especially for AT, treatment beyond SHI standard care was often chosen ($49.8\%$, PT: $31.3\%$). Age, gender, and incentive measures (bonus booklet) influenced WTP per participant.
### Conclusion
This study provides important insights into patient preferences for dental crown treatment in Germany. For our participants, aesthetic for AT and PT as well as out-of-pocket payments for PT play an important role in decision-making. Overall, they are willing to pay more than the current out-of-pockt payments for what they consider to be better crown treatments. Findings may be valuable for policy makers in developing measures that better match patient preferences.
## Introduction
Understanding how patients assess various aspects of health care interventions is important for clinical, coverage, and policy decisions. As a result, considering patient preferences in health care decision-making and policy can improve utilization of interventions and public health programs, satisfaction with those, and patient adherence to finally improve effectiveness of health care services [1]. To identify preferences for various attributes of an intervention, stated preference methods such as the discrete-choice approach can be used, particularly to quantify stakeholders‘ preferences in health care. At the same time, this approach offers a mechanism for patients to participate in decision-making and may facilitate shared decision-making. In practice, the discrete-choice approach is also used to estimate willingness-to-pay for attributes, which is especially beneficial in the case of treatments where high co-payments may arise such as oral health services. Preferences of utilization of oral health services from patients’perspective have rarely been studied so far, although chronic and untreated dental diseases (e.g., caries) can lead to serious consequences such as pain, sepsis, reduced quality of life, and work productivity. This places a burden on patients in various aspects of their lives and on healthcare system in terms of capacity [2].
Although in *Germany a* broad range of oral health services is covered, high out-of-pocket payments may occur for patients. Regarding dentures, for instance, statutory health insurance (SHI) covers $60\%$ of the standard care costs, which can be defined as broad coverage compared to other countries [3, 4]. Incentive measures, such as regular dental check-ups within the last five or ten years before respective treatment, may increase the SHI’s coverage [5]. Additionally, patients may take out private dental supplementary insurances to reduce out-of-pocket payments [6], and choose treatments beyond defined standard care. Nevertheless, perceived unmet needs in dental care still exist. Households with highest budget (fifth household budget quintile) take higher out-of-pocket payments than households with lowest budget (first household budget quintile) [7].
While costs of dental treatments seem to play a major role in patients‘ decision-making [8], the choice of a treatment may also be influenced by individual preferences regarding further attributes. Patients may value certain treatment attributes differently [9, 10], such as color of a dental crown. Former studies have investigated patient preferences for dentures [11], caries prevention measures [12], and willingness-to-pay for medical tourism [13]. We focus on a prosthodontic treatment–the placement of a full dental crown–due to high variability in options and costs to be borne by the patients themselves. In Germany, SHI covers a fixed subsidy of $50\%$ ($60\%$ as of $\frac{10}{2020}$) for standard treatment of dental crowns. The remaining $50\%$ ($40\%$) have to be paid out of pocket by patients, plus the difference of costs when choosing superior materials. The attributes out-of-pocket payment and aesthetics vary greatly between different dental crown treatments. These assumed (un)desirable attributes for patients move proportionally against each other, i.e., an aesthetically pleasing dental crown is expensive and vice versa. The SHI alternative may implicate an aesthetically unattractive result to patients (i.e., darker-colored not natural appealing dental crown). To our knowledge, this treatment has not yet been studied for the German health care system using an experimental approach. This study investigates patient preferences presented by benefit allocations to treatment attributes for dental crown treatments in Germany addressing the following research questions:
## Methods
Discrete-choice experiments (DCE) are an established instrument particularly in health sciences [14] for measuring patient preferences in their choice behavior by estimating benefit assignments, and for calculating willingness-to-pay (WTP) taking out-of-pocket payments into account. Although, medical professionals usually recommend a treatment option, due to restricted coverage in oral health care services, final decisions are largely guided by patient preferences. A DCE is best suited to collect data for analyzing preferences of patients and their WTP. This is especially the case as patients have to decide between different treatment opportunities that come along with large variations in out-of-pocket payments. To analyze patient preferences, we therefore conducted a DCE. Furthermore, we analyzed overall (and individual) WTP to compare monetary value of respondents’ willingness-to-pay and the SHI out-of-pocket payment for each attribute. In addition, we conducted regression analyses to calculate the relationship between socio-economic and other characteristics of participants and defined decision variables. Descriptive analyses were used to illustrate quantitative results (S3, S4 and S6 Files).
## Experimental design & questionnaire
Prior to the study, a systematic literature review [15] and focus group interviews [8] were conducted to identify attributes that influence patients in their choice for or against dental treatments. In the DCE dental treatments are presented as a combination of attribute levels in choice sets. As those attributes and levels should be plausible, and clinically relevant, being as realistic as possible [16–18], we used most relevant treatment attributes identified. Levels for aesthetics, compatibility, durability [19], and out-of-pocket payment [5] were determined by research. We differentiate between two teeth areas "posterior teeth" (PT) and "anterior teeth" (AT), since different patient preferences can be assumed for it [20]. Attributes and levels were presented to (potential) participants in the questionnaire (S1 File: Questionnaire). For the attribute aesthetics, an extra document was created for visualization (S2 File: Document "Aesthetics"). Table 1 gives an overview of the treatment attributes and its levels.
**Table 1**
| Attribute | Definition | Levels | Levels.1 |
| --- | --- | --- | --- |
| 1. Aesthetics | In terms of appearance, result of treatment individually perceived as beautiful. This attribute describes the visibility of a dental crown. | ✓ Natural color✓ Lightly visible✓ Strongly visible | ✓ Natural color✓ Lightly visible✓ Strongly visible |
| 2. Compatibility | Intolerance reaction of human body due to dental material in form of an allergic or a local toxic reaction1. | ✓ No risk✓ 1 out of 10,000 people with allergic or local toxic reaction | ✓ No risk✓ 1 out of 10,000 people with allergic or local toxic reaction |
| 3. Durability | Expected length of time from completion of a treatment to another new treatment that is medically or technically necessary. | ✓ 5 years✓ 10 years✓ 15 years✓ 25 years | ✓ 5 years✓ 10 years✓ 15 years✓ 25 years |
| 4. Out-of-pocket payment | Costs that must be paid by patient for dental crown treatment. The co-payment taken by health insurance has already been subtracted here. | Posterior teeth✓ 50 €✓ 150 €✓ 450 €✓ 600 € | Anterior teeth✓ 50 €✓ 200 €✓ 450 €✓ 600 € |
Considering four attributes (x1-x4) with 3, 2, 4, and 4 levels, $$n = 9$$,216 possible choice sets resulted in a full factorial design (3x2x4x4 = 96; 96x96). Since this cannot be answered by an individual participant, a fractional factorial design was used [21] (S1 Table: Design output). Aiming at a $100\%$ D-efficient design [21], and assuming interaction effects between the attributes x1 and x4, and x3 and x4, $$n = 96$$ choice sets were necessary (S2 Table: Calculation of D-efficiency). Ensuring that the questionnaire was manageable for our study participants [1], $$n = 12$$ questionnaire blocks were formed and randomly assigned to the participants [17]. For the design calculation we used the statistical software SAS (version 9.2). The %ChoicEff macro was used to create the experimental design, considering interactions as constraints via restrictions = option, and the %MktEx macro was used to create the choice sets. Blocked design was realized via the %mktblock macro [22].
The paper-based questionnaire was divided into an introductory part including the informed consent and explaining information on attributes and levels (part A), the choice scenarios (choice sets) to be answered by the participants (part B), and questions regarding the participants‘ (socioeconomic) characteristics (part C). Alternatives of dental crown treatment were presented in eight choice sets each, separated in part B1 focusing on PT and B2 on AT. Studies report that experiments including up to $$n = 32$$ scenarios are manageable by participants [1, 23]. This is in line with our study using $$n = 18$$ choice scenarios per questionnaire: $$n = 8$$ choice sets each for PT and AT, plus two additional sets to test reliability ("double question") and validity ("clear question") of responses (T = (2x8)+2 = 18). For the "clear question", the two-alternative-choice sets included only the best vs. worst attribute levels (e.g., 25 vs. 5 years durability). It can be assumed that only participants who understood the questionnaire’s content correctly answered this question appropriately. The participants were asked to select their preferred alternative with its attributes and levels.
In reality, and particularly in health care, individuals face non-binary multiple choices [24]. For this reason, we created choice sets consisting of two unlabeled alternatives (A & B), and the option of "no treatment (opt-out)". An unlabeled design allowed us to assign attributes to the alternatives without being oriented to a defined treatment [25], and intended to reduce a possible bias. Including opt-out was necessary to create real life scenarios, and to explore patients‘ reasons against a treatment [25]. An example of a choice set can be seen in Table 2. Collected participants‘ (socioeconomic) characteristics are, e.g., age, income, insurance and oral health status. Furthermore, importance of the four treatment attributes was assessed for PT and AT via a 5-point Likert scale (dimension: very important–not important).
**Table 2**
| 1. Choice "anterior teeth" | 1. Choice "anterior teeth".1 | 1. Choice "anterior teeth".2 |
| --- | --- | --- |
| Attributes | Treatment A | Treatment B |
| 1. Aesthetics | strongly visible | natural color |
| 2. Compatibility | 1 out of 10,000 people with allergic or local toxic reaction | no risk |
| 3. Durability | 10 years | 25 years |
| 4. Out-of-pocket payment | 50 € | 200 € |
| I choose … | … treatment A.ο | … treatment B.ο |
| | … none of the treatments.ο | … none of the treatments.ο |
## Data collection
The study was conducted in the German federal states Berlin and Brandenburg. Since it is more likely for patients at a more mature age to have experiences with dental crown treatments and to have significant financial resources of their own [13, 26], we addressed people aged 30 and older. We aimed at an equal distribution of urban and rural population. By prevailing conditions, number of districts and counties in these areas are similar [27]. Household incomes in Berlin and Brandenburg are the same overall and are approximately equally distributed among Brandenburg’s counties [28], with incomes in both states close to the national average [29]. Furthermore, we aimed at women and men equally distributed. In compliance with the European General Data Protection Regulation, address data of potential participants, considering the minimum age and an equal gender distribution, were requested according to §34 of the Federal Registration Act from residents’ registration offices of the city of Berlin and selected counties in Brandenburg. For organizational reasons, the study’s catchment area was limited to these states. For the state of Brandenburg, one registration office per district was randomly selected. The number of contacted registration offices was thus limited to at least 19 institutions. Potential participants, either from urban or rural areas, were randomly assigned to one of the twelve questionnaire blocks using the Software RStudio.
We used the rule of thumb by Orme [30] ((n x t x a)/c > = "500 to 1,000") to determine the sample size n for the DCE, creating a (minimum) recommended level of participants. For calculating the number of choice sets (t), the number of treatment alternatives per choice set (a), and the highest number of levels (c) were considered: t(A) = t(B) = 8, $a = 2$, and $c = 3$x4 = 12. The calculated sample size was $$n = 750$$ (minimum $$n = 375$$). Assuming a response rate of questionnaires of $r = 7$% resulting from further studies experiences at our department, the calculated number of questionnaires was $$n = 10$$,715. Rounding up the results and considering an equal distribution of questionnaires among urban and rural areas, a total of $$n = 10$$,752 questionnaires were sent out by mail.
Before the survey start, a pretest was conducted with $$n = 15$$ participants of diverse educational backgrounds and ages. Based on this, a few minor linguistic corrections for the questionnaire’s comprehensibility followed and a processing time of about 20 minutes was set. The final survey included the following documents: [1] questionnaire and cover letter, declaration of participation, and participant information, [2] extra document "Aesthetics", and [3] free-return envelope addressed to the department. As an incentive for participating in pretest and survey, 20 shopping vouchers of 50 € each, were raffled in a lottery. This study was approved by the ethics committee of the Charité Universitätsmedizin Berlin (application no. EA$\frac{4}{109}$/19).
## Data editing & coding
According to predefined criteria, completed questionnaires were included for further consideration if (i) the declaration of participation was confirmed. Questionnaires were excluded, when (ii) failing the plausibility check: (a) "double question" was not answered in the same way, and (b) "clear question" was answered irrationally. Furthermore, questionnaires were excluded if (iii) >$50\%$ of the choice sets were not answered in section B1 and B2, (iv) the participants did not answer the question on their age or with "under 30 years", (v) the question on insurance coverage was not answered, or with "I don’t know", or "private health insurance", and (vi) the question on gender was not answered.
Regarding choice analysis the data set was effect coded which is recommended when an opt-out alternative is used [31]. Negative levels were selected as reference levels (and positive for WTP analysis) assuming to be unattractive for patients, i.e., strongly visible, risk of incompatibility, shortest duration (5 years), and highest costs (600 €). The reference level was coded with a value of -1. For all attributes of the opt-out, the very low value "-9999" was set because sum of benefit values would result in zero [32]. For calculation of the WTP over all participants the cost attribute was not effect coded but had continuous coding for more interpretable values [33]. For further analyses single variables were dummy coded, e.g., gender.
Since individual WTP cannot be estimated within a DCE [34], we defined a variable WTPmax_PT and _AT presenting the highest level value of the attribute out-of-pocket payment for a chosen treatment alternative across all alternatives per participant and teeth area. If opt-out was selected, we considered 0 €. To examine patients‘ behavior with respect to their choice between (I) "treatment" and opt-out, and (II) "SHI standard care" and "treatment beyond SHI standard care (SHI+)", we created further dependent variables as part of the choice analysis and depicted frequencies per participant and teeth area. The variable for "treatment" comprised the frequency of chosing treatment A or B in the choice set and "no treatment" comprised choice of opt-out. In the variable "SHI+" only choice sets with levels >SHI standard care of the attributes aesthetics and out-of-pocket payment were included: (a) lightly visible or natural color, and 450 € or 600 € for PT, and (b) natural color, and 450 € or 600 € for AT. Combinations of these levels are given in some choice sets of each questionnaire block.
## Choice analysis & willingness-to-pay analysis
In Lancaster’s [35] and McFadden’s [36] random utility theory, it is assumed that the actual utility of a choice set is not directly observable. The total utility of a set is composed of observable and non-observable components. It is assumed that an individual chooses the alternative with a combination of attributes from which she or he has the greatest utility over the other selectable alternatives. An indirect utility function was estimated that represents the expected observable utility (V) for a person, and is composed of a combination of (non-)observable random components as error term (ε) [37–39]. We specified the following utility function, in which the participants’ preferences for the attributes are captured and different utility allocations among attributes can be examined as a function of the participants‘ (socioeconomic) characteristics. The utility function V for individual i and alternative j in choice set s is to be expressed, in terms of the attributes of the alternatives (X) and characteristics of the participants (Z), as: Uijs=Vijs+εijs=Xijs'ßj+Zi'γ+εijs We assume the non-observable component is parametrically distributed and thus use a probalistic analysis of individual choice behavior [38]. The probability of choosing between given alternatives (J) is as follows: Pij=ProbUji>UJi∀J≠j=Prob(Vji+εji>VJi+εJi∀J≠j) Assuming that the error term is extreme-distributed, the probability of choosing alternative j, presenting the standard logit specification [38], is: Lij=eßxij∑JeßxiJ letting Vji=ßxij The model of utility for an individual i choosing a treatment alternative j can be estimated as: Uijs=βaestheticsijs+βcompatibilityijs+βdurabilityijs+βoutofpocketpaymentijs+εijs
The utility appears as random, i.e., we cannot predict the choice. However, if we know the distribution of the random element, we can derive the probability of a choice. Depending on the assumptions for ε different analysis models must be applied. According to Bekker-Grob et al. [ 40], different restrictions have to be considered for each model. All analyses were performed using STATA software (version 15).
We first estimated a conditional logit model (CLM) to analyze how attributes determine the treatment choice. Basis for this analysis are constant choice sets per individual with varying attributes levels across alternatives as descriptive variable [41]. The CLM accounts for observed preference heterogeneity by including participants’ characteristics (Z variables) [25]. Assuming that the utility of each alternative depends on its attributes, CLM models the influence of attributes that vary between alternatives on the selection probability regardless of alternatives (A, B, or opt-out) [42]. In addition, interactions in participants‘ (socioeconomic) characteristics can be considered. However, the CLM has some restrictions: it does not account for unobserved heterogeneity resulting from differences in preferences among participants with the same characteristics or random choices [25]. It models the choice between alternatives as a function of the alternatives’ attributes but not of characteristics of the person making the choice [41]. Furthermore, CLM is making strong assumptions, i.e., independence of single, and irrelevant alternatives [Independence of Irrelevant Alternatives (IIA)] [36, 43]. The IIA describes the ratio of choice probability of two alternatives unaware of other alternatives. However, some alternatives vary more, and some are more similar in the choice sets of our study due to its design. These assumptions must be considered when data has panel character, e.g., participants making multiple choices [44], as it is the case in our study. These model characteristics and disadvantages were countered using other models. For CLM analysis the clogit command was used [45].
Second, we considered a mixed logit model (MXL) working with random parameters that vary between individuals to circumvent the IIA. The MXL allows for an estimation in which the independent assumption is violated by assuming that there is no independence in the choice behavior due to multiple choices by individual participants [25, 46]. It considers the alternatives’ attributes and characteristics of the individuals [41]. The MXL estimates a distribution around each mean preference parameter to avoid potential bias in the estimated mean preference weights due to unobserved heterogeneity [47]. In our calculations, we did not include participant characteristics in the explainable component V but used the MXL to estimate random parameters. This allowed us to account for random variation across participants, i.e., heterogeneity, of unobserved participant characteristics [48]. Random heterogeneity is evident from significant standard deviations per model parameter [49]. The calculations were performed using the mixlogit command [50].
Third, we estimated a generalized multinomial logit model (G-MNL) developed by Fiebig et al. [ 51]. This model provides for more flexible distributions, and accounts for unobserved preference heterogeneity by including random parameters into calculation, as well as scale heterogeneity [52]. Scale heterogeneity implies that choice behavior is more random for some individuals than for others [53]. Results of the G-MNL must be interpreted to allow for a more flexible distribution of confounded preference and scale heterogeneity, rather than estimating scale separately [25]. For G-MNL analysis we used the gmnl command [53].
Quality of all models was assessed using Akaike’s and Schwarz’ Bayesian information criterion (AIC and BIC) (and LL–log likelihood) [25]. The AIC estimates the amount of lost information of a model, and the BIC additionally adapts to the sample size [54]. AIC and BIC should therefore be as low as possible [55]. The most suitable model was chosen.
Based on the results of that model the WTP was calculated. WTP represents the amount of a cost attribute an average participant is willing to pay for one unit of an attribute in relation to the reference level [56]. In these linear models where each attribute in the utility function is associated with a single weight, the ratio of the two utility parameters was used to estimate the WTP. The following function calculates the participants‘ WTP, where βia is a coefficient on one focused attribute "a", and βib is a coefficient on the cost attribute "b" [57, 58], which is out-of-pocket payment in our study: WTP=ßia/ßib Furthermore, an estimation on alternative specific constants (ASC) was done via an ASC-logit model (ASCL) allowing us to include the individual characteristics as independent variables in the analysis. The aim was to examine the influence of regulatory instruments, participants‘ (socioeconomic) characteristics, and the importance of attributes on choice for or against a treatment at all ("treatment" vs. opt-out), and a SHI+ treatment. For the latter analysis, only choice sets representing exactly these treatments as a combination of levels were considered. Therefore, the number of choices is lower here. For analysis we used the asclogit command [59].
Regression analyses were performed to determine the influence of participants’ socioeconomic characteristics [age, gender, income, employment status, residence (urban or rural)], treatment attributes, a bonus booklet, supplementary dental insurance, and its combination on the decision variables individual WTPmax, frequency of choosing opt-out and a SHI+ treatment. Analyses were conducted as follows: correlation analysis for refinement of subsequent multiple regression analysis, and graphical presentation of relevant categorial variables.
## Response rate and participants‘ characteristics
We received $$n = 762$$ questionnaires ensuring the response rate of $r = 7.1$%. According to our in-/exclusion criteria, data sets had to be excluded from further consideration. Fig 1 gives an overview on numbers of selected questionnaires and data sets regarding criteria. Finally, $$n = 380$$ data sets could be included in the analysis. We were thus above the minimum required number of participants (see chapt. 2.2). Most of the participants belong to age group "50 to 59 years" ($$n = 103$$, $27.1\%$). The majority is female ($$n = 249$$, $65.5\%$), has a university degree ($$n = 166$$, $43.7\%$), and is employed full-time ($$n = 173$$, $45.5\%$). Medium and low household incomes are most common (Table 3). Most of the participants indicated not having a dental supplementary insurance ($$n = 256$$, $67.4\%$). In contrast, a large proportion of our participants indicated having a bonus booklet ($$n = 329$$, $86.6\%$). Regarding their oral health, a large proportion of the participants stated to have a "good" ($$n = 170$$, $44.7\%$) or "very good" ($$n = 34$$, $9.0\%$) self-perceived status. Some participants ($$n = 49$$, $12.9\%$) had already decided against dental crown treatment in the past due to high costs ($$n = 17$$, $36.2\%$), or they considered a dental crown treatment as "unnecessary" ($$n = 14$$, $29.8\%$). Further reasons include questioning tooth preservation, and allergies (S3 File: Participants‘ reasons against a dental crown). Aditionally, majority of the participants ($$n = 238$$, $62.6\%$) indicated that they would always decide against a strong visible dental crown. Only a few would choose darker colors (golden-metal: $$n = 23$$, $29.9\%$; dark grey metallic: $$n = 6$$, $7.8\%$) (S4 File: Participants‘ decision for dental crown color). On average, it took the participants 18.5 minutes (range: 4–90min) to complete the questionnaire (S3 Table: Results on questions–Questionnaire Part C; S5 File: Codebook of analysis).
**Fig 1:** *Selection of questionnaires and data sets, numbers regarding citeria.* TABLE_PLACEHOLDER:Table 3
## Importance of treatment attributes
Regarding the importance of the four treatment attributes, durability (PT: $$n = 274$$, $72.1\%$; AT: $$n = 263$$, $69.2\%$) and compatibility (PT: $$n = 197$$, $51.8\%$; AT: $$n = 216$$, $56.8\%$) are assessed "very important" by our participants. Also, assessment of out-of-pocket payment is equally given for both teeth areas. The picture changes for the assessment of aesthetics. For PT aesthetics was assessed least as "very important" ($$n = 26$$, $6.8\%$) by our participants, and for AT it is most frequently given ($$n = 290$$, $76.3\%$) (S6 File: Importance of treatment attributes assessed by participants).
## Results of discrete-choice, willingness-to-pay, and regression analysis
Estimations using the different models have produced different AIC and BIC. Lowest coefficient values were calculated for the MXL model (AIC/BIC PT: 5,$\frac{176}{5}$,312; AT: 4,$\frac{692}{4}$,827). Since MXL also makes the realistic acceptable assumption of including random parameters, we considered these results. For this reason, we have also performed the WTP calculation based on the MXL using the wtp command [57]. The "no. of observations" in the (appendix) tables do not refer to the population sample size, but to the dataset rows included in the analyses, and therefore vary between the models. Rows with missings, and of non-relevant choice sets (e.g., SHI levels in ASCL) were excluded. Single G-MNL results are presented in the following (S4 Table: Coefficients of G-MNL estimations, including marginal effects; S5 Table: Coefficients of CLM estimations, including marginal effects).
## (i.) Discrete-choice-models
The participants preferred lightly visible (Coef.: 0.687, $p \leq 0.01$) and natural color PT (Coef.: 1.290, $p \leq 0.01$) compared to a strongly visible color of the teeth. The probability of choosing natural color instead of lightly visible teeth is almost four times larger compared to the reference level, measured by marginal effects [dy/dx: 1.247, $p \leq 0.01$; vs. lightly visible (dy/dx: 0.334, $p \leq 0.01$)]. In addition, the participants preferred a treatment without risk of incompatibility (Coef.: 0.465, $p \leq 0.01$). A durability of 25 years was associated with higher preferences by our participants compared to other levels [Coef.: 1.540, $p \leq 0.01$; e.g., vs. 15 years (Coef.: 0.477, $p \leq 0.01$)]. The participants preferred low out-of-pocket payments, e.g., 50 € (Coef.: 0.776, $p \leq 0.01$) and 150 € (Coef.: 0.965, $p \leq 0.01$).
The results are similar for AT. Lightly visible (Coef.: 1.026, $p \leq 0.01$) and natural color teeth (Coef.: 3.392, $p \leq 0.01$) are assigned higher preferences than compared to strongly visible crowns. Similarly, no-risk treatments regarding compatibility (Coef.: 0.200, $p \leq 0.05$) are assigned higher preferences by the participants, as well as for the longest durability of 25 years (Coef.: 0.835, $p \leq 0.01$).
Comparing the results of both teeth areas, aesthetics of the AT can be considered more important for the participants since the probability of choosing a natural color crown is approximately three times higher than for the PT compared to the reference level according to marginal effects results [dy/dx: 3.449, $p \leq 0.01$; vs. PT (dy/dx: 1.247, $p \leq 0.01$)]. For teeth in both teeth areas, no-risk treatments and the highest possible durability of 25 years are preferred by the participants. The preferred out-of-pocket payment corresponds to the co-payment of current SHI standard care, for PT and AT. Considering all coefficients, the level natural color for AT stands out. Overall, this level of the attribute aesthetics is preferred by the participants in their decision-making. Coefficients of the MXL model for PT can be seen in Table 4 (S6 Table: Coefficients of MXL estimations for anterior teeth, S7 Table: Marginal effects of MXL estimations).
**Table 4**
| Mixed logit model (MXL) | Mixed logit model (MXL).1 | Mixed logit model (MXL).2 | Mixed logit model (MXL).3 | Mixed logit model (MXL).4 | Mixed logit model (MXL).5 | Mixed logit model (MXL).6 | Mixed logit model (MXL).7 | Mixed logit model (MXL).8 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Posterior teeth | Posterior teeth | Posterior teeth | Posterior teeth | Posterior teeth | Posterior teeth | Posterior teeth | Posterior teeth | Posterior teeth |
| Attributes (Ref. negative levels) | Levels | Coef. | Std. Err. | t-value (z) | p-value (P>|z|) | [95% Conf. interval] | [95% Conf. interval] | Sig. |
| Aesthetics | strongly visible–reference level | strongly visible–reference level | strongly visible–reference level | strongly visible–reference level | strongly visible–reference level | strongly visible–reference level | strongly visible–reference level | strongly visible–reference level |
| Aesthetics | lightly visible | 0.687 | 0.103 | 6.680 | 0.000 | 0.485 | 0.888 | *** |
| Aesthetics | natural color | 1.290 | 0.113 | 11.410 | 0.000 | 1.068 | 1.512 | *** |
| Compatibility | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level | 1 out of 10,000 people with allergic or local toxic reaction–reference level |
| Compatibility | no risk | 0.465 | 0.086 | 5.400 | 0.000 | 0.296 | 0.634 | *** |
| Durability | 5 years–reference level | 5 years–reference level | 5 years–reference level | 5 years–reference level | 5 years–reference level | 5 years–reference level | 5 years–reference level | 5 years–reference level |
| Durability | 10 years | 0.159 | 0.121 | 1.310 | 0.189 | -0.078 | 0.395 | |
| Durability | 15 years | 0.477 | 0.112 | 4.270 | 0.000 | 0.258 | 0.695 | *** |
| Durability | 25 years | 1.540 | 0.140 | 11.000 | 0.000 | 1.266 | 1.815 | *** |
| Out-of-pocket payment | 600 €–reference level | 600 €–reference level | 600 €–reference level | 600 €–reference level | 600 €–reference level | 600 €–reference level | 600 €–reference level | 600 €–reference level |
| Out-of-pocket payment | 450 € | 0.191 | 0.113 | 1.690 | 0.091 | -0.030 | 0.411 | * |
| Out-of-pocket payment | 150 € | 0.965 | 0.107 | 9.000 | 0.000 | 0.755 | 1.175 | *** |
| Out-of-pocket payment | 50 € | 0.776 | 0.130 | 5.960 | 0.000 | 0.521 | 1.031 | *** |
| Log likelihood | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) | -2,569.2494 (Iteration 8) |
| Prob > chi2 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| LR chi2(9) | 459.8 | 459.8 | 459.8 | 459.8 | 459.8 | 459.8 | 459.8 | 459.8 |
| No. of observations | 9039 | 9039 | 9039 | 9039 | 9039 | 9039 | 9039 | 9039 |
## (ii.) ASC-logit models
(ii.1) Analysis of choice between "treatment" and "no treatment (opt-out)". For PT, the opt-out alternative, i.e., no treatment, was selected in $25.7\%$ ($$n = 774$$) of the choice scenarios by the participants [AT: $37.2\%$ ($$n = 1$$,122)]. A combination of bonus booklet and supplementary dental insurance increases the likelihood of choosing a treatment (Coef. PT: 0.335, AT: 0.377; $p \leq 0.05$). This is also true for higher aged participants (Coef. PT: 0.119, AT: 0.087; $p \leq 0.01$), being a resident of an urban region (Coef. PT: 0.131, AT: 0.129; $p \leq 0.01$), and for an increased importance of the attribute aesthetics (Coef. PT: 0.154, $p \leq 0.01$; AT: 0.1, $p \leq 0.05$). Furthermore, gender, bonus booklet, and importance of out-of-pocket payments have an impact on our participants’ choice behavior.
(ii.2) Analysis of choice between "SHI standard care" and "treatment beyond SHI standard care (SHI+)". In $31.3\%$ ($$n = 480$$) of the choice scenarios for PT, in which a decision could be made between a SHI+ treatment versus treatment below standard care ($$n = 1$$,533 decisions in total), the participants chose SHI+. This occurred more frequently for AT: in $49.8\%$ ($$n = 449$$) of the decisions ($$n = 902$$ in total) they decided for SHI+. As the importance of the attribute aesthetics increases, the participants decided against SHI standard care for teeth of both teeth areas and favored treatments beyond that (Coef. PT: -0.243, $p \leq 0.01$; AT: -0.18, $p \leq 0.05$). For AT, the participants chose SHI standard care as the importance of out-of-pocket payments increases (Coef.: 0.17, $p \leq 0.1$). Regression analysis was additionally performed with the same dependent variables (see chapt. 3.1.iv.) [ S8 Table: Coefficients of ASCL estimations "treatment" vs. "no treatment (opt-out)", S9 Table: Coefficients of ASCL estimations "SHI standard care" and "treatment beyond SHI standard care (SHI+)"].
## (iii.) Willingness-to-pay
For PT, the out-of-pocket payment of SHI standard care is set at 150 €. For the attributes aesthetics (strongly visible) and compatibility (risk) levels, there is no participants’ willingness-to-pay. Regarding a durability of 15 years, WTP is higher (258 €). Considering attributes‘ most positive "high quality" treatment levels, we see the following: for natural color teeth, the WTP is 380 €, for a treatment without risk of incompatibility, the participants are willing to pay 162 €, and for dental crowns with 25 years durability 508 €. For AT, the out-of-pocket payment is 200 €. The participants would pay 362 € for lightly visible anterior teeth, but there is no willingness-to-pay for a risk of incompatibility. For dental crowns with a durability of 10 years, the participants are willing to pay 73 €, for natural color teeth 914 €, for risk-free treatments 92 €, and 282 € for a durability of 25 years. Comparing both teeth areas, we see that WTP for aesthetics is higher for anterior teeth, especially at the natural color level. On the contrary, WTP for compatibility and durability is higher for PT (S10 Table: WTP analysis framework, and results).
## (iv.) Regression analysis
The results of correlation analyses, and calculation of the variance inflation factors (VIF), led to the exclusion of certain variables (e.g., combination of bonus booklet and dental supplementary insurance) from regression analyses. Statistical significance (applicable to the following reported results) was not given for all calculations.
With an increasing age (Coef. PT: -13.78, $p \leq 0.01$; AT: -13.73, $p \leq 0.05$), WTPmax decreases. Furthermore, with the presence of a bonus booklet (Coef. PT: 114.79, $p \leq 0.01$; AT: 111.66, $p \leq 0.01$), it increases. For AT, also gender (Coef. female: -91.01, $p \leq 0.01$) has an influence. Female participants more often accepted high out-of-pocket payment amounts.
Individual variables are also correlated with the decision against a treatment ("no treatment") for PT and AT: with increasing age (Coef. PT: 0.21, $p \leq 0.01$; AT: 0.21, $p \leq 0.01$), residence in smaller towns (Coef. PT: -0.24, $p \leq 0.05$; AT: -0.25, $p \leq 0.05$), non-existence of a bonus booklet (Coef. PT: -1.21, $p \leq 0.05$; AT: -1.21, $p \leq 0.05$), and having a dental supplementary insurance (Coef. PT: 0.71, $p \leq 0.05$; AT: 0.71, $p \leq 0.05$), the decision against a treatment has been made more frequently. The participants‘ gender (Coef. female: 0.81, $p \leq 0.05$) also plays a role regarding decisions for AT. Female respondents are less likely to decide against a treatment.
The more important co-payment (Coef. PT: -0.21, $p \leq 0.05$; AT: -0.21, $p \leq 0.1$), the less often a treatment outside SHI standard care was chosen. With the existence of a bonus booklet (Coef. PT: 0.96, $p \leq 0.01$; AT: 0.74, $p \leq 0.05$), SHI+ was chosen more often. For PT, the importance of aesthetics (Coef. 0.26, $p \leq 0.01$) also has an influence. The more important aesthetics, the more often SHI+ was chosen by the participants. For AT, gender (Coef. female: -0.51, $p \leq 0.05$), and residence (Coef.: 0.17, $p \leq 0.05$) additionally influenced their decisions. Females and residents of smaller towns were less likely to choose a more cost-intensive treatment (S11 Table: Tables on correlation analysis, VIF values, and regression analysis; S7 File: Regression plots on WTPmax and SHI+ analysis).
## Discussion
This study provides important insights into factors determining patients‘ choice behavior in dental care, while distinguishing between the two teeth areas, PT and AT. The focus of the choice analyses was on highest benefit expectations assigned by the participants to attributes and its levels of dental crown treatment, as well as the participants‘ willingness-to-pay. Further analyses focused on incentive measures provided by SHI and private health insurance, on choice for or against a treatment ("no treatment"), and for a treatment beyond SHI standard care, and the influence of the participants’ (socioeconomic) characteristics in decision-making.
Our results show that aesthetics is an important factor for the participants in their choice of a dental crown treatment. For AT, aesthetics has a higher weight for the participants. Highest benefit allocations are assigned to "natural color", i.e., tooth-colored, dental crowns, which should be indistinguishable from natural teeth in terms of visibility. Results on the importance of aesthetics underline our choice analysis estimates. Furthermore, the importance of aesthetic aspects of AT has already been shown in previous research [60, 61]. For PT, durability and treatment attributes such as functionality [62, 63] might be more meaningful. Nevertheless, even for PT, natural color teeth are preferred over strongly visible. Risk of a local toxic or allergic reaction seems to have rather less weight among the participants. For PT and AT, the coefficient for non-risk in the choice analysis is small. It can be assumed patients accept those risks. These values may result from the experimental design, i.e., extremely preferred (especially for AT) or non-preferred expressions were opposite to the risk attribute level. Besides, this may result from the fact that the probability of occurrence of a local toxic or allergic reaction appears low, also based on the participants’ awareness and experiences (e.g., does not know about allergies, former allergic reactions were mild) [64]. The attribute durability of a dental crown has a great influence on the participants’ decision-making, for both teeth areas. Highest benefit is clearly assigned to the highest duration of 25 years. A long life cycle could mean convenience for patients: lower costs in the long term, fewer visits at the dentist which may be painful, etc. However, present conditions may stand in the way of this patient desire. Dental crowns made of common materials (e.g., SHI standard care restorations) have an average life span of 15 years. For high-quality and -priced dental crowns, durability could be a few years higher (e.g., gold alloys) [65, 66]. Ultimately, the quest for long lifetime materials needs to be realized through further research activities. Out-of-pocket payments play an important role in our participants treatment choice, independently from teeth area. Nevertheless, these would require high co-payments, especially for AT. In the context of choice analysis, it is important to note that this is a benefit allocation. The participants might have associated high costs with other treatment characteristics, such as high quality [67].
We determined the participants’ willingness-to-pay for treatments with a natural color dental crown, i.e., the best possible attribute level. For both teeth areas and both cost attributes, the maximum amount to pay is above the level of SHI standard care. The participants are willing to pay much more for AT. High WTP values are possible due to the design with closed-ended questions [68]. Nevertheless, it should be noted that these values refer to the entirety of the participants and cannot be attributed to an individual participant [34]. Further analysis of the cost attribute revealed that willingness-to-pay per participants decreases with increasing age. Reduced incomes at an older age (e.g., pension), expensive treatments, or a reduced awareness of aesthetically high-quality as well as prioritization of functionality could be a reason [69–71]. Presumably, the older the patient, the more intentional the dentist is in communicating that a dental crown is possibly the last and only alternative for tooth preservation [72]. There may be financial and pragmatic reasons for choosing the SHI standard care alternative. Previous studies have examined inequalities in dental treatment utilization. Besides income, financial wealth is one reason [73–75].
Descriptive analysis showed that a large proportion of the participants owns a bonus booklet, but only a few participants have taken out a dental supplementary insurance. The proportions correspond to those for Germany: only about a quarter of SHI-insured people have private supplementary insurance [76]. Some participants stated they had already rejected dental crowns in the past, for reasons of cost and lack of necessity from their point of view. Combining bonus booklet and supplementary insurance makes patients more likely to choose a treatment than no treatment at all, for both teeth areas. It should be noted that this approach is used to reduce out-of-pocket payments, regardless of which form of care is chosen. However, it should also be noted that private supplementary insurances incur fees and cannot be financed by every patient [77]. If no dental supplementary insurance has made, out-of-pocket payment might remain at a high level. Patients might choose the most inexpensive alternative, including no treatment, although, SHI provides incentive measures. Medical necessity of dental treatments seems to be irrelevant for the participants. One assumption of our questionnaire was, that the dental crown treatment is found to be necessary by a dentist. Patients’ attitude has been reported in former articles [78, 79]. Overall, SHI standard care is accepted by patients, especially in older age groups, when aesthetics takes a back seat, and cost and functionality aspects become more relevant. However, it is apparent that there is a desire for more aesthetically pleasing and long-lasting alternatives. The former point is given especially for AT. Many patients keep a bonus booklet and make use of it (proof of at least 5 years annual check-ups in a row). SHI should target further possibilities and combinations of bonus measures to reduce access barriers to care and improve utilization of routine check-ups that can prevent caries. These measures could be linked to conditions that promote patients’ oral health behavior, as the bonus booklet successfully demonstrates [80].
Some limitations must be mentioned. The study is limited to the states of Berlin and Brandenburg. A region-typical choice behavior is conceivable here. When selecting the regional areas to which the questionnaires were sent, we took care to ensure an equal distribution of household incomes across federal states’ districts and counties. Nevertheless, a large proportion of the participants tended to belong to low household incomes groups. This may have biased the results, particularly regarding financial preferences. Since the sample is small due to the experimental design, descriptive results may not be representative for Germany. In the results of the models, especially ASCL, there is partly no statistical significance, although we have reached the minimum sample size. Accordingly, there are gaps in the answers to the research questions. Although the treatment attributes and their levels were designed to be as realistic as possible, it should be noted that the presentation of the treatment alternatives in the questionnaire probably does not reflect real life decision-making situations of the individual participants: choice scenarios were limited to a few attributes and levels. More factors probably play a role in patients’ treatment decision, including possible medical consequences or the relationship with the dentist [81, 82], and there may be more than two alternatives to choose from. The questionnaire, including $$n = 18$$ choice scenarios, was also very complex. Possibly, participants applied heuristics to simplify decision situations [1]. Also, conditions under which the questionnaires were completed are unclear, e.g., maybe participants were influenced by relatives. Additionally, the participants could not ask questions in case of any ambiguities.
The often non-statistical-significance of the regression analyses results can be explained by the fact that the participant number was small. However, there is no guideline for minimum population numbers for regressions. Study’s focus was a DCE, in which the experimental design allows small populations [57]. Statistical significances were given for most results in the choice analyses. It should also be noted that in the choice analyses, values of the coefficients were sometimes small, i.e., close to the reference level, or close between levels. Results should be interpreted accordingly.
## Conclusion
Dental interventions such as crown treatments, involve difficult decisions on the optimal allocation of resources for health care systems and patients. This study provides important insights into patient preferences for crown treatment in Germany. Findings show that aesthetic for AT and PT as well as out-of-pocket payments for PT play an important role in the decision for dental crown treatments. Overall, participants are willing to pay more out of pocket compared to out-of-pocket payment that arises for SHI standard care, with a considerably higher willingness-to-pay for AT. Having a bonus booklet increased the willingness-to-pay. Although, the findings should be interpreted with caution due to limitations of choice experiments and the regional restriction to two federal states in Germany, it may also be valuable for policy makers and health insurance funds in developing dental health care programs, creating incentive structures, and planning the provision of dental services that better match patient preferences. For further studies, participants from all income groups should be targeted and included in the analysis in equal proportions (e.g., randomization within income groups), and regression analyses should be planned with larger populations.
## References
1. Bridges JFP, Hauber AB, Marshall D, Lloyd A, Prosser LA, Regier DA. **Conjoint analysis applications in health—a checklist: a report of the ISPOR Good Research Practices for Conjoint Analysis Task Force**. *Value Health* (2011.0) **14** 403-13. DOI: 10.1016/j.jval.2010.11.013
2. Peres MA, Macpherson LMD, Weyant RJ, Daly B, Venturelli R, Mathur MR. **Oral diseases: a global public health challenge**. *The Lancet* (2019.0) **394** 249-60. DOI: 10.1016/S0140-6736(19)31146-8
3. Allin S, Farmer J, Quiñonez C, Peckham A, Marchildon G, Panteli D. **Do health systems cover the mouth? Comparing dental care coverage for older adults in eight jurisdictions**. *Health Policy* (2020.0) **124** 998-1007. DOI: 10.1016/j.healthpol.2020.06.015
4. Klingenberger D, Winkelmann J, Henschke C. **Best Oral Health Practice in Europe? Eine Analyse zur Frage der Vergleichbarkeit der Effizienz zahnmedizinischer Versorgungssysteme [Best oral health practice in Europe? Comparing the efficiency of oral health care systems—an analysis] [German]**. *Zahnmed Forsch Versorg* (2021.0). DOI: 10.23786/2021-4-2
5. 5German Federal Joint Committee (G-BA). Festzuschuss-Richtlinie [Directive of fixed allowance in dental care] [German]. Effective: 01.01.2021. 2021. https://www.g-ba.de/downloads/62-492-2341/FZ-RL_2020-12-11_iK-2021-01-01.pdf. Accessed 2 Dec 2021.
6. Ziller S, Eaton KE, Widström E. **The healthcare system and the provision of oral healthcare in European Union member states. Part 1: Germany**. *Br Dent J* (2015.0) **218** 239-44. DOI: 10.1038/sj.bdj.2015.95
7. 7Busse R, Siegel M. Can people afford to pay for health care? New evidence on financial protection in Germany: World Health Organization Europe; 2018.
8. Felgner S, Dreger M, Henschke C. **Reasons for (not) choosing dental treatments—a qualitative study based on patients’ perspective**. *PLoS ONE* (2022.0). DOI: 10.1371/journal.pone.0267656
9. Wilson Leslie, Tracy Kuo Lin Lindsay A. Hampson, Oh Anna, Ting Jie, Gaither Thomas. **Use of conjoint analysis to determine patient preferences for surgical treatment of urethral stricture disease**. *Journal of Participatory Medicine* (2020.0)
10. Mühlbacher A, Bethge S, Tockhorn A. **Präferenzmessung im Gesundheitswesen: Grundlagen von Discrete-Choice-Experimenten**. *Gesundh ökon Qual manag* (2013.0) **18** 159-72. DOI: 10.1055/s-0032-1330500
11. Espelid I, Cairns J, Askildsen JE, Qvist V, Gaarden T, Tveit AB. **Preferences over dental restorative materials among young patients and dental professionals**. *Eur J Oral Sci* (2006.0) **114** 15-21. DOI: 10.1111/j.1600-0722.2006.00282.x
12. Vernazza CR, Wildman JR, Steele JG, Whitworth JM, Walls AWG, Perry R. **Factors affecting patient valuations of caries prevention: Using and validating the willingness to pay method**. *J Dent* (2015.0) **43** 981-8. DOI: 10.1016/j.jdent.2015.05.009
13. Köberlein J, Klingenberger D. **Auslandszahnersatz und Dentaltourismus—Zahlungsbereitschaft und Einflussfaktoren auf die Nachfrage nach ausländischem Zahnersatz in Deutschland. [Foreign dentures and dental tourism—willingness-to-pay and factors influencing the demand for foreign dental prosthesis in Germany]**. *Gesundheitswesen* (2011.0) **73** e111-8. DOI: 10.1055/s-0030-1254174
14. Wang Y, Wang Z, Wang Z, Li X, Pang X, Wang S. **Application of discrete choice experiment in health care: a bibliometric analysis**. *Front Public Health* (2021.0) **9** 673698. DOI: 10.3389/fpubh.2021.673698
15. Felgner S, Schütte F, Schroll CC, Handrock JCF, Henschke C. **Does it change during COVID-19 pandemic?—a systematic review. (Update)**. *PROSPERO* (2021.0)
16. Coast J, Horrocks S. **Developing attributes and levels for discrete choice experiments using qualitative methods**. *J Health Serv Res Policy* (2007.0) **12** 25-30. DOI: 10.1258/135581907779497602
17. Lancsar E, Louviere J. **Conducting Discrete Choice Experiments to Inform Healthcare Decision Making**. *PharmacoEconomics* (2008.0) **26** 661-77. DOI: 10.2165/00019053-200826080-00004
18. Ratcliffe J.. **Assessing patients’ preferences for characteristics associated with homeopathic and conventional treatment of asthma: A conjoint analysis study**. *Thorax* (2002.0) **57** 503-8. DOI: 10.1136/thorax.57.6.503
19. 19Gernet W, Schwenzer N, editors. Zahn-Mund-Kiefer-Heilkunde. Lehrbuchreihe zur Aus- und Weiterbildung / Zahn-Mund-Kiefer-Heilkunde: Zahnärztliche Prothetik [Oral and maxillofacial surgery. Textbook series for education and training / dentistry: dental prosthetics] [German]. 3rd ed. Stuttgart: Thieme; 2007.
20. Qualtrough AJ, Burke FJ. **A look at dental esthetics**. *Quintessence Int* (1994.0) **25** 7-14. PMID: 8190886
21. Reed Johnson F, Lancsar E, Marshall D, Kilambi V, Mühlbacher A, Regier DA. **Constructing experimental designs for discrete-choice experiments: report of the ISPOR Conjoint Analysis Experimental Design Good Research Practices Task Force**. *Value Health* (2013.0) **16** 3-13. DOI: 10.1016/j.jval.2012.08.2223
22. 22Kuhfeld WF. Marketing research methods in SAS: Experimental design, choice, conjoint, and graphical techniques. 2010th ed.
23. Bech M, Kjaer T, Lauridsen J. **Does the number of choice sets matter? Results from a web survey applying a discrete choice experiment**. *Health Econ* (2011.0) **20** 273-86. DOI: 10.1002/hec.1587
24. Ryan M, Amaya-Amaya M, Gerard K. *Using Discrete Choice Experiments to Value Health and Health Care* (2007.0)
25. Ride J, Lancsar E. **Women’s Preferences for Treatment of Perinatal Depression and Anxiety: A Discrete Choice Experiment**. *PLoS ONE* (2016.0) **11** e0156629. DOI: 10.1371/journal.pone.0156629
26. Oscarson N, Lindholm L, Källestål C. **The value of caries preventive care among 19-year olds using the contingent valuation method within a cost-benefit approach**. *Community Dent Oral Epidemiol* (2007.0) **35** 109-17. DOI: 10.1111/j.1600-0528.2007.00306.x
27. 27Berlin-Brandenburg Statistics Office. Regional statistics. Population—Berlin and Brandenburg 2018. [Amt für Statistik Berlin-Brandenburg. Regionalstatistiken. Bevölkerung—Berlin und Brandenburg 2018.]. https://www.statistik-berlin-brandenburg.de/regionalstatistiken/r-gesamt_neu.asp?Ptyp=410&Sageb=12015&creg=BBB&anzwer=6. Accessed 15 Mar 2020.
28. 28Berlin-Brandenburg Statistics Office. Regional statistics. National accounts—Berlin and Brandenburg 2018. [Amt für Statistik Berlin-Brandenburg. Regionalstatistiken. Volkswirtschaftliche Gesamtrechnungen—Berlin und Brandenburg 2018.]. https://www.statistik-berlin-brandenburg.de/regionalstatistiken/r-gesamt_neu.asp?Ptyp=410&Sageb=82000&creg=BBB&anzwer=8. Accessed 15 Mar 2020.
29. 29Destatis—Federal Statistical Office. National Accounts of the federal states (redistribution calculation)—Disposable income of private households: federal states, years. 2017. [Destatis—Statistisches Bundesamt. VGR der Länder (Umverteilungsrechnung)—Verfügbares Einkommen der privaten Haushalte: Bundesländer, Jahre. 2017.]. URL: https://www-genesis.destatis.de/genesis/online?operation=abruftabelleBearbeiten&levelindex=1&levelid=1662814398480&auswahloperation=abruftabelleAuspraegungAuswaehlen&auswahlverzeichnis=ordnungsstruktur&auswahlziel=werteabruf&code=82411–0001&auswahltext=&wertauswahl=225&wertauswahl=1111&werteabruf=Werteabruf#abreadcrumb. Accessed 10 Sep 2022.
30. Orme BK. *Getting started with conjoint analysis: strategies for product design and pricing research* (2010.0)
31. Louviere JJ, Hensher DA, Swait JD, Adamowicz W. *Stated Choice Methods* (2000.0)
32. Bennett J, Blamey RK. *The choice modelling approach to environmental valuation* (2001.0)
33. Donnan JR, Johnston K, Chibrikov E, Marra CA, Aubrey-Bassler K, Najafzadeh M. **Capturing Adult Patient Preferences Toward Benefits and Risks of Second-Line Antihyperglycemic Medications Used in Type 2 Diabetes: A Discrete Choice Experiment**. *Can J Diabetes* (2020.0) **44** 6-13. DOI: 10.1016/j.jcjd.2019.04.014
34. Breidert C, Hahsler M, Reutterer T. **A review of methods for measuring willingness-to-pay**. *Innovative Marketing* (2006.0)
35. Lancaster KJ. **A New Approach to Consumer Theory**. *Journal of Political Economy* (1966.0) **74** 132-57. DOI: 10.1086/259131
36. McFadden Daniel, Zarembka P. *Frontiers in econometrics* (1974.0) 105-142
37. Ben-Akiva ME, Lerman SR. *Discrete choice analysis: Theory and application to travel demand* (1985.0)
38. Kjaer T, Gyrd-Hansen D. **Preference heterogeneity and choice of cardiac rehabilitation program: results from a discrete choice experiment**. *Health Policy* (2008.0) **85** 124-32. DOI: 10.1016/j.healthpol.2007.07.002
39. HORST P. **L.L. Thurstone and the science of human behavior**. *Science* (1955.0) **122** 1259-60. DOI: 10.1126/science.122.3183.1259
40. Bekker-Grob EW de, Ryan M, Gerard K. **Discrete choice experiments in health economics: a review of the literature**. *Health Econ* (2012.0) **21** 145-72. DOI: 10.1002/hec.1697
41. Hoffman SD, Duncan GJ. **Multinomial and conditional logit discrete-choice models in demography**. *Demography* (1988.0) **25** 415-27. DOI: 10.2307/2061541
42. Oluoch-Aridi J, Adam MB, Wafula F, Kokwaro G. **Understanding what women want: eliciting preference for delivery health facility in a rural subcounty in Kenya, a discrete choice experiment**. *BMJ Open* (2020.0) **10** e038865. DOI: 10.1136/bmjopen-2020-038865
43. Long JS, Freese J. *Regression models for categorical dependent variables using Stata* (2014.0)
44. Auspurg K, Liebe U. **Choice-Experimente und die Messung von Handlungsentscheidungen in der Soziologie**. *Köln Z Soziol* (2011.0) **63** 301-14. DOI: 10.1007/s11577-011-0136-3
45. 45Stata.com. Manuals 13: clogit—conditional (fixed-effects) logistic regression. 2013. https://www.stata.com/manuals13/rclogit.pdf. Accessed 23 Aug 2021.
46. Hensher DA, Rose J, Greene WH. **The implications on willingness to pay of respondents ignoring specific attributes**. *Transportation* (2005.0) **32** 203-22. DOI: 10.1007/s11116-004-7613-8
47. Dong D, Xu RH, Wong EL-Y, Hung C-T, Feng Da, Feng Z. **Public preference for COVID-19 vaccines in China: a discrete choice experiment**. *Health Expect* (2020.0) **23** 1543-78. DOI: 10.1111/hex.13140
48. Train KE. *Discrete Choice Methods with Simulation* (2009.0)
49. Fletcher BR, Rowe R, Hollowell J, Scanlon M, Hinton L, Rivero-Arias O. **Exploring women’s preferences for birth settings in England: a discrete choice experiment**. *PLoS ONE* (2019.0) **14** e0215098. DOI: 10.1371/journal.pone.0215098
50. Hole AR. **Fitting mixed logit models by using maximum simulated likelihood**. *The Stata Journal* (2007.0) **7** 388-401
51. Fiebig DG, Keane MP, Louviere J, Wasi N. **The generalized multinomial logit model: accounting for scale and coefficient heterogeneity**. *Marketing Science* (2010.0) **29** 393-421. DOI: 10.1287/mksc.1090.0508
52. Hess S, Rose JM. **Can scale and coefficient heterogeneity be separated in random coefficients models?**. *Transportation* (2012.0) **39** 1225-39. DOI: 10.1007/s11116-012-9394-9
53. Gu Y, Hole AR, Knox S. **Fitting the generalized multinomial logit model in Stata**. *The Stata Journal* (2013.0) **13** 382-97. DOI: 10.1177/1536867X1301300213
54. Burnham KP, Anderson DR. **Multimodel inference**. *Sociological Methods & Research* (2004.0) **33** 261-304. DOI: 10.1177/0049124104268644
55. Vrieze SI. **Model selection and psychological theory: a discussion of the differences between the Akaike information criterion (AIC) and the Bayesian information criterion (BIC)**. *Psychol Methods* (2012.0) **17** 228-43. DOI: 10.1037/a0027127
56. 56Berendsen R. A discrete choice experiment to estimate willingness to pay for a microfinance product in urban Romania [Master Thesis]: Wageningen University; 2015.
57. 57Ryan M, Kolstad JR, Rockers PC, Dolea C. User Guide with Case Studies: How to Conduct a Discrete Choice Experiment for Health Workforce Recruitment and Retention in Remote and Rural Areas. 2012. https://documents.worldbank.org/en/publication/documents-reports/documentdetail/586321468156869931/how-to-conduct-a-discrete-choice-experiment-for-health-workforce-recruitment-and-retention-in-remote-and-rural-areas-a-user-guide-with-case-studies. Accessed 27 Jul 2021.
58. Lopez-Feldman A.. *Introduction to contingent valuation using Stata* (2012.0)
59. 59Stata.com. Manuals 13: asclogit—alternative-specific conditional logit (McFadden’s choice) model. 2013.
60. Exley C, Rousseau N, Donaldson C, Steele JG. **Beyond price: individuals’ accounts of deciding to pay for private healthcare treatment in the UK**. *BMC Health Serv Res* (2012.0) **12** 53. DOI: 10.1186/1472-6963-12-53
61. Al-Dwairi ZN, El Masoud BM, Al-Afifi SA, Borzabadi-Farahani A, Lynch E. **Awareness, attitude, and expectations toward dental implants among removable prostheses wearers**. *J Prosthodont* (2014.0) **23** 192-7. DOI: 10.1111/jopr.12095
62. Trovik TA, Wisth PJ, Tornes K, Bøe OE, Moen K. **Patients’ perceptions of improvements after bilateral sagittal split osteotomy advancement surgery: 10 to 14 years of follow-up**. *Am J Orthod Dentofacial Orthop* (2012.0) **141** 204-12. DOI: 10.1016/j.ajodo.2011.06.039
63. Tudorici T, Feier R, Balcoș C, Forna N. **Socio-demographic factors and the partial edentulism in the adult population from Iasi, Romania**. *Romanian Journal of Oral Rehabilitation* (2017.0) 9
64. Sriram K, Abilasha R. **Awareness of allergic reactions to dental drugs and materials among patients, dentists and dental personnel—a cross sectional study**. *Pharm. Sci. & Res* (2016.0) **8** 1050-7
65. Chisini LA, Collares K, Cademartori MG, Oliveira LJC de, Conde MCM, Demarco FF. **Restorations in primary teeth: a systematic review on survival and reasons for failures**. *Int J Paediatr Dent* (2018.0) **28** 123-39. DOI: 10.1111/ipd.12346
66. Kassardjian V, Varma S, Andiappan M, Creugers NHJ, Bartlett D. **A systematic review and meta analysis of the longevity of anterior and posterior all-ceramic crowns**. *J Dent* (2016.0) **55** 1-6. DOI: 10.1016/j.jdent.2016.08.009
67. Hussey PS, Wertheimer S, Mehrotra A. **The association between health care quality and cost: a systematic review**. *Ann Intern Med* (2013.0) **158** 27-34. DOI: 10.7326/0003-4819-158-1-201301010-00006
68. van der Marjon P, Shiell A, Au F, Johnston D, Tough S. **Convergent validity between a discrete choice experiment and a direct, open-ended method: comparison of preferred attribute levels and willingness to pay estimates**. *Soc Sci Med* (2008.0) **67** 2043-50. DOI: 10.1016/j.socscimed.2008.09.058
69. Narby B, Kronström M, Söderfeldt B, Palmqvist S. **Changes in attitudes toward desire for implant treatment: a longitudinal study of a middle-aged and older Swedish population**. *The International journal of prosthodontics* (2008.0) **21** 481-5. PMID: 19149061
70. Moon RJ, Millar BJ. **Dental aesthetics: a study comparing patients’ own opinions with those of dentists**. *OJST* (2017.0) **07** 225-33. DOI: 10.4236/ojst.2017.74016
71. Müller F, Shimazaki Y, Kahabuka F, Schimmel M. **Oral health for an ageing population: the importance of a natural dentition in older adults**. *Int Dent J* (2017.0) **67** 7-13. DOI: 10.1111/idj.12329
72. Carvalho MA de, Lazari PC, Gresnigt M, Del Bel Cury AA, Magne P. **Current options concerning the endodontically-treated teeth restoration with the adhesive approach**. *Braz Oral Res* (2018.0) **32** e74. DOI: 10.1590/1807-3107bor-2018.vol32.0074
73. Bernabé E, Sheiham A, Sabbah W. **Income, income inequality, dental caries and dental care levels: an ecological study in rich countries**. *Caries Res* (2009.0) **43** 294-301. DOI: 10.1159/000217862
74. Singh A, Peres MA, Watt RG. **The relationship between income and oral health: a critical review**. *J Dent Res* (2019.0) **98** 853-60. DOI: 10.1177/0022034519849557
75. Murakami K, Hashimoto H. **Wealth-related versus income-related inequalities in dental care use under universal public coverage: a panel data analysis of the Japanese study of Aging and Retirement**. *BMC Public Health* (2016.0) **16** 24. DOI: 10.1186/s12889-015-2646-9
76. 76German Private Health Insurance (PKV). Starker Zuwachs: 16,9 Millionen Deutsche haben eine Zahnzusatzversicherung [Strong growth: 16.9 million Germans have supplementary dental insurance] [German]. 25.05.2022. https://www.pkv.de/verband/presse/meldungen-2021/starker-zuwachs-169-millionen-deutsche-haben-eine-zahnzusatzversicherung/. Accessed 25 May 2022.
77. Born G, Baumeister SE, Sauer S, Hensel E, Kocher T, John U. **Characteristics of risk groups with an insufficient demand for dental services. Results of the Study of Health in Pomerania (SHIP) [Merkmale von Risikogruppen einer unzureichenden Inanspruchnahme zahnmedizinischer Leistungen. Ergebnisse der Study of Health in Pomerania (SHIP)]. [Characteristics of risk groups with an insufficient demand for dental services—results of the study of health in Pomerania (SHIP)]**. *Gesundheitswesen* (2006.0) **68** 257-64. DOI: 10.1055/s-2006-926723
78. van den Branden S, van den Broucke S, Leroy R, Declerck D, Hoppenbrouwers K. **Measuring determinants of oral health behaviour in parents of preschool children**. *Community Dent Health* (2013.0) **30** 19-25. PMID: 23550502
79. Spalj S, Slaj M, Athanasiou AE, Govorko DK, Slaj M. **The unmet orthodontic treatment need of adolescents and influencing factors for not seeking orthodontic therapy**. *Coll Antropol* (2014.0) **38** 173-80
80. Stock S, Stollenwerk B, Klever-Deichert G, Redaelli M, Büscher G, Graf C. **Preliminary analysis of short-term financial implications of a prevention bonus program: first results from the German statutory health insurance**. *Int J Public Health* (2008.0) **53** 78-86. DOI: 10.1007/s00038-008-7026-0
81. Atieh MA, Morgaine KC, Duncan WJ. **A qualitative analysis on participants’ perspectives on oral implants**. *Clin Oral Implants Res* (2016.0) **27** 383-91. DOI: 10.1111/clr.12558
82. Oliveira PGdSA, Tavares RR, Freitas JC. **Assessment of motivation, expectations and satisfaction of adult patients submitted to orthodontic treatment**. *Dental Press J Orthod* (2013.0) **18** 81-7. DOI: 10.1590/s2176-94512013000200018
|
---
title: Certainty equivalence-based robust sliding mode control strategy and its application
to uncertain PMSG-WECS
authors:
- Annas Chand
- Qudrat Khan
- Waqar Alam
- Laiq Khan
- Jamshed Iqbal
journal: PLOS ONE
year: 2023
pmcid: PMC9970108
doi: 10.1371/journal.pone.0281116
license: CC BY 4.0
---
# Certainty equivalence-based robust sliding mode control strategy and its application to uncertain PMSG-WECS
## Abstract
This work focuses on maximum power extraction via certainty equivalence-based robust sliding mode control protocols for an uncertain Permanent Magnet Synchronous Generator-based Wind Energy Conversion System (PMSG-WECS). The considered system is subjected to both structured and unstructured disturbances, which may occur through the input channel. Initially, the PMSG-WECS system is transformed into a Bronwsky form, i.e., controllable canonical form, which is composed of both internal and visible dynamics. The internal dynamics are proved stable, i.e., the system is in the minimum phase. However, the control of visible dynamics, to track the desired trajectory, is the main concern. To carry out this task, the certainty equivalence-based control strategies, i.e., conventional sliding mode control, terminal sliding mode control and integral sliding mode control are designed. Consequently, a chattering phenomenon is suppressed by the employment of equivalent estimated disturbances, which also enhance the robustness of the proposed control strategies. Eventually, a comprehensive stability analysis of the proposed control techniques is presented. All the theoretical claims are verified via computer simulations, which are performed in MATLAB/Simulink.
## 1 Introduction
In the current era, due to the depletion of fossil fuels and their environmental impacts, researchers have focused on renewable energy resources (RES). In various renewable energy resources, energy harnesses from wind are getting much importance due to their sustainable and environment-friendly nature [1–3]. The system used for the said purpose is Permanent Magnet Synchronous Generator-based Wind Energy Conversion System (PMSG-WECS).
WECS is either autonomous or grid-connected [4]. However, depending upon the wind speed that is generated using the anemometer data, WECS can be operated within three different regimes, i.e., no load, partial load and constant load [5]. For partial load, the efficiency of the WECS is more crucial. Thus, to maximise the WECS efficiency, in a partial load scenario, the Maximum Power Point Tracking (MPPT) method has been presented. The controller in MPPT acts as a backbone in the operation of MPPT, which captures maximum energy from wind. While its control functioning is directly related to the operating characteristics, economic effective generation and equipment security stability. So far, various kinds of control strategies are proposed for MPPT design, which include [6, 7], PID control, model predictive control [8], neuro-fuzzy control [9, 10], adaptive backstepping control [11, 12], sliding mode control scheme [13–15] and an integral-based terminal sliding mode control strategy [16]. Moreover, interconnection and damping assignment-based control schemes are presented in [17, 18]. The aforementioned control strategies are synthesized for the WECs to get maximum efficiency, i.e., MPPT, while taking into account the PMSG’s entire dynamic. However, each strategy has its own merits and demerits. Recently, a continuous switching-based sliding mode control scheme is presented in [19]. The proposed control strategy, initially, regulates the grid and the generator side converter to track the desired reference speed. Secondly, it alleviates the chattering issue associated with the conventional sliding mode control scheme.
In distant and harsh circumstances, during a long-term ongoing operation, partial failure of electro-mechanical parts, i.e., gearbox, motor, alternator and power electronic converter, is inevitable [20–23]. These partial faults may lead to poor performance of the actuators, which can result in performance degradation and efficient operation of PMSG-WECS. Therefore, keeping in view the safety, high reliability and long life of WECS, the research community have focused on robust methods, for the control design, to ensure excellent working in uncertain situations.
Robust control strategy ensures the system stability and specific performance not only in the nominal scenarios but also in case of external uncertainties [24–28]. It either counteracts the fault with efficient robustness by using a fixed gain controller or implements a fast dynamic compensation control input. Regarding WECS, the robust control techniques used in the literature are signal-based approach [29], hardware redundancy method [30], data-driven techniques [31], Barrier function-based adaptive non-singular sliding mode control approach [32], fractional-order sliding mode control technique [33], convolution neural network method [34], fuzzy method [35], global sliding mode control approach [36] and sliding mode observer method [37, 38]. Using these techniques, the information on the uncertainty, that occurred in PMSG-WECS, is obtained, which is then adjusted by the designed control law.
It is quite evident that, before designing the controller, the robust control methods require the designed engineers to forecast the bounds of the expected faults/uncertainty. During the faulty condition, a robust control scheme works according to the dynamics of the system, which need to be adjusted according to the specific fault dynamics. Its benefit is the simplicity of the control law, which can ensure a system’s stability and attain predetermined efficiency irrespective of a fault. During severe uncertain situations, the controller must be able to find out the exact gain parameters. As the complexity of the system increases, the design process of the controller becomes more difficult. In [39], a certainty equivalence-based super-twisting algorithm (CESTA) is implemented on a diesel engine for the diagnosis of match uncertainties and later it is counteracted. In [40], the certainty equivalence-based integral sliding mode control (CEISMC) technique is being utilized to diagnose and mitigate the actuator fault for the diesel engine.
It is worth mentioning that our contribution to this work is in three folds. Firstly, the dynamics of the PMSG-WECS are modelled via Park transformation and later transformed into a controllable canonical form, which is a feasible structure and assists us in the design of the control strategy. Secondly, a certainty equivalence-based conventional sliding mode control (CSCMC), terminal sliding mode control (TSMC) and integral sliding mode control are designed. The stability analysis of the designed control strategies is provided in a comprehensive manner. Moreover, the effectiveness of the designed control strategies is demonstrated in MATLAB/Simulink. In addition, the comparative analysis of the aforementioned control schemes is carried out as a third contribution.
This paper is organized as follows; Standalone PMSG-WECS modelling is presented in section 2. Section 3 presents the controllable canonical form and the investigation of zero dynamics in a nonlinear PMSG-WECS system. The design of CECSMC, CETSMC and CEISMC along with Lyapunov stability analysis are outlined in section 4. In section 5, simulations and discussion are presented whereas section 6 concludes the paper. In last, a declaration of conflicting interests is provided in section 7.
## 2 Mathematical modelling of standalone PMSG-WECS
This section is composed of two subsystems, i.e., rotor blade modelling and PMSG modelling, which is further connected to a load. Both are discussed comprehensively in the forthcoming subsections.
## 2.1 Aerodynamic modelling of wind turbine
As a result of fast-moving wind, which struck against the wind turbine blades, the linear wind energy is transformed into mechanical energy. The mechanical power generated via this phenomenon is represented as follows [4] Pm=12ρπRt2vw3Cp(λ,θ), [1] where *Pm is* the mechanical power, *Rt is* radius of blades of the wind turbine, ρ is density of the air, vw is the speed of the wind, λ is the tip speed ratio (TSR), θ is the pitch angle, *Cp is* the power coefficient, which depends on the λ and θ.
Assumption 1 *It is* assumed that the pitch angle is constant and is kept at zero i.e., (θ = 0).
A TSR is the ratio of the blade tip speed to the wind speed. The detailed expression of the TSR appears as follows λ=ΩlRtvw, [2] where Ωl is the rotational speed of the wind turbine blades at a low-speed shaft. As clearly seen in Fig 1, the mechanical output power of the wind turbine increases according to the wind speed. For every wind speed curve, there is a specific peak power point. By joining all these peak power points, a curve is formed which is known as optimal regime characteristics (ORC) [5]. For every wind speed vw, there is a specific generator speed at which the power coefficient Cp reaches its maximum value, i.e., Cpmax, whenever λ becomes λopt. So, in order to harvest maximum power from wind, the TSR should be kept at its optimal value, λopt, in such a way that the shaft speed Ωh exactly tracks the reference speed, Ωref, which is defined as follows Ωref=λoptvwRt [3]
**Fig 1:** *Mechanical power versus rotor speed.*
The rotor power of PMSG is given as follows Pt=ΓtΩl [4] The aerodynamic torque is given by Γt=12πRt3vw2Cq(λ), [5] where Cq(λ) is the coefficient of torque, which can be expressed as follows Cq(λ)=Cp(λ)λ, [6] where Cq(λ), Cp(λ) and λopt are the designed parameters, which are provided by the manufacturer of the wind turbine. Now, all the necessary components of the wind turbine are modelled. In the subsequent subsection, the PMSG modelling is displayed.
## 2.2 Modelling of the Permanent Magnet Synchronous Generator (PMSG)
As the considered PMSG is a standalone system, the power generated is pre-processed for the sake of compatibility before it is stored in the battery banks for later use. The dq-model of the PMSG, while ignoring the zero dynamics, is given by ddtid=-Rid+Lqiq+udLdddtiq=-Riq-(Ldid+ϕm)+uqLqddtΩh=1Jh(Γt-Γem)=ΓtJh-p(ϕdiq-ϕqid)Γt} [7] where *Jh is* the moment of inertia of the generator shaft. ud, uq, Ωh, Γem, id, iq, Ld, and Lq, R, p, Φd = Ldid + Φm, Φq = Lqiq and Φm are the voltages of the DQ-axis, PMSG speed, the electromagnetic torque, DQ-axis current and rotor inductance, the resistance of the stator, pole pair, DQ and the linkage flux, respectively. The under study system is non-salient synchronous generator so Ld = Lq = L. The torque of the generator can be written as Γem=pΦmiq [8] Consider the following assumptions while modeling PMSG.
Assumption 2 Stator winding should have sinusoidal distribution with having an electrical and magnetic symmetry. In addition, the iron losses are not considered.
Remark 1 The dynamics of the electronic circuit are neglected, due to its faster nature as compared to the dynamics of the PMSG-WECS.
The nonlinear dynamics of the SISO PMSG-WECS presented in [7] and [8] can be formulated as x˙1=-Rsx1+p(Lq-Lch)x2x3-Rinix1(Ld+Lch)x˙2=-Rsx2-p(Ld+Lch)x1x3-Rinix2(Lq+Lch)+pΦmx3x˙3=d1vw2i+d2vwx3i2+d3x32i3-pΦmx2Jh} [9] The matrix [x1 x2 x3]T = [id iq Ωh]T ∈ ℜ3, indicates the states vector of the PMSG model. Ωh = Ωl × i is the generator speed, Lch and Rch are the inductance and resistance of the load, respectively. i is the mechanical transmission ratio, *Rch is* considered as a control input. Ld and Lq represent the stator’s dq-axes inductance while id and iq stand for the dq-axes stator’s current, respectively.
## 3 Controllable canonical form
The dynamics of the PMSG-WECS model, in generic form, can be expressed as x˙=f(x)+g(x)uy=h(x)} [10] The variable x ∈ ℜn represents the state vector. The control input is described by u ∈ ℜm, while f(x) and g(x) are the smooth nonlinear vectors which are expressed as f(x)=[-Rsx1+p(Lq-Lch)x2x3(Ld+Lch)-Rsx2-p(Ld+Lch)x1x3(Lq+Lch)+pΦmx3d1vw2i+d2vwx3i2+d3x32i3-pΦmx2Jh], g(x)=[-x1(Ld+Lch)-x2Lq+Lch0] where u=Rch The output is represented by y = h(x) = x3 = Ωh, which describes the angular speed of the rotor shaft. As the objective is to control the output, thus, [9] is transformed into a controllable canonical structure, i.e., input-output form, via the following transformation. z1=y=h(x)=x3=Ωhz2=Lfh(x)=∂h(x)∂x.f(x)z3=Lf2h(x)=x1x2} [11] *It is* quite obvious that the relative degree ‘r’ of the considered system is one less than the system order, i.e., (r < n) as $$n = 3$.$ So the input-output form can be expressed as z˙1=z2z˙2=Lf2h(x)+LgLfh(x)u} [12] z˙3=-m4m1(k1z3m1m4+k2z1m1m4+k3z3m1um4)+(z3m1m4)(m42m12)(-l1m1m4l2m1z3z1m4-l3z1+l4m1um4) [13] The system under consideration, when converted to an input-output form, contains an internal dynamic, i.e., zero-dynamic. The stability of the zero-dynamics is quite crucial to be discussed.
Remark 2 The Eq [13] remains no more dependent on the control input. This system is affected only by the control-driven states i.e., z1 and z3. Its zero dynamics will be discussed in subsequent paragraphs.
The typical plant parameters and the derived parameters are given in Tables 1–3. Note that the nonlinear system [12], is driven by the applied control input u whereas system [13], with states (z1, z3), represents the internal dynamics whose stability, zero-dynamics stability, will be discussed in the following subsection.
## 3.1 Stability analysis of the zero-dynamics
The dynamics of the nonlinear system are subdivided into two subsystems, i.e., visible dynamics system and internal dynamics system [41]. To find out the zero dynamics, choosing z1 = z2 = 0 in [13] and simplifying it, one comes with z˙3=-z3[-h1+m1-m2K1K4] [14] Owing to Table 1, the constant -h1+m1-m2K1K4=τ is positive with numerical value 190.21. z3˙=-τz3 [15] This equation shows that the zero dynamics are strongly asymptotically stable. Thus, the system under study is the minimum phase.
Remark 3 The system developed in [12] and [13], in practice, experiences a different kinds of disturbances.
Hence, the system [12] and [13], in practical form can be described as follows z˙1=z2z˙2=Lf2h(x)+LgLfh(x)(u(1-G)+F(x,t))z3˙=-τz3} [16] where G represents the health of the input channel. If $G = 0$, it means that the system’s input channel is healthy and 0 < G < 1 indicates the unhealthy nature of the input channel. In addition, F(x, t) represents the uncertainties about which the following assumption is made.
Assumption 3 Assume that the uncertain terms can be subdivided into structured and unstructured terms, i.e., F(x,t)=fst(x)+fun(x,t) [17] The bound of the unstructured faults/uncertainties are defined as |fun(x,t)|≤ϑ, [18] where ϑ is a positive constant. fst(x)=ΔΨ(x) [19] The partially known structured uncertainty in [19] can be expressed as the product of an unknown constant parameter Δ and a known base function Ψ(x). Δ can be any parametric change, which occurs in the internal parameters of the wind system. Structured faults/uncertainties may be unknown plant parameters like resistance values or friction coefficients whereas unstructured faults may represent external disturbances. The system [16] represents a complete model of WECS-PMSG. In the next section, the control design will be focused on.
## 4 Certainty equivalence-based sliding mode control strategy
In this section, a synthetic structure of sliding modes and adaptive control is proposed. Conventional sliding mode control (SMC) scheme claims invariance property subjected to the design of the sliding surface and the gains of the discontinuous part. However, it may result in high chattering phenomena which could be dangerous for the actuators and the system’s health. Therefore, a certainty equivalence-based sliding mode control protocol is proposed. The beauty of this strategy is that the robustness of the controller remains higher and the chattering is eliminated or suppressed, which is not possible in the conventional SMC. The design is outlined in the following subsection.
## 4.1 Certainty equivalence-based conventional SMC design
It is quite worth mentioning that the main task of the current work is the extraction of maximum power from the WECS, which can be done by following a reference signal. Thus, reference tracking is the ultimate objective.
Assumption 4 *It is* assumed that the reference speed is of class C1.
Now, by defining the error as follows e=z1-z1ref,e˙=z˙1-z˙1ref [20] To pursue the design, a sliding surface/manifold, in terms of error variable, is defined as follows s=e˙+c1e, [21] where c1 is a positive constant. The time derivative of [21] along [20] and [16], becomes s˙=Lf2h(x)+LgLfh(x)(u+F(x,t))-z¨1ref+c1e˙ [22] Substituting the match faults/uncertainties from [18] and [19] in [22], the following expression is obtained s˙=Lf2h(x)+LgLfh(x)(u+fst(x)+fun(x,t))-z¨1ref+c1e˙ [23] The final control law, u composed of an equivalent ueq and discontinuous ud control laws, which can be written as u=uo+ueq+ud [24] Invoking [24] in [23], one gets s˙=Lf2h(x)+LgLfh(x)(uo+ueq+ud+fst(x)+fun(x,t))-z¨1ref+c1e˙ [25] To calculate the equivalent control input, the uncertain terms and s˙ must be equal to zero, which gives us the following expression. ueq=-1LgLfh(x)(Lf2h(x)-z¨1ref+c1e˙) [26] In order to mitigate the effects of the structured uncertainties, an equivalent cancellation law uo is proposed as follows uo=-Δ^Ψ(x) [27] where Δ^ represents the estimated value of the unknown parameter. The discontinuous control law ud is designed as follows ud=-1LgLfh(x)(M1s+M2sign(s)), [28] where M1 and M2 are the positive gains. The obtained control law is as follows u=-Δ^Ψ(x)-1LgLfh(x)(Lf2h(x)-z¨1ref+c1e˙)-1LgLfh(x)(M1s+M2sign(s)) [29] The aforementioned final control law enforces the sliding mode along the sliding surface given in [21]. To prove the closed loop stability, i.e., sliding mode enforcement, consider a Lyapunov candidate function (LCF) as follows $V = 12$s2+12γΔ˜2, [30] where γ > 0 and Δ˜ = Δ − Δ^ is the error between the actual and estimated parameter. Now, consider the time derivative of [30], one has V˙=ss˙+1γΔ˜(Δ˙-Δ^˙) [31] Substituting [23] into [31], we get V˙=s(Lf2h(x)+LgLfh(x)(u+fst(x)+fun(x,t))-z¨1ref+c1e˙)+1γΔ˜(0-Δ^˙) [32] Using Assumption 3, the following expression is obtained V˙=s(Lf2h(x)+LgLfh(x)u+LgLfh(x)ΔΨ(x)+LgLfh(x)fun(x,t)-z¨1ref+c1e˙)+1γΔ˜(0-Δ^˙) [33] Using values of ueq, uo and ud, one gets the following expression V˙=s(Lf2h(x)+LgLfh(x)ueq+LgLfh(x)ud-LgLfh(x)Δ^Ψ(x)+LgLfh(x)ΔΨ(x)+LgLfh(x)fun(x,t) V˙=s(-M1s-M2sign(s)+LgLfh(x)fun(x,t))+sLgLfh(x)Ψ(x)(Δ-Δ^)-1γΔ˜Δ^˙ V˙≤-M1s2-|s|(M2-Γmϑ)+Δ˜(sLgLfh(x)Ψ(x)-1γΔ^˙) [34] Now, choosing Δ^˙=γsLgLfh(x)Ψ(x), the derivative of LCF can be written as V˙≤-M1s2-|s|(M2-Γmϑ)+0 [35] Now, choose the following expression M2-Γmϑ≥η>0, [36] one gets V˙≤-M1s2-η|s| [37] The inequality 37 proves the negative definiteness of the LCF. Hence, it is confirmed that sliding mode enforcement is achieved in finite time, i.e., s → 0, subjected to the conditions, i.e., M2 ≥ η + Γmϑ and Δ^˙=γsLgLfh(x)Ψ(x). This proves the theorem.
Remark 4 The value of the adaptation gain parameter Δ^ will remain close to zero where there is no structured uncertainty in the framework. However, when some fault affects the framework, the value of the adaptation gain parameter increases according to the magnitude of the fault. A non-zero value of the adaptation parameter indicates the presence of disturbances.
The above-mentioned strategy is developed with terminal SMC which is discussed in the subsequent subsection.
## 4.2 Certainty equivalence-based terminal sliding mode control strategy
To pursue the design of a certainty equivalence-based TSMC scheme, consider the tracking error and its time derivative, the terminal sliding manifold [42] is defined as s=e˙+αe+βepq, [38] where α> β > 0, p and q are odd positive numbers such that 0<pq<1.
Remark 5 The difference between [21] and [38] is simply the addition of the new term on the right-hand side of [38]. The beauty of this manifold is that, as a sliding mode is enforced, the error dynamics converge to zero in finite time instead of asymptotic convergence which results in high precision as compared to CSMC.
Taking the time derivative of [38] along [20] and [16], the following expression is obtained. s˙=Lf2h(x)+LgLfh(x)(u+F(x,t))-z¨1ref+αe˙+βpqe(pq-1)e˙ [39] Substituting [18] and [19] into [39], it yields s˙=Lf2h(x)+LgLfh(x)(u+fst(x)+fun(x,t))-z¨1ref+αe˙+βpqe(pq-1)e˙ [40] The overall control law appears as follows u=uo+ueq+ud, [41] where known terms will be canceled by ueq and matched faults will be handled by ud and uo. Invoking [41] in [40], one gets the following expression. s˙=Lf2h(x)+LgLfh(x)(uo+ueq+ud+fst(x)+fun(x,t))-z¨1ref+αe˙+βpqe(pq-1)e˙ [42] Ignoring the disturbances, uo and ud in [42], one gets ueq=-1LgLfh(x)(Lf2h(x)-z¨1ref+αe˙+βpqe(pq-1)e˙) [43] In order to cancel the effects of structured uncertainties, an equivalent cancellation law uo is proposed as follows uo=-Δ^Ψ(x), [44] where Δ^ represents the estimated value of the unknown parameter. The discontinuous control law ud appears as follows ud=-1LgLfh(x)(M1s+M2sign(s)), [45] The overall control law can be written as u=-Δ^Ψ(x)-1LgLfh(x)(Lf2h(x)-z¨1ref+αe˙+βpqe(pq-1)e˙)-1LgLfh(x)(M1s+M2sign(s)) [46] This control law enforces the sliding mode along the sliding surface. Consequently, the system’s output tracks the desired reference in a finite time. The stability analysis of the current control scheme and the aforementioned control strategy is quite similar. The only difference is the few additional terms in the sliding manifold and the equivalent control law. Thus, the details are avoided here. Again, the same strategy is developed with integral SMC which is presented below.
## 4.3 Certainty equivalence-based integral sliding mode control strategy
To pursue the design of Certainty equivalence-based ISMC, the integral sliding manifold [40] is defined as follows s=e˙+c2e+v, [47] where c2 is a positive constant and v is an integral term that results in the elimination of reaching phase. The time derivative of [47] along [20] and [16] produces the following equation. s˙=Lf2h(x)+LgLfh(x)(u+F(x,t))-z¨1ref+c2e˙+v˙ [48] Now, using [18] and [19] in [48], we get the following expression. s˙=Lf2h(x)+LgLfh(x)(u+fst(x)+fun(x,t))-z¨1ref+c2e˙+v˙ [49] The overall control law appears as follows u=uo+ueq+ud [50] Substituting [50] in [49], yields s˙=Lf2h(x)+LgLfh(x)(uo+ueq+ud+fst(x)+fun(x,t))-z¨1ref+c2e˙+v˙ [51] To calculate the regularizing/equivalent control input, assuming the uncertain terms and s˙ equal to zero, which results in ueq=-Lf2h(x)LgLfh(x) [52] This selection of ueq decouples the system and gives desired output for a nominal plant with no faults. Taking v˙ as v˙=z¨1ref-c2e˙ [53] The selection of v˙ and v[0] = −s[0] confirms the elimination of reaching phase and, thus, sliding mode is initiated from the very initial time. In order to cancel the effects of the structured uncertainties, an equivalent cancellation law uo is proposed as follows uo=-Δ^Ψ(x), [54] where Δ^ represents the estimated value of the unknown parameter. The discontinuous control law ud appears as follows ud=-1LgLfh(x)(M1s+M2sign(s)), [55] where M1 and M2 are positive gains. The overall control law can be written as u=-Δ^Ψ(x)-1LgLfh(x)(Lf2h(x))-1LgLfh(x)(M1s+M2sign(s)) [56] The final control law 56 enforces the sliding mode 47 along the sliding surface, given in [49], in a finite instant of time. The schematic of the overall closed-loop system, i.e., a variable speed wind turbine (VSWT), a gearbox, power electronic converters and a PMSG coupled with a VSWT, is represented in Fig 2.
**Fig 2:** *Schematic of the overall system, i.e., PMSG-based WECS.*
## 5 Simulation results and discussion
In this section, the effectiveness of the designed control approaches, proposed for the maximum power extraction, is verified in MATLAB/Simulink. Moreover, the block diagram of the control strategies, in detail is depicted in Fig 2. The assessment, of the proposed control laws, is presented for the following two cases.
Note that all the matched disturbances are injected at time t ≥ 1 sec. The input channel is choked to $30\%$, i.e., the input channel is $70\%$ healthy and $30\%$ faulty. The unknown part of the structured fault is $2\%$ which is estimated by adaptation law. The parametric variations for inductance and inertia are carried out at times 5 ≤ t ≤ 15 and 30 ≤ t ≤ 50.
## 5.1 Case 1
The simulation results demonstrated in Fig 3 illustrate the estimation of unknown parts, i.e., matched uncertainties, which is subjected to the system via input channel. The aforesaid task is performed by using the proposed control strategies. Fig 3 represents the estimation of the matched uncertainties, corresponding to CEISMC, CETSMC, and CECSMC. It is quite obvious that CEISMC best estimates the unknown uncertainties having known bounds whereas CETSMC also perform quite efficiently but CECSMC possess a steady-state error (SSE), which exists thereafter. Fig 4 depicts the tracking profile achieved by the designed control schemes. The tracking profile is actually the difference between the reference speed of the generator, i.e., Ωref, versus the actual speed of the wind turbine, i.e., Ωh. In Fig 4, *It is* pretty clear that CEISMC accurately tracks the desired reference speed with a negligible steady-state error. In contrast, the steady-state errors that correspond to CETSMC and CECSMC are quite maximum and show no decline with the passage of time. The Optimal Regime Characteristics (ORC) is basically a combination of maximum power points at different wind speeds. Thus, in Fig 5, a nonlinear relationship between the actual speed of the wind turbine and the power produced by the generator is demonstrated. It is obvious that the characteristics achieved by both CEISMC and CETSMC lie closer to the Optimal Regime Characteristics (ORC) while CECSMC didn’t achieve the optimal results.
**Fig 3:** *Delta estimation versus time.* **Fig 4:** *High speed shaft rotational speed.* **Fig 5:** *Low-speed shaft rotational speed versus low-speed shaft power.*
In Fig 6, the tip speed ratio that corresponds to a high-speed shaft power is portrayed. It is quite evident that the CEISMC strategy, in contrast to CETSMC and CECSMC, successfully achieved the tip speed ratio, i.e., 7, which is an optimal operating point. In addition, the tip speed ratio at low-speed shaft power, achieved by the designed control schemes, is presented in Fig 7. Again, the CEISMC strategy proved itself as the best candidate to achieve the optimal tip speed ratio at low shaft power. Fig 8 portrays the different tip speed ratio, i.e., λ, versus time that are achieved by the proposed control schemes. In the case of CEISMC scheme, the attained tip speed ratio, i.e., 7, is quite close to the optimal value. However, in CETSMC and CECSMC, it oscillates around 7 and never stabilises itself at any constant position. In Fig 9, the profile of power coefficient Cp that are accomplished by the designed control schemes is demonstrated. The desirable maximum power coefficient, i.e., Cpmax, for the VSWT system is $47\%$, which is quite efficiently attained by CEISMC technique. However, in the case of CETSMC and CECSMC, a little bit variations are observed, which degrades its effectiveness. As the matched faults are injected in the system at t ≥ 1, so it is quite obvious in Figs 4–9, that despite the faulty situation there is no degradation observed in any of the aforesaid performance. It means that the designed control strategies has sufficient effectiveness to overcome the uncertain condition and thus, the objective is fulfilled, i.e., MPPT is achieved. In Figs 6–9, it is analysed that the designed control schemes reduced the chattering phenomenon. The CEISMC strategy quite effectively mitigates the chattering phenomena while the CETSMC and CECSMC algorithms possess a little bit of chattering, which is dangerous for the actuator health. All the presented simulations are performed for the system, which possess modelled dynamics along with matched uncertainties. So, it is concluded that the proposed control schemes handled the stochastic nature of wind speed, counteract the matched faults, reduced the chattering effect, and efficiently achieved the MPPT.
**Fig 6:** *Tip speed ratio versus high speed shaft power.* **Fig 7:** *Tip speed ratio versus low speed shaft power.* **Fig 8:** *Tip speed ratio versus time.* **Fig 9:** *Power coefficient versus time.*
## 5.2 Case 2
In this subsection, the efficacy of the designed control strategies for the system that is subjected to un-modelled dynamics, i.e., variable load and inertia, are discussed. The system, which is under consideration, is also exposed to uncertainties that are entering via input channel, i.e., matched uncertainties. In Fig 10, the estimation of matched uncertainties, i.e., Δ^, by using the designed strategies are illustrated. It is pretty obvious in Fig 10 that the CEISMC scheme accurately tracks the desired reference trajectory and has a negligible steady-state error, i.e., $2\%$. While the CETSMC quite closely tracks the desired reference trajectory and lies in the vicinity but still has a sufficient amount of tracking error whereas the CECSMC strategy doesn’t track the desired trajectory and possesses the steady-state error, which exists thereafter. In Fig 11, the difference between the actual speed of the generator, i.e., Ωref, and the references trajectory, i.e., Ωh, is depicted. It is quite clear that the CEISMC scheme tracks the reference trajectory in a short time span with a minimum steady-state error. However, the designed strategies, i.e., CEISMC and CETMC, either suffer from long settling time or maximum steady-state error. Fig 12 demonstrates the nonlinear plot comparative analysis between the actual speed of the wind turbine and the power produced by generator. The comparative profiles that are achieved by both CEISMC and CETMC schemes best matches the optimal regime characteristics, which outshine its supremacy in the achievement of MPPT. However, the performance achieved by CECSMC is out of the way and is not suitable for the attainment of an efficient MPPT.
**Fig 10:** *Delta estimation versus time.* **Fig 11:** *High speed shaft rotational speed.* **Fig 12:** *Low shaft rotational speed versus low-speed shaft power.*
Fig 13 shows the comparative illustration of the wind turbine’s power and its tip speed ratio at a high-speed shaft while Fig 14 also represents the same parameters but at a low shaft speed. It is evident in both scenarios that the CEISMC strategy quite efficiently attains the optimal tip speed ratio, i.e., 7, for the stochastic nature of wind speed. However, by employing CETSMC and CECSMC strategies, the tip speed ratio deviates from the optimal value. In Fig 15, the tip speed ratios, i.e., λ, versus time, corresponding to the designed control strategies, are depicted. It is clearly portrayed that the CEISMC scheme accomplishes the optimal tip speed ratio, i.e., 7. while the tip speed ratios that correspond to CETSMC and CECSMC strategies, oscillate around 7 and don’t stabilise at any constant position. Fig 16 depicts the comprehensive profile of the power coefficient Cp. It demonstrates the power coefficient through-out the course of the simulation. It can be obviously seen that the value of Cp lie in the close vicinity of Cpmax despite all the fluctuations in the wind speed, un-modelled dynamics and matched faults. The desirable maximum power coefficient Cpmax for the VSWT system is $47\%$, which is accurately achieved by CEISMC without any oscillations. However, in the case of CETSMC and CECSMC schemes, the undesirable oscillations are observed, which is an unwanted phenomenon. The designed control strategies play a vital role in the mitigation of matched faults. It can be clearly seen in Figs 11–16 that uncertainties are mitigated while ensuring closed-loop stability. Also, the MPPT is achieved. The control schemes, i.e., CETSMC and CECSMC, are a little bit affected by unmodeled dynamics. However, CEISMC remained unaffected and over-perform. So, it is concluded that the proposed control scheme, i.e., CEISMC, is an appealing candidate for the achievement of MPPT in WECS.
**Fig 13:** *Tip speed ratio versus high speed shaft power.* **Fig 14:** *Tip speed ratio versus low speed shaft power.* **Fig 15:** *Tip speed ratio versus time.* **Fig 16:** *Power coefficient versus time.*
## 6 Conclusion
A certainty equivalence-based robust CSMC, TSMC and ISMC have been presented in this work to extract maximum power from a wind energy system, termed PMSG-WECS. The considered system is exposed to both structure and unstructured uncertainties/faults, i.e., load and inertia. Initially, the system, i.e., PMSG-WECS, is converted into controllable canonical form and stability of the zero dynamics is guaranteed. Secondly, certainty equivalence-based robust control laws are designed. The said control strategies efficiently attain the desired performance, i.e., regarding MPPT, for the WECS. The chattering issue, which can affect the actuator’s health, is sufficiently reduced in the control inputs. To comparatively analyse the performance of the designed control techniques, CSMC asymptotically achieves the sliding mode enforcement. As a result, the system output trajectory effectively tracks the desired reference path. While TSMC acquires a finite time error convergence along with suppressed chattering. The aforementioned both control strategies are pretty sensitive to uncertainties in the reaching phase, therefore, an ISMC scheme is proposed. In contrast, an ISMC strategy attains a sliding mode from the initial point, thus eliminating a reaching phase. The absence of reaching phase efficiently improves the robustness of the system to the un-modelled dynamics as well as the sliding mode enforcement is accomplished in a finite time instant. The overall control strategies are numerically developed and stability analysis is guaranteed via the Lyapunov candidate function. The theoretical claims are certified via computer simulations performed in MATLAB/Simulink. The results obtained are discussed thoroughly and it is concluded that ISMC outshines all the reported techniques in the overall performance.
## References
1. 1Jahanpour-Dehkordi M, Vaez-Zadeh S, Ghadamgahi A. An Improved Combined Control for PMSG-Based Wind Energy Systems to Enhance Power Quality and Grid Integration Capability. In: 2019 10th International Power Electronics, Drive Systems and Technologies Conference (PEDSTC). IEEE; 2019. p. 566–571.
2. Ayadi M, Derbel N. *Modeling, Identification and Control Methods in Renewable Energy Systems* (2019.0) 167-191
3. Guerrero JM, Lumbreras C, Reigosa D, Fernandez D, Briz F, Charro CB. **Accurate rotor speed estimation for low-power wind turbines**. *IEEE Transactions on Power Electronics* (2019.0) **35** 373-381. DOI: 10.1109/TPEL.2019.2913781
4. Munteanu I, Bratcu AI, Cutululis NA, Ceanga E. *Optimal control of wind energy systems: towards a global approach* (2008.0)
5. Soufi Y, Kahla S, Bechouat M. **Feedback linearization control based particle swarm optimization for maximum power point tracking of wind turbine equipped by PMSG connected to the grid**. *International journal of hydrogen energy* (2016.0) **41** 20950-20955. DOI: 10.1016/j.ijhydene.2016.06.010
6. Baloch MH, Wang J, Kaloi GS. **Stability and nonlinear controller analysis of wind energy conversion system with random wind speed**. *International Journal of Electrical Power & Energy Systems* (2016.0) **79** 75-83. DOI: 10.1016/j.ijepes.2016.01.018
7. Moradi H, Vossoughi G. **Robust control of the variable speed wind turbines in the presence of uncertainties: A comparison between**. *Energy* (2015.0) **90** 1508-1521. DOI: 10.1016/j.energy.2015.06.100
8. Song D, Yang J, Dong M, Joo YH. **Model predictive control with finite control set for variable-speed wind turbines**. *Energy* (2017.0) **126** 564-572. DOI: 10.1016/j.energy.2017.02.149
9. Asghar AB, Liu X. **Adaptive neuro-fuzzy algorithm to estimate effective wind speed and optimal rotor speed for variable-speed wind turbine**. *Neurocomputing* (2018.0) **272** 495-504. DOI: 10.1016/j.neucom.2017.07.022
10. Mehmood Y, Aslam J, Ullah N, Alsheikhy AA, Din EU, Iqbal J. **Robust fuzzy sliding mode controller for a skid-steered vehicle subjected to friction variations**. *PLoS one* (2021.0) **16** e0258909. DOI: 10.1371/journal.pone.0258909
11. Ayadi M, Derbel N. **Nonlinear adaptive backstepping control for variable-speed wind energy conversion system-based permanent magnet synchronous generator**. *The International Journal of Advanced Manufacturing Technology* (2017.0) **92** 39-46. DOI: 10.1007/s00170-017-0098-3
12. Ali N, Alam W, Pervaiz M, Iqbal J. **Nonlinear Adaptive Backstepping Control Of Permanent Magnet Synchronous Motor**. *Revue Roumaine Des Sciences Techniques-Serie Electrotechnique Et Energetique* (2021.0) **66** 9-14
13. Alam W, Mehmood A, Ali K, Javaid U, Alharbi S, Iqbal J. **Nonlinear control of a flexible joint robotic manipulator with experimental validation**. *Strojniški vestnik-Journal of Mechanical Engineering* (2018.0) **64** 47-55
14. Alam W, Khan Q, Ali Riaz R, Akmeliawati R. **Glucose–insulin stabilization in type-1 diabetic patient: A uniform exact differentiator–based robust integral sliding mode control approach**. *International Journal of Distributed Sensor Networks* (2019.0) **15** 1550147719833573. DOI: 10.1177/1550147719833573
15. Alam W, Khan Q, Riaz RA, Akmeliawati R. **Arbitrary-order sliding mode-based robust control algorithm for the developing artificial pancreas mechanism**. *IET Systems Biology* (2020.0) **14** 307-313. DOI: 10.1049/iet-syb.2018.5075
16. Chatri C, Ouassaid M, Labbadi M, Errami Y. **Integral-type terminal sliding mode control approach for wind energy conversion system with uncertainties**. *Computers & Electrical Engineering* (2022.0) **99** 107775. DOI: 10.1016/j.compeleceng.2022.107775
17. Belkhier Y, Shaw RN, Bures M, Islam MR, Bajaj M, Albalawi F. **Robust interconnection and damping assignment energy-based control for a permanent magnet synchronous motor using high order sliding mode approach and nonlinear observer**. *Energy Reports* (2022.0) **8** 1731-1740. DOI: 10.1016/j.egyr.2021.12.075
18. Belkhier Y, Achour A, Bures M, Ullah N, Bajaj M, Zawbaa HM. **Interconnection and damping assignment passivity-based non-linear observer control for efficiency maximization of permanent magnet synchronous motor**. *Energy Reports* (2022.0) **8** 1350-1361
19. Majout B, Bossoufi B, Bouderbala M, Masud M, Al-Amri JF, Taoussi M. **Improvement of PMSG-Based Wind Energy Conversion System Using Developed Sliding Mode Control**. *Energies* (2022.0) **15** 1625. DOI: 10.3390/en15051625
20. Zappalá D, Sarma N, Djurović S, Crabtree C, Mohammad A, Tavner P. **Electrical & mechanical diagnostic indicators of wind turbine induction generator rotor faults**. *Renewable energy* (2019.0) **131** 14-24. DOI: 10.1016/j.renene.2018.06.098
21. Chou JS, Ou YC, Lin KY, Wang ZJ. **Structural failure simulation of onshore wind turbines impacted by strong winds**. *Engineering Structures* (2018.0) **162** 257-269. DOI: 10.1016/j.engstruct.2018.02.006
22. Spinato F, Tavner PJ, Van Bussel G, Koutoulakos E. **Reliability of wind turbine subassemblies**. *IET Renewable Power Generation* (2009.0) **3** 387-401. DOI: 10.1049/iet-rpg.2008.0060
23. Yang S, Bryant A, Mawby P, Xiang D, Ran L, Tavner P. **An industry-based survey of reliability in power electronic converters**. *IEEE transactions on Industry Applications* (2011.0) **47** 1441-1451. DOI: 10.1109/TIA.2011.2124436
24. Iqbal J, Ullah MI, Khan AA, Irfan M. **Towards sophisticated control of robotic manipulators: An experimental study on a pseudo-industrial arm**. *Strojniški vestnik-Journal of Mechanical Engineering* (2015.0) **61** 465-470. DOI: 10.5545/sv-jme.2015.2511
25. Ali N, Rehman AU, Alam W, Maqsood H. **Disturbance observer based robust sliding mode control of permanent magnet synchronous motor**. *Journal of Electrical Engineering & Technology* (2019.0) **14** 2531-2538. DOI: 10.1007/s42835-019-00256-0
26. Alam W, Ahmad S, Mehmood A, Iqbal J. **Robust sliding mode control for flexible joint robotic manipulator via disturbance observer**. *Interdisciplinary Description of Complex Systems: INDECS* (2019.0) **17** 85-97. DOI: 10.7906/indecs.17.1.11
27. Alattas KA, Vu MT, Mofid O, El-Sousy FF, Fekih A, Mobayen S. **Barrier Function-Based Nonsingular Finite-Time Tracker for Quadrotor UAVs Subject to Uncertainties and Input Constraints**. *Mathematics* (2022.0) **10** 1659. DOI: 10.3390/math10101659
28. Ali K, Mehmood A, Iqbal J. **Fault-tolerant scheme for robotic manipulator—Nonlinear robust back-stepping control with friction compensation**. *Plos one* (2021.0) **16** e0256491. DOI: 10.1371/journal.pone.0256491
29. Wu C, Guo C, Xie Z, Ni F, Liu H. **A signal-based fault detection and tolerance control method of current sensor for PMSM drive**. *IEEE Transactions on Industrial Electronics* (2018.0) **65** 9646-9657. DOI: 10.1109/TIE.2018.2813991
30. Chen L, Edwards C, Alwi H. **Sensor redundancy based FDI using an LPV sliding mode observer**. *IET Control Theory & Applications* (2018.0) **12** 1956-1963. DOI: 10.1049/iet-cta.2017.0818
31. Li M, Yu D, Chen Z, Xiahou K, Ji T, Wu Q. **A data-driven residual-based method for fault diagnosis and isolation in wind turbines**. *IEEE Transactions on Sustainable Energy* (2019.0) **10** 895-904. DOI: 10.1109/TSTE.2018.2853990
32. Mobayen S, Alattas KA, Fekih A, El-Sousy FF, Bakouri M. **Barrier function-based adaptive nonsingular sliding mode control of disturbed nonlinear systems: A linear matrix inequality approach**. *Chaos, Solitons & Fractals* (2022.0) **157** 111918. DOI: 10.1016/j.chaos.2022.111918
33. Pouzesh M, Mobayen S. **Event-triggered fractional-order sliding mode control technique for stabilization of disturbed quadrotor unmanned aerial vehicles**. *Aerospace Science and Technology* (2022.0) **121** 107337. DOI: 10.1016/j.ast.2022.107337
34. Wang J, Dou Y, Wang Z, Jiang D. **Multi-fault diagnosis method for wind power generation system based on recurrent neural network**. *Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy* (2019.0) **233** 604-615
35. Bakri AE, Boumhidi I. **Fuzzy model-based faults diagnosis of the wind turbine benchmark**. *Procedia Computer Science* (2018.0) **127** 464-470. DOI: 10.1016/j.procs.2018.01.144
36. Saghafi Zanjani M, Mobayen S. **Anti-sway control of offshore crane on surface vessel using global sliding mode control**. *International Journal of Control* (2022.0) **95** 2267-2278. DOI: 10.1080/00207179.2021.1906447
37. Saha S, Haque ME, Tan CP, Mahmud MA. **Sensor fault resilient operation of permanent magnet synchronous generator based wind energy conversion system**. *IEEE Transactions on Industry Applications* (2019.0) **55** 4298-4308. DOI: 10.1109/TIA.2019.2910777
38. Alam W, Khan Q, Riaz RA, Akmeliawati R, Khan I, Nisar KS. **Gain-scheduled observer-based finite-time control algorithm for an automated closed-loop insulin delivery system**. *IEEE Access* (2020.0) **8** 103088-103099. DOI: 10.1109/ACCESS.2020.2997776
39. Murtaza G, Bhatti AI, Butt YA. **Super twisting controller-based unified FDI and FTC scheme for air path of diesel engine using the certainty equivalence principle**. *Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering* (2018.0) **232** 1623-1633
40. Murtaza G, Bhatti AI, Butt YA. **Unified FDI and FTC Scheme for Air Path Actuators of a Diesel Engine Using ISM Extended with Adaptive Part**. *Asian Journal of Control* (2020.0) **22** 117-129. DOI: 10.1002/asjc.1871
41. Eghtesad M, Bazargan-Lari Y, Assadsangabi B. **Stability analysis and internal dynamics of MIMO GMAW process**. *IFAC Proceedings Volumes* (2008.0) **41** 14834-14839. DOI: 10.3182/20080706-5-KR-1001.02511
42. Abolvafaei M, Ganjefar S. **Maximum power extraction from a wind turbine using second-order fast terminal sliding mode control**. *Renewable Energy* (2019.0) **139** 1437-1446. DOI: 10.1016/j.renene.2019.03.044
|
---
title: 'Association between impaired glucose metabolism and long-term prognosis at
the time of diagnosis of depression: Impaired glucose metabolism as a promising
biomarker proposed through a machine-learning approach'
authors:
- Dong Yun Lee
- Yong Hyuk Cho
- Myoungsuk Kim
- Chang-Won Jeong
- Jae Myung Cha
- Geun Hui Won
- Jai Sung Noh
- Sang Joon Son
- Rae Woong Park
journal: European Psychiatry
year: 2023
pmcid: PMC9970146
doi: 10.1192/j.eurpsy.2023.10
license: CC BY 4.0
---
# Association between impaired glucose metabolism and long-term prognosis at the time of diagnosis of depression: Impaired glucose metabolism as a promising biomarker proposed through a machine-learning approach
## Abstract
### Background
Predicting the course of depression is necessary for personalized treatment. Impaired glucose metabolism (IGM) was introduced as a promising depression biomarker, but no consensus was made. This study aimed to predict IGM at the time of depression diagnosis and examine the relationship between long-term prognosis and predicted results.
### Methods
Clinical data were extracted from four electronic health records in South Korea. The study population included patients with depression, and the outcome was IGM within 1 year. One database was used to develop the model using three algorithms. External validation was performed using the best algorithm across the three databases. The area under the curve (AUC) was calculated to determine the model’s performance. Kaplan–Meier and Cox survival analyses of the risk of hospitalization for depression as the long-term outcome were performed. A meta-analysis of the long-term outcome was performed across the four databases.
### Results
A prediction model was developed using the data of 3,668 people, with an AUC of 0.781 with least absolute shrinkage and selection operator (LASSO) logistic regression. In the external validation, the AUCs were 0.643, 0.610, and 0.515. Through the predicted results, survival analysis and meta-analysis were performed; the hazard ratios of risk of hospitalization for depression in patients predicted to have IGM was 1.20 ($95\%$ confidence interval [CI] 1.02–1.41, $$p \leq 0.027$$) at a 3-year follow-up.
### Conclusions
We developed prediction models for IGM occurrence within a year. The predicted results were related to the long-term prognosis of depression, presenting as a promising IGM biomarker related to the prognosis of depression.
## Introduction
Depression severely restricts individual psychosocial functions and lowers the quality of life. It leads to national problems such as increased suicide rates and medical expenses because of its chronicity. The World Health Organization cited major depressive disorder as the third cause of the global burden of disease in 2008 and predicted that depression would rank first by 2030 [1]. Numerous factors such as biological markers and poor habits [2] are linked to the onset and recovery of depression. Variable clinical patterns, unpredictable progression and prognosis, and insufficient therapeutic response make depression treatment challenging for clinicians. Remission rates with antidepressants are also overall low (~$27\%$ as per the STAR*D trial) [3], and $20\%$–$25\%$ of patients with depression are at risk of chronic depression [4]. Thus, previous studies have tried to improve treatment outcomes of depression, and evidence has revealed that early intervention of depression is not only associated with better treatment response and long-term outcomes but also with slow disease progression [5–8]. These studies gradually focused on exploring various variables that can predict prognosis in the early stages of depression and achieving personalized treatment through targeted treatment strategies [9–11].
Previous studies have suggested that measuring metabolic markers may be a promising way of predicting long-term clinical outcomes in depression [12]. Recent studies have revealed the relationship among depression, suicidal behavior, insulin resistance (IR), or impaired glucose metabolism (IGM), and evidence of their interactions is accumulating [13–16]. Specifically, IGM is characterized by glucose metabolic disturbance and is defined as prediabetes mellitus (DM) and DM [17]. It can be measured based on hemoglobin A1C (HbA1c) levels in the blood and fasting blood sugar; thus, its clinical utility is high. Many studies have reported that IGM has a bidirectional association with depression. A study reported that reduced serotonin levels were associated with elevated blood glucose levels, Insulin Resistance (IR), and depressed mood [18]. Some previous studies have found that higher glucose levels are associated with dysthymia and higher HbA1c concentrations with recurrent or psychotic depression [19]. In addition, a study in adults with type 2 DM (T2DM) found that certain antidiabetic drugs were associated with a lower risk of depression [20]. Recently, a cross-sectional study correlated IR with depression severity as an endophenotype of depression [13]. Despite these studies, no consensus has yet been reached to the extent that the association between IGM and depression is applicable to clinical practice for establishing patient care strategies.
Machine-learning (ML)-based predictive models are becoming increasingly popular by combining huge data into one model. For depression, conventional regression methods have limitations in prediction; not only well-known demographic factors or factors related to typical treatment but also various comorbidity with physical disease and generally polypharmacy are common [19, 21]. By contrast, ML-based methods have successfully predicted depression persistence, chronicity, severity [22], treatment response, and first and new onset of depressive episodes [17, 18, 23].
This study aimed to investigate whether IGM could be utilized as a biomarker that reflects the clinical severity and prognosis of depression. Initially, we attempted to develop a model that predicts IGM occurrence at the time of the first diagnosis of depression through an ML algorithm. Subsequently, using multicenter and longitudinal data, we intended to analyze and validate whether the IGM occurrence predicted by the model is related to the short-term and long-term prognosis of depression.
## Data source
This study used data from approximately 6 million patients across the four electronic health record databases in South Korea: Ajou University School of Medicine (AUSOM), Daegu Catholic Medical Center (DCMC), Wonkwang University Hospital (WKUH), and Kyung Hee University Hospital at Gangdong (KHNMC) (Supplementary Material S1). The clinical data included diagnoses, observations, provider visits, procedures performed, and medications filled. The databases were formatted according to the Observational Medical Outcomes Partnership–Common Data Model version 5.3.1, maintained by the Observational Health Data Sciences and Informatics (OHDSI), and de-identified [24]. The database of AUSOM was used in model development, and the other three databases were used to validate the developed model. After the development and validation of the model, all databases were used in the survival analysis.
This study was approved by the Institutional Review Board of the Ajou University Hospital (AJOUIRB-MDB-2022-255). Informed consent was not required owing to the use of de-identified data. Access to DCMC, WKUH, and KHNMC databases during the external validation process was allowed under the IRB mutual recognition agreement (research-free zone agreement).
## Study population and outcome
The study population included patients with a new depressive episode. The index date was defined as the patient’s first diagnosis of depressive disorder. To verify their first diagnosis of depressive disorder, at least 1 year of observation before the index date was required. Within the 1-year observation period before the index date, relevant covariates on each patient were collected to predict their future diagnosis of IGM. Patients who were treated for depression, those who had antidepressant prescriptions, and had undergone psychiatric procedures after the index date were included. Also, patients who had at least 1 year of follow-up after the index date were included. For the IGM prediction, patients who had at least one measure of HbA1c or fasting glucose within 1 year after the index date were included. As exclusion criteria, patients with diagnosis of bipolar disorder, schizophrenia, and psychosis on or before the index date were excluded. Regarding DM, a previous history of DM, DM complications, and exposures to antidiabetic drugs were excluded.
The primary outcome for the predictive models was IGM within 1 year after the index date. IGM was defined as pre-DM or T2DM and measured by HbA1c or fasting glucose. For IGM, HbA1c levels were defined as ≥$5.6\%$, and fasting plasma glucose as ≥100 mg/dL [25]. All patients with depression were followed up for 1 year. If IGM occurred within this 1-year period, the observation was stopped on the day that the IGM diagnosis was coded. Thus, the predictive models were developed using the primary outcome. After that, patients were divided into “predicted to have IGM” and “predicted not to have IGM” groups through a predictive model at the time of the index date. Further details of the cohort definitions and code lists are presented in Supplementary Materials S2–S3.
## Model development
We used the patient-level prediction framework of the OHDSI to develop and validate the predictive models. This framework consisted of standardized model development and validation processes that require defining predictable problems and selecting the study population, outcomes, population settings, predictors, and statistical algorithms [26]. The predictive variables for model training were extracted and dichotomized for existence within short-term (30 days) and long-term (365 days) intervals before the index. The variables included patient age, sex, month of the index visit, diagnoses, drug exposures, and procedures. Through this process, 22,904 candidate variables were generated. The models were developed across multiple algorithms, including least absolute shrinkage and selection operator (LASSO)-penalized regression, random forest, and extreme gradient boosting (XGBoost) via threefold cross-validation. The algorithm with the best performance was selected for the final model according to the value of the area under the receiver operating characteristic curve (AUROC).
## External validation
External validation was conducted to confirm the validity of the model’s performance using the databases of DCMC, WKUH, and KHNMC. Specifically, we evaluated the performance of the final model to other databases in the same setting as in the model development.
## Follow-up and long-term outcome measurements
The patients were followed up 3 years after the index date. During the follow-up, risk of hospitalization for depression in patients who were predicted to have IGM compared with patients who were predicted not to have IGM. Risk of hospitalization for depression was defined as hospitalization caused by the exacerbation of depressive episodes. In addition, rehospitalization after discharge for the first diagnosis was considered [27]. To distinguish between existing hospitalization and rehospitalization, only hospitalization after at least a 2-week washout period was defined as an outcome. The outcomes were binarized into hospitalization and non-hospitalization based on the occurrences recorded in the databases.
## Statistical analysis
Descriptive statistical analyses were appropriately performed. Baseline characteristics are presented as counts with proportions for categorical variables and as median with interquartile range for continuous variables. The chi-square test was used to compare categorical variables between populations. Accuracy, AUROC, and area under the precision and recall curve (AUPRC) were calculated to evaluate the performance of the prediction models. We used the maximal Youden index to select the optimal cutoff value in the prediction model [28].
Moreover, we verified whether the group predicted by the final model was related to the actual IGM occurrence. The final model was used to estimate their predicted IGM at the internal validation dataset, and patients with a relatively high probability of IGM were then labeled as predicted to have IGM. If the patients in the internal validation dataset were predicted to have IGM, they were classified as “predicted to have IGM,” and others were classified as “predicted not to have IGM.” The Kaplan–Meier survival analysis and log-rank test were used to analyze the difference in the occurrence of IGM within 1 year after the index date in the group predicted to have IGM versus the group predicted not to have IGM.
After model development and external validation, Kaplan–Meier and Cox survival analyses for the long-term outcomes were performed to assess the risk of hospitalization for depression in patients who have IGM, as determined by the final model. Then, a meta-analysis was performed to calculate the summary hazard ratio (HR) estimates across four databases.
All p-values <0.05 were considered statistically significant. All analyses were conducted using R software version 3.6 (R Foundation for Statistical Computing, Vienna, Austria), OHDSI’s Health Analytics Data to Evidence Suite packages, and open-source statistical R packages.
## Baseline characteristics
A total of 481 outcomes in 3,668 patients from AUSOM were used for model development, and for the external validation, 543 outcomes in a total of 5,716 patients (DCMC, $$n = 2$$,129; WKUH, $$n = 2$$,717; and KHNMC, $$n = 870$$) were used. Table 1 shows the baseline characteristics of the study population in AUSOM. The baseline characteristics of other databases are presented in Supplementary Tables S1–S3. Among the 3,668 patients with depression in the AUSOM database, 481 ($13.1\%$) experienced IGM within 1 year after the diagnosis of depression. No significant differences were found in age, sex, medical history except hypertension, and psychiatric history between the groups. The proportion of hypertension was significantly lower in with IGM group ($p \leq 0.01$). Middle-aged (40–59 years) and female patients were the most predominant in the study population. Hypertension and anxiety disorder were frequent diagnoses (hypertension, $15.2\%$ and $9.1\%$; anxiety disorder, $15.4\%$ and $14.1\%$, respectively).Table 1.Baseline characteristics for study population with or without IGM in AUSOM.VariableWithout IGM($$n = 3$$,187)With IGM($$n = 481$$) p-valueAge group, n (%)0.34[3]0.95 <20217 (6.8)31 (6.4) 20–39629 (19.7)109 (22.7) 40–591,355 (42.5)205 (42.6) ≥60986 (30.9)136 (28.3)Sex, n (%) Male947 (29.7)135 (28.1)0.47[1]0.49Medical history, n (%) Chronic liver disease43 (1.3)2 (0.4)2.28[1]0.13 Renal impairment48 (1.5)4 (0.8)0.92[1]0.33 Hyperlipidemia83 (2.6)5 (1.4)3.72[1]0.05 Obesity63 (1.9)10 (2.1)0.01[1]1.00 Hypertension487 (15.2)44 (9.1)12.21[1]<0.01* Rheumatoid arthritis32 (1.0)2 (0.5)0.99[1]0.31Psychiatric history, n (%) Anxiety disorder492 (15.4)68 (14.1)0.45[1]0.50 Sleep disorder310 (9.7)33 (6.8)3.72[1]0.05 Neurodevelopmental disorder55 (1.7)7 (1.5)0.05[1]0.81 Note:, chi-square value and degree of freedom. Abbreviations: AUSOM, Ajou university school of medicine; IGM, impaired glucose metabolism.*indicates statistical significance ($p \leq 0.05$).
## Prediction models
Figure 1 shows the performance of the ML model in the internal validation set of AUSOM, including LASSO, random forest, and XGBoost. The best-performing model, selected by comparing the average AUROC from the threefold validation, was a logistic regression with LASSO. We defined LASSO as the final model, which showed an AUROC of 0.781 ($95\%$ CI 0.742–0.820) on the internal validation dataset. The accuracy and AUPRC of the final model were 0.667 and 0.338, respectively. The performance metrics are shown in Supplementary Table S4.Figure 1.Receiver operating characteristic (ROC) curve of models predicting impaired glucose metabolism. ( A) ROC curve for the models according to algorithms. ( B) ROC curve for internal and external validations. The performance of the models using the area under the receiver operating characteristic curve is compared.
Table 2 shows the top 10 important predictors. The feature importance analysis showed that a normal range of blood glucose levels before depression diagnosis was the most important predictor across the three algorithms. Drug exposures such as antipsychotics were important predictors in the prediction models. Three models consistently considered the category of the blood test as important predictors. Unlike other models, the LR with LASSO model included the category of image test as a predictor. In Supplementary Table S5, the predictors that increase the IGM prediction risk and those that decrease the risk are indicated in red and blue, respectively. Table 2.Top 10 important predictors of the prediction models for impaired glucose metabolism. RankLR with LASSOXGBoostRandom forest1Normal range of blood glucose level within 1 year before diagnosisNormal range of blood glucose level within 1 year before diagnosisNormal range of blood glucose level within 1 year before diagnosis2Normal range of blood calcium level within 1 year before diagnosisMeasurement of blood aspartate aminotransferase level within 1 year before diagnosisMeasurement of blood alanine aminotransferase level within 1 year before diagnosis3Measurement of urinalysis within 1 month before diagnosisMeasurement of blood triglyceride within 1 year before diagnosisMeasurement of blood cholesterol level within 1 year before diagnosis4Measurement of blood platelet within 1 year before diagnosisNSAIDs prescription within 1 month before diagnosisMeasurement of blood alanine aminotransferase level within 1 year before diagnosis5CT within 1 month before diagnosisMeasurement of urinalysis within 1 year before diagnosisMeasurement of blood aspartate aminotransferase level within 1 year before diagnosis6Measurement of C reactive protein within 1 month before diagnosisAntipsychotics prescription within 1 year before diagnosisNormal range of blood bilirubin level within 1 year before diagnosis7Drugs for peptic ulcer prescription at diagnosisSSRI prescription at diagnosisMeasurement of uric acid level within 1 year before diagnosis8Measurement of blood cholesterol level within 1 year before diagnosisHMG CoA reductase inhibitors prescription within 1 month before diagnosisMeasurement of creatine level within 1 year before diagnosis9Antipsychotics prescription within 1 year before diagnosisMeasurement of blood hepatitis B virus test within 1 month before diagnosisMeasurement of blood albumin level within 1 year before diagnosis10Normal range of blood aspartate aminotransferase level within 1 year before diagnosisHaloperidol prescription at diagnosisMeasurement of blood protein level within 1 year before diagnosis Note: The color in the table means the category of features (orange: laboratory test, green: image test, and blue: drug exposure).Abbreviations: CT, computed tomography; LASSO, least absolute shrinkage and selection operator; LR, logistic regression; NSAID, non-steroidal anti-inflammatory drugs; SSRI, selective serotonin reuptake inhibitor; XGBoost, extreme gradient boosting.
## External model validation
The final model was externally validated using the DCMC, WKUH, and KHNMC databases. In the external validation databases, patients experienced IGM at a rate of $8.8\%$ ($\frac{188}{2}$,129) in DCMC, $12.0\%$ ($\frac{327}{2}$,717) in WKUH, and $3.2\%$ ($\frac{28}{870}$) in KHNMC. The external validation performance of the final model regarding AUROC was 0.643 at DCMC, 0.610 at WKUH, and 0.515 at KHNMC.
## Long-term outcomes of ML-predicted IGM
Figure 2 shows the clinical benefit of using the IGM prediction models. In the internal validation dataset of AUSOM, the group predicted to have IGM had a significantly higher occurrence of IGM within 1 year after the index date than the group predicted not to have IGM (log rank, $p \leq 0.001$). Furthermore, patients predicted to have IGM showed significantly worse long-term outcomes. In the overall cohort of AUSOM, survival analysis showed that the risk of hospitalization for depression occurred more frequently in patients who were predicted to have IGM during the 3-year follow-up (log rank, $$p \leq 0.002$$) (Figure 2).Figure 2.Kaplan–Meier curves in the stratified survival analysis. ( A) Impaired glucose metabolism in the internal validation dataset of AUSOM. ( B) Long-term outcome for the 3-year follow-up in the overall cohort of AUSOM.
We further assessed long-term outcomes not only in AUSOM but also in external validation databases. The meta-analytic comparative effect estimates for the risk of hospitalization for depression are presented in Figure 3. The summary HR of risk of hospitalization for depression during the 3-year follow-up was 1.20 ($95\%$ CI 1.02–1.41, $$p \leq 0.027$$) for patients predicted to have IGM.Figure 3.Risk of long-term outcome in 3 years in patients predicted by the machine-learning model to have IGM within 1 year.
## Discussion
We constructed a model to predict the occurrence of IGM within 1 year at the time of depression diagnosis using ML algorithms. By analyzing the longitudinal data of multiple institutions using this prediction model, we identified relationships between IGM prediction and the long-term prognosis of depression. Thus, IGM might be a promising biomarker associated with the prognosis of depression.
Despite being a common psychiatric disease, depression has a low treatment success rate because of the heterogeneity and difficulty in predicting its course [3, 29]. Thus, clinicians desire to identify biomarkers that can reflect the severity or chronicity of depression. Several previous studies have shown a complex relationship between depression and IGM. Knol et al. [ 30] reported in a meta-analysis a $37\%$ increased risk of T2DM development in adults with depression compared with individuals without depression. Several possibilities have been suggested, and there are reports that hypothalamic–pituitary–adrenal axis abnormalities in patients with depression, hypercortisolemia, and immune system abnormalities, including chronic low-grade inflammations, influence the insulin effect [5, 6, 18]. Conversely, IGM including DM is related to the development or exacerbation of depression and the reactivity of antidepressants [31, 32]. The dysfunction of insulin receptors and subsequent signal cascade, which are related to IGM, has a direct effect on neural metabolism and the brain and is associated with depression by causing abnormalities in neurotransmitter metabolisms such as dopamine, serotonin, and norepinephrine [33, 34]. Moreover, some studies have revealed that the successful treatment of depression can correct insulin response, particularly with more serotonergic agents, such as selective serotonin reuptake inhibitors (SSRIs) [35, 36]. However, a recent study reported that low doses of metformin, DPP4 inhibitors, GLP1 analogs, and especially SGLT2 inhibitors were associated with lower odds of depression than non-users of these medications [20]. In summary, bidirectional pathophysiological connections exist between depression and IGM. This connection means that depression and IGM are important factors not only in each other’s pathogenesis but also in each other’s successful treatment and prognosis.
Recently, a nationwide study revealed that glucose disturbance is associated with increased suicidal ideation and suicidal behavior in patients with depression [37]. In another large-scale study, IR was proposed as a promising marker that reflects severity and chronicity in patients with depression [13]. These large-scale cross-sectional studies opened with a prelude to the relationship between IR and depression. Consequently, clinicians are paying attention to predicting IGM including IR in the early stages of diagnosis, and various treatment strategies can be implemented considering the long-term prognosis and treatment reactivity of patients with depression. However, depression and IGM have a complex relationship, and predicting IGM in the early stages of depression is not easy; thus, analysis using large-scale variables is needed because conventional analysis has limitations.
Data-driven ML algorithms are in the spotlight as a breakthrough in the discovery of hidden predictors and known clinically meaningful predictors selected by researchers [10]. Therefore, in this study, an IGM prediction model was developed using a data-driven ML algorithm. Specifically, the data used in this study consisted of a large number of tabular data, which was advantageous for the use of ML algorithms such as XGBoost, LR with LASSO, and random forest, similar to previous studies [38]. The model using LR with LASSO showed the highest performance in this study (Figure 1).
Since this study developed an IGM prediction model through an ML algorithm rather than deep learning, understandable explanations for prediction were obtained. Initially, at the time of diagnosis of depression, antipsychotics, including haloperidol, are commonly prescribed. Moreover, studies have reported that antipsychotics are related to an increase in blood sugar [39]. In addition, several studies have reported that benzodiazepine [40], corticosteroid [41], and peptic ulcer prescription, which are expected to be proton pump inhibitors [42], are associated with increased blood sugar and DM. CT, C reactive protein measurement, hepatitis B virus test, uric acid measurement, and nonsteroidal anti-inflammatory drug prescriptions are also observed as important predictors. They may have been tested for certain symptoms or prescribed drugs as an extension of the immune system’s dysfunction observed in depression and IGM. Ken et al. revealed that chronic low-grade inflammatory reactions in depression lead to apoptosis in pancreatic beta cells, which is related to IGM [43]. The increase in cytokine levels in patients with depression is linked to metabolic disturbance [34]. We suggest the immunological vulnerability in patients at the time of diagnosis of depression was reflected as a predictor. On the contrary, the most important predictor is “normal range of blood glucose level within 1 year before diagnosis” in all algorithms. Thus, if the blood glucose level within 1 year at the time of diagnosis of depression was normal, this time is not enough to observe the progression to IGM within 1 year. Measuring the levels of cholesterol, triglyceride, urinalysis, creatinine, etc., through blood tests also had a negative relationship with predicting IGM. This can be interpreted in the same way as the results of previous studies that continuity of care had some benefits including prevention of chronic diseases including DM [44]. Finally, unlike other drugs, SSRIs showed a negative relationship with predicting IGM. Although this is controversial, SSRIs had a positive effect on blood sugar control among antidepressants [45, 46].
Furthermore, in this study, survival analysis was conducted to determine whether the results of the IGM incidence prediction model were related to the 3-year prognoses of depression. Through a meta-analysis using four other hospital data with the Common Data Model (CDM) database, we identified differences in hospitalization caused by exacerbation of depressive episodes between the two groups divided into IGM prediction models using longitudinal data. This result can be interpreted through the report of previous studies that when depression or anxiety is accompanied by DM, disease burden and emotional distress increase because of poor metabolic control, low rates of blood glucose self-monitoring, and DM complications, which can predict inadequate response to depression treatment [36, 47].
This study has several limitations. First, this study used data from Koreans only; thus, the results cannot be generalized. However, this study showed that the CDM developed through a distributed research network enables a more efficient meta-analysis than in the past without exposing private information. This suggests that a global meta-analysis is possible if the same CDM is established in various countries. Second, this study used longitudinal data, but it has the limitations of a retrospective study. To clarify the relationship between depression and IGM, a prospective study is required. Third, this study did not include social and environmental factors that would be related to depression and IGM in the model development. This is also a limitation of the psychiatric CDM. Thus, developing measurable environmental and sociological variables is necessary. Fourth, model performance was reduced in the external validation of the IGM prediction model. Model performance commonly decreases in the external validation because of the varying characteristics of the enrolled participants, and it is difficult to control them all. Specifically, the external validation performance was low in the analysis of KHNMC compared with AUSOM. The result was assumed to be caused by the varying rates of IGM occurrence, i.e., $13.1\%$ in AUSOM and $3.2\%$ in KHNMC. Furthermore, since there is no overall difference between patients with and without IGM in the baseline characteristics, predicting IGM may be difficult. Fifth, indirect indicators such as depression-related hospitalization were used to determine the relationship between the results of the IGM prediction model and the long-term prognosis of depression. However, several recent studies have derived meaningful results using operational definitions such as this study [27]. Sixth, we included only individuals who were assessed for IGM for the study population. Among patients with depression in the study database, those with IGM assessment had a higher rate of comorbidities than those without IGM assessment. This suggests that the generalization of the results should be cautious.
In summary, we developed an IGM prediction model at the time of depression diagnosis using an ML algorithm and found a relationship between the results of the IGM prediction model and the long-term prognosis of depression using longitudinal data. Thus, we suggest that IGM is likely to be a promising biomarker in predicting the prognosis of depression. Treatment strategies should be established to improve metabolic disturbance, including IGM, and the use of IGM as an evaluation index for lifestyle modification and increased treatment success rate may be expected. Therefore, a more customized and multidimensional approach to the evaluation and treatment of depression would be possible.
## Data Availability Statement
CDM data are designed to support a distributed research network. Thus, access to the data is restricted on internal private networks. Therefore, data are not publicly available.
## Author Contributions
Conceptualization: D.Y.L., Y.H.C.; Data curation: D.Y.L., M.K.; Formal analysis: D.Y.L., M.K.; Funding acquisition: D.Y.L., R.W.P., S.J.S.; Methodology: D.Y.L,; Project administration: D.Y.L.; Resources: C.-W.J., J.M.C., G.H.W., J.S.N., S.J.S., R.W.P.; Writing—original draft: D.Y.L., Y.H.C.; Writing—review and editing: S.J.S., R.W.P.
## Financial Support
This research was funded by the Bio Industrial Strategic Technology Development Program [20003883, 20005021] funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea) and a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health &Welfare, Republic of Korea (Grant Number: HR16C0001, HR22C173401). And Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education [NRF-2020R1I1A1A01072208, 2019R1A5A2026045].
## Conflicts of Interest
The authors declare none.
## Ethical Standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
## References
1. [1]World Health Organization. The global burden of disease: 2004 update
Geneva: World Health Organization; 2008.. *The global burden of disease: 2004 update* (2008)
2. Lee JH, Park SK, Ryoo JH, Oh CM, Mansur RB, Alfonsi JE, Cha DS, Lee Y, McIntyre RS, Jung JY. **The association between insulin resistance and depression in the Korean general population**. *J Affect Disord* (2017) **208** 553. PMID: 27810270
3. Pigott HE. **The STAR* D trial: it is time to reexamine the clinical beliefs that guide the treatment of major depression**. *Can J Psychiatry* (2015) **60** 9-13. PMID: 25886544
4. Penninx BW, Nolen WA, Lamers F, Zitman FG, Smit JH, Spinhoven P, Cuijpers P, de Jong PJ, van Marwijk HW, van der Meer K, Verhaak P. **Two-year course of depressive and anxiety disorders: results from the Netherlands study of depression and anxiety (NESDA)**. *J Affect Disord* (2011) **133** 76-85. PMID: 21496929
5. Ghio L, Gotelli S, Marcenaro M, Amore M, Natta W. **Duration of untreated illness and outcomes in unipolar depression: a systematic review and meta-analysis**. *J Affect Disord* (2014) **152** 45-51. PMID: 24183486
6. Habert J, Katzman MA, Oluboka OJ, McIntyre RS, McIntosh D, MacQueen GM, Khullar A, Milev RV, Kjernisted KD, Chokka PR. **Functional recovery in major depressive disorder: focus on early optimized treatment**. *Prim Care Companion CNS Disord* (2016) **18** 24746
7. Kautzky A, Dold M, Bartova L, Spies M, Kranz GS, Souery D, Montgomery S, Mendlewicz J, Zohar J, Fabbri C, Serretti A. **Clinical factors predicting treatment resistant depression: affirmative results from the European multicenter study**. *Acta Psychiatr Scand* (2019) **139** 78-88. PMID: 30291625
8. Kupfer DJ, Frank E, Perel JM. **The advantage of early treatment intervention in recurrent depression**. *Arch Gen Psychiatry.* (1989) **46** 771. PMID: 2774846
9. Liu X, Li P, Ma X, Zhang J, Sun X, Luo X, Zhang Y. **Association between plasma levels of BDNF and GDNF and the diagnosis, treatment response in first-episode MDD**. *J Affect Disord* (2022) **315** 190. PMID: 35908604
10. Dinga R, Marquand AF, Veltman DJ, Beekman AT, Schoevers RA, van Hemert AM, Penninx BW, Schmaal L. **Predicting the naturalistic course of depression from a wide range of clinical, psychological, and biological data: a machine learning approach**. *Transl Psychiatry* (2018) **8** 1-11. PMID: 29317594
11. Ju Y, Horien C, Chen W, Guo W, Lu X, Sun J, Dong Q, Liu B, Liu J, Yan D, Wang M. **Connectome-based models can predict early symptom improvement in major depressive disorder**. *J Affect Disord* (2020) **273** 442. PMID: 32560939
12. Vogelzangs N, Beekman AT, Boelhouwer IG, Bandinelli S, Milaneschi Y, Ferrucci L, Penninx BW. **Metabolic depression: a chronic depressive subtype? Findings from the InCHIANTI study of older persons**. *J Clin Psychiatry* (2011) **72** 11748
13. Watson KT, Simard JF, Henderson VW, Nutkiewicz L, Lamers F, Rasgon N, Penninx B. **Association of insulin resistance with depression severity and remission status: defining a metabolic endophenotype of depression**. *JAMA Psychiat* (2021) **78** 439
14. Ceretta LB, Réus GZ, Abelaira HM, Jornada LK, Schwalm MT, Hoepers NJ, Tomazzi CD, Gulbis KG, Ceretta RA, Quevedo J. **Increased prevalence of mood disorders and suicidal ideation in type 2 diabetic patients**. *Acta Diabetol* (2012) **49** 227
15. Han SJ, Kim HJ, Choi YJ, Lee KW, Kim DJ. **Increased risk of suicidal ideation in Korean adults with both diabetes and depression**. *Diabetes Res Clin Pract* (2013) **101** e14. PMID: 23871574
16. Bot M, Pouwer F, De Jonge P, Tack CJ, Geelhoed‐Duijvestijn PH, Snoek FJ. **Differential associations between depressive symptoms and glycaemic control in outpatients with diabetes**. *Diabet Med* (2013) **30** e115. PMID: 23181742
17. King M, Walker C, Levy G, Bottomley C, Royston P, Weich S, Bellon-Saameno JA, Moreno B, Švab I, Rotar D, Rifel J. **Development and validation of an international risk prediction algorithm for episodes of major depression in general practice attendees: the PredictD study**. *Arch Gen Psychiatry* (2008) **65** 1368. PMID: 19047523
18. Wang JL, Patten S, Sareen J, Bolton J, Schmitz N, MacQueen G.. **Development and validation of a prediction algorithm for use by health professionals in prediction of recurrence of major depression**. *Depress Anxiety* (2014) **31** 451. PMID: 24877248
19. Gold SM, Köhler-Forsberg O, Moss-Morris R, Mehnert A, Miranda JJ, Bullinger M, Steptoe A, Whooley MA, Otte C. **Comorbid depression in medical diseases**. *Nat Rev Dis Primers* (2020) **6** 1-22. PMID: 31907359
20. Wium-Andersen IK, Osler M, Jørgensen MB, Rungby J, Wium-Andersen MK. **Diabetes, antidiabetic medications and risk of depression–a population-based cohort and nested case-control study**. *Psychoneuroendocrinology* (2022) **140** 105715. PMID: 35338947
21. Correll CU, Detraux J, De Lepeleire J, De Hert M. **Effects of antipsychotics, antidepressants and mood stabilizers on risk for physical diseases in people with schizophrenia, depression and bipolar disorder**. *World Psychiatry* (2015) **14** 119. PMID: 26043321
22. Kessler RC, van Loo HM, Wardenaar KJ, Bossarte RM, Brenner LA, Cai T, Ebert DD, Hwang I, Li J, de Jonge P, Nierenberg AA. **Testing a machine-learning algorithm to predict the persistence and severity of major depressive disorder from baseline self-reports**. *Mol Psychiatry* (2016) **21** 1366. PMID: 26728563
23. Chekroud AM, Zotti RJ, Shehzad Z, Gueorguieva R, Johnson MK, Trivedi MH, Cannon TD, Krystal JH, Corlett PR. **Cross-trial prediction of treatment outcome in depression: a machine learning approach**. *Lancet Psychiatry* (2016) **3** 243. PMID: 26803397
24. Makadia R, Ryan PB. **Transforming the premier perspective hospital database into the observational medical outcomes partnership (OMOP) common data model**. *EGEMS* (2014) **2** 1110. PMID: 25848597
25. Mansur RB, Rizzo LB, Santos CM, Asevedo E, Cunha GR, Noto MN, Pedrini M, Zeni M, Cordeiro Q, McIntyre RS, Brietzke E. **Impaired glucose metabolism moderates the course of illness in bipolar disorder**. *J Affect Disord* (2016) **195** 57-62. PMID: 26866976
26. Reps JM, Schuemie MJ, Suchard MA, Ryan PB, Rijnbeek PR. **Design and implementation of a standardized framework to generate and evaluate patient-level prediction models using observational healthcare data**. *J Am Med Inform Assoc* (2018) **25** 969. PMID: 29718407
27. Moncrieff J, Crellin NE, Long MA, Cooper RE, Stockmann T. **Definitions of relapse in trials comparing antipsychotic maintenance with discontinuation or reduction for schizophrenia spectrum disorders: a systematic review**. *Schizophr Res* (2020) **225** 47-54. PMID: 31604607
28. Fluss R, Faraggi D, Reiser B. **Estimation of the Youden index and its associated cutoff point**. *Biom J: J Math Meth Biosci* (2005) **47** 458
29. Kennedy SH, Giacobbe P. **Treatment resistant depression—advances in somatic therapies**. *Ann Clin Psychiatry* (2007) **19** 279. PMID: 18058285
30. Knol MJ, Twisk JW, Beekman AT, Heine RJ, Snoek FJ, Pouwer F. **Depression as a risk factor for the onset of type 2 diabetes mellitus. A meta-analysis**. *Diabetologia* (2006) **49** 837. PMID: 16520921
31. Gonzalez JS, Peyrot M, McCarl LA, Collins EM, Serpa L, Mimiaga MJ, Safren SA. **Depression and diabetes treatment nonadherence: a meta-analysis**. *Diabetes Care* (2008) **31** 2398. PMID: 19033420
32. Bogner HR, Morales KH, de Vries HF, Cappola AR. **Integrated management of type 2 diabetes mellitus and depression treatment to improve medication adherence: a randomized controlled trial**. *Ann Fam Med* (2012) **10** 15-22. PMID: 22230826
33. Boucher J, Kleinridders A, Kahn CR. **Insulin receptor signaling in normal and insulin-resistant states**. *Cold Spring Harb Perspect Biol.* (2014) **6** a009191. PMID: 24384568
34. Leonard BE, Wegener G. **Inflammation, insulin resistance and neuroprogression in depression**. *Acta Neuropsychiatrica* (2020) **32** 1-9
35. Silva N, Atlantis E, Ismail K. **A review of the association between depression and insulin resistance: pitfalls of secondary analyses or a promising new approach to prevention of type 2 diabetes?**. *Curr Psychiatry Rep* (2012) **14** 8-14. PMID: 22094982
36. Holt RI, De Groot M, Golden SH. **Diabetes and depression**. *Curr Diab Rep* (2014) **14** 1-9
37. Chen SW, Li X, Lang X, Li J, Zhang XY. **Metabolic parameters and thyroid hormones in relation to suicide attempts in patients with first-episode and drug-naive major depressive disorder with comorbid glucose disturbances: a large cross-sectional study**. *Eur Arch Psychiatry Clin Neurosci.* (2022) **s00406** 1-9
38. Sharma A, Verbeke WJ. **Improving diagnosis of depression with XGBOOST machine learning model and a large biomarkers Dutch dataset (n = 11,081)**. *Front Big Data.* (2020) **3** 15. PMID: 33693389
39. Holt RI. **Association between antipsychotic medication use and diabetes**. *Curr Diab Rep* (2019) **19** 1-10. PMID: 30637535
40. Zumoff B, Hellman L. **Aggravation of diabetic hyperglycemia by chlordiazepoxide**. *JAMA* (1977) **237** 1960. PMID: 576708
41. Hwang JL, Weiss RE. **Steroid‐induced diabetes: a clinical and molecular approach to understanding and treatment**. *Diabetes Metab Res Rev* (2014) **30** 96-102. PMID: 24123849
42. Czarniak P, Ahmadizar F, Hughes J, Parsons R, Kavousi M, Ikram M, Stricker BH. **Proton pump inhibitors are associated with incident type 2 diabetes mellitus in a prospective population‐based cohort study**. *Br J Clin Pharmacol* (2022) **88** 2718. PMID: 34907592
43. Kan C, Silva N, Golden SH, Rajala U, Timonen M, Stahl D, Ismail K. **A systematic review and meta-analysis of the association between depression and insulin resistance**. *Diabetes Care* (2013) **36** 480. PMID: 23349152
44. Koopman RJ, Mainous AG, Baker R, Gill JM, Gilbert GE. **Continuity of care and recognition of diabetes, hypertension, and hypercholesterolemia**. *Arch Intern Med* (2003) **163** 1357. PMID: 12796073
45. Tharmaraja T, Stahl D, Hopkins CW, Persaud SJ, Jones PM, Ismail K, Moulton CD. **The association between selective serotonin reuptake inhibitors and glycemia: a systematic review and meta-analysis of randomized controlled trials**. *Psychosom Med* (2019) **81** 570. PMID: 31136376
46. Deuschle M. **Effects of antidepressants on glucose metabolism and diabetes mellitus type 2 in adults**. *Curr Opin Psychiatry* (2013) **26** 60. PMID: 23187087
47. Ducat L, Philipson LH, Anderson BJ. **The mental health comorbidities of diabetes**. *JAMA* (2014) **312** 691. PMID: 25010529
|
---
title: Somatic comorbidities of mental disorders in pregnancy
authors:
- Vahe Khachadourian
- Arad Kodesh
- Stephen Z. Levine
- Emma Lin
- Joseph D. Buxbaum
- Veerle Bergink
- Sven Sandin
- Abraham Reichenberg
- Magdalena Janecka
journal: European Psychiatry
year: 2023
pmcid: PMC9970155
doi: 10.1192/j.eurpsy.2023.1
license: CC BY 4.0
---
# Somatic comorbidities of mental disorders in pregnancy
## Abstract
### Background
Mental and physical health conditions are frequently comorbid. Despite the widespread physiological and behavioral changes during pregnancy, the pattern of comorbidities among women in pregnancy is not well studied. This study aimed to systematically examine the associations between mental and somatic disorders before and during pregnancy.
### Method
The study used data from mothers of a nationally representative birth cohort of children born in Israel (1997–2008). We compared the risk of all major somatic disorders (International Classification of Diseases, Ninth Revision) in pregnant women with and without a mental disorder. All analyses were adjusted for maternal age, child’s birth year, family socioeconomic status, and the total number of maternal encounters with health services around pregnancy period.
### Results
The analytical sample included 77,030 mother–child dyads, with 30,083 unique mothers. The mean age at child’s birth was 29.8 years. Prevalence of diagnosis of mental disorder around pregnancy in our sample was $4.4\%$. Comorbidity between mental and somatic disorders was two times higher than the comorbidity between pairs of different somatic disorders. Of the 17 somatic disorder categories, seven were positively associated with mental health disorders. The highly prevalent comorbidities associated with mental disorders in pregnancy included e.g. musculoskeletal (OR = 1.30; $95\%$ CI = 1.20–1.42) and digestive system diseases (OR = 1.23; $95\%$ CI = 1.13–1.34).
### Conclusions
We observed that associations between maternal diagnoses and mental health stand out from the general pattern of comorbidity between nonmental health diseases. The study results confirm the need for screening for mental disorders during pregnancy and for potential comorbid conditions associated with mental disorders.
## Introduction
Multimorbidity, defined as co-occurrence of two or more diseases in the same person, is a major health problem affecting a substantial portion of the population [1]. Over the past decades, the prevalence of multimorbidity has been on the rise [2, 3]. Multimorbid health conditions negatively affect the quality of life of the patient, are costly to treat [4] and harder to manage than individual conditions [5, 6]. Moreover, the coexistence of multiple diseases may have health effects that are greater than the sum of the effects of individual diseases, which is especially problematic during pregnancy due to potential long-term adverse effects on both mother and the child [7–10].
Although the comorbidity between mental and somatic health conditions has received increasing attention over the past years [11, 12], still little is known about the full spectrum of mental and somatic comorbidities during pregnancy. Most studies have focused on specific pairs of health conditions, providing valuable insights about such comorbidity (e.g., depression and diabetes) [13, 14]. However, the comorbidity between a wider range of mental and somatic disorders in pregnancy has never been investigated systematically.
The knowledge about comorbidity between mental and somatic disorders from non-pregnant populations may not necessarily translate to pregnancy period. Pregnancy is a unique state with profound physiological and behavioral changes [15]. Women with pre-pregnancy chronic medical illness require special healthcare, because medication regimes and the natural course of mental and somatic disorders may change during this time. For example, pregnancy is a critical period of the onset of cardiovascular conditions, endocrine disorders, and blood diseases (e.g., hypertensive disorders of pregnancy, diabetes gravidarum, and anemia) [16–18]. In parallel, mood and anxiety disorders are highly prevalent in women in their reproductive ages, including during the perinatal period [19–21].
Both mental and somatic conditions during pregnancy have been associated with adverse outcomes in offspring [22, 23] (e.g., risk of infections, asthma, obesity, cognitive performance, and psychiatric disorders). There is an increasing awareness that many disorders may have at least partly fetal origins [24]. Consequently, an improved understanding of the comorbidity between mental and somatic disorders around the pregnancy period may not only offer novel directions into the diagnosis and management of maternal health disorders, but may also shed new light on the determinants of health outcomes for the child.
Using a nationally representative birth-cohort study, we investigate the spectrum of associations between maternal mental and all categories of somatic disorders in the period just before and during pregnancy—without assuming any causal relationships between these comorbidities. The overarching aims of the current study were to investigate the burden of somatic comorbidities in pregnant women with mental disorders, and examine the associations between mental and somatic disorders in pregnancy. Finally, we wanted to contextualize our findings by comparing the pattern of mental–somatic comorbidities to those that occur between somatic disorders.
## Study design and population
We conducted a cohort study using a population-based sample from a large health maintenance organization (HMO) in Israel (Meuhedet), which has been described previously [25, 26]. Briefly, per legislation in Israel, citizens are required to obtain medical insurance from one of the existing HMOs that cannot prohibit a citizen joining on the grounds of socioeconomic status (SES), ethnicity, geographic location, health conditions, and health needs. The equivalent health plans and fee structure across HMOs, along with the regulations prohibiting HMOs from refusing a citizen membership, minimize the risk of ascertainment bias in our sample.
The cohort included all children born between January 1, 1997 and December 31, 2008, including (a) randomly selected $19.5\%$ of all the births within the Meuhedet HMO during that period, and (b) all siblings of children selected in the first stage of the sampling who were also born 1997 to 2008.
All selected children were linked to their family records, creating mother–child dyads. Since our focus was on maternal rather than child’s health, pregnancies leading to multiple live births were represented by a single mother–child dyad in the analyses.
To assure ascertainment of maternal diagnosis during the 12 months preceding their pregnancies, the analytical sample was restricted to pregnancies leading to a live birth between January 1, 1999 and December 31, 2008. All dyads where maternal age was younger than 13, or older than 55, were excluded from the sample due to potential administrative errors in these records. All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The study protocol was reviewed and approved by the Helsinki Ethics Committee of the Meuhedet and the Institutional Review Board of the University of Haifa. Since the data did not include any individual identifiers, a waiver of informed consent was granted by the reviewing bodies.
## Somatic disorders
The hierarchical organization of diagnostic codes in the International Classification of Diseases, Ninth Revision (ICD-9) has four levels, presenting information from least to most specific diagnosis (Figure 1). For the main analyses, somatic disorders were classified at level 1 ICD-9 codes using data from the Meuhedet Diagnostic Classification Register (e.g., ICD-9: 410.0, acute myocardial infarction of anterolateral wall, was classified with other diseases of the circulatory system; Figure 1). In additional analyses, we further defined more specific categories of somatic disorders, hereafter referred to as specific somatic disorders, according to diagnostic codes at level 3 (e.g., ICD-9: 410.0, acute myocardial infarction of anterolateral wall, was classified with other codes under acute myocardial infarction), as they offer an additional level of detail about the underlying health condition. The time window for ascertaining all diagnoses included a total of 636 days preceding the child’s birth, that is, the entire estimated pregnancy period (270 days), as well as the year preceding the conception, to maximize inclusion of chronic diagnoses in the analyses. Few diagnoses that were not coded according to ICD-9, and the available information did not allow for identifying and assigning them to any equivalent ICD-9, were omitted. Figure 1.Example of the hierarchical organization of the ICD-9 taxonomy. ICD-9 categories are organized from the most general (level 1, top row), through most specific diagnostic codes (level 4, bottom row). Level 3 diagnoses were used in the current study.
## Mental disorder
Using maternal diagnostic codes (per ICD-9) ascertained from the Meuhedet Diagnostic Classification Register, we classified general mental disorder (level 1), hereafter referred to as mental disorder, as the presence of any psychiatric diagnosis (ICD-9: 290–319, including all the subcodes) during the pregnancy period and the preceding year (i.e., the same time window as for the somatic disorders). While for the primary analyses, we pooled all diagnoses and only considered the maternal binary status (yes/no) of having any mental disorder during that time window, in additional analyses, we further defined more specific mental disorders, hereafter referred to as specific mental disorders, using the information from level 3 ICD-9 diagnostic codes, with a prevalence of at least $0.1\%$ in the sample.
## Covariates
The covariates included maternal age at child’s birth, residential SES, and the total number of encounters with health services during the pregnancy period and the preceding year. Residential SES was a summary index based on household census data and was a function of the number of electrical appliances per person and per capita income in the area [27]. Information about SES was ascertained from Central Bureau of Statistics Registry, while the Meuhedet records served as the source for all other covariates.
To account for the varying risk of multiple diagnoses across the lifetime, the analysis adjusted for maternal age at child’s birth. The analyses also adjusted for SES to account for the differential likelihood of ascertaining maternal mental health [28] and other diagnoses [29] across the SES strata. Similarly, the categorized total number of encounters with health services (0–1, 2–5, 6–10, 11–15, 16–20, 21–30, 30+) during the 21-month period before child’s delivery was included in the adjusted analyses as a proxy measure for health-seeking behaviors, and healthcare utilization.
## Primary analysis
The associations between maternal mental and somatic disorders (all at ICD-9 level 1) were assessed using binary logistic regression models, where each somatic disorder served as an outcome in a separate, covariate-adjusted model. To account for potential clustering effects due to possible multiple deliveries among mothers in the study period, we used clustered sandwich estimator, implemented in the clusterSEs package (v2.6.2) [30].
Since the sampling strategy yielded a higher probability for the inclusion of mothers with multiple deliveries, we applied inverse probability selection weighting to account for these differential selection probabilities. The weights were computed based on the number of children born to each mother during the study period. Mothers with more offspring received lower weights in the regression model to account for their higher probability of inclusion in the sample (Supplementary Table S1).
To account for the possible correlation between somatic disorders—potentially driving some of the mental–somatic disorders associations, the initial analysis was followed by multivariable logistic regression models, which in addition to mental disorder and the covariates included all other level 1 somatic disorders other than the one serving as the outcome in that specific model.
## Secondary analyses
We ran a series of secondary analyses to further inform the interpretation of the results from the primary analyses.
## General comorbidity
First, we examined the association between pairs of somatic disorders, allowing us to put the burden of comorbidity between mental and somatic disorders (primary analysis) in the broader context of comorbidity. To this end, we evaluated the associations between all possible pairs of somatic disorders, adjusting for covariates.
Comparative analyses exploring the relationship between all the possible pairs of somatic disorders (level 1) yielded several statistically significant associations (Table 2). On average, each somatic disorder category was positively associated with 3.5 other disorders (somatic and mental disorders), whereas mental disorders were positively associated with 7.0 somatic disorders. Higher comorbidity rate for mental disorders was observed in comparison with both rare and prevalent somatic disorders, indicating that our results were not due to differential statistical power to detect associations. Some of the observed associations between somatic diagnoses were of small magnitude, which despite being statistically significant (due to a high power) could lack clinical relevance. Table 2.Adjusted odds ratio of the associations between mental and somatic disorders. Exposure (ICD-9 level 1 diagnostic category)MHC17C16C15C14C13C12C11C10C9C8C7C6C5C4C3C2C1Outcome ICD-9 level 1 diagnostic category)C11.050.891.07**1.09***0.951.031.011.41***0.93***1.001.12***1.27***1.09**1.10***0.94*0.971.18***N/A C21.060.991.030.940.911.64***0.941.75***0.82***1.080.970.85***1.27***1.14***0.981.12*N/A C31.100.81*0.941.15***0.901.060.980.94*1.09**0.91**1.010.84***1.10**0.981.36***N/A C41.15*0.980.92*1.21***0.960.931.001.09**1.31***1.061.17***0.951.15***0.95N/A C51.22***0.901.12***1.09***0.86*1.041.12***1.10***0.90***0.90***1.15***1.22***1.01N/A C61.13*0.990.991.07*1.031.171.12***1.10***1.26***0.91**1.28***1.03N/A C70.910.871.10***1.29***0.72***1.161.12***0.990.82***0.86***1.20***N/A C81.23***0.921.07*1.57***0.951.131.22***1.08**0.971.04N/A C90.980.86*0.89***1.16***0.960.991.020.971.12***N/A C100.921.020.94*1.041.42***1.120.91***0.90***N/A C111.030.991.27***0.960.881.30***1.08**N/A C121.30***1.96***1.44***1.36***0.871.40***N/A C131.140.841.211.000.99N/A C140.980.541.010.99N/A C151.59***1.041.09**N/A C161.084.12***N/A C171.41**N/A MHN/A Note: Odds ratios are adjusted for SES, maternal age at delivery, total number of encounters with health services during the 21 months period before delivery, year of delivery, and all the somatic and mental disorders presented in this table. Statistically significant positive associations are highlighted in blue and negative associations are highlighted in yellow. Abbreviations: C1, infectious and parasitic diseases; C2, neoplasms; C3, endocrine, nutritional and metabolic diseases, and immunity disorders; C4, diseases of blood and blood-forming organs; C5, diseases of the nervous system and sense organs; C6, diseases of the circulatory system; C7, diseases of the respiratory system; C8, diseases of the digestive system; C9, diseases of the genitourinary system; C10, complications of pregnancy, childbirth, and the puerperium; C11, diseases of the skin and subcutaneous tissue; C12, diseases of the musculoskeletal system and connective tissue; C13, congenital anomalies; C14, certain conditions originating in the perinatal period; C15, symptoms, signs, and ill-defined conditions; C16, injury and poisoning; C17, supplementary classification of external causes of injury and poisoning; MH, mental disorders.*q-value <0.05.**q-value <0.01.***q-value <0.001.
Sensitivity analysis using multiple imputation of missing variables yielded results that were near-identical to ones observed in complete case analyses, suggesting our findings were robust to the missingness pattern present in our dataset (Supplementary Table S3).
## Comorbidity patterns across specific mental disorders
Next, we examined whether specific mental disorders (level 3 ICD-9; e.g., “anxiety, dissociative and somatoform disorders” and “personality disorders”) were associated with different patterns of comorbidity. To this end, we repeated the analysis as specified for the primary analyses, using as an exposure each specific mental disorder with a prevalence of >$0.1\%$ in our analytical sample. These models included the same outcomes and covariates as described for the primary analysis.
## Comorbidity of specific somatic disorders with mental disorder
Finally, we investigated whether the comorbidity patterns observed in our primary analysis are likely underlain by the association between mental disorder (ICD-9 level 1) and specific somatic disorders (ICD-9 level 3). Given the large number of specific somatic disorders, in these analyses we followed a systematic, multistep approach to minimize potential false-positive associations, including: (a) To address sparse data bias [31] all maternal specific somatic disorders with a recorded frequency of less than 10 in the pregnancies where mother either did, or did not receive a diagnosis for a mental disorder were excluded. ( b) Each specific somatic disorder was assessed in a separate model (adjusted univariate models) that adjusted for maternal age, SES, year of birth, and total number of encounters with health services during the pregnancy period. ( c) To address potential inflation of type I error due to multiple testing, p-values were corrected for a false discovery rate (q-value) of $5\%$ [32]. ( d) Each maternal somatic disorder that remained significantly associated with mental disorder after adjusting for multiple testing was evaluated in a model that was additionally adjusted for all the other maternal somatic disorder that remained significantly associated with mental disorder. Supplementary Figure S1 outlines the overview of the analytical strategy for this secondary analysis.
The robustness of our results in respect to the potential effect of missing data on study covariates were examined by sensitivity analyses comparing the results of complete case analyses with results obtained after multivariate imputation by chained equation [33]. We performed 10 imputations using information from all the variables in the model, including all level 1 diagnostic categories and covariates (maternal age at child’s birth, SES, and the total number of maternal diagnoses during the pregnancy period and the preceding year). All analyses were performed using R software (version 4.0.0) [34] including mice, miceadds, and clusterSEs packages.
## Results
The source population included 84,744 mother–child dyads, with children born 1999–2008. After removing the observations with missing values on SES ($$n = 7$$,699) and mother–child dyads with a maternal age at delivery below 13 or above 55 ($$n = 15$$), the analytical sample included 77,030 mother–child dyads, including 30,083 unique maternal IDs. None of the observations had a missing value for maternal age or child’s date of birth. Table 1 presents the analytical sample characteristics and prevalence of specific mental disorders with a minimum prevalence of $0.1\%$. In this population-based sample, pregnancy constituted a unique period with respect to the rates of diagnosis of majority of health conditions—with certain diagnoses (e.g., blood disorders) becoming more, and some less (e.g., musculoskeletal disorders) common in pregnancy, compared to the periods immediately before and after (Figure 2 and Supplementary Figure S2).Table 1.Demographic characteristics of the analytical sample (mother–child dyads).VariablesAny mental disorder ($$n = 3$$,361)No mental disorder ($$n = 73$$,669)Total ($$n = 77$$,030)Maternal age at delivery, mean (SD)31.4 (5.2)29.7 (5.4)29.8 (5.4)*Socioeconomic status* (SES), mean (SD)7.8 (4.3)7.6 (4.3)7.6 (4.3)Delivery year, n (%)1999243 ($7.2\%$)7,563 ($10.3\%$)7,806 ($10.1\%$)2000289 ($8.6\%$)7,597 ($10.3\%$)7,886 ($10.2\%$)2001291 ($8.7\%$)7,684 ($10.4\%$)7,975 ($10.4\%$)2002320 ($9.5\%$)7,908 ($10.7\%$)8,228 ($10.7\%$)2003308 ($9.2\%$)8,038 ($10.9\%$)8,346 ($10.8\%$)2004352 ($10.5\%$)7,877 ($10.7\%$)8,229 ($10.7\%$)2005436 ($13.0\%$)7,522 ($10.2\%$)7,958 ($10.3\%$)2006379 ($11.3\%$)7,207 ($9.8\%$)7,586 ($9.8\%$)2007467 ($13.9\%$)7,343 ($10.0\%$)7,810 ($10.1\%$)2008276 ($8.2\%$)4,930 ($6.7\%$)5,206 ($6.8\%$)Total number of health encounters for somatic disorders, mean (SD)29.2 (15.5)18.6 (12.2)19.1 (12.6)Transient mental disorders due to conditions classified elsewhere (ICD-9: 293); n (%)114 ($0.1\%$)Anxiety, dissociative and somatoform disorders (ICD-9: 300); n (%)1,574 ($2.0\%$)Personality disorders (ICD-9: 301); n (%)78 ($0.1\%$)Mental disorder related special symptoms or syndromes not elsewhere classified (ICD-9: 307); n (%)948 ($1.2\%$)Depressive disorder, not elsewhere classified (ICD-9: 311); n (%)633 ($0.8\%$) Note: Observations based on pregnancies; hence, mothers with multiple pregnancies contribute multiple data points. Figure 2.Mean number of diagnoses around pregnancy period by selected ICD-9 level 1 diagnostic categories (left: diseases of the musculoskeletal system and connective tissue, right: diseases of the blood and blood-forming organs).
Women with a mental disorder had a substantially higher number of health encounters for somatic disorders (including for chronic/recurrent disorders) recorded around pregnancy, compared to women without any mental disorders ($p \leq 0.001$). Presence of mental disorder was associated with a higher number of diagnosis from all categories of somatic disorders (Figure 3).Figure 3.Mean number of somatic disorders (ICD-9 level 1 diagnosis) during the 21-month period before child’s delivery by mental health status.
## Associations between maternal mental and somatic disorders
The associations between maternal mental disorder and all categories of somatic disorders around pregnancy are presented in Figure 4 and Supplementary Table S2. Maternal mental disorder was statistically significantly associated with an increased risk of 8 out of 17 somatic disorder categories. Maternal mental disorder around pregnancy was most strongly associated with diseases of the musculoskeletal system and connective tissue (OR = 1.30; $95\%$ CI = 1.20, 1.42), digestive system diseases (OR = 1.23; $95\%$ CI = 1.13–1.34), and neurological diseases (OR = 1.22; $95\%$ CI = 1.12–1.32). Interrogating the risk of receiving a diagnosis within these categories of somatic disorders in association with specific mental disorders (e.g., depression and personality disorders), we observed a similar pattern of associations (see section “Comorbidity patterns across specific mental disorders” of the Supplementary Material). Detailed results on the associations of mental disorder with specific somatic disorders are presented in Section “Comorbidity of specific somatic disorders with mental disorder” of the Supplementary Material. Figure 4.Associations between maternal somatic and mental disorders. Model 1 was adjusted for SES, maternal age at delivery, total number of encounters with health services during the 21 months period before delivery, and year of delivery. Model 2 was adjusted for all variables in model one as well as all the somatic disorder categories presented in this figure.
## Discussion
Our study adds to a body of literature showing that mental disorder around pregnancy is associated with a host of somatic comorbidities. Remarkably, the burden of somatic comorbidities associated with mental disorders was two times higher than that associated with another somatic disorder. This finding is novel and clinically relevant, as it indicates a higher co-occurrence between mental disorders and somatic complications in pregnancy relative to other types of comorbidites.
In this study, we systematically investigated the associations of somatic disorders around pregnancy (12 months before and during pregnancy) with maternal mental disorder. Initial analyses identified 10 broad somatic disorder categories, including musculoskeletal, neurological, and digestive system diseases, that occur at a higher rate in women with mental disorders in the antenatal period, compared to those without any mental disorder. Further analyses of over 700 specific somatic disorders grouped under those broader categories pointed to more specific associations (e.g. between mental disorders and hypertensive diseases), consistent with the existing literature about comorbidity and co-occurrence of somatic and mental disorders [35, 36].
In addition to the already well-established associations, our analyses yielded indications for potentially novel or less commonly studied associations. The somatic disorders most strongly associated with mental disorder included musculoskeletal, neurological, and digestive system diseases. Previous studies have found co-occurrence of mental and musculoskeletal disorders in working age population and aging women [37–39]. In our study, we observed that a similar pattern of comorbidity exists before and during pregnancy. Similarly, we observed an association between neurological diseases and mental disorders, which have been previously shown to be bidirectionally related [40]. Studies have also demonstrated higher rates of anxiety and depression among individuals suffering from irritable bowel syndrome and ulcerative colitis [41]. In line with these findings, we observed a positive association between digestive system diseases and mental disorders. Although we observed a strong association between mental disorders and symptoms, signs, and ill-defined conditions, this was not unexpected as many mental disorders have somatic manifestations which can be part of their diagnostic criteria [42]. Our results from the secondary analyses pointed to novel associations between mental disorders and lacrimal system disorders, and disorders of fluid, electrolyte, and acid–base balance which can serve as the basis for developing new hypotheses.
Importantly, our data present no support for the hypothesis that the patterns of somatic comorbidities vary between different mental disorders. Although the measures of associations between somatic disorders and personality disorders were not statistically significant (at least partly due to the low prevalence of personality disorders in the study sample), their direction and magnitude were mainly consistent with the direction and magnitude of the observed associations between somatic and other specific mental disorders.
The study results confirm the need for screening for mental disorders during pregnancy, a practice recommended by the obstetric guidelines of the American Obstetric Association [43]. Clinicians should also be aware of the high co-occurrence of somatic diseases (e.g., neurological, gastrointestinal, and musculoskeletal diseases) among patients with mental health diagnosis as comorbidities can worsen the course of disease for all the diseases that are present [10, 44]. For researchers, this study can offer insights into the complex network of associations between maternal health conditions and their impact on pregnancy and child outcomes. From the methodical perspective, our results can also inform selection of confounding variables in studies assessing impact of certain maternal diagnoses on health outcomes in the offspring. Even in situations when data on these measures are not available, pervasive nature of comorbidity should be acknowledged when drawing inferences from associations between maternal health and offspring outcomes.
## Strengths and limitations
The rigor and systematic nature of our approach should not be inferred as evidence for causality between maternal somatic and mental disorders. Although overall, our findings are consistent with the existing literature about high comorbidity between mental and somatic disorders [35, 36], the underlying mechanisms of the observed associations in our sample remain to be explored. While the presence of a mental disorder could cause somatic disorder (and vice-versa), shared underlying mechanisms should also be considered. Teasing out causal relationships will require a different methodology and approach, including careful consideration of confounding variables and establishing temporality between somatic and mental disorders. Since the time window for diagnosis of mental and somatic disorders was defined as pregnancy period and 1 year preceding it, the chronic conditions, particularly those adequately managed without requiring frequent health care visits, were less likely to be ascertained. Although we had adequate power to analyze associations between multiple somatic and mental disorders, there were less-common somatic disorders that were not included due to having a frequency of less than 10 across the mothers with or without mental disorders. Although defining the exposure based on at least one record of the specific diagnosis codes offered a higher sensitivity for identifying maternal diagnoses, it might have affected our specificity (e.g., due to administrative errors, or misdiagnosis). The study had limited data for covariate control. Lack of data on routinely adjusted health behaviors, including smoking, alcohol use, and physical activity did not allow us to investigate the possible behavioral factors that could lead to the observed associations between certain pairs of disorders. Additionally, the SES measure in our dataset was based on residential SES and it did not include household or individual level information. Although, we adjusted for residential SES, it could not capture the full depth of individual and household SES, [45, 46] therefore the SES adjustment in our analyses should be deemed incomplete. The analyses adjusted for the total number of encounters with health services during the 21-month period before delivery, which to a certain extent can account for differential health care access and utilization.
## Conclusions
In conclusion, this study followed a novel approach, supplementing prior research [47] with new insights on associations between a wide range of maternal somatic and mental health diagnoses around pregnancy. We demonstrated the high burden of somatic comorbidities among pregnant women with a diagnosis of mental health disorder, especially musculoskeletal, neurological, and digestive system diseases. The degree of comorbidity was twice as high compared to other pairs of somatic health conditions. Women with mental health problems are at high risk of “no shows” and having less control visits puts them at risk for missing somatic diagnoses. Moreover, their mental health conditions can make it difficult to adequately diagnose and monitor their somatic health, and treat comorbid somatic conditions. Awareness of the high comorbidity rates between mental and somatic disorders is thus of critical importance.
## Data Availability Statement
Data access rules do not permit public sharing of the data. Interested researchers should discuss access options with A.K. and S.Z.L.
## Author Contributions
V.K. and M.J. conceived the study idea; V.K. and M.J. planned the analysis. S.Z.L. and A.K. provided the data. V.K., A.R. and M.J. obtained the funding. V.K carried out the analysis and prepared the firstdraft of the manuscript. M.J. provided supervision. All authors contributed to the design of the study, interpretation of the findings, and critical revisions. All authors approved the version to be published and are accountable for the accuracy and integrity of their areas of involvement in the study.
## Funding Statement
This work was supported in part by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Institute of Environmental Health Sciences, and the National Institute of Neurological Disorders and Stroke (A.R.); by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (A.R., S.S., J.D.B. [Grant Number HD098883]); by grants from the National Institute of Mental Health (M.J., A.R., S.S. [Grant Number MH124817] and V.K. [Award Number T32-MH122394]). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies or the authors’ employers.
## Conflicts of Interest
The authors declare none.
## References
1. Salive ME. **Multimorbidity in older adults**. *Epidemiol Rev* (2013) **35** 75-83. DOI: 10.1093/epirev/mxs009
2. van den Akker M, Vaes B, Goderis G, Van Pottelbergh G, De Burghgraeve T, Henrard S.. **Trends in multimorbidity and polypharmacy in the Flemish-Belgian population between 2000 and 2015**. *PLoS One* (2019) **14** e0212046. DOI: 10.1371/journal.pone.0212046
3. King DE, Xiang J, Pilkerton CS. **Multimorbidity trends in United States adults, 1988–2014**. *J Am Board Fam Med* (2018) **31** 503. DOI: 10.3122/jabfm.2018.04.180008
4. Cortaredona S, Ventelou B. **The extra cost of comorbidity: multiple illnesses and the economic burden of non-communicable diseases**. *BMC Med* (2017) **15** 216. DOI: 10.1186/s12916-017-0978-2
5. Foo KM, Sundram M, Legido-Quigley H. **Facilitators and barriers of managing patients with multiple chronic conditions in the community: a qualitative study**. *BMC Public Health* (2020) **20** 273. DOI: 10.1186/s12889-020-8375-8
6. Hajat C, Stein E. **The global burden of multiple chronic conditions: a narrative review**. *Prev Med Rep* (2018) **12** 284. DOI: 10.1016/j.pmedr.2018.10.008
7. Lahti M, Savolainen K, Tuovinen S, Pesonen A-K, Lahti J, Heinonen K. **Maternal depressive symptoms during and after pregnancy and psychiatric problems in children**. *J Am Acad Child Adolesc Psychiatry.* (2017) **56** 30-39. DOI: 10.1016/j.jaac.2016.10.007
8. Strøm MS, Tollånes MC, Wilcox AJ, Lie RT, Forthun I, Moster D. **Maternal chronic conditions and risk of cerebral palsy in offspring: a national cohort study**. *Pediatrics* (2021) **147** e20201137. DOI: 10.1542/peds.2020-1137
9. Admon LK, Winkelman TNA, Heisler M, Dalton VK. **Obstetric outcomes and delivery-related health care utilization and costs among pregnant women with multiple chronic conditions**. *Prev Chronic Dis* (2018) **15** 170397. DOI: 10.5888/pcd15.170397
10. Leucht S, Burkard T, Henderson J, Maj M, Sartorius N. **Physical illness and schizophrenia: a review of the literature**. *Acta Psychiatr Scand* (2007) **116** 317. DOI: 10.1111/j.1600-0447.2007.01095.x
11. Stefana A, Youngstrom EA, Chen J, Hinshaw S, Maxwell V, Michalak E, Vieta E. **The COVID‐19 pandemic is a crisis and opportunity for bipolar disorder**. *Bipolar Disord* (2020) **22** 641. DOI: 10.1111/bdi.12949
12. Momen NC, Plana-Ripoll O, Agerbo E, Benros ME, Børglum AD, Christensen MK. **Association between mental disorders and subsequent medical conditions**. *N Engl J Med* (2020) **382** 1721. DOI: 10.1056/NEJMoa1915784
13. Birk JL, Kronish IM, Moise N, Falzon L, Yoon S, Davidson KW. **Depression and multimorbidity: considering temporal characteristics of the associations between depression and multiple chronic diseases**. *Health Psychol* (2019) **38** 802. DOI: 10.1037/hea0000737
14. Owens-Gary MD, Zhang X, Jawanda S, Bullard KM, Allweiss P, Smith BD. **The importance of addressing depression and diabetes distress in adults with type 2 diabetes**. *J Gen Intern Med* (2019) **34** 320. DOI: 10.1007/s11606-018-4705-2
15. Moya J, Phillips L, Sanford J, Wooton M, Gregg A, Schuda L.. **A review of physiological and behavioral changes during pregnancy and lactation: potential exposure factors and data gaps**. *J Expo Sci Environ Epidemiol* (2014) **24** 449. DOI: 10.1038/jes.2013.92
16. Beckert RH, Baer RJ, Anderson JG, Jelliffe-Pawlowski LL, Rogers EE. **Maternal anemia and pregnancy outcomes: a population-based study**. *J Perinatol* (2019) **39** 911. DOI: 10.1038/s41372-019-0375-0
17. McIntyre HD, Catalano P, Zhang C, Desoye G, Mathiesen ER, Damm P. **Gestational diabetes mellitus**. *Nat Rev Dis Prim* (2019) **5** 47. DOI: 10.1038/s41572-019-0098-8
18. Ramlakhan KP, Johnson MR, Roos-Hesselink JW. **Pregnancy and cardiovascular disease**. *Nat Rev Cardiol* (2020) **17** 718. DOI: 10.1038/s41569-020-0390-z
19. Cena L, Mirabella F, Palumbo G, Gigantesco A, Trainini A, Stefana A. **Prevalence of maternal antenatal anxiety and its association with demographic and socioeconomic factors: a multicentre study in Italy**. *Eur Psychiatry* (2020) **63** e84. DOI: 10.1192/j.eurpsy.2020.82
20. Cena L, Palumbo G, Mirabella F. **Perspectives on early screening and prompt intervention to identify and treat maternal perinatal mental health. Protocol for a prospective multicenter study in Italy**. *Front Psychol* (2020) **11** 365. DOI: 10.3389/fpsyg.2020.00365
21. Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE. **Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the national comorbidity survey replication**. *Arch Gen Psychiatry* (2005) **62** 593-602. DOI: 10.1001/archpsyc.62.6.593
22. Bansil P, Kuklina EV, Meikle SF. **Maternal and fetal outcomes among women with depression**. *J Women’s Health* (2010) **19** 329. DOI: 10.1089/jwh.2009.1387
23. Ng AT, Duan L, Win T, Spencer HT, Lee M-S. **Maternal and fetal outcomes in pregnant women with heart failure**. *Heart* (2018) **104** 1949. DOI: 10.1136/heartjnl-2018-313156
24. Wadhwa PD, Buss C, Entringer S, Swanson JM. **Developmental origins of health and disease: brief history of the approach and current focus on epigenetic mechanisms**. *Semin Reprod Med* (2009) **27** 358. DOI: 10.1055/s-0029-1237424
25. Janecka M, Kodesh A, Levine SZ, Lusskin SI, Viktorin A, Rahman R. **Association of autism spectrum disorder with prenatal exposure to medication affecting neurotransmitter systems**. *JAMA Psychiatry* (2018) **75** 1217. DOI: 10.1001/jamapsychiatry.2018.2728
26. Levine SZ, Kodesh A, Viktorin A, Smith L, Uher R, Reichenberg A, Sandin S. **Association of maternal use of folic acid and multivitamin supplements in the periods before and during pregnancy with the risk of autism spectrum disorder in offspring**. *JAMA Psychiatry* (2018) **75** 176. DOI: 10.1001/jamapsychiatry.2017.4050
27. [27]
Central Bureau of Statistics. Demographic characteristics of the population in localities and statistical areas. Jerusalem, Israel: Central Bureau of Statistics; 1995.. (1995)
28. Hudson CG. **Socioeconomic status and mental illness: tests of the social causation and selection hypotheses**. *Am J Orthopsychiatry* (2005) **75** 3-18. DOI: 10.1037/0002-9432.75.1.3
29. Dalstra JAA, Kunst AE, Borell C. **Socioeconomic differences in the prevalence of common chronic diseases: an overview of eight European countries**. *Int J Epidemiol* (2005) **34** 316. DOI: 10.1093/ije/dyh386
30. Cameron AC, Gelbach JB, Miller DL. **Robust inference with multiway clustering**. *J Bus Econ Stat* (2011) **29** 238. DOI: 10.1198/jbes.2010.07136
31. Greenland S, Mansournia MA, Altman DG. **Sparse data bias: a problem hiding in plain sight**. *BMJ* (2016) **352** i1981. DOI: 10.1136/bmj.i1981
32. Benjamini Y, Hochberg Y.. **Controlling the false discovery rate: a practical and powerful approach to multiple testing**. *J R Stat Soc Ser B.* (1995) **57** 289-300. DOI: 10.1111/j.2517-6161.1995.tb02031.x
33. Zhang Z.. **Multiple imputation with multivariate imputation by chained equation (MICE) package**. *Ann Transl Med* (2016) **4** 30. DOI: 10.3978/j.issn.2305-5839.2015.12.63
34. [34]
R Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2016.. *R: a language and environment for statistical computing* (2016)
35. Stein DJ, Aguilar-Gaxiola S, Alonso J, Bruffaerts R, de Jonge P, Liu Z. **Associations between mental disorders and subsequent onset of hypertension**. *Gen Hosp Psychiatry* (2014) **36** 142. DOI: 10.1016/j.genhosppsych.2013.11.002
36. Lee YB, Yu J, Choi HH, Jeon BS, Kim H-K, Kim S-W. **The association between peptic ulcer diseases and mental health problems: a population-based study: a STROBE compliant article**. *Medicine (Baltimore)* (2017) **96** e7828. DOI: 10.1097/MD.0000000000007828
37. Côté P, Hogg-Johnson S, Cassidy JD, Carroll L, Frank JW. **The association between neck pain intensity, physical functioning, depressive symptomatology and time-to-claim-closure after whiplash**. *J Clin Epidemiol* (2001) **54** 275. DOI: 10.1016/S0895-4356(00)00319-X
38. Heikkinen J, Koivumaa-Honkanen H, Rauma P, Williams LJ, Quirk SE, Leung J, Honkanen RJ. **Comorbidity of mental and musculoskeletal disorders in ageing women: a data linkage study using national registries**. *Maturitas* (2022) **155** 63. DOI: 10.1016/j.maturitas.2021.10.006
39. Lloyd C, Waghorn G, McHugh C. **Musculoskeletal disorders and comorbid depression: implications for practice**. *Aust Occup Ther J* (2008) **55** 23. DOI: 10.1111/j.1440-1630.2006.00624.x
40. Hesdorffer DC. **Comorbidity between neurological illness and psychiatric disorders**. *CNS Spectr* (2016) **21** 230. DOI: 10.1017/S1092852915000929
41. Shah E, Rezaie A, Riddle M, Pimentel M. **Psychological disorders in gastrointestinal disease: epiphenomenon, cause or consequence?**. *Ann Gastroenterol* (2014) **27** 224. PMID: 24974805
42. Bridges KW, Goldberg DP. **Somatic presentation of DSM III psychiatric disorders in primary care**. *J Psychosom Res* (1985) **29** 563. DOI: 10.1016/0022-3999(85)90064-9
43. Kendig S, Keats JP, Hoffman MC, Kay LB, Miller ES, Simas TAM. **Consensus bundle on maternal mental health: perinatal depression and anxiety**. *Obstet Gynecol* (2017) **129** 422. DOI: 10.1097/AOG.0000000000001902
44. Sartorious N. **Comorbidity of mental and physical diseases: a main challenge for medicine of the 21st century**. *Shanghai Arch Psychiatry* (2013) **25** 68. DOI: 10.3969/j.issn.1002-0829.2013.02.002
45. Demissie K, Hanley JA, Menzies D, Joseph L, Ernst P.. **Agreement in measuring socio-economic status: area-based versus individual measures**. *Chronic Dis Can* (2000) **21** 1-7. PMID: 10813687
46. Shavers VL. **Measurement of socioeconomic status in health disparities research**. *J Natl Med Assoc.* (2007) **99** 1013. DOI: 10.13016/avw3-9cvx
47. Kodesh A, Levine SZ, Khachadourian V, Kodesh A, Levine SZ, Khachadourian V, Rahman R, Schlessinger A, O’Reilly PF. **Maternal health around pregnancy and autism risk: A diagnosis-wide, populationbased study**. *Psychological Medicine* (2020) **52** 4076-4084. DOI: 10.1101/2020.05.19.20089581
|
---
title: Does diet or macronutrients intake drive the structure and function of gut
microbiota?
authors:
- Yuhang Li
- Yujie Yan
- Hengguang Fu
- Shiyu Jin
- Shujun He
- Zi Wang
- Guixin Dong
- Baoguo Li
- Songtao Guo
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC9970161
doi: 10.3389/fmicb.2023.1126189
license: CC BY 4.0
---
# Does diet or macronutrients intake drive the structure and function of gut microbiota?
## Abstract
Shift of ingestive behavior is an important strategy for animals to adapt to change of the environment. We knew that shifts in animal dietary habits lead to changes in the structure of the gut microbiota, but we are not sure about if changes in the composition and function of the gut microbiota respond to changes in the nutrient intake or food items. To investigate how animal feeding strategies affect nutrient intakes and thus alter the composition and digestion function of gut microbiota, we selected a group of wild primate group for the study. We quantified their diet and macronutrients intake in four seasons of a year, and instant fecal samples were analyzed by high-throughput sequencing of 16S rRNA and metagenomics. These results demonstrated that the main reason that causes seasonal shifts of gut microbiota is the macronutrient variation induced by seasonal dietary differences. Gut microbes can help to compensate for insufficient macronutrients intake of the host through microbial metabolic functions. This study contributes to a deeper understanding of the causes of seasonal variation in host-microbial variation in wild primates.
## Introduction
What factors cause or have interaction with gut microbiota is a key and hot issue in animal evolutionary adaptation and original of human diet health. It is reported that the composition and structure of animal gut microbiota changes with host diet, which is shown in macroscopic indicators such as diversity of the microbes (Hooper et al., 2012; Morgan et al., 2012; Markle et al., 2013; Amato et al., 2019). Moreover, changes of gut microbiota show seasonal fluctuation in wildlife and humans, and literatures also widely indicated that seasonal dietary changes lead to the reconfiguration of gut microbiota of hosts, or at least both aspects have strong interaction. For example, seasonal cycling in the gut microbes following the dietary fluctuation has been reported in the Hadza hunter-gatherers in Tanzania (Smits et al., 2017), the western lowland gorillas and chimpanzees (Hicks et al., 2018), and red squirrels (Ren et al., 2017).
While, recent studies have shown that diets of invertebrates and vertebrates may be determined by the nutrient components in foods (Ruohonen et al., 2007; Simpson and Raubenheimer, 2012). The intake of each macronutrient should be stable no matter whether an animal is a vegetarian, carnivore, or omnivore and how complex its food composition is (Raubenheimer et al., 2009, 2015; Machovsky-Capuska et al., 2016). Non-human primates can take in stable proportions of macronutrients from foods with complex and diverse components, which suggests that they have stable requirements for three macronutrients. The nutrient intakes of non-human primates are influenced by both the type and the amount of food consumed. Studies on folivorous primates including species of the subfamily Colobus (Chapman et al., 2003) and the subfamily Indriidae (Indri) (Junge et al., 2009; Fleming and John Kress, 2011) show huge variation in macronutrient composition in their foods. Meanwhile, some species such as Mountain gorillas (Gorilla Beringei) can maintain a constant non-protein intake during the period when non-protein nutrients from fruits are scarce by consuming an excessive amount of leaves (Rothman et al., 2011).
As the diet may co-evolve with gut microbiota among different animals, the notion that diet variation can influence gut microbiome composition and structure has been confirmed at the taxonomic level of family (Ley et al., 2008). However, within genera, or ranks below genera, species from the same taxa may vary greatly in diet (Guo et al., 2007). For example, chimpanzees (Pan troglodytes) are considered frugivorous primates, but they feed heavily on leaves and even prey on other animals during the fruitless season (Rothman et al., 2007). The dietary variability makes it difficult to explain the changes in the composition and structure of gut microbes. So far, it has been found that such changes are closely related to the macronutrients consumed on the study of captive animals (Grześkowiak et al., 2015). However, due to the difficulty in quantifying the nutrient intakes in the field and the scarcity of such studies on wild animals, especially endangered wild mammals, we are hampered to figure out the mechanisms of gut microbiota-host co-evolution in many wild species. Therefore, this study aims to investigate how whether diet or nutrients intake affect the structure and composition of gut microbiota and the interaction among these three aspects. This will reveal the mechanisms behind the seasonal shifts in diet in animals that rely heavily on gut microbiota for digestion.
Golden snub-nosed monkeys (Rhinopithecus roxellana) belong to the genus Rhinopithecus in the subfamily Colobinae, and their habitats vary seasonally (Hou et al., 2018). Our previous research on their diet based on time ratio reveals that wild golden snub-nosed monkeys have seasonal diet variation (Guo et al., 2007). Recent studies show that their feeding strategy stabilizes protein intakes and balances energy requirements by regulating carbohydrate and lipid intakes (Hou et al., 2021). Since golden snub-nosed monkeys live in an environment with complex foods and are capable of maintaining a stable amount of macronutrient via various feeding strategies, they are an excellent model for studying the interactions between food consumed, nutrient intakes, and gut microbiota. In view of the reasons above, we propose the following research questions. [ 1] Do seasonal changes in food types lead to changes in the composition and structure of gut microbiota? and [2] Do seasonal variations in nutrient intakes lead to seasonal variations in gut microbiota composition? [ 3] Are there any seasonal variations in gut microbial gene function? Does the gene function correlate with seasonal variations in dietary and nutrient intakes?
## Data collection
Our observation site was in Guanyin Mountain National Nature Reserve (107°51′-108°01′E，33°35′–33°45′N,135.34 km2) on the southern slope of Qinling Mountains, which locates at the northwest of Fuping County, Shaanxi Province, China (Supplementary Figure S1). This region experiences the classic and distinct four seasons throughout the year. The seasons are divided according to climate: spring is from March to May, summer is from June to August, autumn is from September to November, and winter is from December to February (Guo et al., 2007, 2018). We collected feeding data of four season groups (i.e., spring, summer, autumn, and winter). For each season, we chose a month with typical phenological characters, that is, March (spring), June (summer), October (autumn), and December (winter) for data collection.
Our study group of golden snub-nosed monkeys had 78 individuals, all haven been habitualized to the presence of researchers. Adult and juvenile individuals haven been identified by us. Because we needed to collect quantitative observation data, the natural feeding space of the target animals was narrowed. To prevent that their total energy intakes being reduced due to this condition and thus impacting their health, we referred to our previous experience to provision foods (Hou et al., 2021). We provisioned 5 kg of maize grains twice daily at 10 a.m. and 3 p.m. as supplementary food for the group. Maize grains were spread evenly in the feeding grounds.
We randomly chose one individual per day and conducted continuous observations of the focal animal from dawn to dark to record its feeding data. During the observation session, the distance between the observer and the subject was less than 5 m. We recorded the type, quantity, and predefined units of the food and the amount of time feeding (Hou et al., 2018). After the focal individual completed feeding, leftover foods were collected as food samples. All samples were labeled with the information of the collection time, type, and size. Then, they were immediately frozen in liquid nitrogen and sent to the laboratory for storage before the analysis of their nutrient components. We also collected same-day fecal samples for high-throughput sequencing. After the focal individual defecated, we immediately collected the feces with sterile cotton swabs and sterile toothpicks. The sample was then stored in 2 mL centrifuge tubes and frozen in liquid nitrogen before being delivered for testing.
## Nutrient analysis
We used the standard techniques to collect the food samples (Rothman et al., 2012), analyzed the foods nutrients, and calculated the energy values (Jung, 1995) with the same methods used in our previous studies (Guo et al., 2018; Hou et al., 2018). The macronutrients of each food were analyzed for lipid, water-soluble carbohydrate (WSC), starch, neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL), available protein (AP), and ash content. Available proteins are determined by the standard Kjeldahl method (using BUCHI, K-360).
To calculate the daily nutrient intake (DNI), we multiplied the nutrient content of each food item by the recorded amount of that item consumed, then summed these values for all items consumed by that individual on that day. We also calculated the rate of nutrient ingestion per hour (NIH) for each individual by dividing the amount of nutrient ingested by the amount of hours the focal animal was observed. The rate was multiplied by the sunshine duration to estimate the total daily intake (TDNI; Rothman et al., 2008). We lured each monkey with a small portion of food and led it onto a platform scale (accuracy, 0.02 kg; EM-60KAL, A&D, Japan) to record their weight when the readings were stable (Hou et al., 2021). To standardize weight differences between individuals, the calculation was divided by the individual’s estimated metabolic body mass (MBM = M0.762, where M is the body weight in kg).
## DNA extraction and 16S rRNA gene sequencing
The microbial DNA (with a total mass of 1.2–10.0 ng) was isolated from each fecal sample using the MOBIO Pow erSoil DNA Isolation Kit and was quantified with NanoDrop One (Thermo Fisher Scientific, Waltham, MA, United States). The V4 regions of the DNA genes were amplified by using the specific primer 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTW TCTAAT-3′) with 12 bp barcode. Primers were synthesized by Invitrogen (Invitrogen, Carlsbad, CA, United States). The PCR instrument was Bio-Rad S1000 (Bio-Rad Laboratory, CA, United States). The length and concentration of the PCR product were detected by $1\%$ agarose gel electrophoresis. PCR products with bright main strip between were mixed in equidensity ratios according to the GeneTools Analysis Software (Version 4.03.05.0, SynGene). Then, mixture of PCR products was purified with E.Z.N.A. Gel Extraction Kit (Omega, United States). Sequencing libraries were generated using NEBNext® Ultra™ II DNA Library Prep Kit for Illumina® (New England Biolabs, MA, United States) following the manufacturer’s recommendations and index codes were added. The library quality was assessed on the Qubit@ 2.0 Fluorometer (Thermo Fisher Scientific, MA, United States). At last, the library was sequenced on an Illumina Nova6000 platform and 250 bp paired-end reads were generated (Guangdong Magigene Biotechnology Co., Ltd. Guangzhou, China).
## Metagenomic sequencing and gene catalog construction
The sequencing library was created using NEBNext Ultra DNA Library Prep Kit for Illumina (New England Biolabs, Beverly, MA, United States) and indexes were added to attribute sequences to each sample. The DNA sample was fragmented by sonication to a size of 300 bp. DNA fragments were polished at the extremities and were attached to the full-length adapter for Illumina sequencing with further PCR amplification. The library was analyzed for size distribution by Agilent2100 Bioanalyzer (Agilent, United States), and then was sequenced by Illumina HiSeq 2500 platform in Magigene Co., Ltd. (Shenzhen, China).
Quality control was conducted by Trimmomatic (Version 0.38). The reads aligned to the NCBI non-redundant (NR) database were removed with MEGAHIT (Version 1.05). The remaining high-quality reads were used for further analysis. The assembly of reads was conducted using MEGAHIT de novo. For each sample, a series of k-mer (substrings of length k) values (49–87) were used and the optimal one with the longest N50 value was chosen for the remaining scaffolds. The clean data were mapped against scaffolds using MEGAHIT. Unused reads from each sample were assembled using the same parameters.
Genes (minimum length of 100 nucleotides) were predicted on scaftigs longer than 500 bp using Prodigal (Version 2.6.3). Then, a non-redundant gene catalog was constructed with Linclust (Version 2.0) using a sequence identity cut-off of 0.9. To determine the abundance of genes, reads were realigned to the gene catalog with BBMap (Version 37.68). *Only* genes with 2 mapped reads no less than 2 were considered exist in a sample. The abundance of genes was calculated by counting the number of reads and normalizing by gene length. Genes were then searched in Kyoto Encyclopedia of Genes and Genomes (KEGG) database (Kanehisa and Goto, 2000) for annotation.
## Statistical analysis
Kruskal-Wallis test with Bonferroni correction for multiple post-hoc pairwise comparison was used to compare the difference of available protein, fat, and carbohydrate in four seasons. Permutational Multivariate Analysis of Variance (PERMANOVA) was used to compare the difference of Shannon index and Chao1 index. The Analysis of Similarities (ANOSM) was used to compare the result of Principal co-ordinate analysis (PCoA). The Mantel test was used to compare the correlation between food groups and the gut microbial composition in each season. Pearson’s correlation coefficients were calculated to analyze correlations between weighted gene co-expression network analysis (WGCNA) module groupings and traits. Data visualization was performed by R4.1.3 and Cytoscape 3.8.2.
## Seasonal diets
In this study, feeding data of 96 days across 4 months (25 days in spring, 24 days in summer, 24 days in autumn, and 23 days in winter) were collected from the target group. It was observed that the natural foods consumed by wild golden snub-nosed monkeys consisted of 24 plant species belonging to 16 families. Six items of plant including barks, seeds, buds, brunches, leaves, and stems have been observed to be consumed. Throughout the year, the proportion of each plant item consumed by wild snub-nosed monkeys was $33.43\%$ for bark, $3.09\%$ for seed, $1.33\%$ for bud, $3.25\%$ for brunch, $0.17\%$ for stem, and $58.72\%$ for leaf. However, there were huge differences in the amount of the plant items being consumed across four seasons. Herbaceous stems were only taken in spring with a small quantity. Seeds were taken mainly in spring and autumn. Leaves were taken throughout the year. Buds, barks, and brunches were the main food in autumn and winter when leaves become scarce, especially in winter (Supplementary Figure S2A).
In addition, differences have also been found in the plant species consumed between seasons—Photinia beauverdiana, Acer davidii, Dendrobenthamia japonica, Kerria japonica, Ulmus macrocarpa, Quercus aliena, Acer mono Maxim, and *Lonicera japonica* were mainly taken in spring; Cerasus clarofolia, Ailanthus altissima, Juglans mandshurica, and Spiraea blumei were mainly taken in summer; Rubus pungens, Arachis hypogaea, Quercus mongolica, Pinus bungeana, Lonicera hispida, and *Carpinus cordata* were mainly taken in autumn; and Litsea pungens, Quercus dolicholepis, Fargesia qinlingensis, Bothrocaryum controversum, Glechoma longituba, Litchi chinensis, and *Callicarpa nudiflora* were mainly taken in winter (Supplementary Figure S2B). The results of PCA on seasonal percentage data of food mass for 24 plant species and 6 plant items showed that there were significant differences in the plant species consumed by golden snub-nosed monkeys between any two seasons (PERMANOVA, $p \leq 0.05$, Table 1). Also, there were significant differences in plant items consumed between any two seasons except for spring and summer (PERMANOVA, $p \leq 0.05$, Table 2).
## Nutritional properties
We collected 55 types of food from 24 plant species and evaluated their nutritional properties. We measured the energy of metabolic body mass (KJ/MBM) provided by macronutrients, which showed that spring food intake provides 25.93 ± 11.04 kJ/MBM (M ± SE) of available protein, 22.41 ± 10.55 kJ/MBM of fats, and 245.83 ± 106.47 kJ/MBM of carbohydrates; summer food intake provides 43.80 ± 9.72 kJ/MBM of available protein, 36.88 ± 9.21 kJ/MBM of fats, and 411.19 ± 96.55 kJ/MBM of carbohydrates; autumn food intake provides 39.83 ± 22.26 kJ/MBM of available protein, 37.17 kJ/MBM of fats, and 394.68 ± 219.21 kJ/MBM of carbohydrates; winter food intake provides 28.17 ± 10.56 kJ/MBM of available protein, 24.76 ± 9.68 kJ/MBM of fats provides, and carbohydrates provided 278.15 ± 100.48 kJ/MBM. Statistical analysis of available proteins, carbohydrates, and fats provided by foods in four seasons found that they were all significantly different between spring vs. summer and summer vs. winter groups, and that fats also differed in spring vs. autumn group (Figure 1).
**Figure 1:** *Comparison of macronutrients among four seasons (we measured the energy of metabolic body mass (KJ/MBM) provided by macronutrients). (A) The comparison of available protein intake. (B) The comparison of carbohydrate intake. (C) The comparison of fat intake. **p < 0.01, ***p < 0.001. (Kruskal-Wallis test).*
We also divided the sources of macronutrients into natural foods and artificial foods. For natural foods, the energy provided by available proteins decreases in the sequence of summer, winter, autumn, and spring, while the energy provided by carbohydrates and fats increases in the sequence of spring, summer, autumn, and winter. For artificial foods, available proteins, carbohydrates, and fats presented a uniform seasonal pattern throughout the year with energy values decreasing from summer to autumn to winter and to spring (Supplementary Figure S3).
## Microbial compositions
The 16S rRNA gene sequencing of fecal samples revealed that the observed species and the two alpha-diversity indexes reflecting species richness and diversity (Chao1 index and Shannon index, respectively) showed a decreasing in the order of spring, winter, autumn, and summer. That is, the richness of gut microbiota and the diversity of community in golden snub-nosed monkeys were highest in spring and lowest in summer during all seasons. The Chao1 index was significantly different between spring vs. summer, summer vs. autumn, summer vs. winter, and autumn vs. winter (Kruskal-Wallis test, $p \leq 0.05$, Figure 2A), while the Shannon index was only significantly different between spring vs. summer (Kruskal-Wallis test, $p \leq 0.05$, Figure 2B). Principal co-ordinate analysis (PCoA) based on weighted UniFrac distances (Figure 3A) and unweighted UniFrac distances (Figure 3B) of OTUs showed divergence between the groups of different seasons along the first and second principal components. Analysis of Similarities (ANOSM) showed that p values of all groups were less than 0.05, indicating a significant difference between the gut microbiota of four seasons.
**Figure 2:** *Alpha-diversity of gut microbiota. The Chao1 indexes fluctuate significantly between seasons (A), while the Shannon indexes vary only in spring and summer and are stable for the rest (B). *p < 0.05, **p < 0.01, ***p < 0.001 (Kruskal-Wallis test).* **Figure 3:** *Principal co-ordinate analysis based on weighted UniFrac distances (A) and unweighted UniFrac distances (B) between all seasons showed significant difference in structure of seasonal gut microbial composition.*
To further investigate seasonal differences in gut microbiota, species composition was analyzed. Species annotation of the 16S rRNA sequencing showed that most OTUs could be taxonomically assigned to the phylum ($96\%$) and order ($92\%$) level, but assignments decreased substantially at the genus ($38\%$) level. A total of 38 phyla were annotated, of which the top 10 identifiable dominant phylum including Firmicutes, Bacteroidetes, Proteobacteria, Spirochaetes, Tenericutes, Verrucomicrobia, Planctomycetes, Epsilonbacteraeota, Fibrobacteres, and Euryarchaeota accounted for $99\%$ of the total abundance ratio. These formed the core gut microbiota of golden snub-nosed monkeys. When considering them in different seasons, they showed seasonal differences in abundance, with Firmicutes and Euryarchaeota being most abundant in spring and Proteobacteria being least abundant in spring compared to other seasons (Figure 4A). The ratio of Firmicutes to Bacteroidetes (F/B) was highest in spring and lowest in autumn, suggesting a seasonal variation in the capacity of energy absorption by gut microbiota (Supplementary Figure S4).
**Figure 4:** *Analysis of seasonal differences in dominant bacterial populations. Relative abundance of dominant phylum (A) and order (B) in four seasons based on 16S rRNA gene pools. In contrast, macrogenome annotation of dominant bacteria at the level of phylum (C) and order (D). Heat map of proportions at the genus level (E), the darker the color, the higher the proportion of this genus present in the season compared to other seasons.*
A total of 140 orders were annotated. The dominant identifiable bacteria were Clostridiales, Bacteroidales, Aeromonadales, Methanobacteriales, Mollicutes_RF39, Spirochaetales, Verrucomicrobiales, Pirellulales, Erysipelotrichales, and Selenomonadales (Figure 4B). It is noteworthy that the abundance of the Aeromonadales was much lower in spring than in other seasons. Meanwhile, Methanobacteriales, which are associated with methane production, were observed to have abundance much higher in spring and summer than in autumn and winter. They had a particular high abundance in spring. Metagenomic analysis showed that several bacterial taxa with high abundance at the phylum and order levels were consistent with the 16S rRNA study (Figures 4C,D).
At the genus level, there were 352 taxa annotated and the top 100 genera covered nearly $99.9\%$ of the total abundance. The study of these 100 genera found that the gut microbes of golden snub-nosed monkeys were mostly related to hindgut fermentation in ruminant animals. *These* genera include those that can degrade complex polysaccharides such as Methanobrevibacter, Methanosphaera, prevotella_7, Roseburia, Ruminococcaceae_UGG-014, Treponema_2, Clostridium, those that can produce hydrogen efficiently such as christensenellaceae_R-7_group, and those play roles in lipid metabolism such as [Eubacterium]_coprostanoligenes_group, Blautia, Dorea, lactobacillus, Dialister, and Phascolarctobacterium (Figure 4E).
## Gene function prediction of gut microbiota
We performed KEGG annotation using metagenome data to find out the main functions of the gut microbiota in golden snub-nosed monkeys. According to the function prediction based on the KEGG database, we identified 395 metabolic pathways. Among these pathways, gut microbes were mainly involved in the nucleotide metabolism, carbohydrates metabolism, glycans metabolism and biosynthesis, amino acids metabolism, energy metabolism, lipids metabolism, terpenoids and polyketides metabolism, as well as cofactors and vitamins metabolism. Moreover, some functions annotated concerning macronutrients showed relatively high abundance such as glycolysis/gluconeogenesis, pyruvate metabolism, starch and sucrose metabolism, pentose phosphate pathway in carbohydrate metabolism and glycerophospholipid metabolism, glycerolipid metabolism, and fatty acid synthesis in lipid metabolism.
Inter-seasonal KEGG enrichment analysis demonstrated that there were 288, 210, 237, 98, 71, and 78 differentially expressed genes and were enriched in 20, 21, 20, 17, 16, and 14 KEGG pathways in spring vs. summer, spring vs. autumn, spring vs. winter, summer vs. autumn, summer vs. winter, and autumn vs. winter groups, respectively. The seasonal function variation of these differentially expressed genes can be presented in terms of metabolism, organic systems, and environmental information processing. Specifically, the enriched pathways were similar in all season groups. These pathways included biosynthesis of siderophore group nonribosomal peptides (ko01053); protein digestion and absorption (ko04974); flavonoid biosynthesis (ko00941); stilbenoid, diarylheptanoid, and gingerol biosynthesis (ko00945); cardiac muscle contraction (ko04260), phenylpropanoid biosynthesis (ko00940); arachidonic acid metabolism (ko00590); atrazine degradation (ko00791); flavone and flavonol biosynthesis (ko00944); linoleic acid metabolism (ko00591), fatty acid elongation (ko00062); alpha-Linolenic acid metabolism (ko00592); and bile secretion (ko04976) (Figure 5).
**Figure 5:** *Analysis of KEGG annotation in groups with significant differences in macronutrient intakes to predict the function of gut microbial, including spring vs. summer group (A), autumn vs. spring group (B), spring vs. winter (C), autumn vs. summer (D), summer vs. winter group (E), and autumn vs. winter (F).*
## Correlation between gut microbiota and food types
Based on the significant seasonal differences of natural food types and gut microbiota in golden snub-nosed monkeys, we conducted a correlation analysis between food types and OTUs in four seasons. The results demonstrated that *Dendrobenthamia japonica* and Zea mays were correlated with gut microbiota in spring (Mantel test, $p \leq 0.05$, Figure 6A), and *Callicarpa nudiflora* was correlated with gut microbiota in summer (mantel test, $p \leq 0.05$, Figure 6B). In autumn and winter, there was no correlation between food types and gut microbiota (Figures 6C,D).
**Figure 6:** *Heat map of the correlation between food groups and gut microbes of spring (A), summer (B), autumn (C), winter (D). Darker color refers to that r value is closer to 1; thicker line refers to higher r value between food groups and gut microbes. The color of the line segment shows the p value between food group and gut microorganism (Blue indicates p value <0.05, grey indicates p value >0.05).*
## WGCNA on the hub OTUs
The present study used 3,638 OTUs for weighted gene co-expression network analysis (WGCNA). OTUs that exist over half of the sample in each season were selected and the network was constructed in one step. To define the adjacency matrix based on the criterion of approximate scale-free topology (Supplementary Figure S5), the network type was set as sign and the soft threshold parameter set to 10 with a minimum module size of 30 and the module detection sensitivity DeepSplit of 3. Modules that are correlated above 0.75 would be merged (Supplementary Figure S6). The clustering results showed that a total of 814 OTUs were parsed into 5 different modules. The gray module refers to ones that were not classified. The correlation between module eigenvalue and trait was calculated. The module-trait relationship heatmap demonstrated the correlation coefficient between module eigenvalues and traits. The green module refers to OTUs that were significantly correlated with three macronutrients (fat, carbohydrate, and available protein) at the same time, and the blue module was significant correlation with fat ($p \leq 0.05$, Figure 7A).
**Figure 7:** *Identification of key module and hub OTUs based on WGCNA. (A) Correlation between module eigenvalues and traits of golden snub-nosed monkey. Depth of color corresponds to depth of correlation and p value of each module. (B) Network graph of the hub OTUs. Each node represented the OTUs whose betweenness centrality value was in the top 10%, and its color represented the corresponding module, the size of each node represented the betweenness centrality value, the size of each line thickness represented the weight value between nodes (OTUs). (C) Visualization of full weighted networks of 22 candidate hub OTUs in green module associated with three different nutrients (fat, protein, and carbohydrate). (D) Visualization of full weighted networks of 9 candidate hub OTUs in blue module associated with fat.*
In addition, OTUs with betweenness centrality at top $10\%$ in WGCNA were selected and the network graph was constructed in Cytoscape. The results showed that green module took the largest proportion and had richer network relationships (Figure 7B). Therefore, the green and blue modules were selected for hub gene analysis. We calculated the correlation between the module membership (MM) and the genes significance (GS) and nutritional traits. It was found that the relationship between MM and GS for these modules was relatively strong, particularly for those in the green module (r > 0.4, $p \leq 0.05$, Supplementary Figure S7). We also identified 22 candidate hub OTUs that are correlated with fat, carbohydrate and available protein in the green module, and 9 candidate hub OTUs that are correlated with fat in the blue module (threshold values of MM > 0.6 and GS < 0.1, respectively). The network graphs were constructed accordingly (Figures 7C,D). OTU_472, OTU_2009, OTU_226, OTU_81, OTU_67, and OTU_349 in the green module and OTU_44 in the blue module were found to be the most important hub OTUs. They belonged to the family Ruminococcaceae (OTU_472 and OTU_81), Lachnospiraceae (OTU_2009, OTU_226, and OTU_349) in the order Clostridiales and family Muribaculaceae (OTU_67), and Prevotellaceae (OTU_44) in the order Bacteroidales, respectively.
## Discussion
Our results reveal that there are remarkable seasonal differences in both food items and macronutrients intake from food for wild golden snub-nosed monkeys. However, seasonal feeding strategies cannot fully explain the composition and fluctuation of their gut microbiota. On the contrary, the changes in carbohydrates, fats, and available proteins present similar trends with the changes of gut microbiota. Our findings suggest that the key factor that shapes the composition and function of wild golden snub-nosed monkeys’ gut microbiota is macronutrient intakes rather than food types. This differs from the view that diverse gut microbiotas in wild mammals are the result of the seasonal changes in dietary habits or food types in the previous studies.
## The effects of nutrients on gut microbial composition
Previous studies suggest that seasonal changes in the composition and function of mammalian gut microbiota may be related to seasonal changes in the host’s diet. A study on wild blue sheep (Pseudois nayaur) found that changes in the composition of the animals’ gut microbiota is due to seasonal shifts in dietary habits (Zhu et al., 2020). Giant panda (Ailuropoda melanoleuca) gut microbes would produce more single-chain fatty acids during the shoot-eating season compared to the leaf-eating season (Huang et al., 2022). For wild geladas (Theropithecus gelada), the gut microbes in the rainier periods are mainly cellulolytic or fermentative bacterial that specialized in digestion grass, while during dry periods the gut is dominated by bacteria that break down starches found in underground plant parts (Baniel et al., 2021).
Meanwhile, many studies also suggest that the community structure and function of gut microbiota are influenced by the macronutrients consumed by host. In human guts, Bacteroides will be the dominant microbes when diets are rich in proteins and fats, while *Prevotella is* central to the diet rich in carbohydrates (Devkota et al., 2012; Henao-Mejia et al., 2012). Shifts in nutrient intakes have also been found to lead to changes of gut microbial composition in captive mammals. For example, high-fat diet with a high cholesterol intake resulted in dysbiosis of gut microbes and downregulation of microbial tryptophan metabolism in mice (Zhang et al., 2021a). Increasing different types of carbohydrates in the feed could lead to changes in the abundance of gut microbes in pigs (Lyu et al., 2020; Wang et al., 2021). Gut microbes in dogs and cats regulated their growth, reproduction, and homeostasis per se by breaking down nutrients that were not digestible by host digestive enzymes (Oh et al., 2021).
The analysis of the composition of gut microbiota in this study shows that both alpha-diversity and beta-diversity exhibited differences in richness and diversity in different seasons. Firmicutes was the most dominant phylum in all seasons, followed by Bacteroidetes and Proteobacteria. The results also indicate that Firmicutes can interact with other gut microbes to influence the absorption of nutrients. The previous study reveals that a high proportion of Firmicutes means that the host is able to get more energy from the food (Ley et al., 2006). When metabolic disorder produces dysbiosis that disturb the stability of the gut microbes, it is usually accompanied by an increase in Proteobacteria (Shin et al., 2015). In our study, golden snub-nosed monkeys have the least nutrient intakes in spring, and the gut microbiota show high abundance in Firmicutes and low abundance in Proteobacteria. This reflects that gut microbiota of host would be more stable in spring and the host could absorb more energy from low nutrient intakes to sustain life activities. Based on the above findings at the phylum level, we speculate that these gut microbes flourish to compensate for the low nutrient intakes in spring through microbial action to maximize energy utilization.
At the order level, samples in spring show an increase in Methanobacteriales and a decrease in Aeromonadales. Methanogens can reduce intestinal gas accumulation (Kengen et al., 1994) and maintain an anaerobic environment in the hindgut, facilitate the metabolic of polysaccharides, and improve the utilization efficiency of energy (Samuel and Gordon, 2006; Samuel et al., 2007). And more our study finds that the abundance of *Aeromonadales is* much lower in spring than in other seasons, while the increased abundance of Aeromonadales has been verified to be the reason for intestinal inflammation. Aeromonadales-related lipopolysaccharides disrupt the intestinal mucosal barrier and cause the increase of intestinal permeability, thereby causing inflammation (Zhang et al., 2021b).
At the genus level, there is an increase in the genus Methanosphaera, Methanobrevibacter, Shuttleworthia, Ruminococcaceae_UCG-014, Treponema_2, Ruminiclostridium, and Ruminococcaceae_NK4A214_group in spring when proteins, fats, and carbohydrates are consumed at the lowest amount. The methanogenic bacteria Methanosphaera, Methanobrevibacter, and Shuttleworthia can convert hydrogen and formate into methane. When enriched simultaneously with efficient hydrogen-producing bacteria such as Christensenellaceae_R-7_group (Morotomi et al., 2012), they are able to work synergistically to improve the efficiency of gut fermentation of starch and other polysaccharides (Samuel and Gordon, 2006). A large number of methanogenic bacteria can increase the calories obtained from food and enhance the absorption and utilization of nutrients (Mizrahi et al., 2021). They also promote the production of short-chain fatty acids by other fermenting bacteria and stimulate the production of fats (Zhang et al., 2009; Basseri et al., 2012). Ruminococcaceae_UCG-014 can produce butyrate, an important energy source for colon cells. In the meantime, they can increase short-chain fatty acids and affect host appetite and satiety through different mechanisms, delaying gastric emptying and thus energy absorption (Canfora et al., 2015; Guo et al., 2022). Ruminiclostridium is positively correlated with acetate content in the cecum, providing more energy to the host (Zhang et al., 2021c). As to autumn when diets are high-carbohydrate and high-fat, Prevotella, Phascolarctobacterium, and Lactobacillus were the dominant genus. Prevotella is capable of breaking down non-cellulosic polysaccharides and pectins (Flint et al., 2012). Both Phascolarctobacterium and Lactobacillus are probiotics that can break down fats (Lv et al., 2019; Rodrigues et al., 2021). Although our results show that the change in gut microbiota from spring to autumn is not due to a change in diet structure. However, there are diverse polysaccharides and fats that could be potentially utilized in autumn. Therefore, we suggest that differences in nutrient intakes may be a significant factor that shaped the composition of gut microbial communities during animal growth and development.
## The effects of nutrients on gut microbial function
The analysis for gene function prediction based on KEGG database shows that the gut microbiota in wild golden snub-nosed monkeys mainly take part in metabolism and synthesis of lipid, carbohydrate, protein, amino acids, and other secondary metabolites. We should point out that most of the functional genes in the metagenome in this study appear to be involved in carbohydrate metabolism, but these pathways are not enriched in any seasonal grouping. This result may be due to the fact that we fed the monkeys equal amounts of maize throughout the year, which maintained their energy provided by large amounts of carbohydrates at a relatively constant level. Therefore, the gut microbes also responded stably to the degradation of carbohydrate.
The macronutrients are found significantly different both between spring and summer groups and between summer and winter groups. Also, there is a significant difference in fat intakes between spring and autumn groups. We found that 71, 228, and 210 differentially expressed genes between summer vs. winter group, spring vs. summer group, and spring vs. autumn group and were significantly enriched in 16, 20, and 21 KEGG pathways, respectively. Six of these pathways are important for the response to changes in the nutrients intake, including protein digestion and absorption (ko04974), fatty acid elongation (ko00062), arachidonic acid metabolism (ko00590), linoleic acid metabolism (ko00591), alpha-linolenic acid metabolism (ko00592), and bile secretion (ko04976) (Figure 6). In fact, there are studies reporting that these pathways play an important role in physiological activities of the host. For instance, bile acid is an amphiphilic molecule with strong surface activity (Yang et al., 2020), which can emulsify fat into chylomicrons to increase the contact area between lipase and fat, and facilitates fat digestion and reduces autologous fat catabolism (Velazquez-Villegas et al., 2018; Hu et al., 2019). Alpha-linolenic acid, linoleic acid, and arachidonic acid are essential fatty acids that animals cannot be synthesized by the body and must come from food (Di Pasquale, 2009; Martin et al., 2016). These results indicate that seasonal differences in these pathways may mainly be due to the differences in nutrient intakes.
Noticeably, in season groups with significant differences in nutrient intakes, besides the pathway associated with macronutrient metabolism, we observed enrichment of multiple secondary metabolite biosynthesis pathways such as Phenylpropanoid biosynthesis, Flavone and flavonol biosynthesis, Flavonoid biosynthesis, and Diarylheptanoid and gingerol biosynthesis. As signals to gut microbes, microbial diet-based metabolites or small molecules are key mediators that affect physiological processes in the host (Koh et al., 2016). They can activate or inhibit endogenous signaling pathways, or act as a source of nutrients for host cells (Sonnenburg and Bäckhed, 2016). The biosynthesis of phenylpropane begins with the shikimate pathway, which initially breaks down glucose by the combined action of the glycolysis and pentose phosphate pathway to produce phosphoenolpyruvate and erythrose-4-phosphate of the synthetic initiation metabolite 3-deoxy-D-arabino-heptulosonic acid 7-phosphate (DAHP) (Chen et al., 2016). Flavone, flavonol, and flavonoid metabolites all appear as intermediates in phenylpropanoid biosynthesis pathway (Dong and Lin, 2021). In addition, key enzymes for the synthesis of resveratrol (stilbenoid), diarylheptanoid, and gingerol are also the central nodes of the phenylpropane pathway (Yin et al., 2022). Therefore, the enrichment of these pathways may be related to the reutilization of host-ingested carbohydrates by gut microbes. This implies that the intake of macronutrients exceeds the digestibility during seasons when foods were abundant. The macronutrients escape primary digestion and become a substrate for microbial metabolism to produce fermentation by-products and affect host physiological health.
The results of WGCNA analysis indicate that the green module has the highest correlation with fat, protein, and carbohydrate, and it also has the most complex network relationships. Twenty out of 22 candidate members of hub OTUs belong to Clostridiales and two belong to Bacteroidales. A total of five families have been annotated, namely Ruminococcaceae, Lachnospiraceae, Muribaculaceae, Peptococcaceae, and family_XIII. Among them, candidate members of Ruminococcaceae (OTU_472, OTU_81, OTU_339, OTU_570, and OTU_598) show consistent inter-seasonal trends with the energy provided by carbohydrates in natural foods in terms of their abundance, this indicates that these OTUs also have the lowest abundance in spring when carbohydrate intake was lowest. Previous studies have proved that Ruminococcaceae can degrade a variety of polysaccharides and dietary fibers. They are also the producers of short-chain fatty acids (SCFAs) (Scott et al., 2008; Louis and Flint, 2009; Hooda et al., 2012). Our assumptions about the relationship between nutrients and gut microbes are consistent with these findings. OTU_472, OTU_2009, OTU_226, OTU_81, OTU_67, and OTU_349 have largest fluctuation with changes in nutrients intakes. These 6 hub OTUs were annotated to Ruminococcaceae, Lachnospiraceae, and Muribaculaceae. They were at the core of the green module and largely affect the network structure of the co-occurrence bacterial taxa network pf the green module. This could be used as an important indicator to assess the gut nutrient absorption of the golden snub-nosed monkeys.
## The evolution of host adaptation
Based on the present finding that macronutrients are responsible for the changes in golden snub-nosed monkeys’ gut microbiota, we consider this is an important mechanism that helps them survive and increase fitness. This can be inferred from the great number of gut microbes and metabolic pathways annotated in the study. In our study, the gut microbiotas of golden snub-nosed monkeys were annotated to 38 phyla, 140 orders, 352 genera, and 395 metabolic pathways. We refer to previous studies that found golden snub-nosed monkeys have more types of gut microbes and metabolic pathways compared to mammals with relatively homogeneous or food-specific diets such as red pandas (Ailurus fulgens) (Kong et al., 2014), koalas (Phascolarctos cinereus) (Barker et al., 2013), amur tigers (*Panthera tigris* altaica) (Ning et al., 2020), and musk deers (Moschus chrysogaster) (Sun et al., 2019). This is unlikely due to the sequencing depth because the Good’s coverage of each bacterial community was >$97\%$. Golden snub-nosed monkeys rely on microbiota functions to obtain sufficient nutrients from foods to cope with the harsh living conditions and variable food types (Liu et al., 2022). We infer that the gut microbiota of golden snub-nosed monkeys has gradually become more diverse and complex during their co-evolution with their hosts to stabilize the host nutrient intakes under seasonal shifts of the diet. The gut microbiota helps the host adapt to broader dietary by enabling the host to digest multiple food types and obtain sufficient nutrients to meet its survival needs.
## Conclusion
Golden snub-nosed monkeys exist significant difference in food consumed and nutrients intake among seasons that the three macronutrients intake showed a consistent trend that they are higher in summer and autumn and lower in spring and winter. We found seasonal dietary differences caused the macronutrient variation is the main reason for seasonal shifts of gut microbiota. Particularly, phyla Firmicutes, Bacteroidetes, and Proteobacteria are significantly dominant in all samples, but the ratio of Firmicutes to Bacteroidetes was correspondingly highest in spring when nutrient intakes were lowest per metabolic body weight. The dominant genera also showed the same seasonal trends: Methanogens and Ruminococcus, which promote nutrient intake efficiency increased in spring when nutrient intakes were lowest. In autumn when high-carbohydrate and high-fat diets were consumed, Prevotella that digest complex polysaccharides had a high abundance. These results demonstrated that gut microbes through microbial metabolic functions help the host to compensate for the insufficient macronutrients intake.
## Data availability statement
The data presented in the study are deposited in the https://www.ncbi.nlm.nih.gov/bioproject/929077, repository accession number PRJNA929077.
## Ethics statement
The animal study was reviewed and approved by the Ethics Committee of Northwest University.
## Author contributions
SG conceived the project idea and designed the study. HF and SH collected the samples and performed the experiments. YL, YY, and ZW performed the bioinformatic analysis of the sequencing data. YL and YY wrote the manuscript and analyzed and interpreted the data. YL, YY, SJ, and SG edited the manuscript. All authors read and approved the final manuscript.
## Funding
This project was sponsored by the Major International Joint Research Program of Natural Science Foundation of China (NSFC) [32220103002], research programs of NSFC [31872247], the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB 31020302), and the Qinghai Province High-level Innovative “Thousand Talents” Program.
## Conflict of interest
Author GD was employed by Guangdong Chimelong Group Co., Ltd.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1126189/full#supplementary-material
## References
1. Amato K. R., Sanders J. G., Song S. J., Nute M., Metcalf J. L., Thompson L. R.. **Evolutionary trends in host physiology outweigh dietary niche in structuring primate gut microbiomes**. *ISME J.* (2019) **13** 576-587. DOI: 10.1038/s41396-018-0175-0
2. Baniel A., Amato K. R., Beehner J. C., Bergman T. J., Mercer A., Perlman R. F.. **Seasonal shifts in the gut microbiome indicate plastic responses to diet in wild geladas**. *Microbiome* (2021) **9** 26. DOI: 10.1186/s40168-020-00977-9
3. Barker C. J., Gillett A., Polkinghorne A., Timms P.. **Investigation of the koala (**. *Vet. Microbiol.* (2013) **167** 554-564. DOI: 10.1016/j.vetmic.2013.08.025
4. Basseri R. J., Basseri B., Pimentel M., Chong K., Youdim A., Low K.. **Intestinal methane production in obese individuals is associated with a higher body mass index**. *Gastroenterol. Hepatol.* (2012) **8** 22-28
5. Canfora E. E., Jocken J. W., Blaak E. E.. **Short-chain fatty acids in control of body weight and insulin sensitivity**. *Nat. Rev. Endocrinol.* (2015) **11** 577-591. DOI: 10.1038/nrendo.2015.128
6. Chapman C. A., Chapman L. J., Rode K. D., Hauck E. M., Mcdowell L. R.. **Variation in the nutritional value of primate foods: among trees, time periods, and areas**. *Int. J. Primatol.* (2003) **24** 317-333. DOI: 10.1023/A:1023049200150
7. Chen Q., Man C., Li D., Tan H., Xie Y., Huang J.. **Arogenate dehydratase isoforms differentially regulate anthocyanin biosynthesis in**. *Mol. Plant* (2016) **9** 1609-1619. DOI: 10.1016/j.molp.2016.09.010
8. Devkota S., Wang Y., Musch M. W., Leone V., Fehlner-Peach H., Nadimpalli A.. **Dietary-fat-induced taurocholic acid promotes pathobiont expansion and colitis in Il10**. *Nature* (2012) **487** 104-108. DOI: 10.1038/nature11225
9. Di Pasquale M. G.. **The essentials of essential fatty acids**. *J. Diet Suppl.* (2009) **6** 143-161. DOI: 10.1080/19390210902861841
10. Dong N. Q., Lin H. X.. **Contribution of phenylpropanoid metabolism to plant development and plant-environment interactions**. *J. Integr. Plant Biol.* (2021) **63** 180-209. DOI: 10.1111/jipb.13054
11. Fleming T. H., John Kress W.. **A brief history of fruits and frugivores**. *Acta Oecol.* (2011) **37** 521-530. DOI: 10.1016/j.actao.2011.01.016
12. Flint H. J., Scott K. P., Duncan S. H., Louis P., Forano E.. **Microbial degradation of complex carbohydrates in the gut**. *Gut Microbes* (2012) **3** 289-306. DOI: 10.4161/gmic.19897
13. Grześkowiak Ł., Endo A., Beasley S., Salminen S.. **Microbiota and probiotics in canine and feline welfare**. *Anaerobe* (2015) **34** 14-23. DOI: 10.1016/j.anaerobe.2015.04.002
14. Guo L., Guan Q., Duan W., Ren Y., Zhang X. J., Xu H. Y.. **Dietary goji shapes the gut microbiota to prevent the liver injury induced by acute alcohol intake**. *Front. Nutr.* (2022) **9** 929776. DOI: 10.3389/fnut.2022.929776
15. Guo S. T., Hou R., Garber P. A., Raubenheimer D., Righini N., Ji W. H.. **Nutrient-specific compensation for seasonal cold stress in a free-ranging temperate colobine monkey**. *Funct. Ecol.* (2018) **32** 2170-2180. DOI: 10.1111/1365-2435.13134
16. Guo S. T., Li B. G., Watanabe K.. **Diet and activity budget of**. *Primates* (2007) **48** 268-276. DOI: 10.1007/s10329-007-0048-z
17. Henao-Mejia J., Elinav E., Jin C., Hao L., Mehal W. Z., Strowig T.. **Inflammasome-mediated dysbiosis regulates progression of NAFLD and obesity**. *Nature* (2012) **482** 179-185. DOI: 10.1038/nature10809
18. Hicks A. L., Lee K. J., Couto-Rodriguez M., Patel J., Sinha R., Guo C.. **Gut microbiomes of wild great apes fluctuate seasonally in response to diet**. *Nat. Commun* (2018) **9** 1786. DOI: 10.1038/s41467-018-04204-w
19. Hooda S., Boler B. M., Serao M. C., Brulc J. M., Staeger M. A., Boileau T. W.. **454 pyrosequencing reveals a shift in fecal microbiota of healthy adult men consuming polydextrose or soluble corn fiber**. *J. Nutr.* (2012) **142** 1259-1265. DOI: 10.3945/jn.112.158766
20. Hooper L. V., Littman D. R., Macpherson A. J.. **Interactions between the microbiota and the immune system**. *Science* (2012) **336** 1268-1273. DOI: 10.1126/science.1223490
21. Hou R., Chapman C. A., Rothman J. M., Zhang H., Huang K., Guo S.. **The geometry of resource constraint: an empirical study of the golden snub-nosed monkey**. *J. Anim. Ecol.* (2021) **90** 751-765. DOI: 10.1111/1365-2656.13408
22. Hou R., He S., Wu F., Chapman C. A., Pan R., Garber P. A.. **Seasonal variation in diet and nutrition of the northern-most population of**. *Am. J. Primatol.* (2018) **80** e22755. DOI: 10.1002/ajp.22755
23. Hu M. M., He W. R., Gao P., Yang Q., He K., Cao L. B.. **Virus-induced accumulation of intracellular bile acids activates the TGR5-β-arrestin-SRC axis to enable innate antiviral immunity**. *Cell Res.* (2019) **29** 193-205. DOI: 10.1038/s41422-018-0136-1
24. Huang G., Wang L., Li J., Hou R., Wang M., Wang Z.. **Seasonal shift of the gut microbiome synchronizes host peripheral circadian rhythm for physiological adaptation to a low-fat diet in the giant panda**. *Cell Rep.* (2022) **38** 110203. DOI: 10.1016/j.celrep.2021.110203
25. Jung H. J. G.. **Nutritional ecology of the ruminant, second edition**. *J. Nutr.* (1995) **125** 1025. DOI: 10.1093/jn/125.4.1025
26. Junge R. E., Williams C. V., Campbell J.. **Nutrition and behavior of lemurs**. *Vet. Clin. North Am. Exot. Anim. Pract.* (2009) **12** 339-348. DOI: 10.1016/j.cvex.2009.01.011
27. Kanehisa M., Goto S.. **KEGG: Kyoto encyclopedia of genes and genomes**. *Nucleic Acids Res.* (2000) **28** 27-30. DOI: 10.1093/nar/28.1.27
28. Kengen S. W., De Bok F. A., Van Loo N. D., Dijkema C., Stams A. J., De Vos W. M.. **Evidence for the operation of a novel Embden-Meyerhof pathway that involves ADP-dependent kinases during sugar fermentation by Pyrococcus furiosus**. *J. Biol. Chem.* (1994) **269** 17537-17541. DOI: 10.1016/S0021-9258(17)32474-2
29. Koh A., De Vadder F., Kovatcheva-Datchary P., Bäckhed F.. **From dietary fiber to host physiology: short-chain fatty acids as key bacterial metabolites**. *Cells* (2016) **165** 1332-1345. DOI: 10.1016/j.cell.2016.05.041
30. Kong F., Zhao J., Han S., Zeng B., Yang J., Si X.. **Characterization of the gut microbiota in the red panda (**. *PLoS One* (2014) **9** e87885. DOI: 10.1371/journal.pone.0087885
31. Ley R. E., Hamady M., Lozupone C., Turnbaugh P. J., Ramey R. R., Bircher J. S.. **Evolution of mammals and their gut microbes**. *Science* (2008) **320** 1647-1651. DOI: 10.1126/science.1155725
32. Ley R. E., Turnbaugh P. J., Klein S., Gordon J. I.. **Human gut microbes associated with obesity**. *Nature* (2006) **444** 1022-1023. DOI: 10.1038/4441022a
33. Liu R., Amato K., Hou R., Gomez A., Dunn D. W., Zhang J.. **Specialized digestive adaptations within the hindgut of a colobine monkey**. *Innovation* (2022) **3** 100207. DOI: 10.1016/j.xinn.2022.100207
34. Louis P., Flint H. J.. **Diversity, metabolism and microbial ecology of butyrate-producing bacteria from the human large intestine**. *FEMS Microbiol. Lett.* (2009) **294** 1-8. DOI: 10.1111/j.1574-6968.2009.01514.x
35. Lv X. C., Guo W. L., Li L., Yu X. D., Liu B.. **Polysaccharide peptides from Ganoderma lucidum ameliorate lipid metabolic disorders and gut microbiota dysbiosis in high-fat diet-fed rats**. *J. Funct. Foods* (2019) **57** 48-58. DOI: 10.1016/j.jff.2019.03.043
36. Lyu Z., Wang L., Wang J., Wang Z., Zhang S., Wang J.. **Oat bran and wheat bran impact net energy by shaping microbial communities and fermentation products in pigs fed diets with or without xylanase**. *J. Anim. Sci. Biotechnol.* (2020) **11** 99. DOI: 10.1186/s40104-020-00505-7
37. Machovsky-Capuska G. E., Priddel D., Leong P. H. W., Jones P., Carlile N., Shannon L.. **Coupling bio-logging with nutritional geometry to reveal novel insights into the foraging behaviour of a plunge-diving marine predator**. *New Zeal. J. Mar. Fresh.* (2016) **50** 418-432. DOI: 10.1080/00288330.2016.1152981
38. Markle J. G., Frank D. N., Mortin-Toth S., Robertson C. E., Feazel L. M., Rolle-Kampczyk U.. **Sex differences in the gut microbiome drive hormone-dependent regulation of autoimmunity**. *Science* (2013) **339** 1084-1088. DOI: 10.1126/science.1233521
39. Martin S. A., Brash A. R., Murphy R. C.. **The discovery and early structural studies of arachidonic acid**. *J. Lipid Res.* (2016) **57** 1126-1132. DOI: 10.1194/jlr.R068072
40. Mizrahi I., Wallace R. J., Moraïs S.. **The rumen microbiome: balancing food security and environmental impacts**. *Nat. Rev. Microbiol.* (2021) **19** 553-566. DOI: 10.1038/s41579-021-00543-6
41. Morgan X. C., Tickle T. L., Sokol H., Gevers D., Devaney K. L., Ward D. V.. **Dysfunction of the intestinal microbiome in inflammatory bowel disease and treatment**. *Genome Biol.* (2012) **13** R79. DOI: 10.1186/gb-2012-13-9-r79
42. Morotomi M., Nagai F., Watanabe Y.. **Description of christensenella minuta gen. Nov., sp. nov., isolated from human faeces, which forms a distinct branch in the order Clostridiales, and proposal of Christensenellaceae fam. Nov**. *Int. J. Syst. Evol. Micr.* (2012) **62** 144-149. DOI: 10.1099/ijs.0.026989-0
43. Ning Y., Qi J., Dobbins M. T., Liang X., Wang J., Chen S.. **Comparative analysis of microbial community structure and function in the gut of wild and captive Amur tiger**. *Front. Microbiol.* (2020) **11** 1665. DOI: 10.3389/fmicb.2020.01665
44. Oh S. F., Praveena T., Song H., Yoo J. S., Jung D. J., Erturk-Hasdemir D.. **Host immunomodulatory lipids created by symbionts from dietary amino acids**. *Nature* (2021) **600** 302-307. DOI: 10.1038/s41586-021-04083-0
45. Raubenheimer D., Machovsky-Capuska G. E., Chapman C. A., Rothman J. M.. **Geometry of nutrition in field studies: an illustration using wild primates**. *Oecologia* (2015) **177** 223-234. DOI: 10.1007/s00442-014-3142-0
46. Raubenheimer D., Simpson S. J., Mayntz D.. **Nutrition, ecology and nutritional ecology: toward an integrated framework**. *Funct. Ecol.* (2009) **23** 4-16. DOI: 10.1111/j.1365-2435.2009.01522.x
47. Ren T., Boutin S., Humphries M. M., Dantzer B., Gorrell J. C., Coltman D. W.. **Seasonal, spatial, and maternal effects on gut microbiome in wild red squirrels**. *Microbiome* (2017) **5** 163. DOI: 10.1186/s40168-017-0382-3
48. Rodrigues R. R., Gurung M., Li Z., García-Jaramillo M., Greer R., Gaulke C.. **Transkingdom interactions between**. *Nat. Commun.* (2021) **12** 101. DOI: 10.1038/s41467-020-20313-x
49. Rothman J. M., Chapman C. A., Van Soest P. J.. **Methods in primate nutritional ecology: a user’s guide**. *Int. J. Primatol.* (2012) **33** 542-566. DOI: 10.1007/s10764-011-9568-x
50. Rothman J. M., Dierenfeld E. S., Hintz H. F., Pell A. N.. **Nutritional quality of gorilla diets: consequences of age, sex, and season**. *Oecologia* (2008) **155** 111-122. DOI: 10.1007/s00442-007-0901-1
51. Rothman J. M., Plumptre A. J., Dierenfeld E. S., Pell A. N.. **Nutritional composition of the diet of the gorilla (**. *J. Trop. Ecol.* (2007) **23** 673-682. DOI: 10.1017/S0266467407004555
52. Rothman J. M., Raubenheimer D., Chapman C. A.. **Nutritional geometry: gorillas prioritize non-protein energy while consuming surplus protein**. *Biol. Lett.* (2011) **7** 847-849. DOI: 10.1098/rsbl.2011.0321
53. Ruohonen K., Simpson S. J., Raubenheimer D.. **A new approach to diet optimisation: A re-analysis using European whitefish (**. *Aquaculture* (2007) **267** 147-156. DOI: 10.1016/j.aquaculture.2007.02.051
54. Samuel B. S., Gordon J. I.. **A humanized gnotobiotic mouse model of host-archaeal-bacterial mutualism**. *Proc. Natl. Acad. Sci. U.S.A.* (2006) **103** 10011-10016. DOI: 10.1073/pnas.0602187103
55. Samuel B. S., Hansen E. E., Manchester J. K., Coutinho P. M., Henrissat B., Fulton R.. **Genomic and metabolic adaptations of**. *Proc. Natl. Acad. Sci. U. S. A.* (2007) **104** 10643-10648. DOI: 10.1073/pnas.0704189104
56. Scott K. P., Martin J. C., Mrazek J., Flint H. J.. **Transfer of conjugative elements from rumen and human**. *Appl. Environ. Microb.* (2008) **74** 3915-3917. DOI: 10.1128/AEM.02807-07
57. Shin N. R., Whon T. W., Bae J. W.. **Proteobacteria: microbial signature of dysbiosis in gut microbiota**. *Trends Biotechnol.* (2015) **33** 496-503. DOI: 10.1016/j.tibtech.2015.06.011
58. Simpson S. J., Raubenheimer D.. *The Nature of Nutrition: A Unifying Framework From Animal Adaptation to Human Obesity* (2012)
59. Smits S. A., Leach J., Sonnenburg E. D., Gonzalez C. G., Lichtman J. S., Reid G.. **Seasonal cycling in the gut microbiome of the Hadza hunter-gatherers of Tanzania**. *Science* (2017) **357** 802-806. DOI: 10.1126/science.aan4834
60. Sonnenburg J. L., Bäckhed F.. **Diet–microbiota interactions as moderators of human metabolism**. *Nature* (2016) **535** 56-64. DOI: 10.1038/nature18846
61. Sun Y. W., Sun Y. J., Shi Z. H., Liu Z. S., Zhao C., Lu T. F.. **Gut microbiota of wild and captive alpine musk deer (**. *Front. Microbiol* (2019) **10** 3156. DOI: 10.3389/fmicb.2019.03156
62. Velazquez-Villegas L. A., Perino A., Lemos V., Zietak M., Nomura M., Pols T. W. H.. **TGR5 signalling promotes mitochondrial fission and beige remodelling of white adipose tissue**. *Nat. Commun.* (2018) **9** 245. DOI: 10.1038/s41467-017-02068-0
63. Wang X., Tsai T., Wei X., Zuo B., Davis E., Rehberger T.. **Effect of lactylate and bacillus subtilis on growth performance, peripheral blood cell profile, and gut microbiota of nursery pigs**. *Microorganisms* (2021) **9** 803. DOI: 10.3390/microorganisms9040803
64. Yang F., Mao C., Guo L., Lin J., Ming Q., Xiao P.. **Structural basis of GPBAR activation and bile acid recognition**. *Nature* (2020) **587** 499-504. DOI: 10.1038/s41586-020-2569-1
65. Yin W., Wang X., Liu H., Wang Y., Nocker S., Tu M.. **Overexpression of**. *Hortic. Res* (2022) **9** uhab064. DOI: 10.1093/hr/uhab064
66. Zhang X., Coker O. O., Chu E. S., Fu K., Lau H. C. H., Wang Y. X.. **Dietary cholesterol drives fatty liver-associated liver cancer by modulating gut microbiota and metabolites**. *Gut* (2021a) **70** 761-774. DOI: 10.1136/gutjnl-2019-319664
67. Zhang H., Dibaise J. K., Zuccolo A., Kudrna D., Braidotti M., Yu Y.. **Human gut microbiota in obesity and after gastric bypass**. *Proc. Natl. Acad. Sci. U. S. A.* (2009) **106** 2365-2370. DOI: 10.1073/pnas.0812600106
68. Zhang K., He C., Xu Y., Zhang C., Li C., Jing X.. **Taxonomic and functional adaption of the gastrointestinal microbiome of goats kept at high altitude (4800 m) under intensive or extensive rearing conditions**. *FEMS Microbiol. Ecol.* (2021c) **97** 3. DOI: 10.1093/femsec/fiab009
69. Zhang J., Meng H., Kong X., Cheng X., Ma T., He H.. **Combined effects of polyethylene and organic contaminant on zebrafish (**. *Environ. Pollut.* (2021b) **277** 116767. DOI: 10.1016/j.envpol.2021.116767
70. Zhu Z. L., Sun Y. W., Zhu F., Liu Z. S., Pan R. L., Teng, L. W.. **Seasonal variation and sexual dimorphism of the microbiota in wild blue sheep (Pseudois nayaur)**. *Front. Microbiol* (2020) **11** 1260. DOI: 10.3389/fmicb.2020.01260
|
---
title: 'Prevalence of physical health conditions and health risk behaviours in people
with severe mental illness in South Asia: multi-country cross-sectional survey'
authors:
- Gerardo A. Zavala
- Asiful Haidar-Chowdhury
- Krishna Prasad-Muliyala
- Kavindu Appuhamy
- Faiza Aslam
- Rumana Huque
- Humaira Khalid
- Pratima Murthy
- Asad T. Nizami
- Sukanya Rajan
- David Shiers
- Najma Siddiqi
- Kamran Siddiqi
- Jan R. Boehnke
journal: BJPsych Open
year: 2023
pmcid: PMC9970179
doi: 10.1192/bjo.2023.12
license: CC BY 4.0
---
# Prevalence of physical health conditions and health risk behaviours in people with severe mental illness in South Asia: multi-country cross-sectional survey
## Abstract
### Background
People with severe mental illness (SMI) die earlier than the general population, primarily because of physical disorders.
### Aims
We estimated the prevalence of physical health conditions, health risk behaviours, access to healthcare and health risk modification advice in people with SMI in Bangladesh, India and Pakistan, and compared results with the general population.
### Method
We conducted a cross-sectional survey in adults with SMI attending mental hospitals in Bangladesh, India and Pakistan. Data were collected on non-communicable diseases, their risk factors, health risk behaviours, treatments, health risk modification advice, common mental disorders, health-related quality of life and infectious diseases. We performed a descriptive analysis and compared our findings with the general population in the World Health Organization (WHO) ‘STEPwise Approach to Surveillance of NCDs’ reports.
### Results
We recruited 3989 participants with SMI, of which $11\%$ had diabetes, $23.3\%$ had hypertension or high blood pressure and $46.3\%$ had overweight or obesity. We found that $70.8\%$ of participants with diabetes, high blood pressure and hypercholesterolemia were previously undiagnosed; of those diagnosed, only around half were receiving treatment. A total of $47\%$ of men and $14\%$ of women used tobacco; $45.6\%$ and $89.1\%$ of participants did not meet WHO recommendations for physical activity and fruit and vegetable intake, respectively. Compared with the general population, people with SMI were more likely to have diabetes, hypercholesterolemia and overweight or obesity, and less likely to receive tobacco cessation and weight management advice.
### Conclusions
We found significant gaps in detection, prevention and treatment of non-communicable diseases and their risk factors in people with SMI.
## Severe mental Illness
Severe mental illnesses (SMIs) are conditions such as schizophrenia, bipolar disorder and psychotic depression that are debilitating, persistent and associated with serious functional impairment. People with SMI die on average 10–20 years earlier than the general population, and this ‘mortality gap’ is widening.1 Although suicide accounts for $15\%$ of deaths, an estimated $80\%$ of the observed premature mortality is attributable to physical disorders (physical multimorbidity), most commonly non-communicable diseases (NCDs).2
## Physical health in people with SMI
The excess disease burden from physical multimorbidity in people with SMI may be explained by a combination of factors associated with these mental disorders, including clustering of and predisposition to health risk behaviours (e.g. tobacco and alcohol use, lack of physical activity and poor diet), side-effects of medication, social determinants of poor health (e.g. stigma and poverty) and barriers to accessing healthcare.3 Our current understanding of the distribution and determinants of physical multimorbidity in people with SMI is based mostly on evidence from high-income countries. A few small studies from low- and middle-income countries (LMICs) show similar patterns, but with an even shorter life expectancy and higher mortality for people with SMI.4,5 These studies indicate that physical multimorbidity in SMI may be at least as much of a challenge in LMICs as in high-income countries.1 In South Asia, the prevalence of both mental disorders and NCDs has been increasing rapidly.6 This increase is coupled with limited access to essential health services and a widespread neglect of the physical health needs of people with SMI by policy makers and healthcare services.7 The overall burden of disease resulting from physical multimorbidity in this population is, therefore, likely to be high and is set to rise further, with a corresponding increase in within-country and global health inequalities. Despite these concerns, there is a lack of empirical studies originating in South Asia on the distribution and determinants of physical multimorbidity in people with SMI.8 Addressing multimorbidity in LMICs is a global priority, recognised in global policies to help achieve the United Nations Sustainable Development Goals.9 A detailed understanding of the prevalence of physical multimorbidity and current access to health advice and treatments for physical disorders in people with SMI in LMICs can inform appropriate service provision and contribute to achieving these goals.
## Aims
In the current study, we aim to (a) estimate the prevalence of physical health conditions and health risk behaviours; (b) assess access to physical healthcare and health risk modification advice in people with SMI attending mental health services in Bangladesh, India and Pakistan; and (c) compare the findings with those of the general population.
## Method
We conducted a cross-sectional survey of patients with a clinical diagnosis of SMI recruited at three national specialist mental health institutions in South Asia: the National Institute of Mental Health in Dhaka, Bangladesh; the National Institute of Mental Health and Neurosciences in Bangalore, India and the Institute of Psychiatry in Rawalpindi, Pakistan. Further details of the methods are reported in the published protocol,10 and are summarised below.
## Sample size
We aimed to build as large a sample as possible within the resources available over the study period, with an initial target of 1500 participants at each site. As an indicative example of precision to address some of the key research questions, we used the example of diabetes. For investigating the prevalence of type 2 diabetes, assuming a prevalence estimate of $10\%$, 857 participants per country would provide a precision of ±$2\%$ ($95\%$ confidence interval).
## Eligibility
Consenting adults (aged ≥18 years) with a clinical diagnosis of SMI as defined by the ICD-10 (schizophrenia, schizotypal and delusional disorders (F20–F29); bipolar affective disorder (F30, F31) or severe depression with psychotic symptoms (F32.3, F33.3)), and able to provide informed consent as assessed by the treating clinician, were eligible.
## Confirmation of SMI diagnosis
To increase standardisation across sites and alignment with other studies, each SMI diagnosis was confirmed by trained researchers using the Mini-International Neuropsychiatric Interview (MINI) version 6.0.11 The MINI is a short diagnostic structured interview for mental disorders, designed to allow administration by non-specialists.
## Recruitment of participants
We used stratified random sampling to recruit a sample comprising $80\%$ out-patients and $20\%$ in-patients. This reflects the flow of in- and out-patients in the three mental health hospitals on any given day, which was assessed and protocolised in each site before the data collection.
## Patient and public involvement
A community panel comprising patients, caregivers and advocacy group members ensured community, patient and public involvement. The panel reviewed and piloted the planned survey questionnaire, and advised on its feasibility.
## Data collection
We conducted a face-to-face survey with tablets (Qualtrics, Utah, USA; https://www.qualtrics.com/) to collect information about physical disorders, mental health, health risk behaviours, health-related quality of life, health risk behaviour advice and healthcare utilisation, using, wherever available, validated instruments as described below. The survey was translated into Bangla, Hindi, Kannada and Urdu. Interviewers (including males and females, to accommodate participant preference) used regional dialects where required, consistent with usual clinical practice in these settings. Data were collected between July 2019 and December 2021.
## STEPwise Approach to Surveillance of NCDs
We used the World Health Organization (WHO) STEPwise Approach to Surveillance of NCDs (STEPS) instrument version 3.2 to collect information about NCDs, associated risk factors and behaviours, access to physical healthcare and health risk modification advice.12 STEPS is an international standardised tool that has already been translated, used and validated in the general population in Bangladesh, India and Pakistan, and therefore allows comparisons with the general population within and between countries.13,14 The STEPS survey includes the use of show cards with culturally relevant examples used to aid respondents in classifying health risk behaviours. Categorisation of health conditions and risk behaviours followed the WHO guidelines.15 The STEPS module for NCDs was used to ask participants about medically diagnosed type 2 diabetes, raised blood pressure, heart disease and hypercholesterolemia, and treatments advised by a healthcare worker for these conditions (such as medication and dietary, weight management, smoking cessation or physical activity advice). Questions about lung disease, hepatitis B and C, syphilis, tuberculosis and HIV (which are not part of the STEPS survey) were asked in the same format as for the other chronic physical conditions.
## Health risk behaviours
Current or past use of smoking or smokeless tobacco was recorded.15 The alcohol module was used to categorise participants into lifetime abstainers, abstainers in the past 12 months and current users of alcohol;15 and the diet module was used to record the number of days that respondents consumed fruit and vegetables in a typical week, the number of servings consumed on average per day, and adherence to the WHO recommendations of at least five fruits and vegetables per day.16 The physical activity module was used to record activity for transport purposes (e.g. walking, cycling), vigorous and moderate activity at work, vigorous and moderate activity in leisure time and time spent sitting. In addition, risk behaviours related to sexually transmitted diseases, including multiple sexual partners, unprotected sexual contact and use of injectable drugs, were assessed with three questions adapted from the ten-item HIV Risk Screening Instrument.17 People with SMI were more likely not to meet recommendations for physical activity (Bangladesh: odds ratio 5.4, $95\%$ CI 4.8–6.1, $P \leq 0.001$; India: odds ratio 2.8, $95\%$ CI 2.5–3.2, $P \leq 0.001$; Pakistan: odds ratio 2.3, $95\%$ CI 2.0–2.6, $P \leq 0.001$). However, they were less likely not to meet WHO recommendations for fruit and vegetable intake (Bangladesh: odds ratio 0.6, $95\%$ CI 0.5–0.7, $P \leq 0.001$; India: odds ratio 0.3, $95\%$ CI 0.2–0.4, $P \leq 0.001$; Pakistan: odds ratio 0.5, $95\%$ CI 0.4–0.6, $P \leq 0.001$).
Men with SMI in Bangladesh (odds ratio 0.7, $95\%$ CI 0.6–0.9, $$P \leq 0.001$$) and India (odds ratio 0.4, $95\%$ CI 0.3–0.5, $P \leq 0.001$) were less likely to use tobacco products, whereas the opposite was found in Pakistan (odds ratio 2.2, $95\%$ CI 1.8–2.7, $P \leq 0.001$).
## Physical measurements
Blood pressure was taken with an automated blood pressure measuring instrument (OMRON) following instructions in the WHO STEPS surveillance manual; the average of the second and third readings was used for analysis.15 High blood pressure was defined as a measurement of >$\frac{140}{90}$ mmHg.15 Height, weight and waist circumference were measured for all participants except pregnant women. All measurements were taken in duplicate and the average of the two values was calculated, following the protocols set out in the WHO STEPS surveillance manual.15 We calculated the body mass index (BMI) and classified participants according to the WHO classification: underweight (BMI < 18.49 kg/m2), normal weight (BMI = 18.5–24.9 kg/m2), overweight (BMI = 25–29.9 kg/m2) or obese (BMI ≥ 30 kg/m2). Abdominal obesity was defined as a waist circumference of ≥94 cm for males and ≥80 cm for females.15
## Mental health
In addition to administering the MINI, we collected information relevant to the SMI diagnosis, including duration of illness and type and duration of treatments. The Patient Health Questionnaire-9 (PHQ-9) was used to measure the severity of depressive symptoms, and the Generalised Anxiety Disorder-7 (GAD-7) for severity of anxiety symptoms.
## Health-related quality of life
The EQ-5D-5L was used to measure health-related quality of life.18 We used the English, Urdu and Bangla validated versions, provided by EuroQol.
## Blood tests
A blood sample was taken from consenting participants for haemoglobin, glycated haemoglobin (HbA1c), lipid profile, thyroid function tests, liver function tests and creatinine. The cut-off for high HbA1c was according to the WHO definition of ≥$6.5\%$.19 The prevalence of high total triglycerides was defined as ≥180 mg/dL,20 high serum cholesterol was defined as a low-density lipoprotein cholesterol of ≥100 mg/dL.20 Anaemia was defined as haemoglobin ≤13 g/dL for males and ≤12 g/dL for females.21
## Statistical analysis
Quantitative data were summarised with descriptive statistics, with mean, s.d. and $95\%$ confidence interval for continuous data and counts, and percentage and $95\%$ confidence interval for categorical data. Overall and by site, we described the prevalence of chronic physical conditions; prevalence of risk factors (obesity, high blood pressure and hypercholesterolemia) and risk behaviours (poor diet, physical inactivity, tobacco and alcohol use); severity of common mental disorder symptoms (anxiety, depression) and health-related quality of life; and access to treatment for physical conditions and health risk modification advice.
To compare our findings with those in the latest STEPS reports from Bangladesh,22 India23 and Pakistan,14 we calculated weights by comparing the gender and age distribution reported in these STEPS surveys with the distribution in our data. Because of the multiple differences within countries in the operationalisation of socioeconomic status and the definition of rural and urban populations, we did not weight our sample for sociodemographic and geographic variables.
Weights were applied with the complex sample module in SPSS version 26.0 for Windows, and we calculated the odds of people with SMI having an NCD, related risk factors, engaging in health risk behaviours, being screened, being treated and receiving risk modification advice compared with the STEPS survey participants in Bangladesh, India and Pakistan,14,22,23 using Stata version 17.0 for Windows. Results were presented as odds ratios from cross-tabulations of STEPS and weighted survey data. Significance levels were adjusted via Bonferroni correction for multiple hypothesis testing (adjusted level $P \leq 0.006$).
## Ethics statement
Trained researchers provided verbal and written study information to patients and their relatives or caregivers, highlighting that participation was voluntary, the decision would not affect care and consent could be withdrawn at any stage without providing a reason. Written consent was obtained (a thumbprint was accepted where a signature could not be provided). No assessments were conducted where the patient appeared reluctant, even if consent had previously been obtained. The study was approved by the ethics committees of the Department of Health Sciences, University of York, UK (HSRGC-$\frac{3}{17}$); the Centre for Injury Prevention and Research, Bangladesh (CIPRB/ERC/2OI $\frac{8}{003}$); the Institute Ethics Committee, National Institute of Mental Health and Neurosciences, India (BEH.SC.DIV $\frac{20}{19}$); the Health Ministry Screening Committee, India (HMSC$\frac{12}{18}$); and the National Bioethics Committee, Pakistan (4-18/NBC-$\frac{413}{19}$). All study procedures complied with legislation and guidance for good practice governing the participation in research of people who may lack capacity (Mental Capacity Act (UK) 2005). Participants did not receive financial inducements to participate, but results of physical health measurements and blood tests were shared with them and with the treating clinician. This study is registered with the ISRCTN registry under identifier ISRCTN88485933 (https://doi.org/10.1186/ISRCTN88485933). All participants consented and signed an informed consent form.
## Results
We approached 5801 people with SMI in the three sites and 3989 ($58.8\%$) participated in the survey (1500 in Bangladesh, 1175 in India and 1314 in Pakistan). Most of the participants in Bangladesh ($94\%$) and Pakistan ($70\%$) were recruited before the COVID-19 pandemic (July 2019 and March 2020), and most of the participants in India ($86\%$) were recruited after the COVID-19 had begun (February 2021–Dec 2021). The details of participants that were not eligible are provided in Figure 1. Fig. 1Participant flow chart. MINI, Mini-International Neuropsychiatric Interview.
Participant characteristics are shown in Table 1. The proportion of in-patients was $20\%$ for Bangladesh and India, and $10\%$ for Pakistan. In-patients were more likely to be ineligible because of not having the capacity to sign an informed consent or answer the questionnaire. On average, $60.1\%$ of the sample was male, and the mean age was 35.8 years; the Bangladesh cohort was younger than the cohort in India and Pakistan. Almost a third ($32.0\%$) were unemployed, with a higher proportion in Bangladesh ($39.7\%$) than India ($19.4\%$) and Pakistan ($22.2\%$). About half ($50.8\%$) were educated only up to or less than primary school level. Table 1General characteristics of the participantsBangladesh ($$n = 1500$$)India ($$n = 1175$$)Pakistan ($$n = 1314$$)Overall ($$n = 3989$$)n (%) [$95\%$ CI]n (%) [$95\%$ CI]n (%) [$95\%$ CI]n (%) [$95\%$ CI]General characteristicsGender (female)585 (39.0) [36.6–41.5]527 (44.9) [42.0–47.7]518 (39.4) [36.8–42.1]1630 (40.9) [39.3–42.4]Age (years),a mean (s.d.)31.5 (10.8) [31.0–32.1]38.8 (11.2) [38.1–39.4]38.1 (12.3) [37.4–38.8]35.8 (11.4) [35.5–36.2]Education No formal education151 (10.1) [8.6–11.7]141 (12.0) [10.3–14.0]257 (19.6) [17.5–21.8]549 (13.8) [12.7–14.9] Primary842 (56.1) [53.6–58.6]401 (34.1) [31.5–36.9]234 (17.8) [15.8–20]1477 (37.0) [35.6–38.5] Secondary/higher507 (33.8) [31.4–36.2]632 (53.8) [50.9–56.6]821 (62.5) [59.8–65.1]1960 (49.1) [47.6–50.6] Refused to answerb(<$1\%$)(<$1\%$)(<$1\%$)(<$1\%$)Monthly household income over past year, USD,a mean [$95\%$ CI]224 [206–242]305 [252–357]198 [187–209]237 [221–253]Cement/concrete roof427 (28.5) [26.2–30.8]751 (63.9) [61.1–66.6]1034 (78.8) [76.5–80.9]2212 (55.5) [54.1–56.9]Electricity in the household1465 (97.7) [96.8–98.3]1171 (99.7) [99.1–99.9]1292 (100.0) [Not applicable]3928 (99.0) [99.7–99.3]Flush toilet in the household866 (57.7) [55.2–60.2]1046 [89] [87.1–90.7]1269 (100.0) [Not applicable]3181 (80.7) [79.5–81.7]Occupationb Non-government employee153 (10.2) [8.8–11.8]207 (17.6) [15.5–19.9]377 (28.7) [26.3–31.2]737 (18.5) [17.3–19.7] Government employee18 (1.2) [0.8–1.9]44 (3.7) [2.8–5.0]53 (4.0) [3.1–5.2]115 (2.9) [2.4–3.4] Self-employed252 (16.8) [15.0–18.8]271 (23.1) [20.7–25.6]68 (5.2) [4.1–6.5]591 (14.8) [13.8–15.9] Retiredb(<$2\%$)(<$2\%$)(<$2\%$)35 (0.9) [0.6–1.2] Non-paidb(<$2\%$)(<$2\%$)(<$2\%$)25 (0.6) [0.4–0.9] Student118 (7.9) [6.6–9.3]42 (3.6) [2.7–4.8]25 (1.9) [1.3–2.8]185 (4.6) [4–5.3] Homemaker345 (23.0) [20.9–25.2]371 (31.6) [29.0–34.3]466 (35.5) [32.9–38.1]1182 (29.6) [28.2–31.1] Unemployed (able to work)301 (20.1) [18.1–22.2]137 (11.7) [9.9–13.6]268 (20.4) [18.3–22.7]706 (17.7) [16.6–18.9] Unemployed (unable to work)294 (19.6) [17.7–21.7]90 (7.7) [6.3–9.3]23 (1.8) [1.2–2.6]407 (10.2) [9.3–11.1] Did not wish to answerb(<$1\%$)(<$1\%$)(<$1\%$)(<$1\%$)Currently married/living with partner818 (54.5) [52.0–57.0]711 (60.5) [57.7–63.3]747 (56.8) [54.2–59.5]2276 (57.1) [55.5–58.6]Severe mental illness (MINI) Bipolar disorder488 (32.5) [30.2–34.9]464 (39.5) [36.7–42.3]537 (40.9) [38.2–43.6]1489 (37.3) [35.8–38.8] Non-psychosis to affective psychosis935 (62.3) [59.8–64.8]673 (57.3) [54.4–60.1]176 (13.4) [11.6–15.3]1784 (44.7) [43.3–46.1] Major depressive disorder with psychotic features77 (5.1) [4.1–6.4]66 (5.6) [4.4–7.1]605 (46.0) [43.4–48.7]748 (18.8) [17.7–19.8]Patient setting In-patient313 (20.9) [18.9–23.0]264 (22.5) [20.2–24.9]122 (9.3) [7.8–11]699 (17.5) [16.4–18.7] Out-patient1187 (79.1) [77–81.1]911 (77.5) [75.1–79.8]1192 (90.7) [89.0–92.2]3290 (82.5) [81.3–83.6]Duration of SMI ≤2 years436 (29.1) [26.8–31.4]215 (18.3) [16.2–20.6]289 (22.0) [19.8–24.3]940 (23.6) [22.3–24.9] 3–5 years457 (30.5) [28.2–32.8]266 (22.6) [20.3–25.1]320 (24.4) [22.1–26.7]1043 (26.1) [24.8–27.5] 6–10 years332 (22.1) [20.1–24.3]299 (25.4) [23.0–28]299 (22.8) [20.6–25.1]930 (23.3) [22.0–24.7] >10 years271 (18.1) [16.2–20.1]359 (30.6) [28.0–33.3]399 (30.4) [27.9–32.9]1029 (25.8) [24.5–27.2] Do not know or do not remember(<$1\%$)(<$4\%$)(<$1\%$)(<$2\%$)On antipsychotic medication1463 (97.5) [96.6–98.2]1150 (97.9) [96.9–98.6]1253 (95.4) [94.1–96.4]3866 (96.9) [96.3–97.4]Mental healthSeverity of depressive symptoms PHQ-9 scorea10.7 (4.6) [10.4–10.9]5.8 (6.8) [5.4–6.2]12.9 (6.8) [12.5–13.2]10.0 (6.1) [9.8–10.2] None or minimal (0–4)121 (8.1) [6.8–9.6]665 (56.6) [53.7–59.4]177 (13.5) [11.7–15.4]963 (24.1) [23–25.3] Mild (5–9)487 (32.5) [30.1–34.9]224 (19.1) [16.9–21.4]271 (20.6) [18.5–22.9]982 (24.6) [23.3–26] Moderate or severe (≥10)892 (59.5) [57.0–61.9]286 (24.3) [22.0–26.9]866 (65.9) [63.3–68.4]2044 (51.2) [49.8–52.7]Severity of anxiety symptoms GAD-7 scorea8.1 (3.9) [7.9–8.3]4.6 (5.4) [4.2–4.9]9.9 (5.1) [9.6–10.2]7.6 (4.8) [7.5–7.8] None or minimal (0–4)275 (18.3) [16.5–20.4]729 (62.0) [59.2–64.8]201 (15.3) [13.4–17.3]1205 (30.2) [28.9–31.5] Mild (5–9)703 (46.9) [44.3–49.4]236 (20.1) [17.9–22.5]431 (32.8) [30.3–35.4]1370 (34.3) [32.9–35.8] Moderate or severe (≥10)522 (34.8) [32.4–37.2]210 (17.9) [15.8–20.2]682 (51.9) [49.2–54.6]1414 (35.4) [34.2–36.9]Health-related quality of life Visual analogue scalea69.9 (13.4) [69.2–70.6]76.4 (17.6) [75.4–77.4]54.9 (26.3) [53.4–56.3]66.9 (19.7) [66.3–67.5] Mobility471 (31.4) [29.1–33.8]252 (21.4) [19.2–23.9]708 (53.9) [51.2–56.6]1431 (35.9) [34.5–37.3] Self-care526 (35.1) [32.7–37.5]230 (19.6) [17.4–21.9]618 (47.0) [44.3–49.7]1374 (34.4) [33.0–35.9] Usual activities684 (45.6) [43.1–48.1]364 (31.0) [28.4–33.7]727 (55.3) [52.6–58]1775 (44.5) [43.0–46] Pain/discomfort993 (66.2) [63.8–68.6]443 (37.7) [35.0–40.5]1006 (76.6) [74.2–78.8]2442 (61.2) [59.8–62.6] Anxiety/depression1363 (90.9) [89.3–92.2]511 (43.5) [40.7–46.3]959 (73.0) [70.5–75.3]2833 (71.0) [69.7–72.3]Confidence intervals were calculated using bootstrap sampling procedure ($$n = 1000$$) for binomial and continuous variables and Goodman's method for multinomial proportions. MINI, Mini-International Neuropsychiatric Interview; SMI severe mental illness; PHQ-9, Patient Health Questionnaire-9; GAD-7, Generalised Anxiety Disorder-7.a. Values presented as mean (s.d.).b. Data not reported because of low numbers for statistical disclosure control.
## SMI, anxiety and depressive symptoms and health-related quality of life
The most common SMI diagnosis was non-affective psychosis ($44.7\%$), followed by bipolar disorder ($37.3\%$) and depression with psychotic symptoms ($18.8\%$). Non-affective psychosis was the most common diagnosis in Bangladesh ($62.3\%$) and India ($57.3\%$), whereas depression with psychotic symptoms was the most common diagnosis in Pakistan ($46.0\%$). Almost $97\%$ of participants were on antipsychotic medication.
A majority of participants reported having depressive ($75.8\%$) and anxiety ($69.7\%$) symptoms in the ‘moderate or severe’ category. The prevalence of ‘moderate or severe’ depressive symptoms was lower in India ($24.3\%$) than in Bangladesh ($59.5\%$) and Pakistan ($65.9\%$). Similarly, a smaller proportion reported ‘moderate or severe’ anxiety symptoms in India ($17.9\%$) than Bangladesh ($34.8\%$) and Pakistan ($51.9\%$).
The mean EQ-5D-5L (health-related quality of life) visual analogue scale (0–100) score was 66.9 overall, 69.9 for Bangladesh, 76.4 for India and 54.9 for Pakistan. A total of $45\%$ of the participants reported problems in carrying out their usual activities; and around $60\%$ reported pain/discomfort (Table 1).
## Physical disorders, risk factors, health risk behaviours and healthcare
As seen in Table 2, $11\%$ of participants had type 2 diabetes (self-report of clinician diagnosis or those with an HBA1c >$6.5\%$), $1\%$ had chronic respiratory disorders, $3.2\%$ had cardiovascular diseases, $2.3\%$ had tuberculosis and $2.0\%$ had chronic hepatitis. Table 2Non-communicable and communicable diseases and health risk behaviours in people with severe mental illnessBangladesh ($$n = 1500$$)India ($$n = 1175$$)Pakistan ($$n = 1314$$)Overall ($$n = 3989$$)n (%) [$95\%$ CI]n (%) [$95\%$ CI]n (%) [$95\%$ CI]n (%) [$95\%$ CI]Non-communicable diseasesType 2 diabetes diagnosed by HbA1c test (≥$6.50\%$) or self-report of clinician diagnosisa127 (8.8) [7.4–10.3]164 (16.1) [14.0–18.5]121 (9.4) [7.9–11.1]412 (11.0) [10.0–12.0]Type 2 diabetes (self-report of clinician diagnosis only)42 (2.8) [2.1–3.8]94 (8.0) [6.6–9.7]76 (5.8) [4.6–7.2]212 (5.3) [4.7–6.1]Heart disease (angina or a stroke) (self-report of clinician diagnosis)20 (1.3) [0.9–2.1]33 (2.8) [2–3.9]73 (5.6) [4.4–6.9]126 (3.2) [2.7–3.7]Lung condition (self-report of clinician diagnosis)b(<$3\%$)(<$3\%$)(<$3\%$)41 (1.0) [0.8–1.4]Non-communicable disease risk factorsHigh blood pressure (≥$\frac{140}{90}$ mmHg) or self-report of clinician diagnosisc242 (16.1) [14.4–18.1]241 (20.6) [18.3–23.0]444 (33.8) [31.3–36.4]927 (23.3) [22.0–24.6]High blood pressure (self-report of clinician diagnosis only)89 (5.9) [4.8–7.2]85 (7.2) [5.9–8.9]293 (22.3) [20.1–24.6]467 (11.7) [10.8–12.7]Hypercholesterolemia (tested plus self-report of clinician diagnosis)d659 (46.0) [43.4–48.5]470 (47.5) [44.4–50.6]716 (56.0) [53.3–58.7]1845 (49.9) [48.2–51.5]Hypercholesterolemia (self-report of clinician diagnosis only)14 (0.9) [0.6–1.6]37 (3.1) [2.3–4.3]53 (4.0) [3.1–5.2]104 (2.6) [2.2–3.1]Body mass index, kg/m²e24.0 (4.5) [23.8–24.2]25.6 (5.3) [25.3–25.9]26.1 (7.0) [25.7–26.4]25.2 (5.7) [24.9–25.3]Body mass index categories Underweight (<18.5)129 (8.6) [7.3–10.2]66 (5.7) [4.5–7.2]113 (8.7) [7.2–10.3]308 (7.8) [7.0–8.6] Normal weight (18.5–24.9)817 (54.6) [52.0–57.1]505 (43.5) [40.7–46.4]498 (38.2) [35.6–40.8]1820 (45.9) [44.4–47.5] Overweight (25–29.9)405 (27.1) [24.9–29.4]384 (33.1) [30.4–35.8]404 (31.0) [28.5–33.5]1193 (30.1) [28.7–31.5] Obese (≥30)146 (9.8) [8.3–11.4]206 (17.7) [15.6–20.1]290 (22.2) [20.0–24.6]642 (16.2) [15.1–17.4]Waist circumference, cme83.7 (11.6) [83.1–84.3]89.8 (22.9) [88.5–91.2]90.7 (27.8) [89.2–92.2]87.8 (21.4) [87.1–88.5]High waist circumference, women (≥80 cm)393 (67.5) [63.6–71.2]380 (73.6) [69.7–77.3]374 (73.8) [69.8–77.4]1147 (71.5) [69.2–73.6]High waist circumference, men (≥94 cm)128 (14.0) [11.9–16.4]247 (38.6) [34.9–42.4]298 (37.9) [34.6–41.4]673 (28.7) [27.0–30.6]Communicable diseasesHepatitis B or hepatitis C (self-report of clinician diagnosis)b(<$5\%$)(<$5\%$)(<$5\%$)78 (2.0) [1.6–2.4]Tuberculosis (self-report of clinician diagnosis)b(<$5\%$)(<$5\%$)(<$5\%$)92 (2.3) [1.9–2.8]HIV (self-report of clinician diagnosis only)b(<$1\%$)(<$1\%$)(<$1\%$)(<$1\%$)Physical activityTime spent on vigorous physical activity, min/daye69.7 [15.4–22.5]48.7 [10.3–15.8]76.4 [15.3–23.6]66.8 [15.3–19.4]Vigorous physical activity (>0 min/day)196 (13.1) [11.5–14.9]181 (15.4) [13.4–17.6]178 (13.5) [11.8–15.5]555 (13.9) [12.9–15.0]Time spent on moderate physical activity, min/daye32.1 [28.2–34.8]31.2 [25.3–37.3]19.5 [17.0–25.1]28.8 [25.5–30.6]Moderate physical activity (>0 min/day)600 (40.0) [37.5–42.5]289 (24.6) [22.2–27.1]289 (22.0) [19.8–24.3]1178 (29.5) [28.2–30.9]Time spent walking/cycling, min/daye34.5 [30.3–38.7]9.5 [7.7–11.3]27.4 [23.4–31.3]24.8 [22.7–26.9]Cycling/walking activity (>0 min/day)896 (59.7) [57.2–62.2]273 (23.2) [20.9–25.7]792 (60.3) [57.6–62.9]1961 (49.2) [47.7–50.6]Prevalence of low physical activity (total physical activity MET mins per week <600)889 (59.3) [56.8–61.7]389 (33.1) [30.5–35.9]541 (41.2) [38.5–43.9]1819 (45.6) [44.1–47.1]DietDo not meet WHO recommendations (fewer than five servings of fruits or vegetables per day)1262 (84.1) [82.2–85.9]1122 (95.5) [94.1–96.5]1171 (89.1) [87.3–90.7]3555 (89.1) [88.1–90]Tobacco useCurrently smoke, men (daily)392 (42.8) [39.7–46.1]130 (20.1) [17.2–23.3]252 (31.7) [28.5–35]774 (32.8) [31–34.7]Currently smoke, women (daily)b(<$5\%$)(<$5\%$)(<$5\%$)33 (2.0) [1.4–2.8]Currently use smokeless tobacco, men148 (16.2) [13.9–18.7]118 (18.2) [15.4–21.4]325 (40.8) [37.5–44.3]591 (25.1) [23.4–26.8]Currently use smokeless tobacco, women112 (19.1) [16.2–22.5]52 (9.9) [7.6–12.7]49 (9.5) [7.2–12.3]213 (13.1) [11.5–14.8]Any form of tobacco, men460 (50.3) [47–53.5]201 [31] [27.6–34.7]442 (55.5) [52.1–59]1103 (46.8) [44.8–48.7]Any form of tobacco, women112 (19.1) [16.2–22.5]55 (10.4) [8.1–13.4]68 (13.1) [10.5–16.3]235 (14.4) [12.8–16.2]Risk behaviours for infectious diseasesTwo or more sexual partners in the past 10 years137 (9.1) [7.8–10.7]19 (1.6) [1.0–2.5]197 (15.2) [13.3–17.2]353 (8.9) [8.1–9.8]Used condom when having sex in the past 10 years (among those who responded yes for two or more sexual partners) Never (report denominators) ($$n = 353$$)26 (19.0) [13.2–26.5]4 (21.1) [7.9–45.4]119 (60.4) [53.4–67]149 (42.2) [37.6–47] Sometimes or always ($$n = 353$$)111 (81.0) [73.5–86.8]15 (78.9) [54.6–92.1]78 (39.6) [33–46.6]204 (57.8) [53–62.4]Ever injected street drugs, steroids or vitamins with a needleb(<$1\%$)(<$1\%$)39 (3.0) [2.2–4.1]52 (1.3) [1.0–1.7]Confidence intervals were calculated using bootstrap sampling procedure ($$n = 1000$$) for binomial and continuous variables and Goodman's method for multinomial proportions. HbA1C, glycated haemoglobin; MET, metabolic equivalents; WHO, World Health Organization.a. The denominators for HbA1c are 2267 overall, 1370 for Bangladesh and 897 for Pakistan.b. Data not reported because of low numbers for statistical disclosure control.c. The denominators for hypertension are 2343 overall, 1422 for Bangladesh and 921 for Pakistan.d. The denominators for hypercholesterolemia are 2247 overall, 1357 for Bangladesh and 890 for Pakistan.e. Values presented as mean (s.d.).
Overall, $46.3\%$ of participants were overweight or obese; most women ($71.5\%$) and a high proportion of men ($28.7\%$) had a high waist circumference. Underweight was also prevalent in $7.8\%$ of the participants.
Almost a quarter ($23.3\%$) either reported a diagnosis of hypertension or had high measured blood pressure (>$\frac{140}{90}$ mmHg): $16.1\%$ in Bangladesh, $20.6\%$ in India and $33.8\%$ in Pakistan. Almost half ($49.9\%$) were found to have hypercholesterolemia based on either previous reported diagnosis or high levels of low-density lipoprotein cholesterol. A total of $35\%$ of participants had anaemia; this was higher in Bangladesh ($44.9\%$, $95\%$ CI 42.3–$47.4\%$) than India ($31.6\%$, $95\%$ CI 28.7–$34.5\%$) and Pakistan ($28.3\%$, $95\%$ CI 25.9–$30.8\%$) (Supplementary Material available at https://doi.org/10.1192/bjo.2023.12). Most people with hypercholesterolemia ($94.4\%$) and almost half with diabetes ($49.2\%$) and with high measured blood pressure ($48.5\%$) were previously unaware of their condition and were detected during the survey through cholesterol, HbA1c and blood pressure measurements, respectively.
Almost half of men ($46.8\%$) consumed either smoking or smokeless tobacco, and $32.8\%$ reported smoking tobacco daily. Smoking rates were $42.8\%$ in Bangladesh, $20.1\%$ in India and $31.7\%$ in Pakistan. A total of $19\%$ of women reported using tobacco in Bangladesh, $10.4\%$ in India and $13.1\%$ in Pakistan. Around half of participants ($45.6\%$) did not meet the WHO recommendations for physical activity (of 600 metabolic equivalents); and $89.1\%$ of the participants reported to not meet the WHO recommended levels of fruit and vegetable intake (at least five servings). Less than $6.2\%$ of males and $0.4\%$ of females reported consuming alcohol in the past month (data not provided in the tables). Less than $9\%$ of the sample reported to have more than two sexual partners in the past 10 years.
As shown in Table 3, only $56.8\%$ of the participants had been previously tested for any NCDs or NCD risk factors: $52.5\%$ for hypertension, $26.7\%$ for type 2 diabetes and $9.0\%$ for hypercholesterolemia. *In* general, a low proportion of participants received treatment for physical conditions or to address risk factors. Of those with a self-reported NCD or an NCD risk factor, only $48.5\%$ reported receiving related treatment or health risk modification advice. The provision of relevant treatment was highest in those reporting type 2 diabetes ($74.5\%$, $95\%$ CI 68.2–$80.0\%$), followed by hypertension ($43.9\%$, $95\%$ CI 39.5–$48.4\%$) and hypercholesterolaemia ($34.6\%$, $95\%$ CI 25.9–$44.5\%$). Only $42.8\%$ received any type of advice to modify health risk behaviours; the proportion of participants that received any type of health risk modification advice was highest in India ($81.7\%$), followed by Pakistan ($54.2\%$) and Bangladesh ($23.8\%$). Among those who consumed tobacco, only $28.1\%$ had been advised to quit. Table 3Proportion of people with severe mental illness screened, diagnosed and treated for non-communicable diseases and their risk factors, including health risk behaviour modification adviceBangladesh ($$n = 1500$$)India ($$n = 1175$$)Pakistan ($$n = 1314$$)Overall ($$n = 3989$$)n (%) [$95\%$ CI]n (%) [$95\%$ CI]n (%) [$95\%$ CI]n (%) [$95\%$ CI]Screened, diagnosed and treated for NCDs and NCD risk factors (self-reported)Type 2 diabetes Ever had blood glucose measured by doctor or healthcare provider334 (22.3) [20.2–24.4]361 (30.7) [28.1–33.4]372 (28.3) [25.9–30.8]1067 (26.7) [25.4–28.1] Ever diagnosed with type 2 diabetes by doctor or healthcare provider42 (12.6) [9.4–16.6]94 [26] [21.8–30.8]76 (20.4) [16.6–24.8]212 (19.9) [17.6–22.4] Have received treatment for diabetes (among those with type 2 diabetes)a30 (71.4) [55.8–83.2]75 (79.8) [70.3–86.8]53 (69.7) [58.4–79.1]158 (74.5) [68.2–80.0] Unaware of having type 2 diabetes (among those with type 2 diabetes)a85 (66.9) [58.2–74.6]70 (42.7) [35.3–50.4]45 (37.2) [29–46.2]200 (48.5) [43.9–53.2]Hypertension Ever had blood pressure measured by doctor or healthcare provider788 (52.5) [50–55.1]506 (43.1) [40.3–45.9]800 (60.9) [58.2–63.5]2094 (52.5) [51–54] Ever diagnosed with hypertension by doctor or healthcare provider89 (11.3) [9.3–13.7]85 (16.8) [13.8–20.3]293 (36.6) [33.4–40]467 (22.3) [20.6–24.1] Have received treatment (among those diagnosed with hypertension)a45 (50.6) [40.2–60.9]47 (55.3) [44.6–65.6]113 (38.6) [33.1–44.3]205 (43.9) [39.5–48.4] Unaware of having hypertension (among those diagnosed with hypertension)a153 (63.2) [56.9–69.1]156 (64.7) [58.5–70.5]151 (34.0) [29.7–38.6]460 (49.6) [46.5–52.7]Hypercholesterolemia Ever had cholesterol measured by doctor or healthcare provider61 (4.1) [3.2–5.2]174 (14.8) [12.9–17]126 (9.6) [8.1–11.3]361 (9.0) [8.2–10] Ever diagnosed with hypercholesterolemia by doctor or healthcare provider14 (23.0) [14.0–35.3]37 (21.3) [15.8–28]53 (42.1) [33.7–50.9]104 (28.8) [24.4–33.6] Have received treatment (among those diagnosed with hypercholesterolemia)a,b(<$40\%$)(<$40\%$)17 (32.1) [20.8–46]36 (34.6) [25.9–44.5] Unaware of having hypercholesterolemia (among those diagnosed with hypercholesterolemia)a645 (97.9) [96.4–98.7]433 (92.1) [89.3–94.2]663 (92.6) [90.4–94.3]1741 (94.4) [93.2–95.3]Any NCD or NCD risk factor (type 2 diabetes, lung disease, heart disease, hypertension or hypercholesterolemia) Ever been tested for any NCD or NCD risk factorc836 (55.7) [53.2–58.2]574 (48.9) [46–51.7]856 (65.1) [62.5–67.7]2266 (56.8) [55.3–58.3] Ever been diagnosed with any NCD or NCD risk factorc132 (15.8) [13.5–18.4]193 (33.6) [29.9–37.6]380 (44.4) [41.1–47.7]705 (31.1) [29.3–33] Received treatment from doctor or other healthcare worker for any of these (among those with NCD or NCD risk factor)69 (52.3) [43.7–60.7]111 (57.5) [50.4–64.3]162 (42.6) [37.7–47.7]342 (48.5) [44.9–52.2] Unaware of having any NCD or NCD risk factor (among those diagnosed) (denominators: Bangladesh 824, Pakistan 932, India 658, overall 2414)a692 [84] [81.3–86.3]465 (70.7) [67.1–74]552 (59.2) [56–62.3]1709 (70.8) [69–72.5]Screened, diagnosed and treated for communicable diseases (self-reported)Ever been tested for hepatitis B or C25 (1.7) [1.1–2.5]96 (8.3) [6.8–10]228 (17.4) [15.4–19.5]350 (8.8) [8–9.7]Diagnosed with hepatitis B or C among those testedb(<$25\%$)(<$8\%$)65 (28.5) [23–34.7]77 (22.0) [18–26.5]Ever been tested for tuberculosis36 (2.4) [1.7–3.3]24 (2.0) [1.4–3]190 (14.5) [12.7–16.5]250 (6.3) [5.6–7]Diagnosed with tuberculosis among those tested (<$52\%$)(<$26\%$)68 (35.8) [29.2–42.9]92 (36.8) [31–43]Ever been tested for HIV25 (1.7) [1.1–2.5]71 (6.0) [4.8–7.6]12 (0.9) [0.5–1.6]108 (2.7) [2.3–3.3]Received treatment for any chronic communicable disease (among those diagnosed with a chronic communicable disease)56 (100.0) [100.0–100.0]7 (58.3) [29.5–82.4]137 (92.6) [87–95.9]200 (92.6) [88.5–95.3]Health risk behaviour adviceQuit tobacco or do not start127 (9.8) [8.3–11.5]72 (17.1) [13.8–21]102 (14.1) [11.7–16.8]301 (12.3) [11.1–13.7]Quit tobacco among those who currently smoke or use smokeless tobacco108 (22.2) [18.7–26.1]51 (49.5) [39.9–59.1]86 (30.5) [25.4–36.1]245 (28.1) [25.3–31.1]Reduce salt in diet74 (5.7) [4.6–7.1]86 (20.4) [16.8–24.6]138 (19.0) [16.3–22.1]298 (12.2) [11–13.5]Reduce salt in diet among those diagnosed with hypertension32 (15.5) [11.1–21.1]40 (39.2) [30.2–49.1]110 (41.8) [36–47.9]182 (31.8) [28.2–35.6]Eat at least five servings of fruit and/or vegetables each day213 (16.4) [14.5–18.5]248 (58.9) [54.1–63.5]238 (32.8) [29.5–36.3]699 (28.6) [27–30.3]Reduce fat in diet116 (8.9) [7.5–10.6]98 (23.3) [19.5–27.6]192 (26.5) [23.4–29.8]406 (16.6) [15.2–18.1]Start or do more physical activity142 (10.9) [9.4–12.8]286 (67.9) [63.3–72.2]177 (24.4) [21.4–27.7]605 (24.8) [23.3–26.3]*Maintain a* healthy body weight or lose weight96 (7.4) [6.1–9]187 (44.4) [39.7–49.2]102 (14.1) [11.7–16.8]385 (15.8) [14.5–17.2]*Maintain a* healthy body weight or lose weight among those with overweight or obesity50 (10.2) [7.8–13.2]137 (59.1) [52.6–65.2]83 (20.0) [16.5–24.2]270 (23.8) [21.6–26.1]Reduce sugary beverages in diet74 (5.7) [4.6–7.1]74 (17.6) [14.2–21.5]90 (12.4) [10.2–15]238 (9.7) [8.6–11]Reduce sugary beverages in diet among those with type 2 diabetes21 (17.9) [12–26]38 (49.4) [38.3–60.5]39 (47.6) [36.9–58.4]98 (35.5) [30.3–41.1]Any type of lifestyle advice309 (23.8) [21.6–26.2]344 (81.7) [77.7–85.1]393 (54.2) [50.6–57.8]1046 (42.8) [41.1–44.6]Healthcare utilisationVisited a doctor or other healthcare worker in the past 12 months1297 (86.5) [84.6–88.1]421 (35.8) [33.1–38.6]725 (55.2) [52.5–57.8]2443 (61.2) [59.9–62.6]Confidence intervals were calculated using bootstrap sampling procedure ($$n = 1000$$) for binomial and continuous variables and Goodman's method for multinomial proportions. NCD, non-communicable disease.a. People that self-reported not to have type 2 diabetes, hypertension and hypercholesterolemia or had not been tested, but were positive on the test performed during the current survey.b. Data not reported because of low numbers for statistical disclosure control.c. Includes type 2 diabetes, hypertension and hypercholesterolemia.
## Comparison between people with SMI and the general population (STEPS survey)
The results for the comparisons between our data and country STEPS reports are summarised in Table 4. Table 4Odds of people with severe mental illness having non-communicable diseases, related risk factors and health risk behaviours and receiving healthcare screening and advice compared with the general population (severe mental illness data weighteda)BangladeshIndiaPakistanSTEPS, yes/totalSMI, yes/totalOdds ratio [$95\%$ CI]P-valuebSTEP, yes/totalSMI, yes/totalOdds ratio [$95\%$ CI]P-valuebSTEPS, yes/totalSMI, yes/totalOdds ratio [$95\%$ CI]P-valuebNon-communicable diseasesType 2 diabetes (diagnosed by HbA1c test (≥$6.50\%$) or self-report of clinician diagnosis)c$\frac{586}{7056192}$/14411.7 [1.42–2.03]<$\frac{0.001887}{9540163}$/10141.87 [1.55–2.25]<0.001N/A$\frac{117}{1271}$N/AN/AType 2 diabetes (self-report of clinician diagnosis only)$\frac{417}{818571}$/14910.93 [0.71–1.21]$\frac{0.589410}{954093}$/11731.92 [1.50–2.43]<$\frac{0.001250}{735871}$/12971.65 [1.24–2.17]<0.001Cardiovascular diseases (angina or a stroke) (self-report of clinician diagnosis)$\frac{819}{818526}$/14910.16 [0.10–0.24]<$\frac{0.001373}{10}$ $\frac{65934}{11730.82}$ [0.56–1.18]$\frac{0.284363}{735768}$/12971.07 [0.80–1.40]0.637Non-communicable disease risk factorsHypertension (measured in survey or self-report of clinician diagnosis)$\frac{1684}{8019313}$/14911.0 [0.87–1.15]$\frac{0.9953038}{10}$ $\frac{586242}{11700.65}$ [0.56–0.75]<0.001N/A$\frac{420}{1296}$N/AN/AHypertension (self-report of clinician diagnosis only)$\frac{1121}{8185149}$/14910.7 [0.58–0.84]<$\frac{0.001836}{10}$ $\frac{58686}{11730.92}$ [0.72–1.16]$\frac{0.4941096}{7358290}$/12971.65 [1.42–1.91]<0.001Hypercholesterolemia (measured in survey or self-report of clinician diagnosis)$\frac{2002}{7049704}$/14192.48 [2.21–2.79]<0.001N/A$\frac{467}{988}$N/AN/AN/A$\frac{683}{1261}$N/AN/AHypercholesterolemia (self-report of clinician diagnosis only)$\frac{176}{818520}$/14910.62 [0.37–1.0]$\frac{0.041192}{10}$ $\frac{65937}{11731.78}$ [1.21–2.55]$\frac{0.001110}{735748}$/12972.53 [1.76–3.61]<0.001Underweight (BMI < 18.5 kg/m2)$\frac{1038}{7985122}$/14880.6 [0.49–0.73]<$\frac{0.0011999}{10}$ $\frac{40966}{11590.25}$ [0.19–0.33]<$\frac{0.001747}{6613130}$/12860.88 [0.72–1.08]0.215Overweight or obesity (BMI ≥ 25 kg/m2)$\frac{2068}{7985609}$/14881.98 [1.76–2.23]<$\frac{0.0012717}{10}$ $\frac{409592}{11592.96}$ [2.61–3.35]<$\frac{0.0012725}{6613674}$/12861.57 [1.39–1.77]<0.001High waist circumference, women (≥80 cm)$\frac{1687}{4104568}$/7953.58 [3.03–4.25]<0.001N/A$\frac{384}{521}$N/AN/AN/A$\frac{507}{725}$N/AN/AHigh waist circumference, men (≥94 cm)$\frac{556}{3784125}$/6931.28 [1.02–1.59]0.024N/A$\frac{247}{632}$N/AN/AN/A$\frac{209}{548}$N/AN/AHealth risk behavioursLow physical activity (total physical activity MET mins per week <600)$\frac{999}{8118646}$/14915.45 [4.81–6.17]<$\frac{0.0014402}{10}$ $\frac{659783}{11732.85}$ [2.51–3.25]<$\frac{0.0012932}{7064812}$/12972.36 [2.08–2.67]<0.001Do not meet WHO recommendations (fewer than five servings of fruits or vegetables per day)$\frac{7318}{81681257}$/14910.62 [0.53–0.73]<$\frac{0.00110488}{10}$ $\frac{6591120}{11730.34}$ [0.25–0.48]<$\frac{0.0016899}{73391150}$/12970.5 [0.41–0.61]<0.001Smoking, men (daily)$\frac{1773}{3804299}$/6930.87 [0.74–1.03]$\frac{0.0921338}{5818128}$/6400.84 [0.68–1.03]$\frac{0.086876}{3150176}$/5551.21 [0.99–1.47]0.060Smoking, women (daily)d$\frac{44}{4381}$Not reported0.25 [0.03–0.95]$\frac{0.03763}{48414}$/5330.57 [0.15–1.55]$\frac{0.277177}{421637}$/7421.2 [0.81–1.73]0.33Smokeless tobacco use, men$\frac{1023}{3804142}$/6930.70 [0.57–0.86]<$\frac{0.0012124}{5818117}$/6400.39 [0.31–0.48]<$\frac{0.001312}{3150226}$/5556.25 [5.06–7.71]<0.001Smokeless tobacco use, women$\frac{1231}{4381188}$/7980.79 [0.66–0.94]$\frac{0.008581}{484151}$/5330.78 [0.56–1.05]$\frac{0.098198}{421658}$/7421.72 [1.25–2.35]<0.001Any form of tobacco use, men$\frac{2267}{3804367}$/6930.76 [0.65–0.90]$\frac{0.0012979}{5818199}$/6400.43 [0.36–0.51]<$\frac{0.0011121}{3150308}$/5552.26 [1.87–2.72]<0.001Any form of tobacco use, women$\frac{1240}{4381188}$/7980.78 [0.65–0.93]$\frac{0.006629}{484154}$/5330.75 [0.55–1.02]$\frac{0.060367}{421685}$/7421.36 [1.04–1.75]0.016Screening for non-communicable diseases and risk factorsType 2 diabetes Ever had blood glucose measured by doctor or healthcare provider$\frac{2054}{8185434}$/14911.23 [1.08–1.39]$\frac{0.0012509}{9540360}$/11731.24 [1.08–1.42]$\frac{0.001831}{7358360}$/12973.02 [2.61–3.49]<0.001 Ever diagnosed with type 2 diabetes by doctor or healthcare provider$\frac{417}{205471}$/4340.77 [0.57–1.02]$\frac{0.060410}{250993}$/3601.78 [1.36–2.32]<$\frac{0.001250}{83171}$/3600.57 [0.42–0.78]<0.001 Have received treatment (among those with type 2 diabetes)$\frac{244}{41755}$/712.44 [1.32–4.71]$\frac{0.002158}{41074}$/936.21 [3.54–11.29]<$\frac{0.001186}{25051}$/710.88 [0.47–1.68]0.664 Unaware of having type 2 diabetes (among those with type 2 diabetes)e$\frac{301}{586121}$/1921.61 [1.14–2.29]$\frac{0.005477}{88770}$/1630.64 [0.45–0.91]0.011N/A$\frac{47}{117}$N/AN/AHypertension Ever had blood pressure measured by doctor or healthcare provider$\frac{5738}{8185886}$/14910.62 [0.56–0.70]<$\frac{0.0015505}{10}$ $\frac{586507}{11730.7}$ [0.62–0.79]<$\frac{0.0014025}{7358811}$/12971.38 [1.22–1.56]<0.001 Diagnosed with hypertension by doctor or healthcare provider (among those with blood pressure previously measured)$\frac{1121}{5738149}$/8860.83 [0.69–1.01]$\frac{0.056836}{550586}$/5071.14 [0.88–1.46]$\frac{0.2881096}{4025290}$/8111.49 [1.26–1.75]<0.001 Have received treatment (among those with hypertension)$\frac{301}{112188}$/1493.9 [2.7–5.7]<$\frac{0.001134}{83647}$/866.3 [3.9–10.3]<$\frac{0.001580}{1096119}$/2900.62 [0.47–0.81]<0.001 Unaware of having hypertension (among those with hypertension)e$\frac{864}{1684164}$/3131.04 [0.81–1.34]$\frac{0.7232202}{3038157}$/2420.7 [0.53–0.94]0.011N/A$\frac{130}{420}$N/AN/AHypercholesterolemia Ever had cholesterol measured by doctor or healthcare provider$\frac{377}{818575}$/14911.1 [0.84–1.42]$\frac{0.475682}{10}$ $\frac{659175}{11732.57}$ [2.14–3.08]<$\frac{0.001456}{7357118}$/12971.51 [1.22–1.88]<0.001 Diagnosed with hypercholesterolemia by doctor or healthcare provider (among those with cholesterol previously measured)$\frac{176}{37720}$/750.42 [0.23–0.74]$\frac{0.001192}{68237}$/1750.68 [0.45–1.03]$\frac{0.062110}{45648}$/1182.16 [1.37–3.37]<0.001 Have received treatment (among those diagnosed with hypercholesterolemia)d$\frac{71}{176}$Not reported0.63 [0.19–1.87]$\frac{0.37074}{19214}$/370.97 [0.43–2.11]$\frac{0.93648}{11016}$/480.65 [0.30–1.38]0.225 Unaware of having hypercholesterolemia (among those with hypercholesterolemia)e$\frac{1898}{2002684}$/7031.97 [1.19–3.43]0.006N/A$\frac{430}{467}$N/AN/AN/A$\frac{636}{683}$N/AN/AAdvice on health risk behavioursQuit or do not take up tobacco$\frac{807}{3977106}$/12810.35 [0.28–0.44]<$\frac{0.0011865}{10}$ $\frac{65971}{4200.96}$ [0.73–1.25]$\frac{0.7541832}{735674}$/7070.35 [0.27–0.45]<0.001Quit tobacco among those who currently use smoke or smokeless tobacco$\frac{434}{63890}$/4610.11 [0.08–0.15]<$\frac{0.001522}{298550}$/1024.5 [3.0–6.9]<$\frac{0.001419}{82361}$/2160.38 [0.27–0.53]<0.001Start or do more physical activity$\frac{764}{3977149}$/12810.55 [0.46–0.67]<0.001N/A$\frac{286}{420}$N/AN/A$\frac{2060}{7356176}$/7070.85 [0.71–1.02]0.078Maintain a healthy body weight or lose weight$\frac{660}{3977111}$/12810.48 [0.38–0.59]<0.001N/A$\frac{187}{420}$N/AN/A$\frac{1964}{7356105}$/7070.48 [0.37–0.59]<0.001STEPS, STEPwise Approach to Surveillance of NCDs; SMI, severe mental illness; HbA1C, glycated haemoglobin;, N/A: Not available in the STEPS survey; BMI, body mass index; MET, metabolic equivalents; WHO, World Health Organization.a. Data from the general population was extracted from the STEPS 2018 survey in Bangladesh and India, and 2014 survey in Pakistan; data from the SMI survey were weighted by age and gender according to the distribution of the STEPS report.b. After Bonferroni correction for multiple testing, the $P \leq 0.05$ significance level was corrected to $P \leq 0.006.$c. Blood glucose ≥126 mg/dL for the STEPS survey and HbA1c ≥$6.50\%$ for the SMI survey.d. Data not reported because of low numbers for statistical disclosure control.e. People that self-reported not to have the condition or had not been previously tested but tested positive in assessments performed for the current survey or the STEPS survey in Bangladesh.
## Prevalence of NCDs and NCD risk factors
People with SMI in Bangladesh (odds ratio 1.7, $95\%$ CI 1.4–2.0, $P \leq 0.001$) and India (odds ratio 1.8, $95\%$ CI 1.5–2.2, $P \leq 0.001$) were more likely to have type 2 diabetes compared with the general population, and people with SMI in Bangladesh were more likely to have hypercholesterolemia (odds ratio 2.4, $95\%$ CI 2.2–2.7, $P \leq 0.001$) compared with the general population. Blood samples were not collected in the STEPS survey in Pakistan and for cholesterol in India, therefore these comparisons are not available.
People with SMI were more likely to be overweight or obese (BMI > 25 kg/m2) compared with the general population (Bangladesh: odds ratio 1.9, $95\%$ CI 1.7–2.2, $P \leq 0.001$; India: odds ratio 2.9, $95\%$ CI 2.6–3.3, $P \leq 0.001$; Pakistan: odds ratio 1.5, $95\%$ CI 1.3–1.7, $P \leq 0.001$). People with SMI in Bangladesh were less likely to be underweight (odds ratio 0.6, $95\%$ CI 0.4–0.7, $P \leq 0.001$), but there were no differences in India and Pakistan.
## Screening and diagnosis
Compared with the general population, people with SMI were more likely to be screened for type 2 diabetes in Bangladesh (odds ratio 1.2, $95\%$ CI 1.0–1.3, $$P \leq 0.001$$), India (odds ratio 1.2, $95\%$ CI 1.0–1.4, $$P \leq 0.001$$) and Pakistan (odds ratio 3.0, $95\%$ CI 2.6–3.4, $P \leq 0.001$).
People with SMI in Bangladesh (odds ratio 0.6, $95\%$ CI 0.5–0.7, $P \leq 0.001$) and India (odds ratio 0.70, $95\%$ CI 0.6–0.7, $P \leq 0.001$) were less likely to be screened for hypertension, whereas the opposite was found in Pakistan (odds ratio 1.3, $95\%$ CI 1.2–1.5, $P \leq 0.001$). Among those screened, people with SMI in Pakistan were more likely to have hypertension (odds ratio 1.4, $95\%$ CI 1.2–1.7, $P \leq 0.001$), whereas no differences were found in Bangladesh and India.
Regarding hypercholesterolemia, people with SMI in India (odds ratio 2.5, $95\%$ CI 2.1–3.0, $P \leq 0.001$) and Pakistan were more likely to be screened (odds ratio 1.5, $95\%$ CI 1.2–1.8, $P \leq 0.001$), and those that were screened in Pakistan were more likely to have hypercholesterolemia (odds ratio 2.1, $95\%$ CI 1.3–3.3, $P \leq 0.001$) than people in the general population. In Bangladesh, there was no difference in screening; however, those that were screened were less likely (odds ratio 0.4, $95\%$ CI 0.2–0.7, $$P \leq 0.001$$) to have hypercholesterolemia than the general population.
## Health risk modification advice
People with SMI were less likely to receive advice to quit or not take up tobacco in Bangladesh (odds ratio 0.3, $95\%$ CI 0.2–0.4, $P \leq 0.001$) and Pakistan (odds ratio 0.3, $95\%$ CI 0.2–0.4, $P \leq 0.001$), whereas no differences were found in India. A similar pattern was observed for receiving advice on maintaining healthy body weight (Bangladesh: odds ratio 0.4, $95\%$ CI 0.3–0.5, $P \leq 0.001$; Pakistan: odds ratio 0.4, $95\%$ CI 0.3–0.5, $P \leq 0.001$, Pakistan); this indicator was not available for the STEPS report in India.
## Discussion
This is the first multi-country study from South Asia to report on physical multimorbidity, health risk behaviours and access to related healthcare in people with SMI. We found a high prevalence of physical health conditions, primarily NCDs and related risk factors. We also found that people with SMI were more likely to have NCDs and NCD risk factors (overweight/obesity, hypertension, hypercholesterolemia) and engage in some health risk behaviours (tobacco use), but were less likely to receive risk modification advice than the general population. Many people with SMI in our sample reported that they had never been tested or screened for NCDs or NCD risk factors despite the well-established link between SMI and cardiometabolic conditions.4,5 *Moreover a* large proportion of people with type 2 diabetes, hypertension and hypercholesterolaemia had not been previously diagnosed, and these conditions were only detected on testing during the survey. Most had not received appropriate treatment and risk modification advice for their physical health. Therefore, even in the two major specialist mental health institutes included in our survey, most people with SMI failed to receive adequate screening, prevention and management of NCDs and NCD risk factors.
At the time of the survey, there were no policies or recommendations for people with SMI to attend or visit primary care. The difference within countries in terms of healthcare utilisation (visited a doctor or other healthcare worker in the past 12 months) is most likely a result of the differences in the timing of data collection: most of the sample in India was recruited during the COVID-19 pandemic, whereas most of the sample in Bangladesh and Pakistan was recruited before the COVID-19 pandemic. This finding is supported by a multicentre cross-sectional study in India reporting that people had 2.5 higher odds of not being able to access healthcare services during the COVID-19 pandemic compared with before the COVID-19 pandemic.24 The finding that people with SMI are more likely to have NCD risk factors compared with the general population extends previous findings in high-income countries and LMICs for risk factors such as obesity, hypercholesterolemia and decreased physical activity.25,26 Importantly, it should be noted that psychotropic medication might contribute to some of these adverse risks.27 Almost all survey participants were prescribed antipsychotics, which are associated with tiredness and sedation, an increased risk of obesity and adverse effects on glucose and lipid metabolism. The high prevalence of anaemia among participants is consistent with findings in people with SMI in other LMICs,28 and this has been associated with poor diet and side-effects of mood stabilisers.28 In Pakistan, we found a higher prevalence of tobacco use in people with SMI compared with the general population. This is consistent with other studies in people with SMI,29 where tobacco use has been associated with a greater susceptibility to addiction because of a higher subjective experience of reward and an attempt to self-medicate to mitigate anxiety and depressive symptoms.30 Unexpectedly, the opposite was found in Bangladesh and India. This may be because the STEPS survey for Bangladesh and India reported an unusually high estimate of the prevalence of tobacco use (in men, $70\%$ for Bangladesh and $52\%$ for India). The more reliable Global Adult Tobacco Survey31 for the same period reported a prevalence of $58\%$ in Bangladesh and $43\%$ in India in the same group, which is closer to the figures reported in our study.
The low observed prevalence of alcohol use in both men and women is similar to the STEPS survey reports,14,22,23 and is likely to be explained by religious proscription.
Despite the high prevalence of overweight/obesity, hypercholesterolemia, hypertension and tobacco use, health risk modification advice was provided to less than a quarter of people with SMI, and we found that the odds of receiving such advice was lower in people with SMI than in the general population in Bangladesh and Pakistan. Similar treatment gaps have been reported in high-income countries.25 Although psychiatrists are trained in motivational interviewing, there are attitudinal barriers that make mental health professionals reluctant to engage with patients about their tobacco use.32 Moreover, misconceptions about potential side-effects of tobacco cessation medication, unfounded fears of exacerbating depressive symptoms following quitting and low expectations of patients’ motivation or ability to stop smoking are additional barriers.33 On the other hand, there is high-quality evidence from high-income countries about both the effectiveness and cost benefits of smoking cessation interventions in people with SMI.34 Such approaches need to be adopted in South Asia, where tobacco use is common. Similarly, lifestyle interventions have shown promise to reduce weight and improve metabolic risk factors, and are recommended as an essential part of the management of SMI in these countries.35 An important study finding is the differences in the proportion of participants with moderate or severe depressive and anxiety symptoms within the countries (with the highest in Pakistan and lowest in India). This may be explained by the type of patient flow in each hospital: the National Institute of Mental Health and *Neurosciences is* a tertiary care, exclusive neuropsychiatric setting, and the proportion of patients with schizophrenia and bipolar disorders is likely to be higher compared with a general hospital psychiatry unit, whereas the proportion of patients with depression is higher in the Institute of Psychiatry in Rawalpindi, Pakistan, as compared with private mental health facilities. These differences may also be related to the higher proportion of participants in the depression with psychosis category in Pakistan.
The prevalence of tuberculosis was three times higher than in the general population. This is consistent with previous findings in LMICs and the clustering of tuberculosis risk factors reported in people with SMI.36 In contrast, the prevalence of HIV37 and hepatitis B and C38 were similar to those reported in the general population – a surprising finding considering the several risk factors for blood-borne viruses that have been reported to cluster in people with SMI.
Although most of the comparisons between people with SMI and the general population are in line with clinical expectations and previous findings,26 there were some anomalous results. These include the lower odds of people with SMI with a self-reported clinical diagnosis of type 2 diabetes, hypertension and hypercholesterolemia in Bangladesh. This may be because of ‘diagnostic overshadowing’, where the presence of a mental disorder means clinicians do not look for physical health problems, or failure to recall such diagnosis by patients. The lower education and socioeconomic levels for participants from Bangladesh (compared with India and Pakistan) may have contributed to the latter.39 We report findings from the first large-scale effort to document physical multimorbidity in people with SMI attending specialist services in three South Asian countries. We used standardised tools for data collection (i.e. STEPS, EQ-5D-5L, PHQ-9, GAD-7) that allowed us to compare our findings with those in the general population. Data were collected by trained researchers having experience of working with this population. Finally, we gathered objective data on physical conditions (including blood tests), and reported on both previously diagnosed and undiagnosed conditions.
Of the several limitations that need to be mentioned, the first is although we have used findings from studies in the general population to compare and discuss our findings, caution needs to be exercised in such comparisons, since our sample was collected from mental health hospitals and the analyses were only adjusted by gender and age and other. Moreover, we need to be mindful of the time lag between these studies, during which a number of parameters of interest might have changed.
Second, we relied on blood results from each mental health institution's laboratory, but we did not standardise these tests between laboratories. Third, there are methodological considerations that should be considered when making comparisons between the countries: recruitment in India was done during the COVID-19 pandemic, which needs to be considered when comparing the country estimates, since the pandemic might have affected the physical and mental health outcomes as well as healthcare utilisation of the SMI population in India; in-patients were more likely to be excluded because a ‘lack of capacity to answer’, and these patients with more severe SMI symptoms have also shown to have more physical health problems,40 which may be associated with an underestimation of physical ill conditions in our sample; there was a lower proportion of in-patients in Pakistan than in Bangladesh and India, and in-patients are known to have more severe mental health symptoms and are more likely to have physical health conditions,40 which might lead to an underestimation of the prevalence of mental and physical health conditions in Pakistan. Fourth, the information from some of the questions from the STEPS survey are easy to recall (e.g. diabetes diagnosis), whereas some other may be more difficult (e.g. time performing physical activity or receipt of health risk modification advice) and prone to recall bias. To the best of our knowledge, there is no information on the performance of the access to healthcare and health risk modification advice questions in the STEPS survey that could provide further information about the risk of bias. Fifth, since the sample was drawn from tertiary care, the findings may not be representative of the total SMI population in each country. However, unlike mental health services in high-income countries, tertiary care services in South Asian countries accept self-referral without the need for primary or secondary care referral, and often function as ‘the first port of call’ for people with SMI. They also attract patients from both urban and rural areas. Therefore, the study population is likely to be similar to the overall population of people with SMI in these countries.
The high prevalence of physical health conditions and health risk behaviours in SMI compared with the general population, and their underdetection even in specialist centres, merits attention to improve early identification, prevention and management, in line with international recommendations and guidance. Given many of these physical health monitoring and management guidelines for SMI are based on evidence from (and developed in) high-income countries, they may not necessarily be applicable to low-resource settings in LMICs. Our findings can help to identify and contextualise the priority areas for LMICs, and to develop more appropriate guidance for such settings. In view of challenging resource limitations, interventions to address health risk behaviours that are brief and delivered by non-specialist personnel need to be tested in these settings. Integration of physical healthcare with mental healthcare that has been envisioned at all levels of mental healthcare delivery needs to be actioned and scaled up.8 Representative community-based studies may further answer questions related to regional differences in physical health conditions and health risk behaviours.
In conclusion, people with SMI in South Asia have a high prevalence of NCDs, which may be attributable to the associated clustering of several health risk factors and behaviours in this population. There is an unmet need to address physical multimorbidity in people with SMI in South Asia. Policy makers and healthcare professionals working with people with SMI need to recognise the extent and importance of physical multimorbidity in this vulnerable group, and prioritise the prevention, screening and treatment of NCDs in people with SMI.
## Data availability
The data-sets used and/or analysed during the current study are available from the corresponding author, G.A.Z., on reasonable request.
## Author contributions
N.S. conceived the study. A.H.-C., A.T.N., F.A., H.K., K.P.-M., K.S., P.M., R.H. and S.R. provided important contextual information. A.H.-C., F.S., F.A., J.R.B., K.P.-M., K.S. and N.S. developed the methodology and selected the outcomes. A.H.-C., G.A.Z. and J.R.B. developed the analysis plan. A.H.-C., G.A.Z., H.K., J.R.B. and S.R. revised the data and conducted the statistical analysis. A.H.-C., D.S., G.A.Z., H.K., J.R.B., K.A., K.S. and N.S. wrote the manuscript. All authors revised and approved the manuscript.
## Funding
This research was funded by the National Institute for Health Research (NIHR) (grant number GHRG $\frac{17}{63}$/130; awarded to N.S.), using UK aid from the UK Government to support global health research. The views expressed in this publication are those of the author(s) and not necessarily those of the NIHR or the UK Department of Health and Social Care.
## Declaration of interest
The authors declare that they have no competing interests. D.S. is an expert advisor to the UK National Institute for Health and Care Excellence (NICE) centre for guidelines; the views expressed are the authors’ and not those of the NICE.
## References
1. Liu NH, Daumit GL, Dua T, Aquila R, Charlson F, Cuijpers P. **Excess mortality in persons with severe mental disorders: a multilevel intervention framework and priorities for clinical practice, policy and research agendas**. *World Psychiatry* (2017) **16** 30-40. PMID: 28127922
2. Chan JKN, Wong CSM, Or PCF, Chen EYH, Chang WC. **Risk of mortality and complications in patients with schizophrenia and diabetes mellitus: population-based cohort study**. *Br J Psychiatry* (2021) **219** 375. PMID: 33407970
3. Fond G, Pauly V, Leone M, Llorca PM, Orleans V, Loundou A. **Disparities in intensive care unit admission and mortality among patients with schizophrenia and COVID-19: a national cohort study**. *Schizophr Bull* (2021) **47** 624. PMID: 33089862
4. Fekadu A, Medhin G, Kebede D, Alem A, Cleare AJ, Prince M. **Excess mortality in severe mental illness: 10-year population-based cohort study in rural Ethiopia**. *Br J Psychiatry* (2015) **206** 289. PMID: 25657358
5. Teferra S, Shibre T, Fekadu A, Medhin G, Wakwoya A, Alem A. **Five-year mortality in a cohort of people with schizophrenia in Ethiopia**. *BMC Psychiatry* (2011) **11** 165. PMID: 21985179
6. Misra A, Tandon N, Ebrahim S, Sattar N, Alam D, Shrivastava U. **Diabetes, cardiovascular disease, and chronic kidney disease in South Asia: current status and future directions**. *BMJ* (2017) **357** j1420. PMID: 28400361
7. Shavitt S, Cho YI, Johnson TP, Jiang D, Holbrook A, Stavrakantonaki M. **Culture moderates the relation between perceived stress, social support, and mental and physical health**. *J Cross Cult Psychol* (2016) **47** 956
8. **Making more of multimorbidity: an emerging priority**. *Lancet* (2018) **391** 1637. PMID: 29726322
9. 9United Nations. Transforming Our World: The 2030 Agenda for Sustainable Development. United Nations, 2015 (https://sdgs.un.org/2030agenda).. *Transforming Our World: The 2030 Agenda for Sustainable Development* (2015)
10. Zavala GA, Prasad K, Aslam F, Barua D, Hewitt AHC, Huque R. **Prevalence of physical health conditions and health risk behaviours in people with severe mental Illness in South Asia: protocol for a cross-sectional study (IMPACT SMI survey)**. *BMJ Open* (2020) **10** e037869
11. Lecrubier Y, Sheehan DV, Weiller E, Amorim P, Bonora I, Harnett Sheehan K. **The Mini International Neuropsychiatric Interview (MINI). A short diagnostic structured interview: reliability and validity according to the CIDI**. *Eur Psychiatry* (1997) **12** 224
12. Bonita R, Winkelmann R, Douglas KA, de Courten M, McQueen DV, Puska P. *Global Behavioral Risk Factor Surveillance* (2003) 9-22
13. Tripathy JP, Thakur JS, Jeet G, Chawla S, Jain S. **Alarmingly high prevalence of hypertension and pre-hypertension in North India-results from a large cross-sectional STEPS survey**. *PLoS One* (2017) **12** e0188619. PMID: 29267338
14. Rafique I, Saqib MAN, Munir MA, Qureshi H, Rizwanullah KSA, Khan SA. **Prevalence of risk factors for noncommunicable diseases in adults: key findings from the Pakistan STEPS survey**. *East Mediterr Health J* (2018) **24** 33-41. PMID: 29658619
15. 15World Health Organization (WHO).
WHO STEPS Surveillance Manual: The WHO STEPwise Approach to Chronic Disease Risk Factor Surveillance / Noncommunicable Diseases and Mental Health, World Health Organization. WHO, 2005 (https://apps.who.int/iris/handle/10665/43376).. (2005)
16. Bazzano LA. *Dietary Intake of Fruit and Vegetables and Risk of Diabetes Mellitus and Cardiovascular Diseases* (2005)
17. Gerbert B, Bronstone A, McPhee S, Pantilat S, Allerton M. **Development and testing of an HIV-risk screening instrument for use in health care settings**. *Am J Prev Med* (1998) **15** 103. PMID: 9713665
18. 18EuroQol Group. EQ-5D-5L User Guide: Basic Information on How to Use the EQ-5D-5L Instrument (Version 2.0) 2013 04/03/2015. EuroQol Group, 2009 (https://euroqol.org/eq-5d-instruments/eq-5d-5l-about/).. (2009)
19. Lorenzo C, Williams K, Hunt KJ, Haffner SM. (2007)
20. 20World Health Organization (WHO). Global Status Report on Noncommunicable Diseases 2014. WHO, 2014 (https://apps.who.int/iris/handle/10665/148114).. *Global Status Report on Noncommunicable Diseases 2014* (2014)
21. 21World Health Organization (WHO). Iron Deficiency Anemia. Assessment, Prevention, and Control. A Guide for Programme Managers. WHO, 2001 (https://www.who.int/publications/m/item/iron-children-6to23--archived-iron-deficiency-anaemia-assessment-prevention-and-control).. (2001)
22. 22National Institute of Preventive and Social Medicine. National STEPS Survey for Non-Communicable Diseases Risk Factors in Bangladesh 2018. World Health Organization, 2018 (https://apps.who.int/iris/bitstream/handle/10665/332886/STEPS-BAN-eng.pdf).. (2018)
23. 23National Centre for Disease Informatics and Research, Indian Council of Medical Research, Ministry of Health & Family Welfare. National Noncommunicable Disease Monitoring Survey (NNMS) 2017–18. Indian Council of Medical Research and National Centre for Disease Informatics and Research, 2020 (https://www.ncdirindia.org/nnms/resources/Initial_Content.pdf).. (2020)
24. Raman R, Rajalakshmi R, Surya J, Ramakrishnan R, Sivaprasad S, Conroy D. **Impact on health and provision of healthcare services during the COVID-19 lockdown in India: a multicentre cross-sectional study**. *BMJ Open* (2021) **11** e043590
25. Holt R, Afzal M, Ahmad B, Afsheen N, Aslam F, Ali A. **Prevalence of overweight and obesity in people with severe mental illness: systematic review and meta-analysis**. *Front Endocrinol (Lausanne)* (2021) **44**
26. Soundy A, Wampers M, Probst M, De Hert M, Stubbs B, Vancampfort D. **Physical activity and sedentary behaviour in outpatients with schizophrenia: A systematic review and meta-analysis**. *Int J Ther Rehabil* (2013) **20** 588
27. Correll CU, Detraux J, De Lepeleire J. **Effects of antipsychotics, antidepressants and mood stabilizers on risk for physical diseases in people with schizophrenia, depression and bipolar disorder**. *World Psychiatry* (2015) **14** 119. PMID: 26043321
28. Korkmaz S, Yıldız S, Korucu T, Gundogan B, Sunbul ZE, Korkmaz H. **Frequency of anemia in chronic psychiatry patients**. *Neuropsychiatr Dis Treat* (2015) **11** 2737. PMID: 26543367
29. Evins AE, Cather C, Laffer A. **Treatment of tobacco use disorders in smokers with serious mental illness: toward clinical best practices**. *Harv Rev Psychiatry* (2015) **23** 90. PMID: 25747922
30. Peckham E, Bradshaw TJ, Brabyn S, Knowles S, Gilbody S. **Exploring why people with SMI smoke and why they may want to quit: baseline data from the SCIMITAR RCT**. *J Psychiatr Ment Health Nurs* (2016) **23** 282. PMID: 26147943
31. 31World Health Organization (WHO). Global Adult Tobacco Survey (GATS): Bangladesh Factsheet 2009. WHO, 2009 (https://apps.who.int/iris/bitstream/handle/10665/331143/ban_gats_fs_2009-eng.pdf).. (2009)
32. Rogers ES, Gillespie C, Smelson D, Sherman SE. **A qualitative evaluation of mental health clinic staff perceptions of barriers and facilitators to treating tobacco use**. *Nicotine Tob Res* (2018) **20** 1223. PMID: 29059344
33. Kerr S, Woods C, Knussen C, Watson H, Hunter R. **Breaking the habit: a qualitative exploration of barriers and facilitators to smoking cessation in people with enduring mental health problems**. *BMC Public Health* (2013) **13** 221. PMID: 23497231
34. Gilbody S, Peckham E, Bailey D, Arundel C, Heron P, Crosland S. **Smoking cessation for people with severe mental illness (SCIMITAR+): a pragmatic randomised controlled trial**. *Lancet Psychiatry* (2019) **6** 379. PMID: 30975539
35. Sheals K, Tombor I, McNeill A, Shahab L. **A mixed-method systematic review and meta-analysis of mental health professionals’ attitudes toward smoking and smoking cessation among people with mental illnesses**. *Addiction* (2016) **111** 1536. PMID: 27003925
36. Doherty AM, Kelly J, McDonald C, O'Dywer AM, Keane J, Cooney J. **A review of the interplay between tuberculosis and mental health**. *Gen Hosp Psychiatry* (2013) **35** 398-406. PMID: 23660587
37. Khan AA, Khan A. **The HIV epidemic in Pakistan**. *J Pak Med Assoc* (2010) **60** 300. PMID: 20419976
38. Ali SA, Donahue RMJ, Qureshi H, Vermund SH. **Hepatitis B and hepatitis C in Pakistan: prevalence and risk factors**. *Int J Infect Dis* (2009) **13** 9-19. PMID: 18835208
39. Vellakkal S, Subramanian SV, Millett C, Basu S, Stuckler D, Ebrahim S. **Socioeconomic inequalities in non-communicable diseases prevalence in India: disparities between self-reported diagnoses and standardized measures**. *PLoS One* (2013) **8** e68219. PMID: 23869213
40. Kasckow JW, Twamley E, Mulchahey JJ, Carroll B, Sabai M, Strakowski SM. **Health-related quality of well-being in chronically hospitalized patients with schizophrenia: comparison with matched outpatients**. *Psychiatry Res* (2001) **103** 69-78. PMID: 11472791
|
---
title: The effect of fish oil supplementation on resistance training-induced adaptations
authors:
- Jeffery L. Heileson
- Steven B. Machek
- Dillon R. Harris
- Sara Tomek
- Leticia C. de Souza
- Adam J. Kieffer
- Nicholas D. Barringer
- Andrew Gallucci
- Jeffrey S. Forsse
- LesLee K. Funderburk
journal: Journal of the International Society of Sports Nutrition
year: 2023
pmcid: PMC9970203
doi: 10.1080/15502783.2023.2174704
license: CC BY 4.0
---
# The effect of fish oil supplementation on resistance training-induced adaptations
## ABSTRACT
### Background
Resistance exercise training (RET) is a common and well-established method to induce hypertrophy and improvement in strength. Interestingly, fish oil supplementation (FOS) may augment RET-induced adaptations. However, few studies have been conducted on young, healthy adults.
### Methods
A randomized, placebo-controlled design was used to determine the effect of FOS, a concentrated source of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), compared to placebo (PL) on RET-induced adaptations following a 10-week RET program (3 days·week−1). Body composition was measured by dual-energy x-ray absorptiometry (LBM, fat mass [FM], percent body fat [%BF]) and strength was measured by 1-repetition maximum barbell back squat (1RMSQT) and bench press (1RMBP) at PRE (week 0) and POST (10 weeks). Supplement compliance was assessed via self-report and bottle collection every two weeks and via fatty acid dried blood spot collection at PRE and POST. An a priori α-level of 0.05 was used to determine statistical significance and Cohen’s d was used to quantify effect sizes (ES).
### Results
Twenty-one of 28 male and female participants (FOS, $$n = 10$$ [4 withdrawals]; PL, $$n = 11$$ [3 withdrawals]) completed the 10-week progressive RET program and PRE/POST measurements. After 10-weeks, blood EPA+DHA substantially increased in the FOS group (+$109.7\%$, $p \leq .001$) and did not change in the PL group (+$1.3\%$, $$p \leq .938$$). Similar between-group changes in LBM (FOS: +$3.4\%$, PL: +$2.4\%$, $$p \leq .457$$), FM (FOS: −$5.2\%$, PL: $0.0\%$, $$p \leq .092$$), and %BF (FOS: −$5.9\%$, PL: −$2.5\%$, $$p \leq .136$$) were observed, although, the between-group ES was considered large for FM ($d = 0.84$). Absolute and relative (kg·kg [body mass]−1) 1RMBP was significantly higher in the FOS group compared to PL (FOS: +$17.7\%$ vs. PL: +$9.7\%$, $$p \leq .047$$; FOS: +$17.6\%$ vs. PL: +$7.3\%$, $$p \leq .011$$; respectively), whereas absolute 1RMSQT was similar between conditions (FOS: +$28.8\%$ vs. PL: +$20.5\%$, $$p \leq .191$$). Relative 1RMSQT was higher in the FOS group (FOS: +$29.3\%$ vs. PL: +$17.9\%$, $$p \leq .045$$).
### Conclusions
When combined with RET, FOS improves absolute and relative 1RM upper-body and relative 1RM lower-body strength to a greater extent than that observed in the PL group of young, recreationally trained adults.
## Background
The preservation and promotion of skeletal muscle mass and strength is critical for physical performance and healthy physiology throughout the lifespan [1]. Resistance exercise training (RET) may be one of the best- and well-established strategies to influence these parameters [2]. Muscle protein synthesis (MPS), an important determinant of muscle mass and commensurate strength enhancements, is stimulated by RET [3]. The most common and widely recognized nutritional strategy to augment RET-induced adaptations, including MPS, is the intake of dietary protein (1.2-1.6 g·kg−1), especially the provision of essential amino acids [4]. Recently, long-chain omega-3 polyunsaturated fatty acids (LC n-3 PUFA), primarily eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA), have been investigated for their roles in MPS and skeletal muscle health [5].
The incorporation of EPA and DHA into skeletal muscle phospholipid has been shown to enhance both nutrient and mechanically sensitive anabolic signaling proteins known to regulate MPS [6]. Moreover, fish oil supplementation (FOS), a concentrated source of EPA and DHA, augments the anabolic response to nutrient stimuli in healthy young and middle-aged men and women through the activation of the mTOR-p70s6k signaling pathway leading to a ~ $50\%$ increase in MPS [7]. A study in resistance-trained young men demonstrated that FOS augmented the anabolic response elicited by protein feeding alone and with the addition of RET relative to placebo (PL), as indicated by a ~ $30\%$ and $35\%$ increase in MPS, respectively, in the absence of a concomitant increase in kinase signaling activity [8]. Moreover, these findings have corroborated similar alterations in protein signaling following skeletal muscle EPA and/or DHA incorporation [9,10]. As such, it appears that FOS may attenuate signaling cascades without compromising functional outcomes such as muscular hypertrophy and strength. Unlike targeted pharmaceutical interventions, LC n-3 PUFAs may act via several other mechanisms that may influence RET-induced adaptations such as muscle quality associated factors such as muscle fiber type transition or enhanced neuromuscular recruitment; muscle protein breakdown; improved insulin signaling; enhanced cell membrane fluidity; and modulation of inflammatory cytokines [5,11–13].
While plausible mechanisms exist, it is unclear if FOS influences functional skeletal muscle outcomes such as the promotion of hypertrophy, strength, and fat mass (FM) reduction in young adults following a RET program [14,15]. As noted by Anthony et al. [ 16] and James et al. [ 17], n-3 PUFA research is plagued with methodological flaws that may render some interpretations of primary outcomes tenuous at best. Since the influence of EPA and DHA on physiology is mediated by incorporation into tissue phospholipid membranes, the aforementioned authors proposed the use of a membrane-centric hypothesis to resolve these issues in n-3 PUFA research [16]. For example, two studies investigating the effects of RET and n-3 PUFA supplementation on skeletal muscle outcomes in young adults did not measure membrane nor blood LC n-3 PUFA status, likely employing both suboptimal EPA and DHA dosing (<1 g·d−1) and duration (4-8 weeks) protocols to meaningfully influence skeletal muscle incorporation [18,19]. Furthermore, as stated by Rossato et al. [ 15] and others [14], it is apparent from the lack of available data that concurrent RET and FOS trials should be undertaken, appropriately, to determine if LC n-3 PUFAs augment skeletal muscle functional outcomes in young adults.
Therefore, the aim of this study was to determine if there is an augmented response of FOS (3.85 g∙d−1 EPA+DHA), compared to PL, on body composition (LBM, FM, %BF) and 1-repetition maximum (1RM) strength during a 10-week RET program in young adults. Based on the current literature, we hypothesized that an adequate dose of FOS alongside RET, would facilitate increased LBM and strength, as well as a reduction in FM to a greater magnitude than RET plus PL. If so, targeted RET and FOS interventions can be developed.
## Study design
A randomized, single-blind, parallel-group design was used to examine the effects of FOS compared to placebo (PL) on body composition and strength during a 10-week RET program in young adults. Regarding blinding, one of the outcome assessors was aware of the allocations; however, all outcomes were conducted with other investigators present (see the Strength Testing section). A schematic overview of the study design is depicted in Figure 1. Figure 1.Schematic Overview of the Study Design and Resistance Exercise Training Protocol.
## Subjects
Twenty-eight young male ($$n = 12$$) and female ($$n = 16$$) adults were recruited from the local central *Texas area* and university population for this study. All participants met the following criteria: 1) between 18 and 40 years, 2) free from neuromuscular/musculoskeletal disorders and known chronic diseases (heart disease, type-2 diabetes mellitus, etc.), 3) did not regularly consume ergogenic or fish oil supplements within 6 months of starting the study, 4) consumed <2 servings of fatty fish per week, 5) did not take anabolic steroids or selective androgen receptor modulators, 6) reported both being recreationally trained (defined as RET twice per week for at least 6 months) and familiar (i.e. conducted both movements weekly for at least 6 months) with the barbell back squat and barbell bench press, and 7) have a body fat percentage (%BF) ≤26 in males and ≤36 in females. After briefing all study details, eligible subjects signed university-approved written consent forms. This study was approved by the Institutional Review Board for Human Subjects at Baylor University (#1630023-6). Of the initial 28 participants recruited, 7 withdrew from the study. Figure 2 outlines subject recruitment, randomization, and reasons for drop out. Figure 2.CONSORT Participant Flow Diagram.
Since there is little evidence to suggest sex differences in the anabolic response from RET with similar training statues [20], we opted to include male and female subjects. While hormonal variation does appear to uniquely impact women physiologically, there is no clear evidence that the menstrual cycle or oral contraceptive use significantly influences physical performance [21,22]. Furthermore, recent data suggests that RET-induced changes in hypertrophy and strength are minimally influenced by sex [23]. While some evidence suggests a differential response after FOS based on sex in older adults [24], this finding was not replicated in a recent trial [11] and has not been observed in young adults. Based on a recent dose-response study, we also anticipated similar changes in the red blood cell EPA+DHA following omega-3 supplementation, regardless of sex or age [25].
## Supplementation protocol
Groups were supplemented with FO (4.5 g∙d−1 [2.275 g∙d−1 EPA + 1.575 g∙d−1 DHA], 7 capsules; Nordic Naturals, ProOmega, Watsonville, CA, USA) or PL (safflower oil, 4.5 g∙d−1, 5 capsules; NOW, Bloomingdale, IL, USA) for 10-weeks. Based on recent data on skeletal muscle phospholipid incorporation of EPA+DHA following supplementation [6,26], our FOS dose needed to be >3 g∙d−1. To account for possible missed doses, we opted to provide a slightly higher dose (3.85 g∙d−1 EPA+DHA). Supplements were distributed every two weeks to encourage compliance. Adherence was confirmed by verbal confirmation and upon visual inspection of the bottles by the same laboratory technician. Additionally, blood fatty acid status was determined via fatty acid dried blood spot (DBS) as muscle and blood EPA+DHA content are highly correlated [6]. The supplements were packaged in similar bottles and the capsules were similar in size and shape in an attempt to blind the subject to the allocation. To assess blinding, at the end of the study, subjects were asked to guess which group they were in.
## Resistance training procedures
As illustrated in Figure 1, subjects conducted a partially supervised 10-week full-body RET protocol in three nonconsecutive sessions/days per week (two unsupervised and one supervised) consisting of seven exercises per session of 3-4 sets of 8-12 repetitions with 90-120 s rest intervals. See Figure 1 for the complete RET protocol. Briefly, the following exercises were performed in order: barbell back squat, leg press, leg extension/leg curl, barbell bench press, shoulder press, seated cable row, and wide-grip lat pulldown. Similar exercise regimes have been used previously to study various hypertrophy and strength outcomes [27,28]. One RET session per week was supervised by trained exercise physiologist and research personnel, while the other two RET sessions were completed by the participant for a total of 10 supervised sessions and 20 unsupervised sessions over 10-weeks. Subjects were prohibited from performing additional RET or high-intensity anaerobic training until the completion of the study.
Initial training loads were selected based on $70\%$ of the subject’s baseline 1RMs. The load was adjusted for all exercises based on the subject’s ability to reach momentary concentric failure between 8-12 repetitions. The load was decreased at the next training session if the subject completed less than eight repetitions on the final set or increased if the subject was able to complete all 12 repetitions on the final set. Load adjustments were approximately 5-$10\%$ for each exercise.
## Measurements
Body Composition. Total body mass (kg) and height (cm) were determined using a standard scale with a stadiometer (Seca 703, Hamburg, Germany). At PRE and POST, body composition (%BF, FM, and LBM) was obtained under laboratory conditions (e.g. fasted, voided bladder, and same time of day) using dual-energy x-ray absorptiometry (DXA, Horizon DXA™, Hologic®, Bedford, MA). The same technician performed each DXA to minimize variability. Based on previous studies in our lab, the accuracy of the DXA is ± 2 to $3.7\%$ as compared to hydrodensitometry [29].
Strength Testing. Lower- and upper-body strengths were assessed by 1RM testing via the back squat (1RMSQT) and bench press (1RMBP) exercises, respectively. Subjects conducted a 5-minute bike warm-up then a self-directed dynamic warm-up for an additional 5-minutes that included one set of 10 repetitions with an unloaded 20.4 kg barbell. Subjects then completed a standardized warm-up protocol used previously in our lab consisting of 10 repetitions at approximately $50\%$ 1RM, 5 repetitions at $70\%$ 1RM, 3 repetitions at $80\%$ 1RM, and 1 repetition at $90\%$ 1RM [30]. All warm-ups were followed by 2-minute rest intervals. Subjects then performed sets of one repetition with increasing weight for 1RM determination. During the testing phases, 3-minute rest intervals were employed and all 1RM determinations were made within five attempts. For the 1RMSQT, subjects were required to reach parallel, in which the top of the thigh is discernably parallel to the floor, for a repetition to be considered successful. For the 1RMBP, the lift was deemed successful if the subject kept five points of contact (bench: head, upper back, buttocks; ground: both feet), touched the barbell to their chest (no pause), and executed a full lock-out. While we do not have lab-specific reliability measures for the 1RMBP and 1RMSQT, at least two study personnel were available for spotting and standards verification. All strength testing sessions were supervised, and standards additionally verified by the lead technician and one additional technician – all National Strength and Conditioning Association Certified Strength and Conditioning Specialists. The average PRE and POST 1RMSQT and 1RMBP were reported as absolute and relative to body weight (1RM [kg]∙body mass [kg]−1) values.
Volume Load. Volume load data, calculated as sets x reps x load, were obtained from bench press and back squat for each session. Combined volume load and volume load data for back squat and bench press separately from each week were averaged across 3-RET sessions and used for data analysis.
## Dietary records
Subjects were required to submit three-day food logs (two weekdays, one weekend day) before and after the RET intervention using the MyFitnessPal mobile or desktop application (MyFitnessPal; San Francisco, CA, USA). Additionally, LC n-3 PUFA intake was assessed using a food frequency questionnaire (FFQ) that has been validated against whole blood EPA and DHA [31]. While the subject diets were not standardized, they were asked to keep their dietary habits as consistent as possible. Additionally, subjects were asked to consume at least 1.0 g∙kg−1 per day of protein and given gram-specific targets per day by a registered dietitian. Macronutrients (kcals, protein, carbohydrate, fat) and n-3 fatty acids intake (EPA and DHA) were averaged both over the three-day tracking period and per group for analysis by the same registered dietitian. Lastly, macronutrient data were normalized to body mass (kg) for further analysis.
## Fatty acid dried blood spot
Fatty acid dried blood spot (DBS) was obtained to track supplementation compliance and ensure adequate LC n-3 PUFA membrane incorporation. A drop of blood was collected from each participant via finger stick on filter paper pre-treated with a preservation solution (Fatty Acid Preservative Solution, FAPS™) and allowed to dry at room temperature for ~15 minutes. At the conclusion of the study, the DBS were shipped overnight on dry ice to OmegaQuant (Sioux Falls, SD, USA) for fatty acid analysis. Based on their standard laboratory protocol, fatty acids were identified by comparison with a standard mixture of fatty acids characteristic of RBC (GLC OQ-A, NuCheck Prep, Elysian, MN, USA) and used to construct individual fatty acid calibration curves. Fatty acid composition was expressed as a percent of total identified fatty acids. PRE and POST values of EPA, DHA, and the omega-3 index (O3i) were reported. The O3i is defined as the sum of EPA and DHA adjusted by a regression equation ($r = 0.96$) to predict the RBC O3i.
## Statistical analyses
All statistical analyses were performed using IBM SPSS version 28 (Armonk, NY, USA). Data were tested for normality and homogeneity using the Shapiro-Wilks and Levene’s tests, respectively. Baseline characteristics were analyzed using an independent samples t-test. The sample size for this project was 26. This sample size was justified by a priori power analysis in G*power using a target effect size (ES) of $f = 0.35$, alpha of 0.05 and power of 0.80, which determined that 20 subjects were required for participation with an additional number of participants recruited to account for possible attrition. Of note, similar RET investigations with a nutritional intervention used identical per group sample sizes ($$n = 8$$-11) [18,19], even with cohorts including males and females [32]. The primary outcome data (body composition [LBM, FM, %BF], and strength [1RMSQT and 1RMBP]) were analyzed using an ANCOVA on the change scores with baseline values as the covariate. All other data with timepoints (PRE/POST: 1RM strength relative to body weight [kg], fatty acids [O3i, EPA, DHA], dietary variables [kcals, protein, carbohydrate, fats, EPA, and DHA], weeks 1-10: volume load) were analyzed using a two-way repeated measures ANOVA (group x time). If the assumption of sphericity (Mauchly’s test) was violated, the Greenhouse-Geisser correction was used. If significant interaction effects were present, pairwise comparison analyses were used with a Bonferroni adjustment for alpha inflation. Significance was set a priori at $p \leq .05.$ ES values are reported as Cohen’s d to infer the between-group magnitude of differences in change scores. ES values were classified according to Cohen [33] as trivial, < 0.2; small, 0.2 – 0.49; moderate, 0.5 – 0.79; and large, ≥ 0.8. All data presented as mean ± SD, unless otherwise stated.
## Results
Seven subjects dropped out during the study, resulting in a total of 21 subjects (FOS group, $$n = 10$$ [M: 5, F: 5); PL group, $$n = 11$$ [M: 5, F: 6]). Reasons for dropouts are noted in the participant flow diagram (Figure 2). The FO and PL groups had similar ($p \leq .05$) baseline characteristics (Table 1). Table 1.Baseline participant characteristics. Fish Oil($$n = 10$$; 5 men, 5 women)Placebo($$n = 11$$; 5 men, 6 women)p-valueAge (y)28.0 (7.4)30.5 (5.7).403Height (cm)169.7 (9.6)171.8 (8.9).679Weight (kg)75.1 (16.0)79.0 (16.0).906BMI (kg∙m−2)25.8 (3.5)26.6 (4.3).496Body Fat (%)23.9 (6.9)24.9 (8.0).766Training Age (y)1.8 (1.1)2.0 (1.0).652Omega-3 Index (%)4.9 (1.3)4.3 (0.9).209Data are mean (SD). Omega-3 Index = %EPA + %DHA in red blood cells
## Compliance
Supplement. Self-reported supplement compliance was $95.2\%$ for all participants. There was no difference in supplement compliance between groups (FOS: $94.6\%$, PL: $95.8\%$, F [1,19] = 0.331, $$p \leq .572$$). Fifty-seven percent of subjects (12 of 21; FOS: 6 of 10, PL: 6 of 11) were unable to ascertain their allocated group. Only two subjects in the FOS group reported experiencing ‘fishy burps’. No other symptoms or adverse effects were reported.
RET Protocol. Overall attendance for those who completed the study was similar between groups (supervised: F [1,19] = 0.022, $$p \leq .883$$; unsupervised; F [1,19] = 0.1118, $$p \leq .734$$). Participants in the PL and FOS groups had an $94.6\%$ and $95.0\%$ attendance for the RET supervised sessions, respectively, and a self-reported unsupervised session attendance of $95.0\%$ and $94.0\%$, respectively.
Fatty Acid Dried Blood Spot. The baseline average O3i for all subjects was $4.58\%$ ± 1.12 (FOS: $4.9\%$ ± 1.3, PL: $4.3\%$ ± 0.9). There were no baseline group differences in the O3i (F [1,19] = 1.688, $$p \leq .209$$) or whole blood values of EPA (F [1,19] = 0.309, $$p \leq .585$$) and DHA (F [1,19] = 1.829, $$p \leq .192$$). As noted in Figure 3, the O3i did not change in the PL group ($1.3\%$, $$p \leq .938$$) and significantly increased from PRE to POST in the FOS group ($109.7\%$, $p \leq .001$). Similarly, whole blood EPA and DHA did not significantly change in the PL group ($14.7\%$, $$p \leq .869$$; −$0.8\%$, $$p \leq .952$$, respectively) and significantly increased in the FOS group ($613.0\%$, $p \leq .001$; $69.9\%$, $p \leq .001$, respectively). At the individual level, all subjects in the FOS group increased their O3i. Figure 3.Participant Omega-3 Index Before (PRE) and After (POST) 10-Weeks of Supplementation. Black line with whiskers indicates mean ± SD.*significantly different than PRE ($p \leq .001$), #significant difference between groups ($p \leq .001$).
## Dietary intake
All dietary data, normalized by body mass (kg) for the macronutrients, are reported in Table 2. In brief, there were no significant differences between groups nor over time in self-reported calorie and macronutrient intake. Dietary intake of EPA and DHA was similar between groups and did not change from PRE to POST. Table 2.Nutritional analysis. GroupPre,Mean ± SDPost,Mean ± SDp (Time)p (GxT)Total energy (kcal)FO1889.2 ± 343.01905.1 ± 462.7.676.994 PL1901.1 ± 520.61916.5 ± 414.7Carbohydrate (g∙kg−1)FO2.6 ± 1.02.6 ± 1.0.985.868 PL2.7 ± 1.42.7 ± 0.9Protein (g∙kg−1)FO1.5 ± 0.31.5 ± 0.2.889.690 PL1.2 ± 0.31.3 ± 0.2Fat (g∙kg−1)FO1.0 ± 0.21.0 ± 0.3.529.938 PL1.0 ± 0.40.9 ± 0.2EPA (mg)FO29.6 ± 28.817.6 ± 18.6.306.114 PL13.9 ± 22.416.6 ± 17.1DHA (mg)FO63.8 ± 59.139.6 ± 41.0.345.104 PL33.3 ± 53.040.0 ± 37.8Amount of EPA and DHA does not include supplementationAbbreviations: FO, fish oil; PL, placebo; EPA, eicosapentaenoic acid; DHA, docosahexaenoic acid
## Volume load
Total volume load over the 10 weeks was similar between conditions (FOS: 42,670 ± 18,925 kg, PL: 43,879 ± 22,765 kg, $$p \leq .897$$). There were no between-group differences in total volume load for the back squat (FOS: 24,888 ± 9,985 kg, PL: 26,197 ± 12,914 kg, F [1,19] = 0.017, $$p \leq .897$$) or bench press (FOS: 17,782 ± 9,173 kg, PL: 17,683 ± 10,033 kg, $$p \leq .897$$). Back squat and bench press volume load significantly increased over 10-weeks ($p \leq .001$). Compared to baseline, back squat volume load was significantly higher at week 3 ($p \leq .001$), week 8 ($$p \leq .016$$), week 9 ($$p \leq .006$$), and week 10 ($p \leq .001$). For bench press, volume load was significantly higher in week 3 ($$p \leq .025$$), week 8 ($$p \leq .049$$), and week 10 ($p \leq .001$) compared to week 1. Volume load data for the back squat and bench press over 10-weeks and between groups are noted in Figure 4. Figure 4.Weekly Volume Load for the a) Back Squat and b) Bench Press. Data are mean ± SD.*significantly different than week 1 for PL (Back Squat: $$p \leq .016$$; Bench Press: $$p \leq .001$$), #significantly different than week 1 for FO (Back Squat: week 3, $$p \leq .003$$; week 10, $p \leq .001$; Bench Press: $$p \leq .031$$). There were no group by time interactions ($p \leq .05$).
## Body composition
As indicated in Table 3, there was no significant between-group differences in LBM (FOS: +$3.4\%$, PL: +$2.4\%$), FM (FOS: −$5.2\%$, PL: $0.0\%$), nor %BF (FOS: −$5.9\%$, PL: −$2.5\%$). Notably, the between-group magnitude was considered moderate and large for %BF (−0.91, $95\%$CI: −2.13, 0.31, $d = 0.74$) and FM (−1.08 kg, $95\%$CI: −2.36, 0.20, $d = 0.84$), respectively, favoring the FOS group. The 0.6 kg difference in LBM, favoring the FOS group, was considered small ($d = 0.36$). Table 3.Data of main study outcomes. OutcomesGroupPre,Mean ± SDPost,Mean ± SDUnadjusted ∆ ± SDBaselineAdjusted ∆ (CI)p (Group)ESSquat 1RM (kg)FO82.9 ± 35.0106.8 ± 38.423.9 ± 8.124.2 (17.5, 30.9).1910.64PL90.9 ± 43.1109.5 ± 48.618.6 ± 12.118.2 (11.8, 24.6)Bench 1RM (kg)FO62.7 ± 37.073.9 ± 40.711.1 ± 6.811.3 (7.7, 14.8).0471.00PL66.3 ± 37.072.7 ± 39.66.4 ± 5.06.3 (2.9, 9.7)LBM (kg)FO55.9 ± 15.657.9 ± 16.71.9 ± 1.92.0 (0.8, 3.1).4570.36PL58.2 ± 14.959.6 ± 15.41.4 ± 1.71.4 (0.3, 2.5)FM (kg)FO17.4 ± 3.816.5 ± 3.1−0.91 ± 1.2−1.0 (−1.9, −0.1).0920.84PL33.3 ± 53.019.3 ± 5.6−0.02 ± 1.60.1 (−0.8, 0.9)BF (%)FO23.9 ± 6.922.4 ± 6.2−1.4 ± 1.5−1.5 (−2.4, −0.6).1360.74PL24.9 ± 8.024.2 ± 7.3−0.63 ± 1.6−0.6 (−1.4, 0.3)Bold indicates a significant p-value ($p \leq .05$) or large effect size (≥ 0.8)Abbreviations: ∆, change; CI, $95\%$ confidence interval (upper bound, lower bound); ES, effect size (Cohen’s d); 1RM, 1-repetition maximum; LBM, lean body mass; FM, fat mass; BF, body fat
## Strength testing
The 1RM strength data are shown in Table 3. Relative to PL, 10-weeks of FOS and RET increased absolute 1RMBP (5.0 kg, $95\%$CI: 0.07, 9.93, $$p \leq .047$$) and tended to increase absolute 1RMSQT (6.0 kg, $95\%$CI: −3.29, 15.31, $$p \leq .191$$). Figure 5 depicts the change in relative 1RM strength from PRE to POST. Briefly, the change in relative 1RMBP and 1RMSQT was significantly higher in the FOS group compared to PL (0.14 kg∙kg−1 vs. 0.06 kg∙kg−1, $$p \leq .011$$ and 0.31 kg∙kg−1 vs. 0.20 kg∙kg−1, $$p \leq 0.045$$, respectively). Figure 5.Individual Data Points of the Change in Relative a) 1RM Bench Press (1RMBP) and b) Back Squat (1RMSQT) Before (PRE) and After (POST) 10-Weeks of Resistance Exercise Training and Supplementation (placebo [PL, $$n = 11$$] or fish oil [FOS, $$n = 10$$])*significant change from PRE for relative 1RMBP (PL, $$p \leq .006$$; FOS, $p \leq .001$) and 1RMSQT (PL, $p \leq .001$; FOS, $p \leq .001$); #significant difference between groups for relative1RMBP (PL: $7.3\%$ vs. FOS: $17.6\%$, $$p \leq .011$$) and 1RMSQT (PL: $17.9\%$ vs. FOS: $29.3\%$, $$p \leq .045$$).
## Discussion
To our knowledge, this is the first study to investigate the effects of FOS compared to PL on skeletal muscle adaptations following a 10-week RET program in young adults, while using a membrane-centric hypothesis. We demonstrated that 3.85 g combined EPA and DHA daily for 10-weeks resulted in significantly greater RET-induced gains in absolute and relative 1RMBP and relative 1RMSQT. Although we found moderate-to-large between-group differences based on ES for absolute 1RMSQT, %BF, and FM, they were statistically equivocal to PL. Contrary to our hypothesis, FOS failed to differentially influence LBM compared to PL.
In young adults, changes in RET-induced muscular strength are often associated with changes in muscle mass [2]. While LBM increased in both groups within the present investigation, 1RM strength – especially relative strength (kg∙kg−1 body mass) – improved to a greater extent in the FOS group compared to PL. Given that MPS is the primary contributor to hypertrophy in young adults [2,34], we hypothesized that strength improvements with FOS would be mediated by concomitant LBM enhancements. However, despite demonstrating similar between-group LBM changes, absolute 1RMBP, relative 1RMBP, and relative 1RMSQT were improved by 4.7 kg ($8\%$), 0.08 kg∙kg−1 ($10.3\%$), and 0.11 kg∙kg−1 ($11.4\%$), respectively, more than PL. While the relationship between muscle mass and strength is well-established, muscle quality, or strength per unit of muscle mass, can improve through several factors both independently or commensurate to LBM [35]. Thus, beyond hypertrophy, other factors such as fiber type distribution, reductions in intramuscular fat, and enhanced neuromuscular activation may have uniquely improved measures of muscular strength [36].
Although speculative, the baseline adjusted between-group LBM difference of 0.6 kg (+$1.2\%$, $p \leq .05$), may be partially explained by enhanced MPS. FOS in the presence of nutrient stimuli has been shown to improve MPS by $50\%$ in young adults [7]. However, this effect appears to be modestly attenuated following an acute bout of RET in a similar demographic [8]. Notwithstanding that trained-individuals experience a somewhat blunted RET-mediated MPS response [37], we expected a more pronounced MPS-associated LBM accrual among our recreationally trained subjects. Nevertheless, MPS changes following an acute RET bout may not correlate with chronic RET-induced skeletal muscle hypertrophy [38,39]. The between-group differences observed in the present investigation were nonetheless statistically equivocal; regardless, it remains plausible that FOS and its subsequent skeletal muscle phospholipid incorporation upregulated muscle protein synthetic machinery, albeit to a much smaller degree than expected. Evidence suggests that the efficacy of omega-3 fatty acids on the activity of anabolic cell signaling pathways and, thus, LBM may be acutely influenced by protein intake. While previous studies have reported substantial increases in MPS (~50-$100\%$) with FOS [6,7,40], McGlory et al. [ 8] recently reported that MPS may not be acutely upregulated by omega-3 fatty acids in the presence of optimal protein intake. Since our participants reported dietary protein intakes ≥ 1.2 g∙kg−1, it is possible that the muscle protein synthetic machinery was saturated to the extent that FOS would not have exerted an additional anabolic response and, by extension, similar rates of LBM accrual.
Other alternate explanations for the FOS-mediated greater magnitudes of 1RM strength improvement despite similar LBM augmentations may be related to muscle quality and its associated factors such as muscle fiber type transition and neuromuscular function. Since relative strength in the bench press and back squat changed to a greater degree than absolute strength compared to PL, FOS most likely influenced mechanisms related to muscle quality. Although performed in elderly populations, many fish oil supplementation studies with concurrent RET have noted similar strength improvements without significant LBM changes [11,24]. Furthermore, a study in resistance-trained young men found that 4 g∙d−1 FOS improved 1RM leg extension, notably despite a loss in muscle mass during a $40\%$ calorie restricted diet [41]. While our study protocol led to hypertrophy in both groups, this illustrates the ability of fish oil supplementation to improve strength even in an otherwise compromised physiological environment. This may partially be explained by fast-twitch muscle fiber hypertrophy [42]. In older adults, 6 weeks of 3.68 g∙d−1 LC n-3 PUFA (1.86 g EPA, 1.54 g DHA) administration alongside RET significantly increased fast-twitch muscle fiber cross-sectional area (fCSA) in the absence of whole-body LBM alterations [42]. These data corroborated similar investigations’ fiber type-specific data following LC n-3 PUFA supplementation [43,44]. While LBM increased to a similar degree between groups, it is plausible, although highly speculative, based on our differential 1RM strength outcomes, that fCSA increased to a greater extent in the FOS group. Our FOS protocol was intentionally EPA-biased (1.4:1, EPA:DHA) since it has been shown to uniquely influence MPS [12]; however, DHA is the prominent fatty acid involved in neuromuscular control [45]. As previously hypothesized by Philpott et al. [ 41], the DHA component of our supplement significantly increased blood DHA (~$70\%$), thus, it is reasonable to assume that neural DHA concomitantly increased. Consequently, Rodacki et al. [ 13] demonstrated greater neural activation (i.e. faster muscular response after a stimulus) following 90 days of combined FOS and RET. Previous authors have reported that neuromuscular adaptations may occur as early as 21-days following FOS, although the effect may be attenuated compared to studies using higher doses or longer durations [46]. In agreement with previous investigations [41], these data may therein support a FOS-mediated neuromuscular enhancement that ultimately influenced the observed strength outcomes.
Much of the evidence to date on RET and FOS is in older adults and largely reports favorable results for LC n-3 PUFA supplementation [11,13,24,42]. To our knowledge, there are only two studies that used a RET protocol combined with n-3 PUFA supplementation alone or as part of a protein-based supplement in young adults [18,19]. Specifically, Georges et al. [ 18] found that 8-weeks of RET with 3 g·d−1 krill oil administration significantly improved both LBM and 1RM bench and leg press from PRE to POST in young trained men; however, the results were nonetheless similar to placebo. Hayward et al. [ 19] presented similar findings in untrained females amidst a combined 4-week RET program and protein-based supplement containing n-3 PUFAs. Although LBM and strength were improved from baseline, no significant differences were noted compared to controls. The present study also reported similar changes in LBM compared to PL; however, our investigation was the first to report greater improvements in strength outcomes. Notably, our observed FOS-mediated strength increases compared to other studies are ostensibly due to the skeletal muscle LC n-3 PUFA incorporation, likely facilitated by our more optimal dosing regimen. In line with our results, a recent systematic review and meta-analysis, albeit in older adults, found that FOS does not increase LBM; however, it does improve strength with or without RET [47].
Although there are significant strengths to the present investigation, such as the inclusion of young females to further our understanding of the effect of FOS and RET across the general young adult population, blood LC n-3 PUFA status measurement, equivalent macronutrient intake, as well as similar volume loads and RET-induced strength changes compared to resistance trained subjects [27,28,48], this study was not without limitations. While the number of subjects enrolled in our study was similar to previous investigations, it’s possible that our study may have been underpowered to detect significant between-group changes in certain outcomes, especially in light of our wide confidence intervals on some measures (Table 3). The DXA is widely considered the reference method to determine changes in body composition when using a standardized protocol; however, our findings could have been strengthened with the use of total body water or more direct methods of site-specific skeletal muscle mass and type quantification (MRI, muscle biopsies, etc.). Lastly, subject diets were not controlled and, although unlikely, it remains possible that dietary fluctuations may have unintentionally influenced the participants’ PRE-to-POST body composition.
## Conclusion
In summary, our data confirms, once again, that RET is key for beneficial skeletal muscle adaptations and body recomposition in young healthy men and women. The addition of fish oil supplementation (4 g∙d−1 [3.85 g EPA+DHA]) to a 10-week RET program may augment absolute 1RMBP and relative 1RMBP and 1RMSQT, and a greater reduction in FM. It is unclear if FOS increases LBM to a greater extent than RET alone. In light of our results and previous findings regarding the efficacy of FOS in young and athletic populations [14,49], FOS may be explored as a feasible and cost-effective nutritional strategy to influence general health and training adaptations, primarily for those with low blood or suboptimal dietary intake of LC n-3 PUFAs. Unlike targeted pharmaceutical interventions, the complex and often unspecified action of LC n-3 PUFAs – especially the notable divergent actions of EPA and DHA – on human physiology can make identification of an underlying mechanism challenging. Nevertheless, the convergence of multiple known contributors, including MPS, muscle quality characteristics, and neuromuscular control, likely contributed to our findings. As Anthony and colleagues [16] antecedently suggest, future research in healthy, young trained personnel is warranted to examine the influence of FOS on RET-induced adaptations using more precise tracer, muscle biopsy, and neuromuscular assessments to ascertain the aforementioned underlying mechanisms. Additionally, these subsequent investigations should explore the differential effects of EPA and DHA on skeletal muscle functional outcomes, potentially benefiting from longer duration (>10 weeks) supplementation protocols and doses equivalent to 3 servings of fatty fish per week.
## Disclosure statement
No potential conflict of interest was reported by the author(s).
## Author contributions
Study Design: JLH, LKF, JSF, NDB, AJK, ST; Funding Acquisition: JLH, LKF, AG; Data Collection: JLH, LKF, SBM, DRH, and LCS; Data Analysis: JLH and ST; Manuscript Writing: JLH, SBM, LKF, DRH, NDB, AJK, JSF; Final Manuscript Review and Approval: JLH, SBM, DRH, LKF, JSF, NDB, AJK, ST, LCS, and AG.
## References
1. Wolfe RR.. **The underappreciated role of muscle in health and disease**. *Am J Clin Nutr* (2006) **84** 475-50. PMID: 16960159
2. Schoenfeld B, Fisher J, Grgic J. **Resistance training recommendations to maximize muscle hypertrophy in an athletic population: position Stand of the IUSCA**. *J Strength Cond* (2021) 1
3. Phillips SM, Tipton KD, Aarsland A. **Mixed muscle protein synthesis and breakdown after resistance exercise in humans**. *Am J Physiol* (1997) **273** E99-107. PMID: 9252485
4. Morton RW, Murphy KT, McKellar SR. **A systematic review, meta-analysis and meta-regression of the effect of protein supplementation on resistance training-induced gains in muscle mass and strength in healthy adults**. *Br J Sports Med* (2018) **52** 376-384. PMID: 28698222
5. McGlory C, Calder PC, Nunes EA. **The influence of omega-3 fatty acids on skeletal muscle protein turnover in health, disuse, and disease**. *Front Nutr* (2019) **6** 144. PMID: 31555658
6. McGlory C, Galloway SDR, Hamilton DL. **Temporal changes in human skeletal muscle and blood lipid composition with fish oil supplementation**. *Prostaglandins Leukot Essent Fatty Acids* (2014) **90** 199-206. PMID: 24726616
7. Smith GI, Atherton P, Reeds DN. **Omega-3 polyunsaturated fatty acids augment the muscle protein anabolic response to hyperinsulinaemia–hyperaminoacidaemia in healthy young and middle-aged men and women**. *Clin Sci* (2011) **121** 267-278
8. McGlory C, Wardle SL, Macnaughton LS. *Physiol Rep* (2016) 4
9. Turvey EA, Heigenhauser GJF, Parolin M. **Elevated n-3 fatty acids in a high-fat diet attenuate the increase in PDH kinase activity but not PDH activity in human skeletal muscle**. *J Appl Physiol* (2005) **98** 350-355. PMID: 15591305
10. Wardle SL, Macnaughton LS, McGlory C. **Human skeletal muscle metabolic responses to 6 days of high-fat overfeeding are associated with dietary n-3PUFA content and muscle oxidative capacity**. *Physiol Rep* (2020) **8** e14529. PMID: 32845565
11. Dalle S, Van Roie E, Hiroux C. **Omega-3 supplementation improves isometric strength but not muscle anabolic and catabolic signaling in response to resistance exercise in healthy older adults**. *J Gerontol A Biol Sci Med Sci* (2021) **76** 406-414. PMID: 33284965
12. Kamolrat T, Gray SR. **The effect of eicosapentaenoic and docosahexaenoic acid on protein synthesis and breakdown in murine C2C12 myotubes**. *Biochem Biophys Res Commun* (2013) **432** 593-598. PMID: 23438435
13. Rodacki CL, Rodacki AL, Pereira G. **Fish-oil supplementation enhances the effects of strength training in elderly women**. *Am J Clin Nutr* (2012) **95** 428-436. PMID: 22218156
14. Heileson JL, Funderburk LK. **The effect of fish oil supplementation on the promotion and preservation of lean body mass, strength, and recovery from physiological stress in young, healthy adults: a systematic review**. *Nutr Rev* (2020) **78** 1001-1014. PMID: 32483626
15. Rossato LT, Schoenfeld BJ, de Oliveira EP. **Is there sufficient evidence to supplement omega-3 fatty acids to increase muscle mass and strength in young and older adults?**. *Clin Nutr* (2020) **39** 23-32. PMID: 30661906
16. Anthony R, Macartney MJ, Peoples GE. **The influence of long-chain omega-3 fatty acids on eccentric exercise-induced delayed muscle soreness: reported outcomes are compromised by study design issues**. *Int J Sport Nutr Exerc Metab* (2021)
17. James MJ, Sullivan TR, Metcalf RG. **Pitfalls in the use of randomised controlled trials for fish oil studies with cardiac patients**. *Br J Nutr* (2014) **112** 812-820. PMID: 24933212
18. Georges J, Sharp MH, Lowery RP. **The effects of krill oil on mTOR signaling and resistance exercise: a pilot study**. *J Nutr Metab* (2018) 7625981. PMID: 29854443
19. Hayward S, Wilborn CD, Taylor LW. **Effects of a High Protein and Omega-3-Enriched Diet with or Without Creatine Supplementation on Markers of Soreness and Inflammation During 5 Consecutive Days of High Volume Resistance Exercise in Females**. *J Sports Sci Med* (2016) **15** 704-714. PMID: 27928217
20. Burd NA, Tang JE, Moore DR. **Exercise training and protein metabolism: influences of contraction, protein intake, and sex-based differences**. *J Appl Physiol (1985)* (2009) **106** 1692-1701. PMID: 19036897
21. Elliott-Sale KJ, McNulty KL, Ansdell P. **The effects of oral contraceptives on exercise performance in women: a systematic review and meta-analysis**. *Sports Med* (2020) **50** 1785-1812. PMID: 32666247
22. McNulty KL, Elliott-Sale KJ, Dolan E. **The effects of menstrual cycle phase on exercise performance in eumenorrheic women: a systematic review and meta-analysis**. *Sports Med* (2020) **50** 1813-1827. PMID: 32661839
23. Roberts BM, Nuckols G, Krieger JW. **Sex differences in resistance training: a systematic review and meta-analysis**. *J Strength Cond Res* (2020) **34** 1692-1701
24. Da Boit M, Sibson R, Sivasubramaniam S. **Sex differences in the effect of fish-oil supplementation on the adaptive response to resistance exercise training in older people: a randomized controlled trial**. *Am J Clin Nutr* (2017) **105** 151-158. PMID: 27852617
25. Flock MR, Skulas‐Ray AC, Harris WS. **Determinants of erythrocyte omega‐3 fatty acid content in response to fish oil supplementation: a dose–response randomized controlled trial**. *J Am Heart Assoc* (2013) 2
26. McGlory C, Gorissen SHM, Kamal M. **Omega-3 fatty acid supplementation attenuates skeletal muscle disuse atrophy during two weeks of unilateral leg immobilization in healthy young women**. *FASEB J* (2019) **33** 4586-4597. PMID: 30629458
27. Schoenfeld BJ, Contreras B, Krieger J. **Resistance training volume enhances muscle hypertrophy but not strength in trained men**. *Med Sci Sports Exerc* (2019) **51** 94-103. PMID: 30153194
28. Schoenfeld BJ, Pope ZK, Benik FM. **Longer interset rest periods enhance muscle strength and hypertrophy in resistance-trained men**. *J Strength Cond Res* (2016) **30** 1448-1460
29. Shelmadine B, Cooke M, Buford T. **Effects of 28 days of resistance exercise and consuming a commercially available pre-workout supplement, NO-Shotgun®, on body composition, muscle strength and mass, markers of satellite cell activation, and clinical safety markers in males**. *J Int Soc Sports Nutr* (2009) **6** 16. PMID: 19656392
30. Machek SB, Cardaci TD, Wilburn DT. **Neoprene knee sleeves of varying tightness augment barbell squat one repetition maximum performance without improving other indices of muscular strength, power, or endurance**. *J Strength Cond Res* (2021) **35** S6-15. PMID: 33201154
31. Heileson JL, Elliott A, Buzzard JA. **A Cross-sectional analysis of whole blood long-chain ω-3 polyunsaturated fatty acids and its relationship with dietary intake, body composition, and measures of strength and power in collegiate athletes**. *J Am Coll Nutr* (2021) **40** 1-7. PMID: 33305991
32. Jacinto JL, Nunes JP, Gorissen SHM. **Whey protein supplementation is superior to leucine-matched collagen peptides to increase muscle thickness during a 10-week resistance training program in untrained young adults**. *Int J Sport Nutr Exerc Metab* (2022)
33. Cohen J. *Statistical power analysis for the behavioral sciences* (1998)
34. Atherton PJ, Smith K. **Muscle protein synthesis in response to nutrition and exercise**. *J Physiol* (2012) **590** 1049-1057. PMID: 22289911
35. Dankel SJ, Buckner SL, Jessee MB. **Correlations do not show cause and effect: not even for changes in muscle size and strength**. *Sports Med* (2018) **48** 1-6
36. Naimo MA, Varanoske AN, Hughes JM. **Skeletal muscle quality: a biomarker for assessing physical performance capabilities in young populations**. *Front Physiol* (2021) **12** 706699. PMID: 34421645
37. Witard OC, Bannock L, Tipton KD. **Making sense of muscle protein synthesis: a focus on muscle growth during resistance training**. *Int J Sport Nutr Exerc Metab* (2022) **32** 49-61. PMID: 34697259
38. Damas F, Phillips SM, Libardi CA. **Resistance training-induced changes in integrated myofibrillar protein synthesis are related to hypertrophy only after attenuation of muscle damage**. *J Physiol* (2016) **594** 5209-5222. PMID: 27219125
39. Mitchell CJ, Churchward-Venne TA, Parise G. **Acute post-exercise myofibrillar protein synthesis is not correlated with resistance training-induced muscle hypertrophy in young men**. *PLoS One* (2014) **9** e89431. PMID: 24586775
40. Smith GI, Atherton P, Reeds DN. **Dietary omega-3 fatty acid supplementation increases the rate of muscle protein synthesis in older adults: a randomized controlled trial**. *Am J Clin Nutr* (2011) **93** 402-412. PMID: 21159787
41. Philpott JD, Bootsma NJ, Rodriguez-Sanchez N. **Influence of fish oil-derived n-3 fatty acid supplementation on changes in body composition and muscle strength during short-term weight loss in resistance-trained men**. *Front Nutr* (2019) **6** 102. PMID: 31380384
42. Brook MS, Din U, Tarum J. **Omega-3 supplementation during unilateral resistance exercise training in older women: a within subject and double-blind placebo-controlled trial**. *Clin Nutr ESPEN* (2021) **46** 394-404. PMID: 34857226
43. Strandberg E, Ponsot E, Piehl-Aulin K. **Resistance training alone or combined with N-3 PUFA-rich diet in older women: effects on muscle fiber hypertrophy**. *J Gerontol A Biol Sci Med Sci* (2019) **74** 489-494. PMID: 30052800
44. Yamazaki H, Nishimura M, Uehara M. **Eicosapentaenoic acid changes muscle transcriptome and intervenes in aging-related fiber type transition in male mice**. *Am J Physiol Endocrinol Metab* (2021) **320** E346-58. PMID: 33225720
45. Salem N, Litman B, Kim H-Y. **Mechanisms of action of docosahexaenoic acid in the nervous system**. *Lipids* (2001) **36** 945-959. PMID: 11724467
46. Lewis E, Radonic P, Wolever T. **21 days of mammalian omega-3 fatty acid supplementation improves aspects of neuromuscular function and performance in male athletes compared to olive oil placebo**. *J Int Soc Sports Nutr* (2015) **12** 28. PMID: 26085822
47. Cornish SM, Cordingley DM, Shaw KA. **Effects of omega-3 supplementation alone and combined with resistance exercise on skeletal muscle in older adults: a systematic review and meta-analysis**. *Nutrients* (2022) **14** 2221. PMID: 35684018
48. Schoenfeld BJ, Peterson MD, Ogborn D. **Effects of low- vs. high-load resistance training on muscle strength and hypertrophy in well-trained men**. *J Strength Cond Res* (2015) **29** 2954-2963. PMID: 25853914
49. Lewis NA, Daniels D, Calder PC. **Are there benefits from the use of fish oil supplements in athletes? A systematic review**. *Adv Nutr* (2020) **11** 1300-1314. PMID: 32383739
|
---
title: Baicalein attenuates bleomycin-induced lung fibroblast senescence and lung
fibrosis through restoration of Sirt3 expression
authors:
- Yuan Ji-hong
- Ma Yu
- Yuan Ling-hong
- Gong Jing-jing
- Xu Ling-li
- Wang Lv
- Jin Yong-mei
journal: Pharmaceutical Biology
year: 2023
pmcid: PMC9970214
doi: 10.1080/13880209.2022.2160767
license: CC BY 4.0
---
# Baicalein attenuates bleomycin-induced lung fibroblast senescence and lung fibrosis through restoration of Sirt3 expression
## Abstract
### Context
Fibroblast senescence was reported to contribute to the pathological development of idiopathic pulmonary fibrosis (IPF), and baicalein is reported to attenuate IPF.
### Objective
This study explores whether baicalein attenuates lung fibrosis by regulating lung fibroblast senescence.
### Materials and methods
Institute of Cancer Research (ICR) mice were randomly assigned to control, bleomycin (BLM), baicalein and BLM + baicalein groups. Lung fibrosis was established by a single intratracheal dose of BLM (3 mg/kg). The baicalein group received baicalein orally (100 mg/kg/day). Sirtuin 3 (Sirt3) siRNA (50 μg) was injected through the tail vein once a week for 2 weeks to explore its effect on the anti-pulmonary fibrosis of baicalein.
### Results
BLM-treated mice exhibited obvious lung fibrosis and fibroblast senescence by showing increased levels of collagen deposition ($27.29\%$ vs. $4.14\%$), hydroxyproline (208.05 vs. 40.16 ng/mg), collagen I (25.18 vs. 9.15 μg/mg), p53, p21, p16, MCP-1, PAI-1, TNF-α, MMP-10 and MMP-12 in lung tissues, which were attenuated by baicalein. Baicalein also mitigated BLM-mediated activation of TGF-β1/Smad signalling pathway. Baicalein restored the BLM-induced downregulation of Sirt3 expression in lung tissues and silencing of Sirt3 abolished the inhibitory role of baicalein against BLM-induced lung fibrosis, fibroblast senescence and activation of TGF-β1/Smad signalling pathway.
### Conclusions
Baicalein preserved the BLM-induced downregulation of lung Sirt3 expression, and thus the suppression of TGF-β1/Smad signalling pathway and lung fibrosis, which might provide an experimental basis for treatment of IPF.
## Introduction
Idiopathic pulmonary fibrosis (IPF), the most common form of idiopathic interstitial pneumonia, causes progressive pulmonary fibrosis, which is characterized by repeated epithelial cell damage, destruction of alveolar structure and pro-fibrotic mediator-induced extracellular matrix (ECM) deposition by myofibroblasts. The occurrence and factors associated with IPF include genetic susceptibility and other risk factors, such as bacterial or viral infection, smoking and environmental pollution (Cao et al. 2021). IPF has a high mortality rate, with the average survival time being 2–4 years from diagnosis (Wolters et al. 2014). Currently, only pirfenidone and nintedanib are approved by the Food and Drug Administration (FDA) (Ma et al. 2022) and recommended in the recent ATS/ERS/JRS/ALAT clinical practice guideline for the treatment of IPF (Raghu et al. 2022). Although the clinical trials with pirfenidone and nintedanib in IPF were reported to reduce the decline in FVC compared to placebo and the antifibrotic therapies are associated with improved survival in patients with IPF, this therapy could not completely stop the deterioration of lung function over time (Ma et al. 2022; Takehara et al. 2022). Therefore, identifying effective therapeutic methods for IPF is urgent.
The formation of fibrotic foci in IPF is mainly due to excessive ECM protein deposition in the alveolar space, with activated myofibroblasts being the main producers of pulmonary ECM (Du S-F et al. 2019; Blokland et al. 2021; Chen T et al. 2021; Wang Q et al. 2021; Zhang et al. 2021; Röhrich et al. 2022; Wu et al. 2022). Wu et al. [ 2022] demonstrated that the expression of checkpoint kinases 1 and 2 (CHK$\frac{1}{2}$) was markedly increased in the lungs, remodelled pulmonary arteries and isolated fibroblasts from IPF patients and animal models. CHK$\frac{1}{2}$ inhibition could interfere with TGF-β1-mediated fibroblast activation, attenuating fibrosis and pulmonary vascular remodelling. Zhang et al. [ 2021] indicated that N-methyladenosine modification expression was upregulated in a bleomycin (BLM)-induced pulmonary fibrosis mouse model, fibroblast-to-myofibroblast (FMT)-derived myofibroblasts and IPF patient lung samples. N6-methyladenosine (m6A) modification contributes to IPF-induced pulmonary fibrosis by regulating FMT. Of note, an increasing number of studies highlight the critical role of fibroblast senescence in the pathological process of IPF, as well as the existence of an increased and persistent number of senescent fibroblasts in lung tissues with IPF (Schafer et al. 2017). Senolytic drugs, including quercetin and dasatinib plus quercetin have been reported to attenuate IPF-induced pulmonary fibrosis and dysfunction in mice by selectively eliminating senescent fibroblasts and/or myofibroblasts (Hecker et al. 2014; Alvarez et al. 2017; Schafer et al. 2017; Hohmann et al. 2019), providing a potential therapeutic method for IPF.
Baicalein (5,6,7-trihydroxyflavone) is a major phenolic flavonoid extracted from the root of *Scutellaria baicalensis* Georgi (Lamiaceae) and widely used to treat multiple diseases, such as Alzheimer’s disease, Parkinson’s disease (Li Y et al. 2017), neuroinflammation (Rui et al. 2020), ischemia–reperfusion-induced brain injury (Yang et al. 2019), hyperuricaemia (Chen Y et al. 2021), sepsis-induced liver injury (Liu A et al. 2015), avian pathogenic Escherichia coli-induced acute lung injury, pulmonary arterial hypertension (Shi et al. 2018) and pulmonary fibrosis (Gao et al. 2013; Sun X et al. 2020), due to its anti-inflammatory, antioxidant and anti-apoptosis effects. Gao et al. [ 2013] reported that baicalein mitigated lung fibrosis in a rat model of IPF by suppressing miRNA (miR)-21 and TGF-β/Smad signalling. Sun et al. [ 2020] showed that baicalein attenuated TGF-β1-mediated collagen production in lung fibroblasts by inhibiting the expression of connective tissue growth factor.
In the present study, a mouse model of BLM-induced IPF was used to explore the protective effects of baicalein. Baicalein could inhibit BLM-induced lung fibroblast senescence and pulmonary fibrosis by suppressing the TGF-β1/Smad signalling pathway. It was also found that baicalein markedly increased sirtuin 3 (Sirt3) expression in the lung tissue of BLM-treated mice, which may contribute to the inhibitory effects of baicalein against BLM-induced lung fibroblast senescence and pulmonary fibrosis.
## Animals and drug administration
Institute of Cancer Research (ICR) mice (male; 8 weeks old) were provided by Shanghai SLAC Laboratory Animal Co. (Shanghai, China) and provided with free access to food and water in a controlled temperature of 23–25 °C. All animal protocols were approved by the Ethics Committee of the Experimental Animals of Shanghai Seventh People’s Hospital (approval no. 2020-AR-053). Mice were anesthetized using sodium pentobarbital (i.p., 30 mg/kg; MilliporeSigma, Burlington, MA) and then subjected to a single intratracheal dose of BLM (3 mg/kg; Selleck Chemicals, Houston, TX) diluted in 50 µL sterile saline. Mice that had been intratracheally administrated 50 µL sterile saline served as the control. In the baicalein group, mice received baicalein orally (100 mg/kg/day; MilliporeSigma, Burlington, MA); the dose was based on a previous study, which showed that baicalein (p.o. 100 mg/kg/day) markedly mitigated BLM-induced lung fibrosis (Gao et al. 2013).
## Experimental groups and drug treatment
The first experiment was performed to explore the effect of baicalein on BLM-mediated pulmonary fibrosis. Mice were randomly divided into four groups: (i) control group ($$n = 7$$), mice were intratracheally administrated sterile saline; (ii) BLM group ($$n = 7$$), mice were intratracheally administrated BLM; (iii) baicalein group ($$n = 7$$), mice were administrated saline intratracheally and baicalein orally; and (iv) BLM + baicalein group ($$n = 7$$), mice were administrated BLM intratracheally and baicalein orally. Mice were sacrificed after 2 weeks of BLM and baicalein administration.
The second experiment aimed to examine whether Sirt3 siRNA could abolish the protective role of baicalein against BLM-induced pulmonary fibrosis. Mice were randomly divided into six groups: (i) control group ($$n = 7$$); (ii) BLM group ($$n = 7$$); (iii) BLM + baicalein group ($$n = 7$$); (iv) Sirt3 siRNA group ($$n = 7$$); (v) BLM + Sirt3 siRNA group ($$n = 7$$); and (vi) BLM + baicalein + Sirt3 siRNA group ($$n = 7$$). Following drug administration, mice in the control siRNA group were injected with 50 μg control siRNA and those in the Sirt3 siRNA group were injected 50 μg Sirt3 siRNA through the tail vein once a week for 2 weeks.
## Sirt3 siRNA administration
In vivo-jetPEI™ was used for control and Sirt3 siRNA administration, according to the manufacturer’s instructions. Briefly, control or Sirt3 siRNA (50 μg) was dissolved in a 100-µL mixture of equal volumes of in vivo-jetPEI™ and $10\%$ glucose, and then injected into the tail vein once a week for 2 weeks. The Sirt3 siRNAs were designed and synthesized by Shanghai GenePharma Co., Ltd. (Shanghai, China) and the sequences were as follows: Sirt3 siRNA sense, 5′-GUCUGAAGCAGUACAGAAAtt-3′ and antisense, 5′-UUUCUGUACUGCUUCAGACaa-3′ (Srivastava et al. 2018); control siRNA sense, 5′-UUCUCCGAACGUGUCACGUTT-3′ and antisense, 5′-ACGUGACACGUUCGGAGAATT-3′ antisense (Tang et al. 2018).
## Masson’s trichrome staining
The left lower pulmonary tissues were fixed in paraformaldehyde and then dehydrated in graded alcohol and embedded in paraffin. Paraffin-embedded sections (5 µm) were subjected to Masson’s trichrome staining (Wuhan Servicebio Technology Co., Ltd., Wuhan, China) to measure the fibrotic areas, according to the manufacturer’s instructions (Du S-F et al. 2019). Image-Pro Plus software version 6.0 (Media Cybernetics, Inc., Rockville, MD) was used to quantify the fibrotic area by manually examining the blue area (Du S-F et al. 2019). The researcher examining the fibrotic area was blinded to group allocation.
## Measurement of hydroxyproline and collagen I content
Cold PBS containing proteinase inhibitor cocktail (Sigma-Aldrich; Merck KgaA, Darmstadt, Germany) was used to homogenize pulmonary tissues, and hydroxyproline content (Winching, Nanjing, China) and collagen I (R&D Systems, Inc., Minneapolis, MN) were examined according to the manufacturer’s instructions.
## Isolation of mouse lung fibroblasts
Mouse pulmonary tissues were cut into pieces and then digested for 90 min as 37 °C with gentle shaking. Dulbecco’s modified Eagle’s medium-prepared digestion solution contained collagenase type III (0.1 U/mL, Worthington, Lakewood, NJ), trypsin ($0.125\%$, Gibco; Thermo Fisher Scientific, Inc., Waltham, MA) and DNase I (0.1 U/mL, Thermo Fisher Scientific, Inc., Waltham, MA). Following filtration, cells were collected for subsequent experiments (Sun X et al. 2015).
## Reverse transcription-quantitative PCR (RT-qPCR)
TRIzol® reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA) was used to extract total RNA from lung fibroblasts and pulmonary tissues, which was then reverse-transcribed to cDNA using superscript reverse transcriptase (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA) with oligodeoxythymidine for mRNAs. A CFX Connect real-time PCR detection system (Bio-Rad Laboratories, Inc., Hercules, CA) was used to perform qPCR with the following primer sequences: Sirt3 (accession no. NM_001127351.1) forward, 5′-AGCAACCTTCAGCAGTATGACATCC-3′ and reverse, 5′-TTTCACAACGCCAGTACAGACAGG-3′; monocyte chemotactic protein-1 (MCP-1; accession no. NM_011333) forward,5′-CCACTCACCTGCTGCTACTCATTC-3′ and reverse, 5′-GTTCACTGTCACACTGGTCACTCC-3′; TNF-α (accession no. NM_001278601.1) forward, 5′-CACCACGCTCTTCTGTCTACTGAAC-3′ and reverse, 5′-TGACGGCAGAGAGGAGGTTGAC-3′; plasminogen activator inhibitor-1 (PAI-1; accession no. NM_008871.2) forward, 5′-TCAATGACTGGGTGGAAAGGCATAC-3′ and reverse, 5′-AGATGTTGGTGAGGGCGGAGAG-3′ MMP-10 (accession no. NM_019471.3) forward, 5′-GCCTACCAATCTGCTCAGCGTATC-3′ and reverse, 5′-TGAAGCCACCAACATCAGGAACAC-3′ reverse; and MMP-12 (accession no. NM_001320076.1) forward, 5′-TCAATGACTGGGTGGAAAGGCATAC-3′ and reverse, 5′-AGATGTTGGTGAGGGCGGAGAG-3′. The 20 μL reaction solution contained 5 μL diluted cDNA, 0.5 μM paired primer, 10 μL SYBRGreen mix (Aidlab Biotechnologies Co., Ltd., Beijing, China) and 4.9 μL DEPC water. The annealing temperature and amplification were set at 60 °C and 40 cycles, respectively. The comparative Cq method (2–ΔΔCq) was performed to measure the relative gene expression (Du JK et al. 2016). mRNA levels were normalized to those of β-actin.
## Western blotting
Cold RIPA lysis buffer (Beyotime Institute of Biotechnology, Shanghai, China) with Protease Inhibitor Cocktail (Roche Diagnostics, Basel, Switzerland) was used to lyse lung fibroblasts and pulmonary tissues. Following determination of protein concentrations using BCA assays, the isolated protein was separated using $10\%$ SDS-PAGE and then transferred to PVDF membranes. Following blocking with non-fat dry milk dissolved in TBS for 2 h at room temperature, the membranes were incubated with BSA-diluted primary antibodies against Sirt3 (1:1000; cat. no. sc-365175; Santa Cruz Biotechnology, Inc., Dallas, TX), p16 (1:1000, cat. no. sc-56330; Santa Cruz Biotechnology, Inc., Dallas, TX), Smad4 (1:1000; cat. no. sc-7966; Santa Cruz Biotechnology, Inc., Dallas, TX), p-Smad2 (p-Smad2; 1:1000; Cell Signaling Technology, Inc., Danvers, MA), p-Smad3 (1:1000; Cell Signaling Technology, Inc., Danvers, MA), p53 (1:1000; cat. no. 10422-1-AP; ProteinTech Group, Inc., Rosemont, IL), p21 (1:1000; cat. no. 27296-1-AP; ProteinTech Group, Inc., Rosemont, IL), αSMA (1:1000, cat. no. sc-8432; Santa Cruz Biotechnology, Inc., Dallas, TX), fibronectin (1:1000, cat. no. sc-8422; Santa Cruz Biotechnology, Inc., Dallas, TX), Smad2 (1:1000; cat. no. sc-393312 Santa Cruz Biotechnology, Inc., Dallas, TX), Smad3 (1:1000; cat. no. sc-101154; Santa Cruz Biotechnology, Inc., Dallas, TX) or β-actin (1:1000; cat. no. sc-81178; Santa Cruz Biotechnology, Inc., Dallas, TX) at 4 °C overnight. Next, the membranes were incubated with a horseradish peroxidase-conjugated secondary antibody (1:1000) for 2 h at room temperature. An Enhanced Chemiluminescence Western Blotting Detection system (Santa Cruz Biotechnology, Inc., Dallas, TX) and a GeneGnome HR scanner (Syngene Europe, Frederick, MD) were used to visualize the bands.
## Statistical analysis
The data are presented as the mean ± SEM. Statistical analysis was performed using SPSS 16.0 (SPSS, Inc., Chicago, IL). Statistical comparisons between two groups were determined using a two-tailed Student’s t-test. One- or two-way ANOVA with a Bonferroni’s post hoc test was performed for comparisons among multiple groups. $p \leq 0.05$ was considered to indicate a statistically significant difference.
## Baicalein mitigates BLM-induced lung fibrosis in the mouse model
The impact of baicalein on BLM-induced lung fibrosis was first explored. As demonstrated in Figure 1(A,B), Masson’s trichrome staining revealed considerable collagen deposition in the pulmonary interstitium, with collagen deposition primarily observed in the alveolar walls along with epithelial thickening and cellular infiltrates. BLM also increased the expression of fibronectin and α-SMA in the protein level of lung tissue (Fig. S1A). Baicalein could significantly mitigate BLM-induced collagen deposition, structural damage and elevated expression of fibronectin and α-SMA in the lung tissue, as evidenced by a decrease in the amount of collagen deposition, relatively normal pulmonary architecture and low expression of fibronectin and α-SMA (Figure 1(A,B), Fig. S1A).
**Figure 1.:** *Baicalein mitigates BLM-induced lung fibrosis. (A, B) Masson’s trichrome staining was performed to measure collagen deposition in pulmonary tissues in control, BLM, baicalein and BLM + baicalein groups. Representative images (A) and changes in the ratio of collagen-deposited areas to lung substance areas (a morphometric measure of pulmonary fibrosis) (B). Hydroxyproline (C) and collagen I (D) contents in pulmonary tissues were examined by ELISA in control, BLM, baicalein and BLM + baicalein groups. Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. BLM: bleomycin.*
Next, the amount of hydroxyproline and collagen I was measured to quantify the amount of collagen deposited in the lungs. As shown in Figure 1(C,D), the amount of hydroxyproline and collagen I was significantly increased in BLM-treated mice and baicalein could reverse the BLM-induced increase in the content of hydroxyproline and collagen I, indicating that baicalein mitigates BLM-induced lung fibrosis.
## Baicalein mitigates BLM-induced lung fibroblast senescence in the mouse model
Next, the effects of baicalein on BLM-induced cell senescence in the lung tissue were explored. As shown in Figure 2(A), the protein levels of senescence effectors, including p53, p21 and p16, were significantly elevated in the lung tissue of BLM-treated mice. The ratio of the SA-β-gal-positive senescent cells was also increased in isolated lung fibroblast of BLM-treated mice (Fig. S1B and C). However, the increased expression of p53, p21 and p16 and the elevated ratio of SA-β-gal-positive senescent cells were markedly reversed by baicalein treatment. In addition, the lung tissue in BLM-treated mice demonstrated markedly increased transcript levels of proinflammatory and profibrotic senescence-associated secretory phenotype (SASP) factors, including MCP-1, PAI-1, TNF-α, MMP-10 and MMP-12, which were partly attenuated by baicalein treatment (Figure 2(B)).
**Figure 2.:** *Baicalein mitigates BLM-induced senescence in lung tissues and isolated lung fibroblasts. (A, B) Protein levels of p53, p21 and p16 and the mRNA levels of SASP factors, including MCP-1, PAI-1, TNF-α, MMP-10 and MMP-12 in lung tissues were measured in control, BLM, baicalein and BLM + baicalein groups. (C, D) Protein levels of p53, p21 and p16 and the mRNA levels of MCP-1, PAI-1, TNF-α, MMP-10 and MMP-12 in isolated lung fibroblasts were measured using western blotting and RT-qPCR, respectively. Representative protein bands were presented on the top of the histograms (A, C). Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. SASP: senescence-associated secretory phenotype; MCP-1: monocyte chemotactic protein-1; PAI-1: plasminogen activator inhibitor-1; TNF-α: tumour necrosis factor-α; MMP: matrix metalloproteinase; RT-qPCR: reverse transcription-quantitative PCR; BLM: bleomycin.*
To further clarify the direct effect of baicalein on lung fibroblast senescence in BLM-treated mice, the lung fibroblasts in BLM-treated mice exhibited an obvious increase in the protein levels of p53, p21 and p16, as well as the transcript levels of proinflammatory and profibrotic SASP factors, and this was reversed by baicalein treatment (Figure 2(C,D)), indicating that baicalein mitigated BLM-induced fibroblast senescence.
## Baicalein ameliorates BLM-induced activation of TGF-β1/Smad signalling in lung tissues
The TGF-β1/Smad pathway has been widely reported to play a crucial role in tissue fibrosis (Tseliou et al. 2014; Tran et al. 2019; Yao et al. 2019; Chale-Dzul et al. 2020; Hussein et al. 2020; Li X-F et al. 2020; Du J-K et al. 2021). As shown in Figure 3(A,B), the transcript levels of TGF-β1 and protein levels of p-Smad2, p-Smad3 and Smad4 were increased in the lung tissue of BLM-treated mice compared with the control, suggesting that BLM activates TGF-β1-Smad signalling in the lung. It was also found that baicalein could suppress the BLM-induced TGF-β1 production, p-Smad2, p-Smad3 and Smad4 expression in the lung.
**Figure 3.:** *Baicalein mitigates BLM-induced TGF-β1/Smad signalling in lung tissue. (A) The mRNA expression levels of TGF-β1 in pulmonary tissues were examined using RT-qPCR. (B) The protein expression levels of p-Smad2, p-Smad3 and Smad4 in pulmonary tissues were examined using western blotting. Representative immunoblots and the corresponding histograms are presented. Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. BLM: bleomycin; TGF-β1: transforming growth factor-β1; Smad: mothers against decapentaplegic homolog; RT-qPCR: reverse transcription-quantitative PCR.*
## Baicalein restores Sirt3 expression in the lung tissues of BLM-treated mice
A previous study indicated that the downregulation of Sirt3 contributes to aging-associated tissue fibrosis by blocking TGF-β expression (Sundaresan et al. 2015). As shown in Figure 4(A,B), BLM treatment resulted in an obvious decrease in the mRNA and protein expression levels of Sirt3 in the lungs. Baicalein could prevent the BLM-induced decrease in the lung tissue expression of Sirt3.
**Figure 4.:** *Baicalein prevents BLM-induced downregulation of Sirt3 in lung tissues. RT-qPCR and western blotting analysis were performed to measure the (A) mRNA and (B) protein Sirt3 expression and in the lung tissues of control, BLM, baicalein and BLM + baicalein groups. Representative immunoblots of Sirt3 and the corresponding histograms are presented. Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. p-: phosphorylated. Sirt3: sirtuin 3; BLM: bleomycin.*
## Sirt3 siRNA abrogates the protective role of baicalein against BLM-induced pulmonary fibrosis
Next, the impact of Sirt3 siRNA on the protective role of baicalein against BLM-induced pulmonary fibrosis was explored. As shown in Figure 5(A), Sirt3 siRNA resulted in an ∼$80\%$ decrease in Sirt3 expression in the lung tissues, and Sirt3 knockdown blocked the protective effects of baicalein against BLM-induced pulmonary fibrosis, as evidenced by the decreased collagen deposition (Figure 5(B)), and the levels of hydroxyproline (Figure 5(C)), collagen I (Figure 5(D)), aSMA and fibronectin (Fig. S2A and B). These results suggested that Sirt3 siRNA could abolish the protective role of baicalein against BLM-induced pulmonary fibrosis.
**Figure 5.:** *Silencing of Sirt3 abolishes the protective effect of baicalein against BLM-induced lung fibrosis. Masson’s trichrome staining was performed to measure collagen deposition in pulmonary tissues. (A, B) Representative images and the ratio of fibrotic areas to the total lung area. (C) Hydroxyproline and (D) collagen I content in pulmonary tissues was examined using ELISA. Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. $$p < 0.01 vs. BLM + baicalein. Sirt3: sirtuin 3; BLM: bleomycin.*
## Sirt3 siRNA abolishes the protective role of baicalein against BLM-induced lung fibroblast senescence and activation of TGF-β1/Smad signalling in lung tissues
Next, the impact of Sirt3 siRNA on the beneficial role of baicalein against BLM-induced lung fibroblast senescence was explored. As shown in Figure 6, Fig. S2C and D, Sirt3 siRNA blocked the protective effects of baicalein against BLM-induced senescence and the alteration of the levels of proinflammatory and profibrotic SASP factors in the lung tissue and isolated lung fibroblasts, as evidenced by the decreased protein levels of p53, p21, p16 and the ratio of the SA-β-gal-positive senescent cells in isolated lung fibroblast, as well as the transcript levels of proinflammatory and profibrotic SASP factors.
**Figure 6.:** *Sirt3 silencing abolishes the protective effect of baicalein against BLM-induced senescence in lung tissue and isolated lung fibroblasts. (A, B, D and E) The protein expression levels of Sirt3, p53, p21 and p16 in pulmonary tissues were measured using western blotting. (C, F) The mRNA levels of MCP-1, PAI-1, TNF-α, MMP-10 and MMP-12 were measured using RT-qPCR. Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. $$p < 0.01 vs. BLM + baicalein. Sirt3: sirtuin 3; BLM: bleomycin; MCP-1: monocyte chemotactic protein-1; PAI-1: plasminogen activator inhibitor-1; TNF-α: tumour necrosis factor-α; MMP: matrix metalloproteinase; RT-qPCR: reverse transcription-quantitative PCR.*
We then explored the impact of Sirt3 siRNA on the role of baicalein against BLM-induced activation of TGF-β1-smad pathway. Results showed that Sirt3 siRNA blocked the inhibitory effect of baicalein against BLM-induced activation of TGF-β1/Smad signalling, as shown by increased transcript levels of TGF-β1 (Figure 7(A)) and protein levels of p-Smad2, p-Smad3 and Smad4 (Figure 7(B,C)) in the lung tissue.
**Figure 7.:** *Sirt3 knockdown abolishes the protective effects of baicalein on BLM-induced TGF-β1/Smad signalling in the lung tissue. (A) The mRNA expression levels of TGF-β1 in pulmonary tissues were examined using RT-qPCR. (B) The protein expression levels of p-Smad2, p-Smad3 and Smad4 in pulmonary tissues were examined using western blotting. Data are presented as the mean ± SEM (n = 7). **p < 0.01 vs. control. ##p < 0.01 vs. BLM. $$p < 0.01 vs. BLM + baicalein. Sirt3: sirtuin 3; BLM: bleomycin; TGF-β1: transforming growth factor-β1; Smad: mothers against decapentaplegic homolog; RT-qPCR: reverse transcription-quantitative PCR.*
## Discussion
IPF is a chronic, progressive, fibrosing interstitial lung disease that affects hundreds of thousands of people worldwide, reducing their quality of life and leading to death from respiratory failure within years of diagnosis. Herein, BLM administration was found to result in significant pulmonary fibrosis and inflammation, and baicalein was found to play a positive role in preventing BLM-induced fibrosis and inflammatory responses in a mouse model. Our findings suggested that the dysregulation of Sirt3-mediated pulmonary fibroblast senescence contributed to BLM-induced lung fibrosis, and Sirt3 knockdown blocked the protective role of baicalein in preventing BLM-induced fibrosis.
Baicalein is a common plant flavonoid with a wide range of beneficial pharmacological properties, including anti-inflammatory (Teng et al. 2020; D’Amico et al. 2021), antioxidative (D'Amico et al. 2019), anti-apoptotic (Liu C et al. 2010; Li X-x et al. 2012; Lin M et al. 2014; Hung et al. 2016; Wang M et al. 2020), antitumorigenic (Li J et al. 2019) and pro-immunoregulatory functions (Shi et al. 2018). In addition, previous studies suggested that baicalein has antifibrotic potential in several types of tissues, including the lung (Gao et al. 2013), kidney (Hu et al. 2017), liver (Sun H et al. 2010) and heart (Kong et al. 2010). Baicalein has been reported to reverse BLM-induced lung fibrosis in rats, which was partly achieved through the inhibition of TGF-β/Smad signalling. Consistent with these findings, it was also found that baicalein could attenuate BLM-induced pulmonary fibrosis in mice. The protective role of baicalein against BLM-induced pulmonary fibrosis was partly independent in inhibiting TGF-β/Smad signalling, as shown by the improved lung architecture, decreased collagen deposition and amounts of hydroxyproline and type I collagen, as well as TGF-β1 production and phosphorylation of Smad2, Smad3 and Smad4 expression in the lung.
Cell senescence refers to a relatively stable state in which cells irreversibly leave the cell cycle and lose their proliferative ability under the action of signal transduction. In recent decades, cell senescence has attracted widespread attention as it increases the morbidity of fibroproliferative pulmonary diseases in elderly individuals (Parimon et al. 2021). Recent studies have demonstrated that epithelial progenitor cell dysfunction and cellular senescence, including epithelial and fibroblast senescence, were associated with the pathological development of IPF (Demaria et al. 2014; Lehmann et al. 2017). Hohmann et al. [ 2019] found that quercetin could attenuate BLM-induced lung fibrosis and injury by inhibiting fibroblast senescence and enhancing FasL- or TRAIL-induced apoptosis. Of note, Cui et al. [ 2018] demonstrated that baicalein could mitigate TGF-β1-mediated lung FMT differentiation through the inhibition of miR-21 expression. In the present study, it was found that baicalein reversed BLM-induced lung fibroblast senescence and increased the transcript levels of proinflammatory and profibrotic SASP factors in mice. However, whether baicalein enhances lung fibroblast apoptosis in BLM-treated mice and regulates the expression of miR-21 is worthy of further study.
Sirt3 dysregulation was reported to be involved in the pathological process of lung fibrosis. Clinically, it was demonstrated that Sirt3 was absent within fibrotic areas, as compared with adjacent areas within the same tissue in the scleroderma and IPF specimens (Sosulski et al. 2017). It was also reported that Sirt3 was deficient in the alveolar epithelial cells of IPF patients (Cheresh et al. 2021); these findings indicated that Sirt3 may have therapeutic potential in the management of lung fibrosis. In animal models of pulmonary fibrosis, the expression of Sirt3 was significantly decreased in the lung tissue of Ad-TGF-β1-treated mice (Sosulski et al. 2017). Sirt3 overexpression attenuated asbestos-mediated lung fibrosis and the beneficial effect of Sirt3 overexpression was associated with decreased lung mtDNA damage and Mo-AM recruitment (Jacobs et al. 2008). Consistent with the study, Sosulski et al. [ 2017] found that Sirt3-deficient mice were susceptible to lung fibrosis, and Sirt3 overexpression mitigated TGF-β1-induced FMT differentiation. The role of Sirt3 in lung fibrosis and FMT differentiation was associated with Smad3 (Schafer et al. 2017). Nevertheless, it was also found that Sirt3 expression was significantly decreased in the lung tissues of BLM-treated mice and Sirt3 downregulation contributed to the pathological process of BLM-mediated pulmonary fibrosis.
Located in the mitochondrial matrix, Sirt3 is a member of the Sirtuin family. It was reported to deacetylate and activate mitochondrial forkhead box class O 3a, which then regulated defective mitochondrial clearance through transcriptional regulation of autophagy-related genes (Jacobs et al. 2008). In addition, Im et al. [ 2015] showed that autophagy was deficient in lung fibroblasts and in the fibrotic lungs of IPF patients. A limitation of the present study was that it did not examine mitochondrial function and autophagy. In addition, no in vitro experiments were performed in the present study, which would have improved its relevance for IPF and/or senescence. Inhibition experiments with baicalein using primary human IPF fibroblasts should be performed, and the robustness of senescence markers can be improved by performing senescence-associated-β-galactosidase staining and measuring SASP factors, in addition to gene expression, using ELISA.
## Conclusions
Baicalein inhibited the BLM-mediated activation of TGF-β1/Smad and lung fibroblast senescence, which was in parallel with the protective roles of baicalein against BLM-mediated lung fibrosis. Furthermore, baicalein preserved the BLM-induced downregulation of lung Sirt3 expression, and thus the suppression of TGF-β1/Smad signalling pathway and lung fibrosis, which might provide an experimental basis for treatment of IPF (Figure 8).
**Figure 8.:** *Schematic diagram of the mechanism through which baicalein attenuates BLM-induced pulmonary fibrosis through upregulating Sirt3 expression. BLM suppressed Sirt3 expression, and Sirt3 dysregulation contributed to pulmonary fibroblast senescence and inflammation, resulting in pulmonary fibrosis, which was markedly mitigated by baicalein administration. BLM: bleomycin; Sirt3: sirtuin 3.*
## Disclosure statement
All authors declare no conflict of interest.
## References
1. Alvarez D, Cardenes N, Sellares J, Bueno M, Corey C, Hanumanthu VS, Peng Y, D'Cunha H, Sembrat J, Nouraie M. **IPF lung fibroblasts have a senescent phenotype**. *Am J Physiol Lung Cell Mol Physiol* (2017) **313** L1164-L1173. PMID: 28860144
2. Blokland KEC, Habibie H, Borghuis T, Teitsma GJ, Schuliga M, Melgert BN, Knight DA, Brandsma CA, Pouwels SD, Burgess JK.. **Regulation of cellular senescence is independent from profibrotic fibroblast-deposited ECM**. *Cells* (2021) **10** 1628. PMID: 34209854
3. Cao X, Li Y, Shi J, Tang H.. **The potentially therapeutic role of EPAC in curbing the process of idiopathic pulmonary fibrosis via differential cellular pathways**. *J Inflamm Res* (2021) **14** 611-619. PMID: 33679138
4. Chale-Dzul J, Perez-Cabeza de Vaca R, Quintal-Novelo C, Olivera-Castillo L, Moo-Puc R.. **Hepatoprotective effect of a fucoidan extract from**. *Int J Biol Macromol* (2020) **145** 500-509. PMID: 31874267
5. Chen T, Guo Y, Wang J, Ai L, Ma L, He W, Li Z, Yu X, Li J, Fan X. **LncRNA CTD-2528L19.6 prevents the progression of IPF by alleviating fibroblast activation**. *Cell Death Dis* (2021) **12** 600. PMID: 34112765
6. Chen Y, Zhao Z, Li Y, Yang Y, Li L, Jiang Y, Lin C, Cao Y, Zhou P, Tian Y. **Baicalein alleviates hyperuricemia by promoting uric acid excretion and inhibiting xanthine oxidase**. *Phytomedicine* (2021) **80** 153374. PMID: 33075645
7. Cheresh P, Kim S-J, Jablonski R, Watanabe S, Lu Z, Chi M, Helmin KA, Gius D, Budinger GRS, Kamp DW.. **SIRT3 overexpression ameliorates asbestos-induced pulmonary fibrosis, mt-DNA damage, and lung fibrogenic monocyte recruitment**. *Int J Mol Sci* (2021) **22** 6856. PMID: 34202229
8. Cui X, Sun X, Lu F, Jiang X.. **Baicalein represses TGF-beta1-induced fibroblast differentiation through the inhibition of miR-21**. *Toxicol Appl Pharmacol* (2018) **358** 35-42. PMID: 30201452
9. D'Amico R, Fusco R, Gugliandolo E, Cordaro M, Siracusa R, Impellizzeri D, Peritore AF, Crupi R, Cuzzocrea S, Di Paola R.. **Effects of a new compound containing palmitoylethanolamide and baicalein in myocardial ischaemia/reperfusion injury**. *Phytomedicine* (2019) **54** 27-42. PMID: 30668378
10. D’Amico R, Genovese T, Cordaro M, Siracusa R, Gugliandolo E, Peritore AF, Interdonato L, Crupi R, Cuzzocrea S, Di Paola R. **Palmitoylethanolamide/baicalein regulates the androgen receptor signaling and NF-kappaB/Nrf2 pathways in benign prostatic hyperplasia**. *Antioxidants* (2021) **10** 1014. PMID: 34202665
11. Demaria M, Ohtani N, Youssef SA, Rodier F, Toussaint W, Mitchell JR, Laberge RM, Vijg J, Van Steeg H, Dolle ME. **An essential role for senescent cells in optimal wound healing through secretion of PDGF-AA**. *Dev Cell* (2014) **31** 722-733. PMID: 25499914
12. Du JK, Cong BH, Yu Q, Wang H, Wang L, Wang CN, Tang XL, Lu JQ, Zhu XY, Ni X.. **Upregulation of microRNA-22 contributes to myocardial ischemia–reperfusion injury by interfering with the mitochondrial function**. *Free Radic Biol Med* (2016) **96** 406-417. PMID: 27174562
13. Du J-K, Yu Q, Liu Y-J, Du S-F, Huang L-Y, Xu D-H, Ni X, Zhu X-Y.. **A novel role of kallikrein-related peptidase 8 in the pathogenesis of diabetic cardiac fibrosis**. *Theranostics* (2021) **11** 4207-4231. PMID: 33754057
14. Du S-F, Wang X-L, Ye C-L, He Z-J, Li D-X, Du B-R, Liu Y-J, Zhu X-Y.. **Exercise training ameliorates bleomycin-induced epithelial mesenchymal transition and lung fibrosis through restoration of H**. *Acta Physiol* (2019) **225** e13177
15. Gao Y, Lu J, Zhang Y, Chen Y, Gu Z, Jiang X.. **Baicalein attenuates bleomycin-induced pulmonary fibrosis in rats through inhibition of miR-21**. *Pulm Pharmacol Ther* (2013) **26** 649-654. PMID: 23523661
16. Hecker L, Logsdon NJ, Kurundkar D, Kurundkar A, Bernard K, Hock T, Meldrum E, Sanders YY, Thannickal VJ.. **Reversal of persistent fibrosis in aging by targeting Nox4–Nrf2 redox imbalance**. *Sci Transl Med* (2014) **6** 231-247
17. Hohmann MS, Habiel DM, Coelho AL, Verri WA, Hogaboam CM.. **Quercetin enhances ligand-induced apoptosis in senescent idiopathic pulmonary fibrosis fibroblasts and reduces lung fibrosis**. *Am J Respir Cell Mol Biol* (2019) **60** 28-40. PMID: 30109946
18. Hu Q, Gao L, Peng B, Liu X.. **Baicalin and baicalein attenuate renal fibrosis**. *Exp Ther Med* (2017) **14** 3074-3080. PMID: 28928802
19. Hung KC, Huang HJ, Wang YT, Lin AM.. **Baicalein attenuates α-synuclein aggregation, inflammasome activation and autophagy in the MPP**. *J Ethnopharmacol* (2016) **194** 522-529. PMID: 27742410
20. Hussein KH, Park KM, Yu L, Kwak HH, Woo HM.. **Decellularized hepatic extracellular matrix hydrogel attenuates hepatic stellate cell activation and liver fibrosis**. *Mater Sci Eng C Mater Biol Appl* (2020) **116** 111160. PMID: 32806289
21. Im J, Hergert P, Nho RS.. **Reduced FoxO3a expression causes low autophagy in idiopathic pulmonary fibrosis fibroblasts on collagen matrices**. *Am J Physiol Lung Cell Mol Physiol* (2015) **309** L552-L561. PMID: 26186945
22. Jacobs KM, Pennington JD, Bisht KS, Aykin-Burns N, Kim H-S, Mishra M, Sun L, Nguyen P, Ahn B-H, Leclerc J. **SIRT3 interacts with the daf-16 homolog FOXO3a in the mitochondria, as well as increases FOXO3a dependent gene expression**. *Int J Biol Sci* (2008) **4** 291-299. PMID: 18781224
23. Kong EKC, Huang Y, Sanderson JE, Chan K-B, Yu S, Yu C-M.. **Baicalein and wogonin inhibit collagen deposition in SHR and WKY cardiac fibroblast cultures**. *BMB Rep* (2010) **43** 297-303. PMID: 20423617
24. Lehmann M, Korfei M, Mutze K, Klee S, Skronska-Wasek W, Alsafadi HN, Ota C, Costa R, Schiller HB, Lindner M. **Senolytic drugs target alveolar epithelial cell function and attenuate experimental lung fibrosis**. *Eur Respir J* (2017) **50** 1602367. PMID: 28775044
25. Li J, Tian C, Xia Y, Mutanda I, Wang K, Wang Y.. **Production of plant-specific flavones baicalein and scutellarein in an engineered**. *Metab Eng* (2019) **52** 124-133. PMID: 30496827
26. Li X-F, Zhang S-H, Liu G-F, Yu S-N.. **miR-363 alleviates detrusor fibrosis via the TGF-beta1/Smad signaling pathway by targeting Col1a2 in rat models of STZ-induced T2DM**. *Mol Ther Nucleic Acids* (2020) **22** 1142-1153. PMID: 33294298
27. Li X-x, He G-r, Mu X, Xu B, Tian S, Yu X, Meng F-r, Xuan Z-h, Du G-h. **Protective effects of baicalein against rotenone-induced neurotoxicity in PC12 cells and isolated rat brain mitochondria**. *Eur J Pharmacol* (2012) **674** 227-233. PMID: 21996316
28. Li Y, Zhao J, Holscher C.. **Therapeutic potential of baicalein in Alzheimer’s disease and Parkinson’s disease**. *CNS Drugs* (2017) **31** 639-652. PMID: 28634902
29. Lin M, Li L, Li L, Pokhrel G, Qi G, Rong R, Zhu T.. **The protective effect of baicalin against renal ischemia–reperfusion injury through inhibition of inflammation and apoptosis**. *BMC Complement Altern Med* (2014) **14** 19. PMID: 24417870
30. Liu A, Wang W, Fang H, Yang Y, Jiang X, Liu S, Hu J, Hu Q, Dahmen U, Dirsch O.. **Baicalein protects against polymicrobial sepsis-induced liver injury via inhibition of inflammation and apoptosis in mice**. *Eur J Pharmacol* (2015) **748** 45-53. PMID: 25533331
31. Liu C, Wu J, Xu K, Cai F, Gu J, Ma L, Chen J.. **Neuroprotection by baicalein in ischemic brain injury involves PTEN/AKT pathway**. *J Neurochem* (2010) **112** 1500-1512. PMID: 20050973
32. Ma H, Liu S, Li S, Xia Y.. **Targeting growth factor and cytokine pathways to treat idiopathic pulmonary fibrosis**. *Front Pharmacol* (2022) **13** 918771. PMID: 35721111
33. Parimon T, Hohmann MS, Yao C.. **Cellular senescence: pathogenic mechanisms in lung fibrosis**. *Int J Mol Sci* (2021) **22** 6214. PMID: 34207528
34. Raghu G, Remy-Jardin M, Richeldi L, Thomson CC, Inoue Y, Johkoh T, Kreuter M, Lynch DA, Maher TM, Martinez FJ. **Idiopathic pulmonary fibrosis (an update) and progressive pulmonary fibrosis in adults: an Official ATS/ERS/JRS/ALAT Clinical Practice Guideline**. *Am J Respir Crit Care Med* (2022) **205** e18-e47. PMID: 35486072
35. Röhrich M, Leitz D, Glatting FM, Wefers AK, Weinheimer O, Flechsig P, Kahn N, Mall MA, Giesel FL, Kratochwil C. **Fibroblast activation protein-specific PET/CT imaging in fibrotic interstitial lung diseases and lung cancer: a translational exploratory study**. *J Nucl Med* (2022) **63** 127-133. PMID: 34272325
36. Rui W, Li S, Xiao H, Xiao M, Shi J.. **Baicalein attenuates neuroinflammation by inhibiting NLRP3/caspase-1/GSDMD pathway in MPTP induced mice model of Parkinson’s disease**. *Int J Neuropsychopharmacol* (2020) **23** 762-773. PMID: 32761175
37. Schafer MJ, White TA, Iijima K, Haak AJ, Ligresti G, Atkinson EJ, Oberg AL, Birch J, Salmonowicz H, Zhu Y. **Cellular senescence mediates fibrotic pulmonary disease**. *Nat Commun* (2017) **8** 14532. PMID: 28230051
38. Shi R, Zhu D, Wei Z, Fu N, Wang C, Liu L, Zhang H, Liang Y, Xing J, Wang X. **Baicalein attenuates monocrotaline-induced pulmonary arterial hypertension by inhibiting endothelial-to-mesenchymal transition**. *Life Sci* (2018) **207** 442-450. PMID: 29969608
39. Sosulski ML, Gongora R, Feghali-Bostwick C, Lasky JA, Sanchez CG.. **Sirtuin 3 deregulation promotes pulmonary fibrosis**. *J Gerontol A Biol Sci Med Sci* (2017) **72** 595-602. PMID: 27522058
40. Srivastava SP, Li J, Kitada M, Fujita H, Yamada Y, Goodwin JE, Kanasaki K, Koya D.. **SIRT3 deficiency leads to induction of abnormal glycolysis in diabetic kidney with fibrosis**. *Cell Death Dis* (2018) **9** 997. PMID: 30250024
41. Sun H, Che QM, Zhao X, Pu XP.. **Antifibrotic effects of chronic baicalein administration in a CCl**. *Eur J Pharmacol* (2010) **631** 53-60. PMID: 20079350
42. Sun X, Chen E, Dong R, Chen W, Hu Y.. **Nuclear factor (NF)-kappaB p65 regulates differentiation of human and mouse lung fibroblasts mediated by TGF-β**. *Life Sci* (2015) **122** 8-14. PMID: 25498897
43. Sun X, Cui X, Chen X, Jiang X.. **Baicalein alleviated TGF β1-induced type I collagen production in lung fibroblasts via downregulation of connective tissue growth factor**. *Biomed Pharmacother* (2020) **131** 110744. PMID: 32932046
44. Sundaresan NR, Bindu S, Pillai VB, Samant S, Pan Y, Huang J-Y, Gupta M, Nagalingam RS, Wolfgeher D, Verdin E. **SIRT3 blocks aging-associated tissue fibrosis in mice by deacetylating and activating glycogen synthase kinase 3β**. *Mol Cell Biol* (2015) **36** 678-692. PMID: 26667039
45. Takehara K, Koga Y, Hachisu Y, Utsugi M, Sawada Y, Saito Y, Yoshimi S, Yatomi M, Shin Y, Wakamatsu I. **Differential discontinuation profiles between pirfenidone and nintedanib in patients with idiopathic pulmonary fibrosis**. *Cells* (2022) **11** 143. PMID: 35011705
46. Tang LX, Wang B, Wu ZK.. **Aerobic exercise training alleviates renal injury by interfering with mitochondrial function in type-1 diabetic mice**. *Med Sci Monit* (2018) **24** 9081-9089. PMID: 30551123
47. Teng C, Lin C, Huang F, Xing X, Chen S, Ye L, Azevedo HS, Xu C, Wu Z, Chen Z. **Intracellular codelivery of anti-inflammatory drug and anti-miR 155 to treat inflammatory disease**. *Acta Pharm Sin B* (2020) **10** 1521-1533. PMID: 32963947
48. Tran BH, Yu Y, Chang L, Tan B, Jia W, Xiong Y, Dai T, Zhong R, Zhang W, Le VM. **A novel liposomal S-propargyl-cysteine: a sustained release of hydrogen sulfide reducing myocardial fibrosis via TGF-β1/Smad pathway**. *Int J Nanomedicine* (2019) **14** 10061-10077. PMID: 31920303
49. Tseliou E, Reich H, de Couto G, Terrovitis J, Sun B, Liu W, Marban E.. **Cardiospheres reverse adverse remodeling in chronic rat myocardial infarction: roles of soluble endoglin and Tgf-β signaling**. *Bas Res Cardiol* (2014) **109** 443
50. Wang M, Dong Y, Wu J, Li H, Zhang Y, Fan S, Li D.. **Baicalein ameliorates ionizing radiation-induced injuries by rebalancing gut microbiota and inhibiting apoptosis**. *Life Sci* (2020) **261** 118463. PMID: 32950576
51. Wang Q, Liu J, Hu Y, Pan T, Xu Y, Yu J, Xiong W, Zhou Q, Wang Y.. **Local administration of liposomal-based Srpx2 gene therapy reverses pulmonary fibrosis by blockading fibroblast-to-myofibroblast transition**. *Theranostics* (2021) **11** 7110-7125. PMID: 34093874
52. Wolters PJ, Collard HR, Jones KD.. **Pathogenesis of idiopathic pulmonary fibrosis**. *Annu Rev Pathol* (2014) **9** 157-179. PMID: 24050627
53. Wu W-H, Bonnet S, Shimauchi T, Toro V, Grobs Y, Romanet C, Bourgeois A, Vitry G, Omura J, Tremblay E. **Potential for inhibition of checkpoint kinases 1/2 in pulmonary fibrosis and secondary pulmonary hypertension**. *Thorax* (2022) **77** 247-258. PMID: 34226205
54. Yang S, Wang H, Yang Y, Wang R, Wang Y, Wu C, Du G.. **Baicalein administered in the subacute phase ameliorates ischemia–reperfusion-induced brain injury by reducing neuroinflammation and neuronal damage**. *Biomed Pharmacother* (2019) **117** 109102. PMID: 31228802
55. Yao Y, Chen R, Wang G, Zhang Y, Liu F.. **Exosomes derived from mesenchymal stem cells reverse EMT via TGF-β1/Smad pathway and promote repair of damaged endometrium**. *Stem Cell Res Ther* (2019) **10** 225. PMID: 31358049
56. Zhang JX, Huang PJ, Wang DP, Yang WY, Lu J, Zhu Y, Meng XX, Wu X, Lin QH, Lv H. **m**. *Mol Ther* (2021) **29** 3436-3448. PMID: 34111558
|
---
title: 'The symphony of open-heart surgical care: A mixed-methods study about interprofessional
attitudes towards family involvement'
authors:
- Anna Drakenberg
- MiaLinn Arvidsson-Lindvall
- Elisabeth Ericsson
- Susanna Ågren
- Ann-Sofie Sundqvist
journal: International Journal of Qualitative Studies on Health and Well-being
year: 2023
pmcid: PMC9970227
doi: 10.1080/17482631.2023.2176974
license: CC BY 4.0
---
# The symphony of open-heart surgical care: A mixed-methods study about interprofessional attitudes towards family involvement
## ABSTRACT
### Purpose
The overall aim of this study was to describe the attitudes towards family involvement in care held by nurses and medical doctors working in open-heart surgical care and the factors influencing these attitudes.
### Methods
Mixed-methods convergent parallel design. A web-based survey was completed by nurses ($$n = 267$$) using the Families’ Importance in Nursing Care-Nurses Attitudes (FINC-NA) instrument and two open-ended questions, generating one quantitative and one qualitative dataset. Qualitative interviews with medical doctors ($$n = 20$$) were conducted in parallel, generating another qualitative dataset. Data were analysed separately according to each paradigm and then merged into mixed-methods concepts. Meta-inferences of these concepts were discussed.
### Results
The nurses reported positive attitudes in general. The two qualitative datasets from nurses and medical doctors resulted in the identification of seven generic categories. The main mixed-methods finding was the attitude that the importance of family involvement in care depends on the situation.
### Conclusions
The dependence of family involvement on the situation may be due to the patient’s and family’s unique needs. If professionals’ attitudes rather than the family’s needs and preferences determine how the family is involved, care runs the risk of being unequal.
## Introduction
Open-heart surgery, e.g., coronary artery bypass grafting, aortic procedures and cardiac valve replacement, is a common treatment for cardiovascular diseases (Head et al., 2017; Stephens & Whitman, 2015). Coronary artery bypass grafting is the most frequent cardiac surgical procedure in high-income countries (Head et al., 2017; McDermott & Liang, 2021), with a rate of approximately 63 per 100 000 inhabitants in the US (McDermott & Liang, 2021) and 44 of 100 000 in western Europe (Head et al., 2017). However, open-heart surgery may lead to life-threatening complications. Because of these risks, patients need to be monitored and cared for in an advanced, highly technological postoperative environment by a multiprofessional team (Stephens & Whitman, 2015). Just as a symphony requires an orchestra, open-heart surgical care requires a multiskilled team of professionals with various perspectives on patient care. A team consisting of cardiothoracic surgeons, anaesthesiologists, intensive care nurses, surgical nurses, physiotherapists, and other healthcare professionals (Stephens & Whitman, 2015) makes interprofessional collaboration essential for care quality and safety (Pomare et al., 2020; World Health Organization, 2010). Patients are often dependent on their families during the rehabilitation phase at home (Bjørnnes et al., 2019). Family involvement in care is recommended internationally (Davidson et al., 2017; Johnson & Abraham, 2012; Shajani & Snell, 2019) and has been known to improve both patient (Eskes et al., 2019; Mackie et al., 2019) and family (Bjørnnes et al., 2019; Joseph et al., 2015) outcomes.
The meaning of family involvement in care has a broad definition in this study. Family includes not only relations by bloodline or law but also emotional relationships (Benzein, Johansson, Årestedt, & Saveman, 2008). Family involvement requires family presence, information sharing, and the facilitation of family members’ participation in shared decision making and basic care activities (Olding et al., 2016). Furthermore, family involvement means that family members should be supported and have their own needs met by health care professionals (Olding et al., 2016).
Postoperative recovery may be enhanced when the family is involved in preventive patient care targeting surgical complications (Eskes et al., 2019), leading to an improvement in family satisfaction with care (Bjørnnes et al., 2019). The risks and consequences of open-heart surgery put a strain on both the patient and the patient’s family in terms of stress and anxiety (Bjørnnes et al., 2019; Joseph et al., 2015; Kemp et al., 2020; Robley et al., 2010). The family plays a key role in rehabilitation following open-heart surgery, a responsibility they are not always prepared for (Bjørnnes et al., 2019). Preoperative anxiety is associated with impaired postoperative recovery for the patient and may double all-cause mortality after open-heart surgery (Joseph et al., 2015). It has been suggested that stress and anxiety can be reduced for both family and patients when the family is involved in the care of their ill family member (Bjørnnes et al., 2019; Mackie et al., 2019). Attitudes of the health care team towards family involvement in care influence how families are treated and involved (Bell, 2013; Benzein, Johansson, Årestedt, & Saveman, 2008; Mackie et al., 2018). An attitude may be defined as a state of believing, valuing, or feeling something that predisposes an action or behaviour (Altmann, 2008). Attitudes predispose how we act, but this does not mean that we always act according to our attitudes (Altmann, 2008). Attitudes can be either conscious or unconscious and therefore cannot be measured directly and may be illuminated in our behaviour (Bakanauskas et al., 2020). Attitudes towards family involvement in care have been known to vary between care contexts and groups of health care professionals (Al Mutair et al., 2013; Barreto et al., 2022; Benzein, Johansson, Årestedt, Berg, et al., 2008; Davis et al., 2014; Dijkman et al., 2021; Jordan et al., 2014; R. Laidsaar-Powell et al., 2017; Rosland et al., 2011; Shin et al., 2017). Nurses’ attitudes towards family involvement have been explored to some extent in the context of cardiology (Gusdal et al., 2017; Luttik et al., 2017), surgical care (Blöndal et al., 2014) and a general nursing context (Benzein, Johansson, Årestedt, Berg, et al., 2008; Østergaard et al., 2020). *Nurses* generally hold positive attitudes towards family involvement in care, but some variations related to nurses´ personal experiences, educational level and context of workplace have been found in previous studies (Barreto et al., 2022). There are also significant differences in nurses’ attitudes between countries (Cranley et al., 2022; Shamali et al., 2022). At times, nurses’ negative attitudes may hinder family involvement, such as holding the belief that the patient comes first and family members take time away from patient care (Mackie et al., 2018). Research on medical doctors’ (MDs) attitudes towards family involvement in care has mostly focused on primary, geriatric and oncology care settings (Dijkman et al., 2021; R. C. Laidsaar-Powell et al., 2013; Rosland et al., 2011; Shin et al., 2017). Aspects of family involvement in care from the perspective of MDs in previous studies are shared decision making, communication and family presence (Dijkman et al., 2021; Jordan et al., 2014; R. C. Laidsaar-Powell et al., 2013; Shin et al., 2017). Surgeons’ interactions with families have been reported to vary (Jordan et al., 2014). For example, some always include the family in their preoperative communications while others never do (Jordan et al., 2014). Studies including both nurses’ and MDs’ attitudes towards family involvement in the care of adult patients are limited (Al Mutair et al., 2013; Davis et al., 2014; R. Laidsaar-Powell et al., 2017). When compared, nurses (R. Laidsaar-Powell et al., 2017), patients and families (Shin et al., 2017) hold more positive attitudes towards family involvement in oncology care than MDs. On the other hand, MDs tend to be more positive than nurses to family involvement in patient-safety practices in the acute care setting (Davis et al., 2014).
Because of the complexity of capturing attitudes, a mixed-methods approach was considered beneficial for the description of attitudes towards family involvement from informants from various professions. The merging of qualitative and quantitative results enables comparisons to be made, and a more complete understanding emerges than that provided by the quantitative and qualitative results alone. The overall aim was to describe the attitudes towards family involvement in care held by nurses and medical doctors working in open-heart surgical care and the factors influencing these attitudes.
The study aim was based on the following research questions. How do nurses and medical doctors working in open-heart surgical care describe their attitudes towards family involvement in care and the factors affecting these attitudes? ( Qualitative)How do nurses working in open-heart surgical care rate the importance of family involvement in nursing care? ( Quantitative)Is age, education and/or previous experiences associated with nurses’ ratings of their attitudes towards family involvement in nursing care? ( Quantitative) To what extent do these attitudes and components converge and diverge? ( Mixed-methods)
## Design
A mixed-methods convergent parallel design was used. This is a type of design in which qualitative and quantitative data are collected in parallel and analysed separately, and then the results are merged and meta-inferences can be made (Creswell & Plano Clark, 2018). The reporting in this study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (von Elm et al., 2014) and the Consolidated Criteria for Reporting Qualitative Research (COREQ) (Tong et al., 2007). Strategies to minimize validity threats in convergent mixed-method design (Creswell & Plano Clark, 2018) is applied in this study. The mixed-methods convergent parallel process is presented in Figure 1. Figure 1.The mixed method convergent design process: Data collection, analysis, merging and meta-inference.
## Setting
Patient autonomy and rights to participate in their own care is emphasized in c 3 § 4 of the Swedish Patient Act (SFS, 2014). In c 5 § 3, the family’s right to be involved in patient care is also established, given that the patient cannot receive information himself or herself and if it is deemed appropriate (SFS, 2014). In Sweden, open-heart surgical care is organized similarly in eight hospitals, seven of which are university hospitals. When a surgical indication is identified, the patient is referred to the cardiothoracic surgical department from the cardiology department. Patients are cared for in a surgical ward, the operating theatre, the intensive-care unit, and at some departments, a step-down unit. These units are staffed with health care professionals with various educational backgrounds, such as nurse assistants; registered nurses (some having a vocational degree and others having a bachelor’s degree); nurses with a postgraduate diploma in surgery, anaesthesiology or cardiology (some holding a master’s degree); physiotherapists (some holding a master’s degree in respiratory therapy or intensive care); occupational therapists; and MDs in training to become specialized in cardiology, cardiothoracic surgery or anaesthesiology. In addition to the abovementioned, resident cardiothoracic anaesthesiologists and surgeons work in all units. All units have close collaboration; for this study, nurses and MDs were included from the intensive care units, step-down units and/or surgical wards since these are places where family members are invited to visit. Family involvement in the operation theatre is an unusual exception in this context and foremost an issue when the patient is a child.
## Nurses
Nurses included in this study were recruited from all eight cardiothoracic departments in Sweden from April-November 2020. Data collection was planned for March-April 2020, but this period was the first wave of the COVID-19 pandemic in Sweden. Therefore, the departments themselves chose when to participate in this study, and they were given the opportunity to adjust the timing of their participation to their workload in relation to the strain of pandemic care. Three clinics agreed to complete the Families’ Importance in Nursing Care-Nurses Attitudes (FINC-NA) instrument from April-September 2020, while the remaining five clinics completed the instrument from September-November 2020. The nurses were instructed to answer while thinking of the normal situation when there were normal routines for family involvement in care.
The inclusion criteria were 1) being employed as a nurse at one of the wards included in the study and 2) caring for patients undergoing elective open-heart surgery and meeting these patients’ families.
After written consent was obtained from the head of each department, email addresses for the nurses were retrieved. Written information about the study was sent out via workplace email to 650 eligible participants. One week after the information was distributed, an electronic questionnaire was sent out via a unique link to each participant. Nurses willing to participate gave their consent by answering and submitting the questionnaire anonymously. Three reminders were sent to the first group, and four reminders were sent to the second group.
## Medical doctors
The MDs included in this study were recruited from three out of the seven cardiothoracic departments situated at university hospitals in Sweden. The MDs were working at the aforementioned university hospitals from April-November 2020.
The inclusion criteria were 1) being employed as an MD at one of the three cardiothoracic centres included in this part of the study and 2) treating patients undergoing elective open-heart surgery and meeting these patients’ families.
Recruitment of participants for the individual interviews started with an oral presentation in two out of three study sites. Due to travel restrictions during spring 2020, it was not possible to visit the third study site as planned to provide group information to the MDs. Written information about the study was sent via workplace email to all cardiothoracic surgeons and anaesthesiologists working at the three study sites ($$n = 133$$). MDs were recruited via convenience sampling, and all MDs interested in participating were included except for one who was no longer employed by the department at the time for data collection. MDs interested in participating answered the email, and an interview was scheduled.
The characteristics of the nurses and MDs are presented in Table I. Table I.*Demographic data* of the participating medical doctors and nurses. Medical doctors($$n = 20$$)Nurses($$n = 267$$)Age in years, mean (SD)52.8 (12.3)44.0 (12.0) Missing n (%)-6 ($2.2\%$)Years of clinical experience mean (SD)26.28 (11.1)17.6 (11.9) Missing n (%)--Sex n (%) Male14 ($70.0\%$)41 ($15.4\%$) Female6 ($30.0\%$)221 ($82.8\%$) Prefer not to say 1 ($.4\%$) Missing-4 ($1.5\%$)Degree n (%) Vocational5 ($25.0\%$)110 ($41.2\%$) Bachelor-70 ($26.2\%$) Master’s (60 credits)4 ($20.0\%$)73 ($27.3\%$) Master’s (120 credits)-9 ($3.4\%$) Doctoral11 ($55.0\%$)1 ($.4\%$) Missing-4 ($1.5\%$)Experience of being a patient n (%) Yes13 ($65.0\%$)138 ($51.7\%$) No6 ($30.0\%$)121 ($45.3\%$) Prefer not to say1 ($5.0\%$)2 ($.7\%$) Missing 6 ($2.2\%$)Experience of being a family member n (%) Yes15 ($75.0\%$)199 ($74.5\%$) No4 ($20.0\%$)64 ($24.0\%$) Prefer not to say1 ($5.0\%$)1 ($.4\%$) Missing-3 ($1.1\%$)Medical specialty n (%) Anesthesiology10 ($50.0\%$)- Cardiothoracic surgery10 ($50.0\%$)-*Postgraduate diploma* in nursing n (%) Yes-160 ($59.9\%$) Intensive care-116 ($43.4\%$) Surgical care-6 ($2.2\%$) Cardiac care-7 ($2.6\%$) Registered nurse anesthetists-4 ($1.5\%$) Other-4 ($1.5\%$) Missing-23 ($9\%$) No-107 ($40.1\%$) Prefer not to say--
## Families’ importance in nursing care – nurses’ attitudes (FINC-NA)
The FINC-NA was originally developed as a generic measurement of nurses’ attitudes towards the importance of families in nursing care, and it was validated in Swedish (Benzein, Johansson, Årestedt, & Saveman, 2008). It has since been refined (Saveman et al., 2011), and the refined version was used in this study. The FINC-NA consists of 26 items and uses a 5-point Likert-type scale ranging from “totally agree” to “totally disagree” (Saveman et al., 2011). The minimum possible score for the total scale is 26, and the maximum is 130. The higher the score, the more supportive nurses’ attitudes towards families in nursing care are. The FINC-NA has four subscales. Family as a resource in nursing care (Fam-RNC) assesses positive aspects of how the family influences the nurse’s work and the importance of family presence for nursing care. Family as a conversational partner (Fam-CP) concerns the nurse’s inclusive work with families when planning care, mapping out who belongs to the family, and communicating and inviting the family to participate in care planning and nursing care. Family as a burden (Fam-B) covers negative aspects of family involvement, such as whether the nurse is stressed by the family or does not have time for families; this scale is reverse scored. Family as own resource (Fam-OR) concerns how the nurse collaborates with and supports the family, enhancing the family’s own resources as a strategy for them to cope with the situation. The α reliability coefficients are.89 for the whole instrument and between.71 to.86 for the four subscales (Saveman et al., 2011). The original authors gave their permission to use the instrument for nurses but not for MDs since it was not developed for use in the MD population.
## Open-ended questions in the instrument
The FINC-NA was concluded with two open-ended questions added by the authors of this study. The open-ended questions for the nurses were intended to ask questions similar to those for the MDs, as parallel questions facilitate the merging of results in a mixed-methods convergent parallel design (Creswell & Plano Clark, 2018).
The two questions were as follows: What does family involvement mean to you?What influences your attitude towards family involvement in care?
The answers to these open-ended questions were described by the participating nurses in a total of 15,919 words, which was the dataset to be included in the qualitative analysis.
## Qualitative interviews
The key structure of the FINC-NA (Saveman et al., 2011) guided the construction of semistructured questions for the qualitative interviews with the MDs. Areas covered by the interview guide were the meaning of family involvement, family members’ role in caregiving, negative and positive aspects of family involvement and what the participant thought influenced his or her attitudes towards family involvement in care.
After conducting two pilot interviews with two MDs not eligible as study participants, the first author received feedback from the co-authors on the interview technique. The pilot interviews were not included for analysis in this study. The original opening question was “What does family involvement mean to you?” This question prompted the participants answer with a general statement regarding semantics of the phrase “family involvement” as opposed to describing their own personal attitudes. The question was therefore altered to “Think back to a situation when family were involved in the care you gave. Please describe the situation.” Prompting questions such as “How was that for you?” were also used.
Twenty MDs participated in qualitative interviews in the study. The first author conducted all interviews. They were held at the place of the participant’s choice, usually in the participant’s private office at the clinic. Seven interviews were held using a videoconferencing platform (i.e., Zoom) due to national travel restrictions during the COVID-19 pandemic. The interviews lasted 24–97 minutes, with a median of 50 minutes. All interviews were audio recorded and transcribed verbatim by a professional transcriber and formed the basis of the analysis.
## Analysis
The analysis was conducted in four separate steps consisting of 1) statistical analysis of the FINC-NA scores, 2) qualitative content analysis of the answers to the open-ended questions, 3) qualitative content analysis of the transcribed interview data and 4) mixed-methods merging and meta-inference using a side-by-side joint display.
## Statistical analysis
The quantitative data were analysed using descriptive statistics. Continuous variables are presented as the median and interquartile range (q1; q3), and categorical variables are presented as numbers and percentages. The Fam-B scale scores were reversed prior to analysis, making all scores indicate attitudes in the same direction from 1 (negative) to 5 (positive). The Mann—Whitney U test was performed for comparisons between groups by sex, educational degree, possession of a postgraduate diploma in nursing, and experience of being a patient or a family member of a patient. The correlations between age of the nurses and subscales and between years of experience as a nurse and subscales were analysed with Spearman’s rank correlation (rs). Missing data ($1.2\%$) were not imputed, and cases with missing data were listwise deleted. To be included in the analysis, a minimum of $60\%$ of the items on the FINC-NA had to be answered. The participants’ academic degrees were dichotomized into basic (vocational and bachelor’s degrees) and advanced (master’s and doctoral degrees) levels. The “prefer not to say” answer to demographic questions was treated as missing in the dichotomous group analysis. All tests were two-sided and conducted at the $5\%$ significance level. Statistical analyses were performed using IBM SPSS Statistics (Version 26) software.
## Qualitative content analysis
All qualitative data were analysed independently by all authors who had diverse preunderstandings. The authors’ clinical backgrounds were in the cardiothoracic care setting as nurses (AD, ASS, SÅ), in general surgical care as a nurse anaesthetist (EE) and in stroke and primary care as a physiotherapist (MLAL). All authors had experience conducting qualitative analysis. Analysis of free text answers in the questionnaire and the transcribed interviews was made separately and followed Elo and Kyngäs’s [2008] description of inductive qualitative content analysis. The first author independently analysed all interviews and all the material from the open-ended questions. All data from the interviews and all the material from the open-ended questions were divided between the other co-authors for individual analysis. Initially, all authors independently analysed the material by open coding, transferring the codes to a coding sheet, and thereafter grouping codes into preliminary categories. After this, the authors’ individual analysis was compared and discussed within the research group until agreement was reached. Hence, all data were analysed individually by two authors until the grouping stage of Elo and Kyngäs’s [2008] analysis process. After discussion, agreement regarding the grouping was reached. At this stage, the first author expanded the analysis through Elo and Kyngäs’s [2008] categorization. The expanded analysis was then discussed within the research group until agreement was reached.
## Mixed-methods merging and meta-inference
This study’s mixed-methods research question was “To what extent do attitudes towards family involvement in care held by nurses and medical doctors working in open-heart surgical care and the factors affecting these attitudes converge and diverge?”. To answer this question, the three datasets were merged, using a side-by-side joint display (Creswell & Plano Clark, 2018). The merging of results was expected to yield greater insight into the phenomenon of attitudes towards family involvement in care held by nurses and MDs than each unit of data separately would. In the instances where merging is the integration procedure, a side-by-side joint display is recommended (Younas & Durante, 2022). The headings and textual descriptions of the generic categories and groupings from the two qualitative datasets were initially contrasted inductively by the first author by their overarching themes. Thereafter, the descriptions of the FINC-NA subscales were contrasted by the first author against the themes identified in the two qualitative datasets. The descriptions of the FINC-NA subscales and median values and results from analytical statistical tests in each subscale and on total scale level were organized according to the preliminary themes and their correspondence to one and another. When displayed for contrast and comparison, the content of the results was merged into concepts of convergent and divergent attitudes. At this stage, the merging and meta-inference was discussed within the research group until agreement was reached. The last step in the mixed-methods convergent parallel design process, the meta-inference of the concepts, is discussed in the discussion section of this article. An example from the side-by-side joint display table used for the mixed-methods merging and meta-inference is presented in Table II. Table II.The mixed method merging and meta-inference of three datasets: An example from the side-by-side joint display analysis.1. Mixed-method concepts2. Qualitative results from interviews with medical doctors2. Qualitative results from open-ended questions answered by registered nurses3. Quantitative results from FINC-NA* answered by registered nurses4. Meta-inferenceSupporting, informing andimproving careGeneric category: The synthesized textual description of the content in Striving for the patient´s bestGrouping: The synthesized textual description of the content in Advocating for the patient and Improving healthQuote: And when everything is perfect, then they motivate the patient in ways I can’t. They have the patient’s entire life as a source of information that I don’t have. ( Physician 12)Generic category: The synthesized textual description of the content in Affecting the quality of careGrouping: The synthesized textual description of the content in Improved quality of careQuote: Support and comfort for the patient, before and after surgery. Involvement in rehabilitation and usually better patient outcome. Being an extra pair of ears, listening to information. Sometimes giving information of importance. Family involvement can sometimes be more time-consuming but is, in my opinion, better for patient outcomes (Nurse 220)Descriptions of sub-scales:Family as a resource in nursing care (Fam-RNC) assesses positive aspects of how the family influences the nurse’s work and the importance of family presence for nursing care. Family as a conversational partner (Fam-CP) concerns the nurse’s inclusive work with families when planning care, mapping out who belongs to the family, and communicating and inviting the family to participate in care planning and nursing care. Number of items in the FINC-NA belonging to the concept† Supporting, informing and improving care: 18 out of 26 itemsMedians on item level: The items with the highest median scores [5] belong in this theme, variation between 3–5 among these 18 items. ConvergentPositive attitudes were convergent between paradigm and interprofessionally, nurses’ and medical doctors’ attitudes are foremost positive to family involvement in open-heart surgical careNote: FINC-NA Families’ Importance in Nursing Care—Nurses’ Attitudes questionnaire †The authors decided which items belonged to which concept after a discussion about how the sub-scale descriptions corresponded to the content of the preliminary themes developed from merging results from the two qualitative datasets.
## Ethical considerations
The study was approved by the Swedish Ethical Review Authority (No 2019–06315), and it conforms with the principles outlined in the Declaration of Helsinki (World Medical Association, 2013). Some of the participants were recruited from the first, fourth and last authors’ workplaces. The first author conducted all interviews. The written information was at times followed up with oral information by the first or fourth author during clinical duty. This oral information was given on the request of the MD without pressure to participate. No author, except for the first author, analysed the interviews of MDs with whom they had a professional relationship. Since the data from the nurses were anonymized, professional relationships between participants and authors could not be determined in this dataset. The timing of this study could be considered burdensome for the participants, considering the workload during the first year of the COVID-19 pandemic. On the other hand, participants might have felt good about the opportunity to reflect and share their thoughts on this issue during a time when several aspects of family involvement were limited.
## Results from the FINC-NA completed by nurses
In total, 267 out of the 650 eligible nurses returned the FINC-NA with at least a $60\%$ completion grade, giving a total response rate of $41\%$. Response rates from the eight clinics varied between 27.9–$55.6\%$. Complete responses (i.e., responses on all 26 items) were returned by 222 out of the 267 included nurses. The median score for the FINC-NA total scale was 93 (Q1 = 85 Q3 = 104). Descriptive results from the total FINC-NA and the four subscales and items are presented in Table III. Table III.Results for the families’ importance in nursing care – nurses’ attitudes questionnaire; item scores for the total population. Number (Missing)Totally disagree/1234Totallyagree/5Median(Q1/Q3)Possible rangeFamilies’ importance in Nursing Care- Nurses’ Attitudes (FINC-NA) total score†222 [45]n (%)n (%)n (%)n (%)n (%)93 ($\frac{85}{104}$)26–130Family as a resource in nursing care (Fam-RNC)250 [17] 38 ($\frac{34}{43}$)10–503) A good relationship with family members gives me job satisfaction266 [1]2 ($.8\%$))4 ($1.5\%$)33 ($12.4\%$)86 ($32.3\%$)141 ($53.0\%$)5.0 ($\frac{4}{5}$) 4) Family members should be invited to take an active part in the patient’s care266 [1]10 ($3.8\%$)36 ($13.5\%$)87 ($32.7\%$)82 ($30.8\%$)51 ($19.2\%$)3.5 ($\frac{3}{4}$) 5) The presence of family members is important to me as a nurse267 [0]3 ($1.1\%$)31 ($11.6\%$)65 ($24.3\%$)99 ($37.1\%$)69 ($25.8\%$)4.0 ($\frac{3}{5}$) 7) The presence of family members gives me a feeling of security265 [2]20 ($7.5\%$)50 ($18.9\%$)111 ($41.9\%$)63 ($23.8\%$)21 ($7.9\%$)3.0 ($\frac{2}{4}$) 10) The presence of family members eases my workload266 [1]13 ($4.9\%$)46 ($17.3\%$)118 ($44.4\%$)75 ($28.2\%$)14 ($5.3\%$)3.0 ($\frac{3}{4}$) 11) Family members should be invited to actively take part in the planning of patient care265 [2]3 ($1.1\%$)13 ($4.9\%$)57 ($21.5\%$)106 ($40.0\%$)86 ($32.5\%$)4.0 ($\frac{3}{5}$) 13) The presence of family members is important for the family members themselves264 [3]0 ($0\%$)2 ($.7\%$)34 ($12.7\%$)95 ($36.0\%$)133 ($50.4\%$)5.0 ($\frac{4}{5}$) 20) Getting involved with families gives me a feeling of being useful259 [8]8 ($3.1\%$)28 ($10.8\%$)85 ($32.8\%$)82 ($31.7\%$)56 ($21.6\%$)4.0 ($\frac{3}{4}$) 21) I gain a lot of worthwhile knowledge from families that I can use in my work264 [3]8 ($3.0\%$)17 ($6.4\%$)49 ($18.6\%$)107 ($40.5\%$)83 ($31.4\%$)4.0 ($\frac{3}{5}$) 22) *It is* important to spend time with families264 [3]2 ($.8\%$)6 ($2.3\%$)43 ($16.3\%$)93 ($35.2\%$)120 ($45.5\%$)4.0 ($\frac{4}{5}$) Family as a conversational partner (Fam-CP)243 [24] 28 ($\frac{24}{32}$)8–401) *It is* important to find out who belongs to the patient’s family264 [3]0 ($0\%$)4 ($1.5\%$)34 ($12.9\%$)73 ($27.7\%$)153 ($58.0\%$)5.0 ($\frac{4}{5}$) 6) I ask family members to take part in discussions when a patient first comes into my care265 [2]33 ($12.5\%$)33 ($12.5\%$)59 ($22.3\%$)79 ($29.8\%$)61 ($23.0\%$)4.0 ($\frac{2.5}{4}$) 9) Discussion with family members when a patient first comes into my care saves time in my future work265 [2]16 ($6.0\%$)20 ($7.5\%$)82 ($30.9\%$)85 ($32.1\%$)62 ($23.4\%$)4.0 ($\frac{3}{4}$) 12) I always find out who belongs to the patient’s family265 [2]9 ($3.4\%$)43 ($16.2\%$)62 ($23.4\%$)81 ($30.6\%$)70 ($26.4\%$)4.0 ($\frac{3}{5}$) 14) I invite family members for a discussion at the end of the care period266 [1]47 ($17.7\%$)70 ($26.3\%$)79 ($29.7\%$)50 ($18.8\%$)20 ($7.5\%$)3.0 ($\frac{2}{4}$) 15) I invite family members to take an active part in the patient´s care262 [5]48 ($18.3\%$)80 ($30.5\%$)89 ($34.0\%$)38 ($14.5\%$)7 ($2.7\%$)3.0 ($\frac{2}{3}$) 19) I invite family members for discussions when the patient’s condition changes/deteriorates263 [4]5 ($1.9\%$)11 ($4.2\%$)43 ($16.3\%$)94 ($35.7\%$)110 ($41.8\%$)4.0 ($\frac{4}{5}$) 24) I invite family members for discussions when planning care259 [8]14 ($5.4\%$)29 ($11.2\%$)69 ($26.6\%$)71 ($27.4\%$)76 ($29.3\%$)4.0 ($\frac{3}{5}$) Family as own resource (Fam-OR)255 [12] 13 ($\frac{11}{16}$)4–2016) I ask families how I can support them265 [2]14 ($5.3\%$)49 ($18.5\%$)84 ($31.7\%$)73 ($27.5\%$)45 ($17.0\%$)3.0 ($\frac{3}{4}$) 17) I encourage families to use their own resources so that they can cope with their situation as far as possible263 [4]15 ($5.7\%$)50 ($19.0\%$)87 ($33.1\%$)84 ($31.9\%$)27 ($10.3\%$)3.0 ($\frac{3}{4}$) 18) I consider family members as cooperating partners264 [3]15 ($5.7\%$)36 ($13.6\%$)74 ($28.0\%$)89 ($33.7\%$)50 ($18.9\%$)4.0 ($\frac{3}{4}$) 25) I see myself as a resource for families so that they can cope as well as possible with their situation259 [8]9 ($3.5\%$)44 ($17.0\%$)83 ($32.0\%$)92 ($35.5\%$)31 ($12.0\%$)3.0 ($\frac{3}{4}$) Number (Missing)Totallyagree/1234Totally disagree/5Median(Q1/Q3)Possible rangeFamily as a burden (Fam-B) reversed score‡254 [13] 15 ($\frac{12}{17}$)4–202) The presence of family members holds me back in my work266 [1]2 ($.7\%$)23 ($8.6\%$)72 ($27.1\%$)104 ($39.1\%$)65 ($24.4\%$)4 ($\frac{3}{4}$) 8) I do not have time to take care of families262 [5]8 ($3.1\%$)48 ($18.3\%$)68 ($26.0\%$)82 ($31.3\%$)56 ($21.4\%$)4 ($\frac{3}{4}$) 23) The presence of family members makes me feel that they are checking up on me262 [5]8 ($3.1\%$)40 ($15.3\%$)78 ($29.8\%$)73 ($27.9\%$)63 ($24.0\%$)4 ($\frac{3}{4}$) 26) The presence of family members makes me feel stressed261 [6]3 ($1.1\%$)30 ($11.5\%$)58 ($22.2\%$)84 ($32.2\%$)86 ($33.0\%$)4 ($\frac{3}{5}$) Note: †Displaying the original items in the instrument Families Importance in Nursing Care- Nurses Attitudes. Published by Saveman et al. [ 2011] in Refinement and psychometric re-evaluation of the instrument Refinement and Psychometric Re-evaluation of the Instrument: Families’ Importance in Nursing Care—Nurses’ Attitudes. Journal of Family Nursing, 17[3], 312–329. https://doi.org/$\frac{10.1177}{1074840711415074.}$ ‡The Fam-B scale was reversed prior to analysis, making ratings on all items indicate attitudes in the same direction, i.e., high score= positive attitude and low score =negative attitude; Q1=First quartile Q3=Third quartile Internal consistency for subscales was reliable, with Cronbach’s alphas between.71 and.88 for the four subscales and.90 for the total scale. Overall FINC-NA scores were high, indicating positive attitudes towards the importance of families in nursing care. There was no significant difference in the FINC-NA total score by age, sex, years of experience as a nurse, academic level, or experience of either being a patient or being a family member to a patient. Female participants held more positive attitudes on the Fam-RNC subscale than male participants ($$p \leq 0.004$$). Nurses with an advanced educational degree (i.e., master’s or doctoral degrees) showed a significantly higher ranking on the reversed score Fam-B subscale than nurses with vocational or bachelor’s degrees ($$p \leq 0.003$$). Having a postgraduate diploma in nursing was associated with a more positive attitude towards families on the Fam-CP ($$p \leq 0.040$$) and Fam-B ($$p \leq 0.013$$) subscales.
A poor correlation to no correlation was found between the nurses’ ages and the subscales (rs= −.12–.08) and between years of experience as a nurse and the subscales (rs= −.05–.16).
Analyses of differences on the subscales and background variables are presented in Table IV. Table IV.Group differences in the Families’ Importance in Nursing Care- Nurses’ Attitudes (FINC-NA), scale and subscale results. High score = positive attitudes on all items and scalesFamilies’ Importance in Nursing Care (FINC-NA) total score possible range of 26–130Family as a resourcein nursing care (Fam-RNC)possible range of 10–50Family as a conversational partner (Fam-CP)possible range of 8–40Family as own resource(Fam-OR)possible range of 4–20Family as a burden (Fam-B) reversed scorepossible range of 4–20nMedianQ1/Q3p†nMedianQ1/Q3p†nMedianQ1/Q3p†nMedianQ1/Q3p†nMedianQ1/Q3p†SexFemale$\frac{1859386}{105.3622073935}$/43.004*$\frac{2012824}{32.1972122824}$/$\frac{32.2572111512}{17.131}$Male$\frac{349383}{98383531}$/$\frac{39382723}{31392723}$/$\frac{31391612}{18}$Experience of being a patientYes$\frac{1109385}{105.6831283733}$/$\frac{43.2751212824}{32.8091301312}$/$\frac{16.2251301512}{17.834}$No$\frac{1069386}{1041153935}$/$\frac{431162824}{311191311}$/$\frac{151171513}{17}$Experience of being familyYes$\frac{1639385}{104.9231853733}$/$\frac{43.2841792825}{32.4891901311}$/$\frac{15.8751911512}{17.062}$No$\frac{559685}{104613935}$/$\frac{44602823}{32611411}$/$\frac{16591412}{16}$Post graduate diplomaYes$\frac{1349487}{104.2251523834}$/$\frac{43.9511462925}{32.040}$*$\frac{1521412}{16.0521541513}$/18.013*No$\frac{889283}{104983833}$/$\frac{43972724}{301031311}$/$\frac{161001412}{16}$Degree‡Basic$\frac{1479384}{104.4131673833}$/$\frac{43.5461632824}{32.8101711311}$/$\frac{15.1501711412}{17.003}$*Advanced$\frac{739487}{105803835}$/$\frac{43772824}{32801412}$/$\frac{16791614}{17}$Note: Q1=First quartile Q3=Third quartile †Mann—Whitney U test *$p \leq 0.05$ ‡Degree//Basic degree= Vocational, Bachelor Advanced degree=Master, Doctoral.
## Results from open-ended questions answered by nurses
Out of the 267 nurses included in the quantitative analysis, 206 answered at least one of the two open-ended questions. Analysis of this material generated three generic categories: affecting the quality of care, including family in their mission, and influential aspects. Quotes illustrating the categories are displayed in Table V. Table V.The content analysis process of nurses’ attitudes towards family involvement in open-heart surgical care. DATASET: OPEN-ENDED QUESTIONS, NURSES QuoteOpen codingGroupingCategorization(final generic category)DATASET: OPEN-ENDED QUESTIONS, NURSES They can promote health and contribute to recovery of the soul in a way I never could because our relationship is built on illness/the operation. They have a relationship reaching beyond the time when the patient is under my care. They experience the time of illness together, and that’s why it is so important that the family isn’t just distant visitors. ( Nurse 70)Recovery of the soulImproved quality of careAFFECTING THE QUALITY OF CARE It can be really wearing when a family member has a negative impact on the patient, for example, when they encourage the patient to be still in bed when the patient really needs to be ambulated or when they feed a patient who needs to practice eating independently. They help “too much”, hindering the patient´s recovery. ( Nurse 178)Hindering postoperative recoveryImpaired quality of care I have mostly had good experiences with family members who cheered up, activated and, if necessary, helped to reorient patients who have been confused, etc. But I have also been with family members who stressed, tired out patients and inhibited reactivation. No one is cast in the same shape, and sometimes the family members help and sometimes not (Nurse 149)Contributes to nursing and recoveryImproved and Impaired quality of care Inhibits postoperative recovery I consider myself a resource, guiding them in (probably) the worst time of their lives, making sure that they know that we are doing everything we can for their relative who is ill. It is unquestionably important to give my time and effort to the family. ( Nurse 194)Important to attend to the family´s needsFamily health and well-beingINCLUDING THE FAMILY IN THEIR MISSION That they are informed about the planning and delivery of care. That they are given the opportunity to share information and opinions related to the care. That they are given the opportunity to participate in care if they wish. ( Nurse 246)The family should be given the opportunity to be involvedPromoting family involvement The culture of the workplace, my own relationship to my family and if there is time to invite the family to participate in care (Nurse 166)Workplace culture, timeOrganizational aspects andExperiencesINFLUENTIAL ASPECTS Personal experiences I can see a change in my attitude due to my experience of being an intensive care nurse specialist. As an ICU-nurse you spend a lot of time bedside; at first, I experienced visitation by family as stressful. I was insecure. But now I really think it enriches the care. For everyone involved! ( Nurse 206)Professional experienceExperiences If the family supports and helps one an another, or if it is more of a burden for the patient to have the family involved, the patients feeling towards his/her family has an impact on my attitude (Nurse 217)The family´s influence on the patient affects attitudeFamily function My attitude towards family involvement is influenced by the contact I get with the family, how they behave in the patient room together with the patient. If I get a good contact with the family quickly, I ʺtake inʺ the family faster and then more easily include them in the care. If, on the other hand, there are some kind of obstacles in our communication (aggression, language barriers, attitude problems, for example) then it takes longer to establish a contact with them. ( Nurse 268)The family’s attitude, acting and communication related to the nurseThe nurse-family relationship
## Affecting the quality of care
The nurses believed that family involvement affected the quality of care, both in a positive and negative way. Information about the patient provided by family members was considered a means for the nurses to give personalized care and for the family to act as the patient’s voice when they could not speak for themselves. Some reported that family involvement eased the nurse’s workload and facilitated nurse—patient communication. Family members were believed to improve recovery after surgery by providing emotional and physical support for the patient. Family involvement was seen as providing the nurse a sense of security, which in turn improved care quality. If family members transferred their own stress to the patient and limited self-care activities, family involvement could be seen as hindering the patient’s recovery. The patient is the nurse´s number one priority, and nursing may become more complicated if the family consumes the nurse’s time, space, and energy. Some nurses believed that family involvement could become an intrusion on patient integrity and autonomy, an impairment of nurse—patient communication, and a safety hazard when participating in care in the intensive care unit.
## Including the family in their mission
Nurses reported that they believed that including the family in their mission was of importance for the sake of the family. Family involvement was seen as improving the family’s health, increasing the family’s sense of coherence, giving the family a sense of security, and enhancing the family´s comprehension of the course of the disease. The nurses considered themselves to have an extended responsibility to care for not only the patient but also the family. Welcoming the family, appreciating the family, and supporting the family members’ own choices regarding their level of involvement was said to be a part of the nurse’s work.
## Influential aspects
The most prominent components influencing nurses’ attitudes were their professional experiences of meeting family members as a nurse and at times their personal experiences of being a family member. Several nurses reported using their own personal preferences regarding family involvement as a guide when caring for families. The patient’s wishes and how the nurse understood the relationship and functioning within the family were also considered to affect nurses’ attitudes towards family involvement in individual situations. The nurse-family relationship was seen as influencing collaboration between the nurse and the family. Organizational conditions, such as the context, policy of care, and attitudes of colleagues, were more general components described. Family members of a person cared for in the intensive care unit for an extended period of time were often described as having a higher priority than the family of a patient who had an elective surgery without complications.
## Results from interviews with the medical doctors
Analysis of the twenty interviews with the MDs generated four generic categories: caring relationship, complicating care, striving for the patient’s best and frames of reasoning. Quotes illustrating the categories are displayed in Table VI. Table VI.The content analysis process of medical doctors’ (MDs) attitudes towards family involvement in open-heart surgical care. DATASET: SEMISTRUCTURED INTERVIEWS, MEDICAL DOCTORS QuoteOpen codingGroupingCategorization(final generic category)DATASET: SEMISTRUCTURED INTERVIEWS, MEDICAL DOCTORS Yes, but sometimes the patient does not make it, despite all our efforts, and then it is, but the family will live on with a lot of questions and speculations about what really happened, if the relative was well taken care of. I believe it is much easier to handle grief if you have been kept well informed continuously and been well taken care of. ( MD 16)Being well-informed eases the family’s sufferingCreating trust through informationCARING RELATIONSHIP So I´ll stay with the one who is afraid or does not want to. And explain to her or him what it is, and I have done this for everyone else, but I´ll explain it again. What they’ll see when they get in there. ( MD 2)Supporting the familyAttention to the relatives Then it’s almost better to isolate the patient because everything that can go wrong is highlighted by the family members, picturing worst case scenario and planning the funeral of someone far from needing one. Their anxiety is contagious, infecting the patient. ( MD 12)Protecting the patient from the familyActing in difficult situationsCOMPLICATING CARE The biggest disadvantage is their anxiety, anxiety and fear, and they do not calm the operated patient; it is a problem. They’re too worried and stuff. ( MD 1)Contagious anxietyChallenging elements Even if they are not physically present, only the knowledge that the wife is waiting at home makes them recover. That’s how it is. ( MD 9)Meaningful involvement at a distanceImproved health for the patientSTRIVING FOR THE PATIENT’S BEST I: And when there are no family members, what do you do then?P: Well, then you must … well, then the team must have a discussion about the bigger issues. And then the care team plays a greater part in listening to the patient’s wishes, collecting information about the best fit for the patient. There is something incredibly sad about that. ( MD 5)The patient without family supportAdvocating for the patient Well, I would say it depends to a great extent on what kind of care. They have, in a cardiothoracic department, not a big role in the actual care except the few, a small percentage I think, really, among the ones passing through, that have an extended hospital stay. In those cases you can really benefit from involving family members. ( MD 4)Family involvement vary, depending on care contextThe structure of careFRAMES OF REASONING But if you have a distanced relationship that will be, well that will persist in the care setting, I don’t think the relationships within the family are all that effected, the pre-existing type of relationship they have, when the patient is hospitalized. That will be the same (Physician 8)The pre-existing family relationship affects family involvementThe family´s needs Well, my thoughts are not that different regarding what I want to do, on the other hand, being new is difficult because you don’t have the experience backing up your arguments (Physician 5)Experience facilitates communication with familiesThe physician´s experiences The family’s’ involvement is also, well it is, it is in my opinion an ethical issue really, and then I’m thinking of autonomy and the principle of autonomy and so on; it is really that, that’s guiding in this case. But that is a bit of a short cut, like it always is with ethical considerations; there are, like, other aspects as well. But I think these aspects are dominating. ( MD 15)Family involvement is an ethical issueEthical approach
## Caring relationship
In the category caring relationship, the MDs described trust as being of great importance for the family and the MD. The MDs aimed for the MD-family relationship to be supportive for the family. The relationship is built on trust, providing the family with honest and situational information. The MDs balanced information regarding patient outcomes, thereby trying to prepare the family for the worst while simultaneously giving them hope and avoiding unnecessary stress. Surgeons as well as anaesthesiologists emphasized the importance of the postoperative phone call undertaken by the surgeon to the family in establishing contact between the MD and the family. Family members were seen as collaborative partners who should be welcomed, supported, and cared for. The MDs described themselves as being attentive to the strain that involvement can place on the family and explained how they encouraged family members to preserve their resources by limiting the family’s bedside presence. Supportive conversations and displays of empathy and understanding for family members were examples of acts of caring for the family.
## Complicating care
The category complicating care entails descriptions of situations when MDs perceived family involvement as challenging and examples of how the MDs acted when family involvement complicated a situation. Cultural differences regarding expectations of health care services and communication were described as demanding. Large families were considered to take space and time and were sometimes burdensome for the MDs. Difficult conversations with families regarding adverse events or death were reported as a heavy responsibility for the MDs. Worry, guilt, and fear of being accused of mistreatment were prominent feelings in connection to difficult conversations with family members expressed by some MDs. A few MDs expressed rare instances when family members could be dangerous for the patient, for example, in domestic abusive relationships. The MDs considered themselves obligated to act when problems with family involvement occurred. Restriction of visitation and shifting focus back to the patient were other actions undertaken by the MDs when problems with family involvement occurred.
## Striving for the patient’s best
The MDs noted that the family usually was striving for the patient’s best by safeguarding the patient’s interests and giving the MD information regarding the patient’s unique situation. Attitudes regarding family members acting as experts on the patient’s needs and wishes were described on a continuum from negative to positive. The family was considered to know the person who was now a patient, thereby possessing knowledge that could improve the patient’s care and recovery. Family members acting as experts on the patient’s wishes were believed by some MDs to hinder the patient from speaking for themselves. Without family support, safeguarding of the patient’s autonomy was considered to become the responsibility of the MDs and the care team. The MDs exemplified how family members could improve patient health and promote patient recovery by being physically, psychologically, and cognitively supportive. The family was viewed as giving the patient healing care and love and motivating the patient to come home.
## Frames of reasoning
The MDs’ frames of reasoning regarding their attitudes towards family involvement in care were set by their experiences, the health-care mandate, the family´s individual needs and the MD’s ethical approach. They did not believe their attitudes had changed over time, but their professional experience had made encounters with family members easier. Some MDs described how personal experiences of being a patient or being a family member had made their attitudes regarding family involvement more positive. Professional experiences of being questioned or reported for clinical errors had in some instances led to more negative attitudes. Policy and praxis regarding health care professionals’ responsibility for the family and the family’s responsibility to be involved, as well as hospital environment and time limitations, were reported to influence MD attitudes.
Some MDs stated that their attitudes towards family involvement varied between care settings. Patients experiencing complications or requiring intensive care were regarded as being in greater need of family involvement than patients undergoing elective heart surgery without complications. Some MDs thought that a person’s attitudes towards family involvement could not be generalized. Descriptions of how families should be involved in an individualized manner were highlighted. The MDs’ perceptions of the family´s experiences and relationships affected the MDs’ beliefs about the individual family’s practice of involvement. Family involvement was considered an ethical aspect of care and a duty for the MDs. A principle of treating others the same way you want to be treated yourself was commonly applied by the MDs in relation to family involvement in care.
## Mixed-methods merged concepts
The integration involved merging the results from the quantitative and qualitative data so that a comparison could be made, and a more complete understanding emerged than provided by the qualitative or the quantitative results alone.
Preliminary themes were initially identified in the qualitative material as 1) positive attitudes, 2) negative attitudes, 3) family care, 4) components affecting attitudes. Thereafter, the descriptions of the FINC-NA subscales, the descriptive and analytical statistics and number of items corresponding to the themes was contrasted against all categories over the two qualitative datasets.
Attitudes from the three datasets were merged into four final concepts, illustrated in a visual side-by-side joint display (Figure 2). Convergent concepts were supporting, informing and improving care and caring for the family. The concepts depending on the situation and impairing care were divergent. The main finding, a concept overarching all concepts, was the attitude that the importance of family involvement in open-heart surgical care depends on the situation. Figure 2.Visual joint display of qualitative, quantitative and mixed-methods results and meta-inference.
## Discussion
In this study, it was found that family involvement was foremost regarded as important for the patients´ health and recovery. These attitudes were merged into the convergent concept of supporting, informing and improving care. Nurses and MDs working in open-heart surgical care considered family involvement to be important for the family members as well as for the patient under certain conditions, as shown in the convergent concept of caring for the family. Components that the participants themselves reported as affecting their attitudes were presented. These self-reported components were not always supported by the statistical analyses of the quantitative material, and they are discussed in the divergent concept depending on the situation. The negative attitudes held by nurses and MDs were not as prominent as their positive attitudes. Nevertheless, these negative attitudes need to be addressed since these areas are of greatest importance for improvement. These aspects were merged into the divergent concept impairing care.
## Convergent concepts: supporting, informing, and improving and caring for the family
The most prominent aspect of nurses’ and MDs’ attitudes towards family involvement in open-heart surgical care was illuminated in the convergent concept supporting, informing and improving care. This concept entails how family involvement was seen as supporting the patient, providing valuable information for the nurses and MDs and improving quality of care and postoperative recovery. Several of these aspects have previously been highlighted by patients and family members (Bélanger et al., 2018; Mackie et al., 2019) and surgeons (Jordan et al., 2014). The predominant positive attitude towards family involvement may also be reflected in the construct of the FINC-NA. In our mixed-methods merging, 18 out of 26 items seemed to “belong” to the concept supporting, informing, and improving reflecting positive attitudes. After the refinement of the FINC-NA, ceiling effects persisted (Saveman et al., 2011). These were suggested by the original author to be explained by the socially undesirable attitude of discarding the importance of family involvement in care (Saveman et al., 2011).
The overall positive rating of attitudes in our study was supported by the qualitative material. One could, however, question the MDs’ emphasis on informing the family and question whether these practices truly are inclusive and an expression of mutual information sharing as perceived by patients and family members. MDs are often aware of how the timing and delivery of information can affect how the receiver retains the information given (Jordan et al., 2014). Family members express how shock, anger (Robley et al., 2010) and information that is unadjusted for the person´s level of health literacy (Mackie et al., 2019) may interfere with the message. Patients and family members have expressed a need for a dedicated professional who can bridge the gap in communication between the family and the care team, as well as between different members within the care team (Mackie et al., 2019).
Family, as a part of the health care mandate, was illuminated in the concept caring for the family. This concept was convergent over all datasets. It was considered to be important to care for family members in addition to the patient. Surgeons, in a previous study, highlighted how they consider communication to alleviate anxiety for family members (Jordan et al., 2014). The concept of caring for the family appeared to be prioritized more when the patient’s hospital stay was prolonged due to the severity of illness or complications. This attitude emerged in the qualitative datasets and could explain why the ratings on the Fam-OR scale were lower than those on the other subscales, being predominantly neutral with a median score of three for three out of four items. The Fam-OR scale concerns the importance of support for the family. The interpretation of Fam-OR scores that was made after the mixed-methods merging—that the importance of supporting the family could depend on the severity of the patients’ condition—might not have been made without the mixed-method approach. An alternative interpretation could have been that support for families was not as important as the other aspects covered by the FINC-NA. Having a postgraduate diploma in nursing was associated with more positive attitudes in our sample. The majority of nurses holding postgraduate diplomas in this study were in intensive care. The association between positive attitudes and postgraduate diplomas could thus be an expression of the attitude that family involvement is more important in the intensive care setting than in the step-down or surgical ward setting. Data specifying whether the nurses worked in the surgical ward, step-down unit or intensive care unit were not collected; therefore, it was not possible to compare these groups. Previous research regarding how the postoperative care experience differs between patients and families of patients undergoing elective emergency surgery is inconclusive but indicates that a prolonged stay in the intensive care unit predisposes them to negative experiences (Göktas et al., 2016) and that family members have specific stress-related information needs in relation to intensive care (Joseph et al., 2015). On the other hand, the need for family involvement is expressed by patients and family members of patients undergoing elective open heart surgery regardless of complications and adverse events (Bjørnnes et al., 2019; Joseph et al., 2015; Kemp et al., 2020). The patient and family experience on this topic seems to need further investigation prior to directing resources in either course.
## Divergent concepts: depending on the situation and impairing care
Components affecting the nurses’ and MDs’ attitudes towards family involvement were merged into the divergent concept of depending on the situation. The MDs and the nurses stated that their attitudes regarding family involvement depended on several factors, such as family functioning, organizational factors, and care setting. The ethical approach was an influential aspect reported by the MDs in this study that is important to investigate further when exploring health care professionals’ attitudes towards family involvement in care since it is not asked about in the FINC-NA. The nurses in this study did not report about the ethical aspect of family involvement in their care in the qualitative data set to the same degree as the MDs. Divergence in this aspect might be due to differences between nurses and MDs views on why and how family involvement is important.
Regarding experiences being influential on the nurses’ and MDs’ attitudes, the concept was divergent. Both nurses and MDs described how having the experience of being a family member influenced their attitude towards family involvement, but there were no statistically significant differences regarding these aspects in the quantitative material. The divergence could be explained by methodology and differences between paradigms. It could be hypothesized that the nurses and MDs participating in this study whose attitudes were affected by their experiences elaborated on this in the qualitative material. In the quantitative material, the experiences of being a patient or family member was reported as either “yes” or “no”, hence both significant and nonsignificant experiences for the persons’ attitudes were reported. For example, the experience of being an adult child to a parent with a severe chronical condition may influence ones’ attitude towards family involvement in care to a greater extent compared to being an adult child to a parent in need of hospital care on one occasion. In the quantitative dataset this was not adjusted for. Another interpretation of this divergent finding is that perhaps nurses’ and MDs’ subjective beliefs about how their personal experiences influence their attitudes cannot be generalized statistically on group level. Evidence regarding whether personal health care experiences affect the overall rating in the FINC-NA from previous studies is contradictory; some have not found this association (Blöndal et al., 2014; Østergaard et al., 2020), while others have (Linnarsson et al., 2015). During the development of the FINC-NA, it was learned that the answers could not be dichotomized because the answer depended on the situation (Benzein, Johansson, Årestedt, & Saveman, 2008), supporting the concept of depending on the situation in this study. The influence of organizational conditions was reported as divergent over the three datasets regarding having a lack of time for families. A few MDs and several nurses reported having a lack of time for families in the qualitative material. This finding was not reflected in the FINC-NA Fam-B subscale, where all items had a median of four, indicating a general attitude of having time for families in this population of nurses as one item specifically asks about this in the Fam-B subscale. Time restraints have previously been described by nurses as a factor influencing their involvement of families in care, stating that their number one priority is the patient (Mackie et al., 2018).
Negative attitudes over the three datasets were merged into the concept of impairing care. How family involvement may impair patient care and recovery was described in the two qualitative datasets, making it convergent between the nurses and MDs. The nurses and MDs also expressed that family involvement could complicate their work. This was not shown in the quantitative results in the Fam-B scale, making the concept divergent between paradigms. There are no questions on the FINC-NA asking about how family involvement can be harmful for the patient, which implies divergence between the quantitative and qualitative material. This is an attitude in need of further exploration in future research. In our study, the MDs described having some difficulties caring for families of various cultural backgrounds. This was interpreted as divergent from the nurses’ experiences, possibly due to differences in data collection strategies. Perhaps, describing negative aspects and experiences would require time for reflection and report building, which is more commonly achieved during a qualitative interview compared to answering a survey. In a meta-synthesis of nurses’ experiences of caring for culturally diverse families in hospitals (Murcia & Lopez, 2016), experiences similar to those expressed by the MDs in this study were found. Difficulties communicating, a lack of space for all family members and the violation of visitation policy are some examples of barriers described by nurses in relation to culturally diverse families (Murcia & Lopez, 2016).
In health care, it is important to provide competent transcultural care (Health Research & Educational Trust, 2013). Patient and family stressors related to open-heart surgery are contextual and have cultural dimensions (Sedaghat et al., 2019). Understanding and flexibility are key when caring for culturally diverse families (Murcia & Lopez, 2016). Transcultural competence involves an understanding and exploration of health-related beliefs in different cultures and contexts (Health Research & Educational Trust, 2013). The illumination of beliefs about illness and health is also a core concept in family-centred care that is applicable to a broad spectrum of care contexts (Shajani & Snell, 2019). Therefore, it could be that education on family-centred care practices, such as illuminating families’ health and illness beliefs, could reduce negative attitudes and facilitate family involvement in care when caring for families from different cultural backgrounds in this context.
Patients and family members have, as previously stated, expressed a need for a dedicated professional who can bridge the gap in communication between the family and the care team. To enhance family involvement in care, this bridging professional would preferably have competency in family-centred care. With a family-centred care approach, the family is treated as a system in which all parts affect one and another (Bell, 2013; Shajani & Snell, 2019). Health and illness are considered a family affair, and the most effective care strategies are those targeting the whole system at once (Bell, 2013; Shajani & Snell, 2019). Competency in family-centred care would diminish the subjective practice of having family involvement depending on the situation from the professionals’ perspective and instead having the family preferences as the main focus when planning and delivering open-heart surgical care. Family-centred care with the involvement of family cannot be accomplished if not approached interprofessionally (Naef et al., 2020). It is therefore important to have a team-based approach if family-centred care is introduced in the context of open-heart surgery.
## Strengths and limitations
This study has strengths and limitations that need to be addressed. By using a joint display in the merging and meta-inference processes, validity of these inferences was strengthened (Creswell & Plano Clark, 2018). The interview guide used in this study contributed to consistency in qualitative data collection and therefore enhanced dependability. Furthermore, the interview guide was influenced by the FINC-NA, and the open-ended questions for the nurses were intended to ask questions similar to those for the MDs. Using parallel questions between different groups in a mixed-methods convergent parallel design facilitates the merging process and is a strategy to minimize validity threats in mixed-method convergent design studies (Creswell & Plano Clark, 2018). Having different populations in the two paradigms made the mixed-methods merging and inference in this study complicated, as previously described in the literature (Creswell & Plano Clark, 2018). On the other hand, the inclusion of different populations for the different data collection strategies facilitated the usage of appropriate quality criteria according to paradigm, important for minimizing validity threats in mixed-methods studies (Creswell & Plano Clark, 2018). The FINC-NA has previously been applied to MDs working in neonatal care, with the substitution of the term “nursing care” for “care” (Naef et al., 2020). Even though the present study shows many similarities in attitudes between nurses and MDs, there are reasons to believe that an instrument developed for nurses’ attitudes towards family involvement in nursing care should not be directly applied to MDs attitudes towards family involvement in care as the instrument is not validated for the MD population. Another potential validity threat in convergent mixed-method studies is “failing to resolve disconfirming results” (Creswell & Plano Clark, 2018, p. 251). The authors of this study consider that suggestions for resolving the inferences of divergent concepts have been presented. However, wether these inferences can be considered valid is debatable.
The multidisciplinary composition of the research group may have contributed to the strengthening of credibility regarding qualitative analysis. Credibility was further enhanced by recruiting study participants from several different departments. Site credibility reduces the risk of having local factors influence the results (Shenton, 2004). Regarding transferability and generalizability, one problem was mutual between the paradigms, that is, selection bias. The aspiration to recruit all 650 patients who fulfilled the inclusion criteria in the cross-sectional study was part of the attempt to enhance the generalizability. The low response rate, however, limited this effect. The low response rate could be explained to some extent by the high strain on nurses during the COVID-19 pandemic in 2020. It is possible that the persons willing to participate in this study were the ones most favourable towards family involvement in the open-heart surgical care setting. This could, on the other hand, also be an advantage in the qualitative paradigm, where interested participants may have been able to provide richer data (Patton, 2015). The MDs participating in this study may also have considered the contribution to research to be important since $55\%$ of them held a doctoral degree. Interviewing via a videoconferencing platform (i.e., Zoom) could have affected the richness of the data. The richness of the data could also have been affected by the fact that the first author had a professional relationship with some of the interviewees. These MDs could have responded in a more positive way, giving more positive self-descriptions or responding in a way they believed that the first author would want them to, thereby leading to socially desirable responses (Malham & Saucier, 2016). However, the elaborative answers from the MDs and their willingness to share their experiences implied that this was not a problem in the present study. The restrictions of family visitations during the COVID-19 pandemic and the heavy workload during the data collection period could have influenced the results of this study. One strategy targeting this limitation was the instructions to both nurses and MDs to answer while having the normal situation in mind.
## Conclusions
This study has provided knowledge regarding attitudes among nurses and MDs towards family involvement in open-heart surgical care. Foremost, positive attitudes, including views on how family involvement improves postoperative recovery, have been illuminated and could be regarded as a basis for family-centred care practices. Some areas of improvement in terms of negative attitudes could be targeted by implementing structured assessment of family functioning and illness beliefs. Competencies in family-centred practices and transcultural care could enhance family involvement in the open-heartcare context. There are organizational demands to further improve family involvement, such as prioritizing time, providing physical space for families and having family-centred policies. Family involvement in open-heart surgical care may be necessary due to the patient’s and family’s unique needs in relation to open-heart surgery. Identifying those needs, as opposed to letting one’s own unconscious personal beliefs determine the extent of family involvement, demands ethical reasoning and family care competency among all members of the team.
## Disclosure statement
No potential conflict of interest was reported by the authors.
## Author controbutions
Study conception and design was performed by AD, EE, SÅ and ASS. AD collected the data. All authors participated in data analysis. The first draft of the manuscript was written by AD, and all authors gave critical feedback and edited previous versions of the manuscript. All authors read and approved the final version of the manuscript.
## Data availability statement
Due to the nature of this research, participants of this study did not agree for their data to be shared publicly, so supporting data is not available.
## References
1. Al Mutair A., Plummer V., O’brien A. P., Clerehan R.. **Attitudes of healthcare providers towards family involvement and presence in adult critical care units in Saudi Arabia: A quantitative study**. (2013) **23** 744-19. DOI: 10.1111/jocn.12520
2. Altmann T. K.. **Attitude: A concept analysis**. (2008) **43** 144-150. DOI: 10.1111/j.1744-6198.2008.00106.x
3. Bakanauskas A. P., Kondrotienė E., Puksas A.. **The theoretical aspects of attitude formation factors and their impact on health behaviour**. (2020) **83** 15-36. DOI: 10.1515/mosr-2020-0002
4. Barreto M. S., Marquete V. F., Camparoto C. W., García‐vivar C., Barbieri‐figueiredo M. D. C., Marcon S. S.. **Factors associated with nurses’ positive attitudes towards families’ involvement in nursing care: A scoping review**. (2022) **31** 3338-3349. DOI: 10.1111/jocn.16226
5. Bélanger L., Desmartis M., Coulombe M.. **Barriers and facilitators to family participation in the care of their hospitalized loved ones**. (2018) **5** 56-65. DOI: 10.35680/2372-0247.1250
6. Bell J. M.. **Family nursing is more than family centered care**. (2013) **19** 411-417. DOI: 10.1177/1074840713512750
7. Benzein E., Johansson P., Årestedt K. F., Berg A., Saveman B. I.. **Families’ importance in nursing care**. (2008) **14** 97-117. DOI: 10.1177/1074840707312716
8. Benzein E., Johansson P., Årestedt K. F., Saveman B. -I.. **Nurses’ attitudes about the importance of families in nursing care**. (2008) **14** 162-180. DOI: 10.1177/1074840708317058
9. Bjørnnes A. K., Moons P., Parry M., Halvorsen S., Tønnessen T., Lie I.. **Experiences of informal caregivers after cardiac surgery: A systematic integrated review of qualitative and quantitative studies**. (2019) **9** e032751. DOI: 10.1136/bmjopen-2019-032751
10. Blöndal K., Zoëga S., Hafsteinsdottir J. E., Olafsdottir O. A., Thorvardardottir A. B., Hafsteinsdottir S. A., Sveinsdóttir H.. **Attitudes of registered and licensed practical nurses about the importance of families in surgical hospital units**. (2014) **20** 355-375. DOI: 10.1177/1074840714542875
11. Cranley L. A., Lam S. C., Brennenstuhl S., Kabir Z. N., Boström A. M., Leung A., Konradsen H.. **Nurses’ attitudes toward the importance of families in nursing care: A multinational comparative study**. (2022) **28** 69-82. DOI: 10.1177/10748407211042338
12. Creswell J. W., Plano Clark V. L.. (2018)
13. Davidson J. E., Aslakson R. A., Long A. C., Puntillo K. A., Kross E. K., Hart J., Cox C. E., Wunsch H., Wickline M. A., Nunnally M. E., Netzer G., Kentish-Barnes N., Sprung C. L., Hartog C. S., Coombs M., Gerritsen R. T., Hopkins R. O., Franck L. S., Skrobik Y., Kon A. A., Curtis J. R.. **Guidelines for family-centered care in the neonatal, pediatric, and adult ICU**. (2017) **45** 103-128. DOI: 10.1097/ccm.0000000000002169
14. Davis R., Savvopoulou M., Shergill R., Shergill S., Schwappach D.. **Predictors of healthcare professionals’ attitudes towards family involvement in safety-relevant behaviours: A cross-sectional factorial survey study**. (2014) **4** e005549. DOI: 10.1136/bmjopen-2014-005549
15. Dijkman B. L., Luttik M. L., Van der Wal-Huisman H., Paans W., van Leeuwen B. L.. **Factors influencing family involvement in treatment decision-making for older patients with cancer: A scoping review**. (2021) **13** 391-397. DOI: 10.1016/j.jgo.2021.11.003
16. Elo S., Kyngäs H.. **The qualitative content analysis process**. (2008) **62** 107-115. DOI: 10.1111/j.1365-2648.2007.04569.x
17. Eskes A. M., Schreuder A. M., Vermeulen H., Nieveen van Dijkum E. J. M., Chaboyer W.. **Developing an evidence-based and theory informed intervention to involve families in patients care after surgery: A quality improvement project**. (2019) **6** 352-361. DOI: 10.1016/j.ijnss.2019.09.006
18. Göktas S. B., Yildiz T., Nargiz S. K., Gur O.. **A comparison of the intensive care experiences of emergency and elective cardiac surgery patients**. (2016) **19** 284-289. DOI: 10.4103/1119-3077.175963
19. Gusdal A. K., Josefsson K., Thors Adolfsson E., Martin L.. **Nurses’ attitudes toward family importance in heart failure care**. (2017) **16** 256-266. DOI: 10.1177/1474515116687178
20. Head S. J., Milojevic M., Taggart D. P., Puskas J. D.. **Current practice of state-of-the-art surgical coronary revascularization**. (2017) **136** 1331-1345. DOI: 10.1161/circulationaha.116.022572
21. Health Research & Educational Trust. (2013) . Becoming a culturally competent health care organization.. (2013)
22. Johnson B. H., Abraham M. R.. (2012)
23. Jordan A. L., Rojnica M., Siegler M., Angelos P., Langerman A.. **Surgeon-family perioperative communication: Surgeons’ self-reported approaches to the “surgeon-family relationship”**. (2014) **219** 958-967. DOI: 10.1016/j.jamcollsurg.2014.05.019
24. Joseph H. K., Whitcomb J., Taylor W.. **Effect of anxiety on individuals and caregivers after coronary artery bypass grafting surgery**. (2015) **34** 285-288. DOI: 10.1097/dcc.0000000000000137
25. Kemp K. A., Naqvi F., Quan H., Paolucci E. O., Knudtson M. L., Santana M. J.. **Eliciting patient experiences about their care after cardiac surgery**. (2020) **3** 427-433. DOI: 10.1016/j.cjco.2020.11.016
26. Laidsaar-Powell R. C., Butow P. N., Bu S., Charles C., Gafni A., Lam W. W., Jansen J., McCaffery K. J., Shepherd H. L., Tattersall M. H., Juraskova I.. **Physician-patient-companion communication and decision-making: A systematic review of triadic medical consultations**. (2013) **91** 3-13. DOI: 10.1016/j.pec.2012.11.007
27. Laidsaar-Powell R., Butow P., Bu S., Fisher A., Juraskova I.. **Oncologists’ and oncology nurses’ attitudes and practices towards family involvement in cancer consultations**. (2017) **26** e12470. DOI: 10.1111/ecc.12470
28. Linnarsson J. R., Benzein E., Årestedt K.. **Nurses’ views of forensic care in emergency departments and their attitudes, and involvement of family members**. (2015) **24** 266-274. DOI: 10.1111/jocn.12638
29. Luttik M., Goossens E., Ågren S., Jaarsma T., Mårtensson J., Thompson D. R., Moons P., Strömberg A.. **Undertaking Nursing Interventions Throughout Europe (UNITE) research group**. (2017) **16** 299-308. DOI: 10.1177/1474515116663143
30. Mackie B. R., Marshall A., Mitchell M.. **Acute care nurses’ views on family participation and collaboration in fundamental care**. (2018) **27** 2346-2359. DOI: 10.1111/jocn.14185
31. Mackie B. R., Mitchell M., Marshall A. P.. **Patient and family members’ perceptions of family participation in care on acute care wards**. (2019) **33** 359-370. DOI: 10.1111/scs.12631
32. Malham P. B., Saucier G.. **The conceptual link between social desirability and cultural normativity**. (2016) **51** 474-480. DOI: 10.1002/ijop.12261
33. McDermott K. W., Liang L.. **Overview of operating room procedures during inpatient stays in U.S. hospitals, 2018**. (2021)
34. Murcia S. E., Lopez L.. **The experience of nurses in care for culturally diverse families: A qualitative meta-synthesis**. (2016) **24** e2718. DOI: 10.1590/1518-8345.1052.2718
35. Naef R., Kläusler-Troxler M., Ernst J., Huber S., Dinten-Schmid B., Karen T., Petry H.. **Translating family systems care into neonatology practice: A mixed method study of practitioners’ attitudes, practice skills and implementation experience**. (2020) **102** 103448. DOI: 10.1016/j.ijnurstu.2019.103448
36. Olding M., McMillan S. E., Reeves S., Schmitt M. H., Puntillo K., Kitto S.. **Patient and family involvement in adult critical and intensive care settings: A scoping review**. (2016) **19** 1183-1202. DOI: 10.1111/hex.12402
37. Østergaard B., Clausen A. M., Agerskov H., Brødsgaard A., Dieperink K. B., Funderskov K. F., Nielsen D., Sorknæs A. D., Voltelen B., Konradsen H.. **Nurses’ attitudes regarding the importance of families in nursing care: A cross‐sectional study**. (2020) **29** 1290-1301. DOI: 10.1111/jocn.15196
38. Patton M. Q.. (2015)
39. Pomare C., Long J. C., Churruca K., Ellis L. A., Braithwaite J.. **Interprofessional collaboration in hospitals: A critical, broad-based review of the literature**. (2020) **34** 509-519. DOI: 10.1080/13561820.2019.1702515
40. Robley L., Ballard N., Holtzman D., Cooper W.. **The experience of stress for open heart surgery patients and their caregivers**. (2010) **32** 794-813. DOI: 10.1177/0193945910361469
41. Rosland A. M., Piette J. D., Choi H., Heisler M.. **Family and friend participation in primary care visits of patients with diabetes or heart failure**. (2011) **49** 37-45. DOI: 10.1097/mlr.0b013e3181f37d28
42. Saveman B. I., Benzein E. G., Engström Å. H., Årestedt K.. **Refinement and psychometric reevaluation of the instrument**. (2011) **17** 312-329. DOI: 10.1177/1074840711415074
43. Sedaghat S., Rostami S., Ebadi A., Fereidooni-Moghadam M.. **Stressors in open-heart surgery patients: A qualitative study**. (2019) **15** 192-200. DOI: 10.22122/arya.v15i4.1840
44. **Patientlag [Swedish patient act]. Socialdepartementet**. (2014)
45. Shajani Z., Snell D.. (2019)
46. Shamali M., Esandi Larramendi N., Østergaard B., Barbieri-Figueiredo M., Brødsgaard A., Canga-Armayor A., Dieperink K. B., Garcia-Vivar C., Konradsen H., Nordtug B., Lambert V., Mahrer-Imhof R., Metzing S., Nagl-Cupal M., Imhof L., Svavarsdottir E. K., Swallow V., Luttik M. L.. **Nurses’ attitudes towards family importance in nursing care across Europe**. (2022). DOI: 10.1111/jocn.16456
47. Shenton A. K.. **Strategies for ensuring trustworthiness in qualitative research projects**. (2004) **22** 63-75. DOI: 10.3233/efi-2004-22201
48. Shin D. W., Cho J., Roter D. L., Kim S. Y., Yang H. K., Park K., Kim H. J., Shin H. Y., Kwon T. G., Park J. H.. **Attitudes toward family involvement in cancer treatment decision making: The perspectives of patients, family caregivers, and their oncologists**. (2017) **26** 770-778. DOI: 10.1002/pon.4226
49. Stephens R. S., Whitman G. J. R.. **Postoperative critical care of the adult cardiac surgical patient. Part I**. (2015) **43** 1477-1497. DOI: 10.1097/ccm.0000000000001059
50. Tong A., Sainsbury P., Craig J.. **Consolidated criteria for reporting qualitative research (COREQ): A 32-item checklist for interviews and focus groups**. (2007) **19** 349-357. DOI: 10.1093/intqhc/mzm042
51. von Elm E., Altman D. G., Egger M., Pocock S. J., Gøtzsche P. C., Vandenbroucke J. P.. **The strengthening the reporting of observational studies in epidemiology (STROBE) statement: Guidelines for reporting observational studies**. (2014) **12** 1495-1499. DOI: 10.1016/j.ijsu.2014.07.013
52. **Framework for action on interprofessional education and collaborative practice**. (2010)
53. **World medical association declaration of helsinki: Ethical principles for medical research involving human subjects**. (2013) **310** 2191-2194. DOI: 10.1001/jama.2013.281053
54. Younas A., Durante A.. **Decision tree for identifying pertinent integration procedures and joint displays in mixed methods research**. (2022). DOI: 10.1111/jan.15536
|
---
title: Vitamin D modulates inflammatory response of DENV-2-infected macrophages by
inhibiting the expression of inflammatory-liked miRNAs
authors:
- Jorge Andrés Castillo
- Silvio Urcuqui-Inchima
journal: Pathogens and Global Health
year: 2022
pmcid: PMC9970239
doi: 10.1080/20477724.2022.2101840
license: CC BY 4.0
---
# Vitamin D modulates inflammatory response of DENV-2-infected macrophages by inhibiting the expression of inflammatory-liked miRNAs
## ABSTRACT
Dengue disease caused by dengue virus (DENV) infection is the most common vector-borne viral disease worldwide. Currently, no treatment is available to fight dengue symptoms. We and others have demonstrated the antiviral and immunomodulatory properties of VitD3 as a possible therapy for DENV infection. MicroRNAs (miRNAs) are small non-coding RNAs responsible for the regulation of cell processes including antiviral defense. Previous transcriptomic analysis showed that VitD3 regulates the expression of genes involved in stress and immune response by inducing specific miRNAs. Here, we focus on the effects of VitD3 supplementation in the regulation of the expression of inflammatory-liked miR-182-5p, miR-130a-3p, miR125b-5p, miR146a-5p, and miR-155-5p during DENV-2 infection of monocyte-derived macrophages (MDMs). Further, we evaluated the effects of inhibition of these miRNAs in the innate immune response. Our results showed that supplementation with VitD3 differentially regulated the expression of these inflammatory miRNAs. We also observed that inhibition of miR-182-5p, miR-130a-3p, miR-125b-5p, and miR-155-5p, led to decreased production of TNF-α and TLR9 expression, while increased the expression of SOCS-1, IFN-β, and OAS1, without affecting DENV replication. By contrast, over-expression of miR-182-5p, miR-130a-3p, miR-125b-5p, and miR-155-5p significantly decreased DENV-2 infection rates and also DENV-2 replication in MDMs. Our results suggest that VitD3 immunomodulatory effects involve regulation of inflammation-linked miRNAs expression, which might play a key role in the inflammatory response during DENV infection.
## Introduction
Dengue disease caused by the transmission of the arthropod-borne dengue virus (DENV) considered a major health problem in developing countries [1]. The World Health Organization (OMS) estimates that about $50\%$ of the world population lives in areas where the DENV transmitter mosquito Aedes spp. is present. Despite the high economic and social burden of dengue disease, there is no specific treatment available, and approved vaccine *Denvaxia is* only partially effective [2]. Furthermore, the precise mechanisms that mediate the development of severe manifestations are not fully understood. However, a number of studies have shown that uncontrolled and exacerbated inflammatory response is responsible for such complications observed in some DENV infected patients [3–5].
An exacerbated and sustained inflammatory response is a hallmark of DENV pathogenesis [3,6]. Thus, development of new therapeutics that modulate inflammatory response while also restricts DENV replication is needed. Vitamin D (VitD3) is a pleiotropic hormone that has wide immunoregulatory effect [7]. VitD3 downregulates the expression of several Toll-like receptors (TLRs) and thus contributes to the amelioration of inflammatory response triggered by interferon-gamma (IFN-γ), lipopolysaccharide (LPS), or lipoteichoic acid [8,9]. These observations suggest that VitD3 could protect from exacerbated inflammatory responses. We and others have previously shown that VitD3 decreases DENV infection and replication and the production of proinflammatory cytokines in vitro [10–13]. In addition, we reported a decreased susceptibility to DENV-2 infection and production of proinflammatory cytokines by monocyte-derived macrophages (MDMs) and monocyte-derived dendritic cells (MoDCs) obtained from healthy individuals who received oral supplementation of VitD3 [14,15]. This body of evidence demonstrates that VitD3 has a wide immunomodulatory and antiviral effect against DENV infection. However, specific mechanisms by which VitD3 regulates the inflammatory response during DENV infection remain unclear.
VitD3 can regulate expression of several genes that harbors Vitamin D response elements [16,17]. One of these regulated genes could encode microRNAs (miRNAs), which are small non-coding RNAs of 20–25 nucleotides that post-transcriptionally regulate the expression of a great number of genes involved in cell development, cell division, metabolism, programmed cell death, and viral pathogenesis [18]. Previous studies have shown that MoDCs treated with VitD3 produce low levels of IL-23 and show low expression of miR-155-5p, while the expression levels of miR-378 are high [19]. Similarly, VitD3 treatment of human adipocytes reduced the expression of miR-146a-5p, miR-150, and miR-155-5p during TNF-α stimulation [20]. Among these VitD3 regulated miRNAs, miR-155-5p appears to play a key role. VitD3 treatment in murine macrophages reduced the expression of miR-155-5p, which in turn increased the activity of SOCS-1 leading to a decreased inflammatory response after TLR4 activation [21]. Further, we observed that differentiation of MDMs in the presence of VitD3 (D3-MDMs) reduced miR-155-5p expression during DENV-2 infection, which was linked to increased expression of SOCS-1 [22]. Finally, we found that MDMs obtained from healthy donors who received VitD3 supplement for 10 days, showed differential expression of a set of miRNAs during DENV-2 infection in vitro, which target immune and cellular stress response genes [23]. The results suggest that VitD3 regulates the expression of miRNAs in inflammatory conditions.
This study aims to evaluate the modulation of inflammation-linked miRNAs by VitD3 in DENV-2-infected D3-MDMs. Further, we determined the role of these miRNAs in the inflammatory response by assessing the expression of pro-inflammatory cytokines, pattern recognition receptors (PRRs), SOCS-1, IFN-I, and IFN-stimulated genes (ISGs) in MDMs under miRNA inhibition conditions. Collectively, our results show that VitD3 treatment can modulate inflammatory response by the regulation of inflammatory-linked miRNAs in DENV-2 infected D3-MDMs, which may be associated with decreased expression of TNF-α and TLR9 and increased expression of SOCS-1, IFN-β, and OAS-1.
## Ethics statement
The protocols for sample collection were approved by the Committee of Bioethics Research of Sede de Investigación Universitaria, Universidad de Antioquia (Medellín, Colombia), and inclusion was preceded by a signed informed consent form, according to the principles expressed in the Declaration of Helsinki.
## Cells and reagents
The mosquito C$\frac{6}{36}$ HT cell line was obtained from the American Type Culture Collection (ATCC) and cultured in Leibovitz L-15 medium (Sigma Aldrich, USA) supplemented with $10\%$ v/v heat-inactivated fetal bovine serum (FBS) (Thermo Scientific, USA), 4 mM L-glutamine, 10 U/mL penicillin, and 0.1 mg/mL streptomycin (Sigma Aldrich, USA), at 34°C in an atmosphere without CO2. BHK-21 cells, obtained from the ATCC, were maintained in D-MEM (Sigma Aldrich, USA), supplemented with $10\%$ v/v FBS, 4 mM L-glutamine, 10 CFU/mL penicillin, and 0.1 mg/mL streptomycin at 37°C with $5\%$ CO2, and used for plaque assays. A conjugated antibody against CD14 (clone M5E2) was purchased from eBioscience (USA).
## Virus stocks and titration
DENV-2 New Guinea C was provided by the Centers for Disease Control and Prevention (CDC, USA). Viral stocks were obtained by inoculating a monolayer of C$\frac{6}{36}$ HT cells in a 75-cm2 tissue culture flask with the virus at a multiplicity of infection (MOI) of 0.05 diluted in L-15 supplemented with $2\%$ FBS. After 3 h of adsorption, fresh L-15 medium supplemented with $2\%$ FBS was added, and the cells were cultured for 5 days at 34°C without CO2. The supernatant was obtained by centrifugation at 1000 × g for 5 min to remove cellular debris and then aliquoted and stored at −70°C until use. Virus titration was performed by quantification of plaque-forming units (PFU) using a plaque assay as described previously [13].
## Blood samples from healthy donors
Venous peripheral blood samples were obtained from healthy individuals, aged 20–40 years, who had not been previously vaccinated against yellow fever virus and were seronegative for the DENV NS1 antigen and DENV IgM/IgG, as determined by the SD BIOLINE Dengue Duo rapid test (Standard Diagnostics). All our experiments were performed with cells from at least six healthy donors.
## Monocyte isolation and monocyte-derived macrophage differentiation (MDMs)
To obtain MDMs, peripheral blood mononuclear cells (PBMCs) were obtained from 50 mL of peripheral blood from healthy individuals with $2\%$ v/v ethylenediaminetetraacetic acid, as described previously [16,17]. Briefly, the PBMCs were separated using density gradient centrifugation and suspended in RPMI-1640 medium (Sigma Aldrich, USA) supplemented with $0.5\%$ autologous heat-inactivated serum (30 min at 56°C). Monocytes were then obtained from the PBMCs by plastic adherence, as described previously [12]. Briefly, 5 × 105 CD14+ cells into 24-well plates (Corning Incorporated Life Sciences, USA) in RPMI-1640 medium supplemented with $0.5\%$ inactivated autologous serum and cultured at 37°C with $5\%$ CO2 to allow enrichment of monocytes through plastic adherence. After 3 h, the non-adherent cells were removed by extensive washing with pre-warmed PBS supplemented with $0.5\%$ FBS. Adherent cells were then cultured in RPMI-1640 medium supplemented with $10\%$ FBS at 37°C with $5\%$ CO2 for 6 days to obtain MDMs. Fresh medium with $10\%$ FBS was replenished every 48 h. The purity of MDMs and D3-MDMs was repeatedly above $90\%$, as measured by the presence of contaminant cell populations, including CD19+, CD3+, and CD56+ in monocytes before differentiation, and measuring CD68+ cells after 6 days of differentiation.
## MDMs differentiation in the presence of VitD3 (D3-MDMs)
Monocytes were differentiated for 6 days in the presence of 1α,25-dihydroxyvitamin D3 (VitD3; Sigma Aldrich, USA), at a concentration of 0.1 nM, which represents the physiological and therapeutical concentration [23,24], as we have described previously [12,13]. The biological activity VitD3 was determined by us previously by the quantification of the transcriptional induction of VitD3 targets genes, such as VDR and CYP24A1 [12,13]. The purity of D3-MDMs was repeatedly above $90\%$, as measured by the presence of contaminant cell populations, including CD19+, CD3+, and CD56+ (non-myeloid cells).
## MDMs and D3-MDMs infection with DENV
Both MDMs and D3-MDMs monolayers were challenged with DENV-2 at an MOI of 5, diluted in 300 µl of RPMI-1640 medium supplemented with $2\%$ FBS. Two hours post-infection (hpi), cells were washed with PBS, and the medium was replenished with RPMI $10\%$ FBS and then cultured at 37°C $5\%$ CO2. At 2, 8, and 24 hpi, monolayers were harvested, and either the percentage of infection was determined by flow cytometry, or total RNA extraction was carried out and used for viral RNA quantification, while the supernatants were used for viral titration by plaque assay and for quantification of cytokine production.
## Quantification of miRNA expression
Total RNA was obtained from MDMs and D3-MDMs, in both mock and DENV-2 infected samples, using the kit Direct-zol RNA miniprep (Zymo Research, USA) following the manufacturer´s instructions. The RNA concentration was quantified using a NanoDrop spectrophotometer (NanoDrop Technologies, USA). MicroRNA cDNA was synthesized from 1 μg total RNA samples using specific miRNA stem-loop primers and TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific, USA). miRNA qPCR analysis was performed in a 15 μl reaction (TaqMan™ Gene Expression Master Mix for miRNAs; Thermo Fisher Scientific, USA), and run on a Bio-Rad CFX PCR System using the following cycle conditions: 95°C for 10 mins followed by 40 cycles of 95°C for 15 secs and 60°C for 1 min. The TaqMan Assays for the following miRNAs were used: miR-182-5p (Assay ID # 000597), miR-146a-5p (Assay ID # 000468), miR-130a-3p (Assay ID # A25576), miR-125b-5p (Assay ID # 000449), miR-155-5p (Assay ID # 002623), and RNU48 (Assay ID # 001006). RNU48 was used as a reference gene to normalize the miRNA. Relative quantification of miRNA expression was evaluated using the 2-ΔΔCT method. Cutoffs for significant changes were set at p-value ≤ 0.05.
## Inhibition and overexpression of miRNAs
To inhibit the expression or overexpression assessment of selected miRNAs, MDMs were transfected with synthetic miR-182-5p, miR-130a-3p, miR-125b-5p and miR-155-5p antisense (inhibition) or mimics (over-expression) at a final concentration of 50 nM/well (Ambion, TX, USA), using DharmaFect (Thermo Scientific, NH, USA) according to manufacturer’s instructions. At 24 h post-transfection, cells were infected with DENV-2 following the procedure described above. At 24 hpi, cell monolayers were harvested and used to assess the percentage of DENV-2 infection and quantification of gene expression. The cell supernatants were used for quantification of viral RNA copies by RT-qPCR, viral titer by plaque assay, and cytokine quantification by ELISA.
## Flow cytometry assays
Flow cytometry was used to assess the frequency of DENV-infected cells [16,17]. Briefly, DENV infection was evaluated through the intracellular detection of DENV E antigen at 24 hpi fixing the cells with fixation/permeabilization buffer (eBioscience, USA). Following washing steps with PBS, cells were stained with the monoclonal antibody 4G2 (Millipore, Germany) for 40 min, followed by 40 min staining with goat anti-mouse IgG-FITC (Thermo Scientific, USA). All data acquisition and analysis were done using the BD FACScan system and FACSDiva software, respectively.
## Quantitation of viral RNA copy number
Total RNA was purified from DENV-2-infected and mock-infected MDMs using TRIzol reagent (Thermo Scientific, USA) following the manufacturer’s instructions. The RNA concentration was quantified using a NanoDrop spectrophotometer (NanoDrop Technologies, USA). Then, cDNA was synthesized using random primers from a standard concentration of 50 ng of RNA and the RevertAid H Minus First-Strand cDNA Synthesis Kit (Thermo Scientific, USA) following the manufacturer’s instructions. Viral RNA copies quantification with cDNA was carried out using specific primers, as described previously [25]. qPCR was performed with Maxima SYBR Green qPCR Master Mix (Thermo Scientific, USA) and analyzed with the CFX96 Touch Real-Time PCR Detection System (Bio-Rad, USA). The quantification of viral RNA copies was based on a standard curve of Ct values of 10-fold serial dilutions of a plasmid encoding the full genome of DENV-2 of known length and concentration, as previously described [26].
## Cytokine production
The levels of IL 6 and TNF-α were assessed in supernatants from MDMs infected with DENV-2 at 24 hpi under miRNA inhibition conditions, using an ELISA assay (BD OptEIA, BD Biosciences, USA), following the manufacturer’s recommendations.
## Quantification of gene expression
The mRNA quantification of TLR3, TLR4, TLR9, RIG-I, IFN-α, IFN-β, protein kinase R (PKR), 2ʹ-5ʹ-oligoadenylate synthetase 1 (OAS1), CAMP, SOCS-1, and ubiquitin was performed in mock and DENV-2-infected MDMs after expression inhibition of selected miRs, using qPCR. Briefly, cDNA was synthesized from total RNA using random primers, a standard concentration of 50 ng of RNA, and the RevertAid H Minus First-Strand cDNA Synthesis Kit (Thermo Scientific, USA). Then, qPCR was performed using Maxima SYBR Green (Thermo Scientific, USA), following the manufacturer’s instructions using specific primers, as described previously [25]. The specificity of the amplification product was determined by melting curve analysis. The relative quantification of each mRNA was normalized to the constitutive gene, ubiquitin, and mock-treated MDMs and D3-MDMs from each time point evaluated (e.g. 2 hours mock MDM vs 2 hours DENV-2-infected MDMs), using the ΔΔCt method and reported as the fold change.
## Statistical analysis
Comparisons between MDMs and D3-MDMs were undertaken using a two-way analysis of variance (ANOVA) along with a Bonferroni test. A value of $p \leq 0.05$ was considered statistically significant. The calculation of these parameters was carried out using GraphPad Prism version 6 (GraphPad Software, USA) software.
## VitD modulates expression of miRNAs in DENV-2 infected MDMs
Previously, we showed that a high supplement of VitD3 regulates expression of miRNAs during DENV-2 infection, which in turn regulated genes involved in stress and immune response [27]. From these differentially expressed miRNAs, we selected to study VitD3-modulated miR-182-5p, miR-130a-3p, and miR-125b-5p, because their expression has been associated with an inflammatory response in various diseases [28–30]. In addition, we included miR-146a-5p since it is overexpressed early during DENV replication, and mediates IFN-I evasion [20,31]. We also studied miR-155-5p as we and others have previously described that VitD3 can regulate its expression, and it has been associated with inflammation [21,22]. Therefore, expression of miR-182-5p, miR-130a-3p, miR-125b-5p, miR-146a-5p and miR-155-5p was quantified in MDMs and D3-MDMs during DENV-2 infection in vitro at 2, 8 and 24 hpi.
DENV-2 induced miRNA expression in MDMs depending on the time of infection. MiR-182-5p expression increased at 8 and 24 hpi, while miR-146a-5p and miR-155-5p increased significantly at 2 and 24 hpi, compared to the mock MDMs (Figure 1a,c, and e, respectively). On the other hand, expression of miR-130a-3p in DENV-2 infected MDMs significantly decreased at 24 hpi as compared to mock-infected MDMs (Figure 1b). Notably, during DENV-2 infection of D3-MDMs, expression of miR-182-5p remained unchanged at 2, 8 and 24 hpi compared to mock-infected MDMs (Figure 1a), the expression of miR-146a-5p did not change at 2 and 24 hpi compared as well with mock-infected MDMs (Figure 1c). Similarly, expression of miR-130a-3p significantly decreased in DENV-2 infected D3-MDMs at 24 hpi as compared to both mock and DENV-2-infected MDMs (Figure 1b). We did not observe differences in the expression of miR-125b-5p and miR-155-5p between MDMs and D3-MDMs in the presence of 0.1 nM of VitD3 and in response to DENV-2 infection Figure 1(d,e). Since it has been reported that VitD3 regulates the expression of these last two miRNAs [21,32], we differentiated D3-MDMs with increasing concentrations of VitD3, followed by infection with DENV-2, and evaluated the expression of miR-125b-5p and miR-155-5p. We observed that miR-125b-5p and miR-155-5p were downregulated in D3-MDMs at VitD3 doses of 10 nM and 1 nM, respectively (Figure 1(f,g). The results suggest that VitD3 modulates expression of miR-125b-5p and miR-155-5p at different concentrations. In summary, our data indicate that VitD3 downregulates the expression of inflammatory-response miRNAs, including miR-182-5p, miR-130-3p, miR-146a-5p, miR-125b-5p, and miR-155-5p in DENV-2-infected D3-MDMs. Figure 1.Differentiation of MDMs in the presence of VitD3 decreases the expression of miR-182-5p, miR-130a-3p, miR-146a-5p, miR-125b-5p and miR-155-5p during DENV-2 infection. The MDMs were differentiated in the presence of VitD3 (0.1 nM) for 6 days (D3-MDMs) and then infected with DENV-2 with an MOI of 5 for 2, 8, or 24 h. Expression of miR-182-5p (a), miR-130a-3p (b), miR-146a-5p (c), miR-125b-5p (d) and miR-155-5p (e) was measured by qPCR in MDMs and D3-MDMs using RNU48 as a housekeeper gene. Data are expressed as fold change relative to mock-treated MDMs and D3-MDMs from each time point. The MDMs were also differentiated with increasing concentrations of VitD3 from 0.1 to 100 nM and then infected with DENV-2 for 24 h. Expression of miR-125b-5p (f) and miR-155-5p (g) was measured by qPCR in D3-MDMs using RNU48 as a housekeeper gene. Figures represent five individual experiments. Differences were identified using a two-way ANOVA with a Bonferroni test for A-E and using a Kruskal–Wallis test for F and G. In both cases a $95\%$ confidence interval was used (***$p \leq 0.001$, **$p \leq 0.01$, *$p \leq 0.05$).
## Inhibition of miR-182-5p, miR-130a-3p or miR-125b-5p expression in DENV-2 infected-MDMs lead to lower production of TNF-α
VitD3 decreased the expression of miR-182-5p, miR-130-3p, miR-125b-5p, and miR-155-5p, which are shown to be associated with inflammatory response in some diseases [20,28–30]. Furthermore, we have previously shown that VitD3 can decrease the production of pro-inflammatory cytokines during DENV-2 infection of macrophages [12,13]. To test whether the decrease of miR-182-5p, miR-130a-3p, miR-125b-5p or miR-155-5p expression by VitD3 contributes to the modulation of inflammatory response of MDMs during DENV-2 infection, we transfected MDMs with anti-sense oligonucleotides against each of these miRNAs. With this strategy, a miRNA-miRNA duplex is expected to be formed inhibiting miRNA function. 24 hours after transfection, MDMs were infected with DENV-2 for an additionally 24 hours, and the production of TNF-α and IL-6 was quantified. Inhibition of each miRNA was first confirmed through qPCR. Transfection resulted in a 10-fold decrease in expression of each miRNA (Supplemental Fig 1A). Further, inhibition of these miRNAs did not affect MDMs viability under DENV-2 infection conditions (Supplemental Fig 1B).
While inhibition of miR-182-5p, miR-130a-3p, and miR-125b-5p significantly decreased production of TNF-α in DENV-2 infected MDMs as compared to MDMs transfected with scramble miR control (Figure 2a), inhibition of miR-155-5p did not affect TNF-α production. IL-6 production on the other hand was not altered by inhibition of the miRNAs (Figure 2b). These results suggest that miR-182-5p, miR-130a-3p, and miR-25b-5p may contribute to the inflammatory response of DENV-2 infected MDMs as they regulate TNF-α production. Figure 2.Inhibition of the expression of selected miRNAs decreases the production of TNF-α in DENV-2 infected MDMs. MDMs were transfected either with a miR scrambled (miR control) or with an anti-sense specific miRNA. 24 hours later, cells were infected with DENV-2 at an MOI of 5, and at 24 hpi, the production of TNF-α (a) and IL-6 (b) was quantified by ELISA. Figures represent four individual experiments. Differences were identified using a Kruskal–Wallis test with a $95\%$ confidence interval was used (***$p \leq 0.001$, **$p \leq 0.01$, *$p \leq 0.05$).
## Inhibition of miRNAs expression in DENV-2 infected MDMs altered TLR9 and SOCS-1 mRNA expression
Macrophages sense DENV infection through an array of PRRs including retinoic acid-inducible gene-I-like receptors (RLRs) such as RIG-I [33,34], and Toll-like receptors (TLRs) [35,36]. Activation of these PRRs can lead to the production of inflammatory cytokines. To test whether regulation of the inflammatory response was mediated by the inhibition of miRNAs and/or the regulation of PRRs, we quantified expression levels of RIG-I, TLR3, TLR4, and TLR9 mRNA under inhibition of selected miRNAs in DENV-2-infected MDMs. We found that inhibition of miR-182-5p, miR-130a-3p, miR-125b-5p or miR-155-5p, was not linked to expression of RIG-I, TLR3, or TLR4 mRNAs in DENV-2 infected MDMs (Figure 3a,b and c). However, inhibition of miR-130a-3p, miR-125b-5p or miR-155-5p decreased expression of TLR9 mRNA in DENV-2 infected MDMs, although for miR-125b-5p and miR-155-5p it was not statistically significant (Figure 3d). We further evaluated the expression of SOCS-1 under miRNAs inhibition conditions, since the protein encoded by this gene has an important role in the negative feedback of proinflammatory cytokine signaling [37]. Inhibition of miR-182-5p or miR-155-5p significantly increased the expression of SOCS-1 mRNA in DENV-2 infected MDMs as compared to infected MDMs treated with scramble control (Figure 3e). Altogether, these results suggest that inflammatory response networks involving miR-130a-3p, miR-125b-5p, and miR155-5p may be associated with TLR9 in MDMs during DENV-2 infection. Also, inhibition of miR-182-5p and miR-155-5p upregulated SOCS-1, which could contribute to the regulation of inflammatory response. Figure 3.Inhibition of miR-130a-3p, miR-125b-5p and miR-155-5p leads to decreased mRNA expression of TLR9, while inhibition of miR-182-5p and miR-155-5p increases the expression of SOCS-1 mRNA in DENV-2 infected MDMs. MDMs were transfected either with a miRNA scrambled negative control or with an anti-sense specific miRNA. 24 hours later, cells were infected with DENV-2 at an MOI of 5. At 24 hpi, mRNA expression of RIG I (a), TLR3 (b), TLR4 (c), TLR9 (d), and SOCS-1 was measured by qPCR in MDMs using the gene encoding RNU48 as a housekeeper gene. Data are expressed as fold change relative to DENV-2 infected MDMs transfected with miRNA scrambled control. Figures represent four individual experiments. Differences were identified using a Kruskal–Wallis test with a $95\%$ confidence interval was used (***$p \leq 0.001$, **$p \leq 0.01$, *$p \leq 0.05$).
## Inhibition of miR-182-5p, miR-130a-3p, miR-125b-5p or miR-155-5p expression increased IFN-I and OAS1 mRNA levels in DENV-2 infected MDMs
We have previously shown that VitD3 induces a partial resistance to DENV-2 infection in MDMs via down-regulation of mannose receptor [12,13]. Also, VitD3 has been shown to increase LL-37 expression [38]. To test whether inhibition of inflammatory-response inked miRNAs regulated by VitD3 could lead to an improved antiviral response, we evaluated expression of IFN-α, IFN-β, PKR, OAS1, and CAMP (LL-37 gene) under miRNAs inhibition in DENV-2 infected MDMs. Inhibition of miR-125b-5p and miR-155-5p expression induced a statistically significant increase of IFN-α mRNA levels in DENV-2 infected MDMs as compared to scrambled miRNA transfected MDMs (Figure 4a). Similarly, inhibition of miR-182-5p, miR-130a-3p, and miR-155-5p significantly increased IFN-β mRNA levels in DENV-2 infected MDMs as compared to MDMs transfected with scrambled miRNA (Figure 4b). While inhibition of miR-182-5p, miR-130a-3p, miR-125b-5p and miR-155-5p did not affect the expression of PKR mRNA (Figure 4c), inhibition of miR-125b-5p and miR-155-5p significantly increased mRNA expression of OAS1, in DENV-2 infected MDMs (Figure 4d). Surprisingly, inhibition of miR-182-5p and miR-155-5p decreased expression levels of CAMP mRNA in DENV-2 infected MDMs as compared to scrambled control (Figure 4e). Overall, the results show that inhibition of the miRNAs increases expression of IFN-I and OAS1 mRNA in DENV-2 infected MDMs. Figure 4.Inhibition of the expression of miRNAs differentially modulates the mRNA levels of IFN-β, OAS-1, and CAMP in DENV-2 infected MDMs. MDMs were transfected either with a miRNA scrambled negative control or with an anti-sense specific miRNA. 24 hours later, cells were infected with DENV-2 at an MOI of 5 and, at 24 hpi the mRNA expression of IFN-α (a), IFN-β (b), PKR (c), OAS1 (d) and CAMP (e) was quantified by qPCR in MDMs using the gene encoding RNU48 as a housekeeper gene. Data are expressed as fold change relative to DENV-2 infected MDMs transfected with miRNA scrambled control. Figures represent four individual experiments. Differences were identified using a Kruskal–Wallis test with a $95\%$ confidence interval was used (***$p \leq 0.001$, **$p \leq 0.01$, *$p \leq 0.05$).
## Inhibition of miR-182-5p, miR-130a-3p, miR-125b-5p or miR-155-5p does not alter DENV-2 replication, whereas its overexpression resulted in inhibition of DENV-2 replication in MDMs
So far, we have shown that inhibition of miR-182-5p, miR-130a-3p, and miR-125b-5p decreased the production of TNF-α, modulated the expression of TLR9, and increased expression of IFN-I and OAS-1. However, the consequence of inhibition of such miRNAs on DENV-2 replication is unknown. Therefore, we tested if inhibiting expression of miRNAs, regulated by VitD3, could also have an impact on DENV-2 replication in MDMs. For this, MDMs were transfected with anti-sense inhibitor of miR-182-5p, miR-130a-3p, miR-125b-5p, and miR-155-5p, and at 24 hours post-trasnfection (hpt), MDMs were challenged with DENV-2. Finally, at 24 hpi viral replication was evaluated.
As shown in Figure 5, inhibition of miR-182-5p, miR-130a-3p, miR-126b-5p, and miR-155-5p had no effect in DENV-2 replication, since the percentage of E+ cells Figure 5(a,b) and viral titer (Figure 5c) were the same compared to control MDMs transfected with scrambled miRNA. In addition, we evaluated if overexpression of miR-182-5p, miR-130a-3p, miR-125b-5p, and miR-155-5p had any effect on DENV-2 replication. For this, MDMs were transfected with RNA mimic oligonucleotides corresponding to each miRNA, and at 24 hpt, MDMs were challenged with DENV-2. Viral replication was evaluated at 24 hpi. Overexpression of each miRNAs was confirmed through qPCR showing that transfection with miR-182-5p, miR-125b-5p, and miR-155-5p mimics resulted in a 100-fold increase, while a 50-fold increase was observed with miR-130a-3p mimic. Viability of MDMs was not affected by overexpression of miRNAs and DENV-2 infection (Supplemental Fig 2). Figure 5.Inhibition of the expression of miR-182-5p, miR-130a-3p, miR-125b-5p and miR-155-5p did not affect DENV-2 replication in MDMs. MDMs were transfected either with a miRNA scrambled negative control or with an anti-sense specific miRNA. 24 hours later, cells were infected with DENV-2 at an MOI of 5 and, at 24 hpi, the percentage of DENV-2 infected cells was evaluated by the staining of viral envelope (E) protein and detected by flow cytometry (a, b). The infectious viral particles production was quantified at 24 hpi using plaque assay (c). Figures represent four individual experiments. Differences were identified using a Kruskal-Wallis test with a $95\%$ confidence interval was used (***$p \leq 0.001$, **$p \leq 0.01$, *$p \leq 0.05$).
Overexpression of miR-182-5p, miR-130a-3p, and miR-155-5p significantly decreased the proportion of DENV-2-infected MDMs Figure 6(a,b), the infectious particle production (Figure 6c), and viral RNA genome in supernatant (Figure 6d), compared to scramble miRNA transfected MDMs. Although overexpression of miR-125b-5p significantly decreased the percentage of DENV-2-infected MDMs, it did not alter the production of infectious particles or the amount of RNA genome in supernatant (Figure 6). Together, our results show that inhibition of expression of miR-182-5p, miR-130a-3p, miR-125b-5p, and miR-155-5p does not affect viral replication; however, overexpression of these miRNAs significantly decreased DENV-2 replication in MDMs. Figure 6.Over-expression of miR-182-5p, miR-130a-3p, and miR-155-5p inhibit DENV-2 replication in MDMs. MDMs were transfected either with a miRNA scrambled negative control or with an anti-sense specific miRNA. 24 hours later, cells were infected with DENV-2 at an MOI of 5 and, at 24 hpi, the percentage of DENV-2-infected MDMS was evaluated by the staining of the viral envelope protein (e) and detected by flow cytometry (a, b). Viral replication was evaluated at 24 hpi by the quantification of infectious viral particles using plaque assay (c) and by the quantification of the genomic equivalent copies using RT-qPCR (d). Figures represent four individual experiments. Differences were identified using a Kruskal–Wallis test with a $95\%$ confidence interval was used (***$p \leq 0.001$, **$p \leq 0.01$, *$p \leq 0.05$).
## Discussion
MiRNAs are small non-coding RNAs that are known to regulate cellular programs including cell division, metabolism, and immune response [18]. Dysregulation of miRNA expression has been shown to impact inflammatory response (reviewed in [39,40]). A body of evidence shows that miRNAs mediate cellular responses to VitD3 [41]. In this study, we evaluated the modulation of miRNAs expression by VitD3 in MDMs during DENV-2 infection. Our data show that VitD3 differentially regulates expression of miR-182-5p, miR-130a-3p, miR-125b-5p, miR-146a-5p, and miR-155-5p, which have been associated with the inflammatory response in certain diseases [28–30] and in DENV pathogenesis [22,31]. Accordingly, inhibition of miR-182-5p, miR-130a-3p, and miR-125b-5p by antisense RNAs decreased the production of TNF-α. Due to the important role of TNF-α in DENV pathogenesis [42] and its potential to induce vascular leakage [43,44], miRNA regulation of this cytokine shed a light on the development of miRNA-based therapy.
Furthermore, we found that inhibition of miR-130a-3p expression decreased mRNA levels of TLR9 during DENV infection. This TLR has been shown to be associated with the development of inflammatory responses in certain diseases (reviewed in [45]). Also, DENV infection activates TLR9 in MoDCs leading to inflammatory response [46]. Previously, we reported that TLR9 expression is reduced in DENV-2-infected D3-MDMs [13]. These results suggest a possible link between VitD3, miR-130a-3p, and TLR9 expression. Further studies are needed to dissect the mechanism behind this link. Finally, we found an increased expression of IFN-I, OAS-1, and SOCS-1 due to inhibition of miRNA expression. Importantly, inhibition of the miRNAs did not affect DENV-2 replication, suggesting that the effect on immune response are independent of viral replication. Altogether, our results suggest that the immunomodulatory effects of VitD3 in MDMs during DENV-2 infection may be mediated by the regulation of the expression of some inflammatory-linked miRNAs.
DENV pathogenesis is mediated by uncontrolled and exacerbated inflammatory responses. Recent studies have detailed the involvement of miRNAs in the regulation of inflammation and antiviral response. For example, stimulation with LPS, Poly (I:C), CpG-ODN or IFN-β induces expression of miR-155 in murine macrophages through JNK signaling pathways [47]. Similarly, miR-125b modulates inflammatory state of macrophages by restricting the expression of B7-H4 protein, an important costimulatory molecule that suppresses T cell function [29]. Our results show that DENV-2 infection of MDMs is accompanied by increased expression of several inflammatory-response associated miRNAs, including miR-182-5p, miR-130a-3p, miR-146a-5p, miR-125b-5p, and miR-155-5p. In agreement with our results, Wang et al. demonstrated that leukotriene B4 positively regulates expression of miR-155, miR146b, and miR-125b, promoting an inflammatory state via suppression of SOCS-1 expression and increasing MyD88 expression [48]. Further, the serine/threonine kinase Akt induced by LPS stimulation in murine macrophages, upregulates the expression of miRNA let-7e, miR-155, miR-181c, and miR-125b [49]. Differential expression of inflammatory miRNAs could contribute to the development of severe dengue, since differential expression of miRNAs, including miR-6499, miR-122, miR-486, miR-383, and miR-146a, has been observed in patients with severe dengue [50–52]. Overall, miRNAs, with special emphasis on miR-155, miR146, and miR-125b, are involved in the inflammatory response of macrophages and could contribute to the exacerbated inflammation observed in DENV infection.
Innate immune response is highly regulated by the activation of transcription factor NF-κB, and its dysregulation can lead to an exacerbated/sustained inflammatory response [53]. MiRNAs have an important role in the NF-κB pathway either by regulating its activation indirectly or by being induced by NF-κB signaling [54]. Thus, inflammatory signals in hepatocytes, adipocytes, and MoDCs lead to NF-κB activation and increased expression of miR-155 and miR-146 [20,55,56]. These findings suggest that expression of these miRNAs is important for induction of inflammatory response and clearance of viral infections. Mann et al. showed a miRNA-based regulatory network in which miR-146a repressed the activation of NF-κB induced by miR-155 in mouse macrophages, representing negative feedback for the resolution of the inflammatory response [57]. In the present study, we observed increased expression of miR-146a-5p and miR-155-5p in DENV-2 infected MDMs. Surprisingly, during DENV-2 infection, the expression kinetics for these two miRNAs was the same, which may suggest that miR-146a-5p could also represent negative feedback for the inflammatory response induced by miR-155 during DENV infection. Further experiments are needed to test this hypothesis. Overall, these results suggest that miR-125b, miR-146a-5p, and miR-155 are involved in NF-κB activation and could contribute to the pathogenesis of DENV infection.
Importantly, we found that VitD3 regulated the expression of inflammatory-linked miRNAs in DENV-2-infected D3-MDMs. Since the progression to severe dengue is associated with an exacerbated inflammatory response, we propose that VitD3 regulates the innate immune during DENV infection via a miRNA-based mechanism. In other diseases with an inflammatory milieu, VitD3 has also been shown to modulate miRNAs expression. In human umbilical vein endothelial cells under treatment with serum albumin and glucose (diabetic-like environment), supplementation of VitD3 down-regulates the expression of several miRNAs [58]. Similar results have been observed in human adipocytes stimulated with TNF-α, in which VitD3 reduces the expression of miR-146a-5p, miR-150, and miR-155-5p [20]. Also, during pregnancy, there is a differential expression of miRNAs between women that show insufficient levels of circulating VitD (<25.5 ng/ml) compared to women with sufficient levels of VitD (>31.7 ng/ml) [59]. Altogether, these results show that VitD3 can regulate expression of various miRNAs involved in inflammatory disorders, including in DENV infection.
Suppressors of cytokine-signaling proteins (SOCS) are members of a family of intracellular cytokine-inducible proteins important for the regulation of inflammatory response [60]. In neutrophils treated with VitD3 and infected with Streptococcus pneumonia, there is increased expression of SOCS-1 and SOCS-3 compared to non-treated neutrophils [51]. We have previously shown that VitD3 treatment of MDMs increases expression of SOCS-1 during DENV-2 infection [13]. This result suggests that the upregulation of SOCS proteins by VitD3 could explain in part, its immunomodulatory activity. Here, we showed that inhibition of miR-182-5p and miR-155-5p, which are downregulated by VitD3, significantly increased the expression of SOCS-1. In agreement with our results, Chen et al. demonstrated that VitD3 treatment in mice restricted miR-155-5p expression, which in turn promoted increased expression of SOCS-1 and an attenuated inflammatory response to LPS [21]. Interestingly, SOCS-1 mRNA-3´UTR have target sequences for miR-155-5p, miR-572, miR-221, and miR-150 [61], which explains its downregulation when miR-155-5p is overexpressed. Of note, Chen et al. found that augmented expression of miR-150 and depressed expression of SOCS-1 was associated with severe dengue in patients infected with DENV [61]. Together, these results demonstrate the inflammatory potential of miR-155-5p and miR-150 during DENV infection by decreasing expression of SOCS-1, which can be further modulated by VitD3 treatment.
Of note, regulation of mRNA levels of TLR9, CAMP, IFN-I, OAS1, and SOCS-1 by miRNAs could not be confirmed at the protein levels, which is a limitation of our study. However, the main mode of action of miRNAs is mRNA decay [62], which would translate in lower quantities of mRNA as we detected by RT-qPCR. If mRNA levels correspond to protein levels should be further studied. For example, it could be that in our model, different levels of IFN-I and OAS1 mRNAs and protein are being produced, which may explain why miRNA inhibition did not affect DENV-2 replication even with higher levels of antiviral IFN-I and OAS1 mRNAs. Another possible explanation for DENV-2 replication levels under increased expression of IFN-I is the evasion of this antiviral system by DENV-2. It has been shown that DENV interferes with IFN-I signaling by blocking the activation of STAT1 and STAT2 through non-structural proteins NS4B and NS5 [63,64]. Given that STATs are required for IFN-I signaling, it could also explain why we did not observe changes in PKR mRNA expression under miRNA inhibition, even though high levels of IFN-α and IFN-β were seen. However, we could not answer why miRNA inhibition did not affect DENV-2 replication levels despite increased levels of IFN-I and OAS1, a question that could be addressed in the future.
The antiviral activity of miRNAs has been extensively described in various infections of plants, fungi, and mammals [65]. For example, miR-24 and miR-93 restrict replication of the vesicular stomatitis virus (VSV) [66], miR-29a inhibits human immunodeficiency virus 1 (HIV-1) [67], and miR-323, miR-491, and miR-654 block Influenza A H1N1 infection [68]. Surprisingly, we found that overexpression of miR-182-5p, miR-130a-3p, miR-155, and to a lesser extent miR-125b-5p, significantly inhibited DENV-2 replication in primary macrophages. Similar to our results, overexpression of miR-155 has been shown to limit DENV replication in vitro in Huh-7 cells and mice [69]. Further experiments revealed that miR-155 targets Bach1 resulting in inhibition of NS2B/NS3 protease activity [69]. These results highlight the key role of miR-155 during DENV infection and suggest that this miRNA could have an antiviral role during viral replication. Interestingly, Su et al. showed that miR-155 overexpression in HepG2 cells resulted in increased expression of antiviral MxA and ISG15 genes, resulting in inhibition of Hepatitis B virus (HBV) replication [70]. Our results show that DENV-2 infection upregulates expression of miR-155 suggesting that in MDMs this antiviral effect is not significant. Whether expression of miR-155 was insufficient for antiviral effect in MDMs, or whether DENV-2 actively blocks miR-155-mediated Bach1 expression requires further studies.
In addition to miR-155, we found that other miRNAs were also restricted DENV-2 replication. Although the antiviral effect of miR-182, miR-130a, or miR-125b against DENV has not been reported, other studies have shown its antiviral activity against other types of viruses. MiR-182 suppresses human cytomegalovirus by the induction of IFN-I response [71]. Over-expression of miR-130a can restrict Hepatitis C virus (HCV) replication by increasing the expression of antiviral ISG15, USP18, MxA, MX1, and OAS3 genes [72,73]. Further, miR-130a has been shown to limit HBV replication by limiting liver transcription factors PGC1α and PPARγ [74]. Similarly, overexpression of miR125b reduces replication of the flavivirus *Japanese encephalitis* virus (JEV) [75]. Similarly, miR-125b inhibits HIV-1 replication since its mRNA-3ʹUTR harbors seed sequences that are targeted by this miRNA [76,77]. Unfortunately, we could not dissect the mechanism of such DENV-2 inhibition. However, based on the information discussed above, we suggest that over-expression of miRNAs may enhance innate immunity-related gene expression, or indirectly, regulate the expression of key factors involved in DENV-2 replication, as we reported previously for miR-133a [78]. Altogether, these results show the antiviral potential of miRNAs against DENV infection.
In conclusion, this study demonstrated that VitD3 regulates the expression of various miRNAs involved in several inflammatory disorders, which could represent a novel immunomodulatory mechanism of VitD3 in addition to other previously reported effects, such as PRRs regulation, decrease of ROS production, and increased expression of SOCS-1 [12–14]. Our data suggest that the mechanisms behind this effect could be regulation of miRNAs expression since their inhibition had the same effect as VitD3 treatment.
## Disclosure statement
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
## References
1. Halstead SB.. **Dengue**. *Lancet* (2007) **370** 1644-1652. PMID: 17993365
2. **Dengue vaccine**. *Weekly Epidemiological Rec* (2016) **2016** 349-364
3. Guzman MG, Harris E. **Dengue**. *Lancet* (2014) **6736** 1-13
4. Halstead SB. **Recent advances in understanding dengue**. *F1000Res* (2019) **8** 1-12
5. Basu A, Chaturvedi UC. **Vascular endothelium: the battlefield of dengue viruses. FEMS immunology and medical microbiology**. *FEMS Immunology and Medical Microbiology* (2008) **53** 287-299. PMID: 18522648
6. Malavige GN, Ogg GS. **Pathogenesis of vascular leak in dengue virus infection**. *Immunology* (2017) **151** 261-269. PMID: 28437586
7. White JH. **Vitamin D signaling, infectious diseases, and regulation of innate immunity**. *Infect Immun* (2008) **76** 3837-3843. PMID: 18505808
8. Verma R, Jung JH, Kim JY. **1,25-Dihydroxyvitamin D3 up-regulates TLR10 while down-regulating TLR2, 4, and 5 in human monocyte THP-1**. *J Steroid Biochem Mol Biol* (2014) **141** 1-6. PMID: 24373795
9. Sadeghi K, Wessner B, Laggner U. **Vitamin D3 down-regulates monocyte TLR expression and triggers hyporesponsiveness to pathogen-associated molecular patterns**. *Eur J Immunol* (2006) **36** 361-370. PMID: 16402404
10. Jadhav NJ, Gokhale S, Seervi M. **Immunomodulatory effect of 1, 25 dihydroxy vitamin D3on the expression of RNA sensing pattern recognition receptor genes and cytokine response in dengue virus infected U937-DC-SIGN cells and THP-1 macrophages**. *Int Immunopharmacol* (2018) **62** 237-243. PMID: 30032048
11. Puerta-Guardo H, Medina F, De la Cruz Hernández SI. **The 1α,25-dihydroxy-vitamin D3 reduces dengue virus infection in human myelomonocyte (U937) and hepatic (Huh-7) cell lines and cytokine production in the infected monocytes**. *Antiviral Res* (2012) **94** 57-61. PMID: 22387385
12. Arboleda Alzate JF, Rodenhuis-Zybert IA, Hernández JC. **Human macrophages differentiated in the presence of vitamin D3restrict dengue virus infection and innate responses by downregulating mannose receptor expression**. *PLoS Negl Trop Dis* (2017) **11** 1-18
13. Castillo JA, Giraldo DM, Hernandez JC. **Regulation of innate immune responses in macrophages differentiated in the presence of vitamin D and infected with dengue virus 2**. *PLoS Negl Trop Dis* (2021) **15** e0009873. PMID: 34634046
14. Giraldo DM, Cardona A, Urcuqui-Inchima S. **High-dose of vitamin D supplement is associated with reduced susceptibility of monocyte-derived macrophages to dengue virus infection and pro-inflammatory cytokine production: an exploratory study**. *Clin Chim Acta* (2018) **478** 140-151. PMID: 29289621
15. Martínez-Moreno J, Hernandez JC, Urcuqui-Inchima S. **Effect of high doses of vitamin D supplementation on dengue virus replication, Toll-like receptor expression, and cytokine profiles on dendritic cells**. *Mol Cell Biochem* (2019) **464** 169-180. PMID: 31758375
16. Yasmin R, Williams RM, Xu M. **Nuclear import of the retinoid X receptor, the vitamin D receptor, and their mutual heterodimer**. *J Biol Chem* (2005) **280** 40152-40160. PMID: 16204233
17. Hii CS, Ferrante A. **The non-genomic actions of vitamin D**. *Nutrients* (2016) **8** 1-14
18. Bartel DP. **MicroRNA Target Recognition and Regulatory Functions**. *Cell* (2009) **136** 215-233. PMID: 19167326
19. Pedersen AW, Holmstrøm K, Jensen SS. **Phenotypic and functional markers for 1alpha,25-dihydroxyvitamin D(3)-modified regulatory dendritic cells**. *Clin Exp Immunol* (2009) **157** 48-59. PMID: 19659770
20. Karkeni E, Bonnet L, Marcotorchino J. **Vitamin D limits inflammation-linked microRNA expression in adipocytes in vitro and in vivo: a new mechanism for the regulation of inflammation by vitamin D**. *Epigenetics* (2018) **13** 156-162. PMID: 28055298
21. Chen Y, Liu W, Sun T. **1,25-Dihydroxyvitamin D promotes negative feedback regulation of TLR signaling via targeting microRNA-155-SOCS1 in macrophages**. *J Iimmunol* (2013) **190** 3687-3695
22. Arboleda JF, Fernandez GJ, Urcuqui-Inchima S. **Vitamin D-mediated attenuation of miR-155 in human macrophages infected with dengue virus: implications for the cytokine response**. *Infect Genet Evol* (2019) **69** 12-21. PMID: 30639520
23. Milani C, Katayama MLH, De Lyra EC. **Transcriptional effects of 1,25 dihydroxyvitamin D3 physiological and supra-physiological concentrations in breast cancer organotypic culture**. *BMC Cancer* (2013) **13** 1-15. PMID: 23282137
24. Smith DC, Johnson CS, Freeman CC. **A phase I trial of calcitriol (1,25-dihydroxycholecalciferol) in patients with advanced malignancy**. *Clin Cancer Res* (1999) **5** 1339-1345. PMID: 10389917
25. Shu P, Chang S, Kuo Y. **Development of group- and serotype-specific one-step SYBR green i-based real-time reverse transcription-PCR assay for dengue virus**. *Journal of Clinical Microbiology* (2003) **41** 2408-2416. PMID: 12791857
26. a SL, Schnurr D, Yagi S. **Quantitative real-time PCR for rhinovirus, and its use in determining the relationship between TCID50 and the number of viral particles**. *J Virol Methods* (2011) **171** 212-218. PMID: 21070809
27. Fernandez GJ, Castillo JA, Giraldo DM. **Vitamin D regulates the expression of immune and stress response genes in dengue virus-infected macrophages by inducing specific microRNAs**. *Microrna* (2021) **10** 240-249. PMID: 34939557
28. Duan X, Li Y, Chen S. **Vitamin D potentiates the inhibitory effect of MicroRNA-130a in hepatitis C virus replication independent of type I interferon signaling pathway**. *Mediators Inflamm* (2015) **2015** 1-6
29. Diao W, Lu L, Li S. **MicroRNA-125b-5p modulates the inflammatory state of macrophages via targeting B7-H4**. *Biochem Biophys Res Commun* (2017) **491** 912-918. PMID: 28754594
30. Song L, Li M, Li J. **TGF- b induces miR-182 to sustain NF- k B activation in glioma subsets Find the latest version : TGF- β induces miR-182 to sustain NF- κ B activation in glioma subsets**. (2012) **122** 3563-3578
31. Wu S, He L, Li Y. **MiR-146a facilitates replication of dengue virus by dampening interferon induction by targeting TRAF6**. *J Infect* (2013) **67** 329-341. PMID: 23685241
32. Giangreco AA, Vaishnav A, Wagner D. **Tumor suppressor microRNAs, miR-100 and −125b, are regulated by 1,25-dihydroxyvitamin D in primary prostate cells and in patient tissue**. *Cancer Prev Res* (2013) **6** 483-494
33. Qin C-F, Zhao H, Liu Z-Y. **Retinoic acid inducible gene-I and melanoma differentiation-associated gene 5 are induced but not essential for dengue virus induced type I interferon response**. *Mol Biol Rep* (2011) **38** 3867-3873. PMID: 21113677
34. Nasirudeen a MA, Wong HH, Thien P. **MDA5 and TLR3 synergistically play an important role in restriction of dengue virus infection**. *PLoS Negl Trop Dis* (2011) **5** 1-11
35. Tsai Y-T, Chang S-Y, Lee C-N. **Human TLR3 recognizes dengue virus and modulates viral replication in vitro**. *Cell Microbiol* (2009) **11** 604-615. PMID: 19134117
36. Wang JP, Liu P, Latz E. **Flavivirus activation of plasmacytoid dendritic cells delineates key elements of TLR7 signaling beyond endosomal recognition**. *Journal of immunology* (1950) **177** 7114-7121
37. Whyte CS, Bishop ET, Ruckerl D. **Suppressor of cytokine signaling (SOCS)1 is a key determinant of differential macrophage activation and function**. *J Leukoc Biol* (2011) **90** 845-854. PMID: 21628332
38. Gombart AF, Borregaard N, Koeffler HP. **Human cathelicidin antimicrobial peptide (CAMP) gene is a direct target of the vitamin D receptor and is strongly up-regulated in myeloid cells by 1,25-dihydroxyvitamin D 3**. *FASEB J* (2005) **19** 1067-1077. PMID: 15985530
39. Li Y, Tan J, Miao Y. **MicroRNA in extracellular vesicles regulates inflammation through macrophages under hypoxia. cell death discovery**. *Cell death Discovery* (2021) **7**. DOI: 10.1038/s41420-021-00670-2
40. Zingale VD, Gugliandolo A, Mazzon E. **Mir‐155: an important regulator of neuroinflammation**. *International Journal of Molecular Sciences. MDPI* (2022). DOI: 10.3390/ijms23010090
41. Zeljic K, Supic G, Magic Z. **New insights into vitamin Danticancer properties: focus on miRNA modulation. Molecular genetics and genomics**. *Molecular genetics and genomics* (2017) **292** 511-524. DOI: 10.1007/s00438-017-1301-9
42. Green S, Vaughn DW, Kalayanarooj S. **Early immune activation in acute dengue illness is related to development of plasma leakage and disease severity**. *The Journal of Infectious Diseases* (1999) **179** 755-762. PMID: 10068569
43. Inyoo S, Suttitheptumrong A, Pattanakitsakul SN. **Synergistic effect of TNF-α and dengue virus infection on adhesion molecule reorganization in human endothelial cells**. *Jpn J Infect Dis* (2017) **70** 186-191. PMID: 27580574
44. Suttitheptumrong A, Mahutchariyakul T, Rawarak N. **Altered moesin and actin cytoskeleton protein rearrangements affect transendothelial permeability in human endothelial cells upon dengue virus infection and tnf-α treatment**. *Viruses* (2021) **14** 13. PMID: 35062217
45. Marshak-Rothstein A. **Toll-like receptors in systemic autoimmune disease**. *Nat Rev Immunol* (2006) **6** 823-835. PMID: 17063184
46. Lai J, Wang M, Huang C. **Infection with the dengue RNA virus activates TLR9 signaling in human dendritic cells**. *EMBO Rep* (2018) **19** e46182. PMID: 29880709
47. O’Connell RM, Taganov KD, Boldin MP. **MicroRNA-155 is induced during the macrophage inflammatory response**. *Proceedings of the national academy of sciences* (2007) **104** 1604-1609. DOI: 10.1073/pnas.0610731104
48. Wang Z, Filgueiras LR, Wang S. **Leukotriene B 4 enhances the generation of proinflammatory microRNAs to promote MyD88-Dependent macrophage activation**. *J Immunol* (2021) **192** 2349-2356
49. Androulidaki A, Iliopoulos D, Arranz A. **The kinase Akt1 controls macrophage response to lipopolysaccharide by regulating microRNAs**. *Immunity* (2009) **31** 220-231. PMID: 19699171
50. Saini J, Bandyopadhyay B, Pandey AD. **High-Throughput RNA sequencing analysis of plasma samples reveals circulating microRNA signatures with biomarker potential in dengue disease progression**. *mSystems* (2020) 5. DOI: 10.1128/msystems.00724-20
51. Tambyah PA, Ching CS, Sepramaniam S. **microRNA expression in blood of dengue patients**. *Ann Clin Biochem* (2016) **53** 466-476. PMID: 26290515
52. de Oliveira LF, de Andrade AAS, Pagliari C. **Differential expression analysis and profiling of hepatic miRNA and isomiRNA in dengue hemorrhagic fever**. *Sci Rep* (2021) **11** 1-9. PMID: 33414495
53. Newton K, Dixit VM. **Signaling in innate immunity and inflammation**. *Cold Spring Harb Perspect Biol* (2012) 4. DOI: 10.1101/cshperspect.a006049
54. Gao Z, Dou Y, Chen Y. **MicroRNA roles in the NF-κB signaling pathway during viral infections**. *Biomed Res Int* (2014) **2014** 19-21
55. Zhang Y, Wei W, Cheng N. **Hepatitis C virus-induced up-regulation of microRNA-155 promotes hepatocarcinogenesis by activating Wnt signaling**. *Hepatology* (2012) **56** 1631-1640. PMID: 22610915
56. Ceppi M, Pereira AM, Dunand-Sauthier I. **MicroRNA-155 modulates the interleukin-1 signaling pathway in activated human monocyte-derived dendritic cells**. *Proceedings of the national academy of sciences* (2009) **106** 2735-2740. DOI: 10.1073/pnas.0811073106
57. Mann M, Mehta A, Zhao JL. **An NF-κB-microRNA regulatory network tunes macrophage inflammatory responses**. *Nat Commun* (2017) **8** 8. PMID: 28364116
58. Zitman-Gal T, Green J, Pasmanik-Chor M. **Vitamin D manipulates miR-181c, miR-20b and miR-15a in human umbilical vein endothelial cells exposed to a diabetic-like environment**. *Cardiovasc Diabetol* (2014) **13** 8. PMID: 24397367
59. Enquobahrie D, Williams M, Qiu C. **Global maternal early pregnacy peripheral blood mRNA and miRNA expression profiles according to plasma 25-hydroxyvitamin D concentrations**. *Journal Maternal Fetal Neonatal Medecine* (2012) **100** 130-134
60. Wilson HM. **SOCS proteins in macrophage polarization and function**. *Front Immunol* (2014) **5** 1-5. PMID: 24474949
61. Chen R, Yang KD, Lee I. **Augmented miR-150 expression associated with depressed SOCS1 expression involved in dengue haemorrhagic fever**. *J Infect* (2014) **69** 366-374. PMID: 24907421
62. oki IH, Tomari Y. **The functions of microRNAs: mRNA decay and translational repression**. *Trends Cell Biol* (2015) **25** 651-665. PMID: 26437588
63. Muñoz-Jordán JL, Laurent-Rolle M, Ashour J. **García-sastre adolfo. inhibition of alpha/beta interferon signaling by the NS4B protein of flaviviruses**. *J Virol* (2005) **79** 8004-8013. PMID: 15956546
64. Morrison J, Laurent-Rolle M, Maestre AM. **Dengue virus co-opts UBR4 to degrade STAT2 and antagonize type I interferon signaling**. *PLoS Pathog* (2013) **9** e1003265. PMID: 23555265
65. Girardi E, López P, Pfeffer S. **On the importance of host MicroRNAs during viral infection**. *Front Genet* (2018) **9** 1-17. PMID: 29387083
66. Otsuka M, Jing Q, Georgel P. **Hypersusceptibility to vesicular stomatitis virus infection in Dicer1-deficient mice is due to impaired miR24 and miR93 expression**. *Immunity* (2007) **27** 123-134. PMID: 17613256
67. Frattari G, Aagaard L, Denton PW. **The role of miR-29a in HIV-1 replication and latency**. *J Virus Erad* (2017) **3** 185-191. PMID: 29057080
68. Song L, Liu H, Gao S. **Cellular microRNAs inhibit replication of the H1N1 influenza A virus in infected cells**. *J Virol* (2010) **84** 8849-8860. PMID: 20554777
69. Su YC, Huang YF, Wu YW. **MicroRNA-155 inhibits dengue virus replication by inducing heme oxygenase-1-mediated antiviral interferon responses**. *FASEB J* (2020) **34** 7283-7294. PMID: 32277848
70. Su C, Hou Z, Zhang C. **Ectopic expression of microRNA-155 enhances innate antiviral immunity against HBV infection in human hepatoma cells**. *Virol J* (2011) **8** 1-11. PMID: 21205327
71. He X, Teng J, Cui C. **MicroRNA-182 inhibits HCMV replication through activation of type I IFN response by targeting FOXO3 in neural cells**. *Exp Cell Res* (2018) **369** 197-207. PMID: 29792850
72. Li S, Duan X, Li Y. **MicroRNA-130a inhibits HCV replication by restoring the innate immune response**. *J Viral Hepat* (2014) **21** 121-128. PMID: 24383925
73. Duan X, Liu X, Li W. **Microrna-130a downregulates HCV replication through an atg5-dependent autophagy pathway**. *Cells* (2019) **8** 338. PMID: 30974864
74. Huang JY, Chou SF, Lee JW. **MicroRNA-130a can inhibit hepatitis B virus replication via targeting PGC1α and PPARγ**. *Rna* (2015) **21** 385-400. PMID: 25595716
75. Huang CW, Tsai KN, Chen YS. **Differential mirna expression profiling reveals correlation of mir125b-5p with persistent infection of Japanese encephalitis virus**. *Int J Mol Sci* (2021) **23** 22. PMID: 35008458
76. Huang J, Wang F, Argyris E. **Cellular microRNAs contribute to HIV-1 latency in resting primary CD4 + T lymphocytes**. *Nat Med* (2007) **13** 1241-1247. PMID: 17906637
77. Mantri CK, Pandhare Dash J, Mantri JV. **Cocaine enhances HIV-1 replication in CD4+ T cells by down-regulating MiR-125b**. *PLoS ONE* (2012) 7. DOI: 10.1371/journal.pone.0051387
78. Castillo JA, Castrillón JC, Diosa-Toro M. **Complex interaction between dengue virus replication and expression of miRNA-133a**. *BMC Infect Dis* (2016) **16** 29. PMID: 26818704
|
---
title: The prevalence of subclinical hypothyroidism in a pre-diabetes population and
an analysis of related factors
authors:
- Xingyu Chang
- Yaqi Wang
- Yi Liu
- Yanyu Shen
- Jiaqing Feng
- Qianqian Liu
- Chenjun Jiang
- Jing Yu
- Xulei Tang
- Gaojing Jing
- Qianglong Niu
- Songbo Fu
journal: Annals of Medicine
year: 2023
pmcid: PMC9970244
doi: 10.1080/07853890.2023.2178668
license: CC BY 4.0
---
# The prevalence of subclinical hypothyroidism in a pre-diabetes population and an analysis of related factors
## Abstract
### Background
To investigate the prevalence and related influencing factors of subclinical hypothyroidism (SCH) in a pre-diabetes (PreDM) population.
### Patients and methods
A multi-stage stratified cluster random sampling method was used to select the adult Han population in Gansu Province for investigation. General data and related biochemical indices were recorded and SPSS software was used for statistical analyses.
### Results
This study selected 2876 patients, including 548 with SCH and 433 with PreDM. In the PreDM population, the levels of thyroid stimulating hormone (TSH), serum phosphorus, TPOAb and TgAb in the SCH group were higher than those in the euthyroid group ($P \leq 0.05$). The level of TPOAb in females of SCH group was higher than that in males ($P \leq 0.05$). The positive rates of TPOAb and TgAb in females were higher than those in males in the total population and SCH population. The prevalence of SCH in the PreDM group under 60 was significantly higher than that in the normal glucose tolerance (NGT) group ($26.02\%$ vs. $20.40\%$, χ2 = 5.150, $P \leq 0.05$). We defined SCH as a TSH level of >4.20 mIU/L. Using this criterion, the prevalence of SCH in the total population of PreDM was higher than that in the NGT population (χ2 = 8.611, $P \leq 0.05$), the prevalence of SCH in the PreDM population generally showed an upward trend. However, we performed a separate analysis considering the accepted impact of age on TSH redefining SCH as TSH >8.86 mIU/L (for individuals over age 65). However, allowing for the expected rise in TSH levels in individuals over age 65, the prevalence of SCH in the elderly over 65 years of age decreased significantly (NGT population from $27.48\%$ to $9.16\%$, PreDM population from $34.18\%$ to $6.33\%$, $P \leq 0.05$). Logistic regression analysis showed that the risk factors for SCH in the PreDM population were female gender, fasting plasma glucose and TSH (all $P \leq 0.05$). Risk factors for SCH in the impaired fasting glucose (IFG) population were female gender, OGTT 2 h, TSH and TPOAb (all $P \leq 0.05$).
### Conclusion
The prevalence of SCH in the PreDM population not considering the known physiological increase in age related TSH was relatively high and was significant in female and the IFG population. However, the effect of age on these findings needs to attract more attention.
## KEY MESSAGES
The prevalence of subclinical hypothyroidism (SCH) in the pre-diabetic population was analysed by cross-sectional survey. There is a great deviation in the diagnosis of SCH in the elderly with physiologically increased thyroid stimulating hormone, which needs to be redefined.
## Introduction
Pre-diabetes (PreDM) is a transitional state between normal glucose tolerance (NGT) and diabetes [1]. In recent years, with the rapid economic development and changes in dietary structure, the number of patients with PreDM has continued to increase. It is estimated that the number of patients worldwide will reach 470 million in 2030, and the prevalence rate of PreDM in China could reach as high as $35.2\%$ [2,3]. Subclinical hypothyroidism (SCH) refers to the increase in serum thyroid stimulating hormone (TSH) levels while thyroid hormone levels remain normal. A study based on 31 provinces found that the prevalence of SCH in China reached $12.93\%$ [4]. This condition is often ignored until it develops into clinical hypothyroidism and increases the risk of cardiovascular and other diseases [5]. Previous studies have found that abnormal glucose metabolism closely relates to the occurrence of SCH, but its mechanism has not yet been fully clarified, and relatively few studies have included PreDM populations [6]. Therefore, this study selected adult residents in Gansu Province to analyse the prevalence and characteristics of SCH in a PreDM population and explore the influence of related indicators and their relationship with TSH. Our findings provide a reference for the diagnosis of SCH in the Chinese population.
## Selection method
A multi-stage stratified cluster random sampling method was used in Gansu Province. From 4 September 2016 to 1 February 2017, adult Han residents living in Lanzhou, Longnan, Dingxi, Baiyin and Linxia for more than 5 years were randomly selected. Following the procedure of ‘registration first and mobilization later’, the registered population at each site should be more than twice the size of the sample to meet the number of samples and the requirements of gender and age structure, and to prevent voluntary entry into the investigation queue by persons outside the site. A total of 2876 subjects were included, including 1463 males and 1413 females. The age range was 18–87 years, and the average age was 42.87 ± 14.99 years. Exclusion criteria: 1, severe heart, liver, renal insufficiency diseases, severe anaemia, or malignant tumours; 2, pregnant women or lactating women; 3, have taken drugs that interfere with blood lipids, blood pressure and thyroid function in the past 3 months. Such drugs included glucocorticoids, metoclopramide and propranolol.
## Clinical data
Under the guidance of professionals, the participants filled out the survey registration form and accurately recorded gender, age, height, weight, body mass index (BMI), waist circumference, heart rate, systolic blood pressure (SBP), diastolic blood pressure (DBP), family history of diabetes, history of hypertension, history of thyroid disease and other general conditions.
We obtained ethical approval and a letter of cooperation from the Medical Ethics Research Committee of the First Affiliated Hospital of China Medical University (AF-SOP-07-1.0-01), which was conducted in accordance with the Declaration of Helsinki [7].
## Biochemical indices
The blood lipid-related indexes: total cholesterol (TC, mmol/L), triglyceride (TG, mmol/L), high-density lipoprotein cholesterol (HDL-C, mmol/L) and low-density lipoprotein cholesterol (LDL-C, mmol/L); uric acid (UA, mmol/L), aspartate aminotransferase, alanine aminotransferase, blood calcium and serum phosphate (mmol/L) using test kit and biochemical analyser (BS-180, Mindray company). Fasting plasma glucose (FPG, mmol/L) and 2 h blood glucose after OGTT load (2 h PG, mmol/L) were determined by glucose oxidase method using test kit and biochemical analyzer (BS-180, Mindray Company). Glycosylated hemoglobin (HbA1c, %) using BioRad reagent, measured by VARIANT II (BioRad Company). Thyroid-stimulating hormone (TSH, mIU/L), Free thyroxine (FT4, pmol/L) anti-thyroid peroxidase antibody (TPOAb, 0–34IU/L) and anti-thyroid globulin antibody (TgAb, IU/L) using ICMA (Roche Company). Urinary iodine (UIC, g/L) was determined by inductively coupled plasma mass spectrometer (7700×, Agilent, USA).
## PreDM diagnostic criteria and grouping
According to the Guidelines for the Prevention and Treatment of Type 2 Diabetes in China (WHO1999) [1] the diagnostic criteria were as follows:NGT: FPG < 6.1 mmol/L and 2 h PG < 7.8 mmol/L.PreDM: [1] impaired fasting glucose (IFG): 6.1 mmol/≤FPG < 7.0 mmol/L and 2 h PG < 7.8 mmol/L. [2] impaired glucose tolerance (IGT): 7.8 mmol/L ≤ 2 h PG < 11.1 mmol/L and FPG < 6.1 mmol/L. [3] IGT combined with impaired fasting glucose (IFG + IGT): 6.1 mmol/L ≤ FPG < 7.0 mmol/L and 7.8 mmol/L ≤ 2h PG < 11.1 mmol/L. 3 Diabetes: FPG ≥ 7.0 mmol/L or 2 h PG ≥ 11.1 mmol/L.
## Subclinical hypothyroidism
In reference to the Guidelines for the Diagnosis and Treatment of Adult Hypothyroidism and An Age-Specific Serum Thyrotropin Reference Range for the Diagnosis of Thyroid Diseases in Older Adults: A Cross-Sectional Survey in China [8,9], the criteria are as follows: normal free thyroxine (FT4: 9.00–22.00 pmol/L) and free triiodothyronine (FT3: 3.1–6.8 pmol/L) levels. We defined SCH as a TSH level of >4.20 mIU/L. However, we performed a separate analysis considering the impact of age on TSH defining SCH as TSH >8.86 mIU/L (age > 65) statistical methods.
## Statistical methods
SPSS software (version 25.0) was used for statistical analyses. Normal distribution measurement data are expressed as (x ± s). Two independent sample t-tests were used for comparisons between the two groups. Count data were described by frequency. The χ2 test was used to analyse the differences in prevalence between the two groups. A logistic regression analysis model was used to analyse the possible risk factors for PreDM and its different subtypes, with a test level of α = 0.05. Non-normal distribution data were expressed as median (median, M), 25th and 75th percentiles (P25 and P75, respectively). The Mann-Whitney U test was used for comparisons between the two groups. All of the comparison results were statistically significant ($P \leq 0.05$).
## Baseline data distribution of the survey population
Baseline data of the 2876 subjects, including region, educational level, occupation and annual household income, are shown in Table 1.
**Table 1.**
| Characteristics | Number of cases | Composition ratio |
| --- | --- | --- |
| Area | | |
| Urban | 1597.0 | 55.53% |
| Rural | 1279.0 | 44.47% |
| Education | | |
| Illiteracy | 346.0 | 12.03% |
| Primary school | 282.0 | 9.81% |
| Junior high school | 455.0 | 15.82% |
| Senior high school/technical secondary school | 468.0 | 16.27% |
| Undergraduate/junior college | 1252.0 | 43.53% |
| Postgraduate | 73.0 | 2.54% |
| Profession | | |
| Worker | 790.0 | 27.47% |
| Farmer | 1079.0 | 37.52% |
| Staff | 664.0 | 23.09% |
| Housework | 53.0 | 1.84% |
| Student | 97.0 | 3.37% |
| Other | 193.0 | 6.71% |
| Annual household income (1000 yuan) | | |
| ≤5 | 62.0 | 2.16% |
| 5–10 | 178.0 | 6.19% |
| 10–30 | 627.0 | 21.80% |
| 30–50 | 570.0 | 19.82% |
| 50–100 | 935.0 | 32.51% |
| >100 | 504.0 | 17.52% |
## Comparison of general data between SCH and euthyroid groups in the PreDM population
There are 433 subjects with PreDM in 2876 and the 426 subjects with PreDM were divided into euthyroid (excluding hyperthyroidism, subclinical hyperthyroidism, hypothyroidism and SCH) and SCH groups. Analysis of the general data of 426 patients with PreDM revealed that the levels of TSH, Serum phosphate, TPOAb and TgAb in the SCH group were higher than those in the euthyroid group (all $P \leq 0.05$). The levels of TSH, TPOAb and TgAb in females were higher than those in males in the euthyroid group, and the level of TPOAb in females was higher than that in males in PreDM group (all $P \leq 0.05$). The corresponding results are shown in Tables 2 and 3.
## Comparison of the prevalence of SCH in PreDM and its subtype population
This study included 2659 participants in the following cohorts: NGT (2226 individuals) and PreDM (433 individuals). This study also included 548 patients with SCH (231 males and 317 females). The prevalence of SCH in the PreDM population was $22.63\%$ ($\frac{98}{433}$). Furthermore, the prevalence of SCH in females was higher than that in males in PreDM and IFG populations ($28.92\%$ vs. $17.03\%$, $46.51\%$ vs. $15.91\%$, all $P \leq 0.05$). The prevalence of SCH in males compared with that in females were statistically significant ($17.20\%$ vs. $24.09\%$, P*<0.001, χ2 =19.273), and as shown in Table 4.
**Table 4.**
| Unnamed: 0 | Male | Male.1 | Female | Female.1 | P* | χ 2 | Total | Total.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | n | n-SCH (%) | n | n-SCH (%) | P* | χ 2 | n (%) | n-SCH (%) |
| PreDM | 229 | 39 (17.03) | 204 | 59 (28.92) | 0.002 | 9.639 | 433 | 98 (22.63) |
| IFG | 44 | 7 (15.91) | 43 | 20 (46.51) | 0.003 | 9.124 | 87 | 27 (31.30) |
| IGT | 150 | 26 (17.33) | 137 | 31 (22.63) | 0.188 | 1.733 | 287 | 57 (19.86) |
| IFG + IGT | 35 | 6 (17.14) | 24 | 8 (33.33) | 0.151 | 2.062 | 59 | 14 (23.73) |
| NGT | 1114 | 192 (17.24) | 1112 | 258 (23.20) | <0.001 | 13.533 | 2226 | 450 (20.22) |
| P # | 0.890 | 0.890 | 0.076 | 0.076 | – | – | 0.276 | 0.276 |
| χ 2 | 0.019 | 0.019 | 3.158 | 3.158 | – | – | 1.188 | 1.188 |
| Total | 1343 | 231(17.20) | 1316 | 317(24.09) | <0.001 | 19.273 | 2659 | 548 (20.61) |
## Distribution of SCH prevalence in different age groups based on differences in glucose metabolism
Totala represents the consideration of the physiological increase of TSH in the elderly, the diagnostic criteria of SCH for people under 65 years old (TSH > 4.20 mIU/L, FT3 and FT4 are normal), and different diagnostic criteria for SCH for people over 65 years old (TSH > 8.86 mIU/L, FT3, FT4 normal) the prevalence of SCH. Totalb represents the prevalence of SCH based on the same diagnostic criteria for SCH (TSH > 4.20 mIU/L, normal FT3 and FT4) for all ages regardless of the effect of age on TSH. [ 1] The prevalence of SCH in the PreDM population under the age of 60 was higher than that in the NGT population ($26.02\%$ vs. $20.40\%$, χ2 =5.150, $P \leq 0.05$). [ 2] When considering the effect of age on TSH, there was no difference in the prevalence of SCH in the total PreDM population compared with the NGT population; when not considering the effect of age on TSH, the prevalence of SCH in the total PreDM population was higher than that in the NGT population ($27.71\%$ vs. $21.29\%$, χ2 = 8.611, $P \leq 0.05$). [ 3] Regardless of whether the effect of age on TSH was considered, the prevalence of SCH in the IFG population was higher than that in the NGT population (all $P \leq 0.05$). [ 4] Among the three PreDM subtypes, the prevalence of SCH in the IFG population was higher ($31.03\%$, $39.51\%$). [ 5] Age-stratified analysis showed that the prevalence of SCH in the PreDM population showed an overall upward trend with an increase age (<60). [ 6] In the ≥61-year-old population, the prevalence of SCH in both NGT and PreDM populations was significantly higher ($17.95\%$a vs. $30.26\%$b, χ2=8.047; $13.16\%$a vs. $32.45\%$b, χ2=12.058, all $P \leq 0.05$). In the population ≥65 years old, the prevalence of SCH in both NGT and PreDM populations was significantly higher compared with considering the effect of age on TSH without considering the effect of age on TSH ($9.16\%$a vs. $27.48\%$b, χ2=14.692; $6.33\%$a vs. $34.18\%$b, χ2=18.966, all $P \leq 0.05$), as shown in Table 5 and Figure 1.
**Figure 1.:** *Distribution of SCH prevalence in different age groups under different glucose metabolism.* TABLE_PLACEHOLDER:Table 5.
## Thyroid antibodies in different populations
In the total population and SCH population, the positive rate of thyroid-associated antibody in females was higher than that of males (all $P \leq 0.05$). In the PreDM population, the positive rate of both TPOAb (+)and TgAb (+) in females was higher than that of males ($P \leq 0.05$). There was no significant difference in the positive rate of thyroid-associated antibody between male and female in the SCH combined with PreDM, as shown in Table 6 and Figure 2.
**Figure 2.:** *Thyroid antibodies in different populations.* TABLE_PLACEHOLDER:Table 6.
## Logistic regression analysis of risk factors for SCH in PreDM and IFG populations
In the PreDM and IFG populations, with or without SCH as the dependent variable, the independent variables were screened by single-factor analysis, and the independent variables with statistical significance were further evaluated in a multi-variate analysis. The results showed that the risk factors for SCH in the PreDM population were female gender, FPG, TSH (all $P \leq 0.05$). The risk factors for SCH in the IFG population were female gender, OGTT 2 h, TSH, TPOAb (all $P \leq 0.05$), as shown in Table 7.
**Table 7.**
| Unnamed: 0 | PreDM | PreDM.1 | PreDM.2 | IFG | IFG.1 | IFG.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | P | OR | 95% Cl | P | OR | 95% Cl |
| Gender | 0.024 | 0.297 | 0.104–0.851 | 0.018 | 0.238 | 0.072–0.784 |
| FPG | 0.010 | 1.667 | 1.132–2.454 | 0.684 | 0.620 | 0.062–6.202 |
| OGTT 2 h | 0.217 | 0.891 | 0.742–1.070 | 0.043 | 1.894 | 1.020–3.514 |
| TSH | <0.001 | 5.012 | 3.456–7.268 | 0.003 | 4.325 | 2.390–6.515 |
| TPOAb | 0.465 | 1.001 | 0.998–1.003 | 0.032 | 1.016 | 1.001–1.030 |
## Discussion
Abnormal glucose metabolism and thyroid diseases are common factors that affect the health of Chinese residents. In the context of rapid social and economic development and population ageing, their prevention and treatment should be paid great attention [10]. SCH is closely related to PreDM, and the incidence rate of diabetes complicated with thyroid dysfunction is $12.5\%$–$51.6\%$, with SCH being the most common [11].
The prevalence of SCH in the PreDM population under the age of 60 in Gansu was $26.02\%$, which was significantly higher than that in the NGT population ($20.40\%$). However, the prevalence of SCH in the whole age group, considering that TSH levels increase physiologically with age (the SCH diagnostic criteria for people over 65-years-old consider TSH > 8.86 mIU/L alone), the prevalence of SCH in the total PreDM population is no higher than that in the NGT population. However, if the effect of age on TSH was not considered (the diagnostic criteria for SCH in all populations were TSH > 4.20 mIU/L), the prevalence of SCH in the total PreDM population was significantly higher than that in the NGT population ($27.71\%$ vs. $21.29\%$), it can be seen that there is a large floating difference in the prevalence of SCH in the elderly population. The diagnostic criteria of SCH for the elderly are still controversial. The review of Professor Biondi et al. showed that the serum TSH level of elderly patients may exceed the upper limit of the traditional reference range of 4–5 mU/L, which may lead to an overestimation of SCH in this age group. True prevalence of SCH in the above population [12]. This view is consistent with our study, which found that the prevalence of SCH was higher in people aged ≥65 years (NGT: $27.48\%$, PreDM: $34.18\%$) when the effect of age on TSH was not considered; considering the physiological increase in TSH caused by age, the prevalence of SCH in people aged ≥65 years was significantly lower (NGT: $9.16\%$, PreDM $6.33\%$). Some studies suggest that mildly elevated serum TSH in the elderly is not associated with increased morbidity and mortality [13], suggesting that we should update the diagnostic criteria of TSH according to the reference range of TSH in the elderly population in this region. When calculating the prevalence of SCH in the elderly, it is necessary to reconsider the threshold of TSH to avoid misdiagnosis of SCH in the population, so that the calculation of the prevalence of SCH is more reasonable, and it is helpful for the diagnosis and treatment of clinicians. However, in either case, the prevalence of SCH in Gansu PreDM population was much higher than that in Egypt ($16.30\%$) [14] and Beijing ($9.72\%$) [15], which may be related to many factors such as geographical environment and living habits (iodine consumption) in Gansu. At the same time, we found that among the three subtypes of PreDM, the prevalence of IFG population was slightly higher, and regardless of whether the effect of age on TSH was considered, the prevalence was higher than that of SCH in the NGT population.
In PreDM population, we found that the levels of TSH, TPOAb and TgAb in SCH group were higher than those in the euthyroid group, which may be because people with abnormal glucose metabolism are more prone to autoimmune thyroiditis (AIT) and thyroid cell destruction. There is increasing evidence that the imbalance of thyroid hormones and antibodies is related to the pathogenesis of type 1 diabetes [16]. Whether it is related to type 2 diabetes needs further study. At the same time, the leptin levels in patients with abnormal glucose metabolism is higher, which may affect the hypothalamus–pituitary–thyroid axis in vitro and in vivo through the Janus activated kinase (JAK)-2/signal transduction and transcription activation (STAT) 3 pathway, thereby stimulating the synthesis of TSH and affecting thyroid function [10]. Further gender-stratified analyses of the PreDM population found that the prevalence of SCH in females was always higher than that in males, in agreement with the findings of El-Eshmawy et al. [ 14] who also observed that SCH in PreDM was more common in females. The logistic regression analysis model showed that females with PreDM have a higher risk of developing SCH, which suggests that females are at greater risk for SCH and need more clinical attention.
AIT is the most common human organ-specific autoimmune disease, Hashimoto’s thyroiditis accounts for the vast majority. Its incidence is ∼$10\%$ in the general population, with a male to female ratio of 1:10. TgAb and TPOAb are the serum markers of AIT [17]. Our research found that, in the gender comparison of thyroid antibody positive rate, the positive rate of thyroid-related antibody in females was higher than that in males in the general population and in the SCH population, that is, the prevalence of AIT in females was higher. Research found that patients with abnormal glucose metabolism had higher thyroid autoantibody (TPOAb and TgAb) positive rate compared with the control group [18]. Most studies believe that the prevalence of AIT in females is high, the high positive rate of thyroid-related antibodies lead to an increase in the prevalence of SCH compared with males. However, we found that there was no difference in the positive rate of thyroid antibodies between females and males in the population of PreDM complicated with SCH. We speculate that the reason why the prevalence of SCH in the PreDM population is higher than that in the NGT population may be related to dysglycemia and diabetes-related AIT disease [19]. Abnormal glucose metabolism may interfere with thyroid metabolism by disturbing the level of thyroid hormone in plasma [20]. In addition, it may also be due to the small sample size of this population and certain limitations in this study. Our analysis is cross sectional and not longitudinal. And self-limited thyroiditis could account for the finding of elevated TSH in a subfraction of the cases in our report.
It is necessary to comprehensively pay attention to the relationship between glucose metabolism and thyroid hormone levels. However, our research still has some shortcomings, for example, the accuracy of the measurement of TSH level at a single time point to reflect the real situation of thyroid function level is a question worth considering. However, in view of the actual situation and ethical requirements, we have not been able to measure it multiple times, which is what we need to pay attention to next. Second, with the increase of age, the prevalence of glucose metabolism and thyroid diseases will increase. However, due to the interaction between the two diseases, we cannot determine the causal relationship. If we consider that TSH level increases physiologically with age, then the conclusion that TSH of people with abnormal glucose metabolism is higher than that of people with NGT needs further research.
## Conclusion
In conclusion, the prevalence of SCH is high in the PreDM population in Gansu Province, and further subdivision of SCH severity may provide the next research direction for studying the relationship between SCH and PreDM. Improvement of blood glucose levels in PreDM patients, early SCH screening and corresponding interventions may have a positive effect on reducing the prevalence of SCH in PreDM patients.
## Ethical approval
We obtained ethical approval and a letter of cooperation from the Medical Ethics Research Committee of the First Affiliated Hospital of China Medical University (AF-SOP-07-1.0-01), and our study was conducted in accordance with the Declaration of Helsinki. All of the study participants provided written informed consent and were informed of the confidentiality, purpose, and importance of their information.
## Author contributions
Songbo Fu: contributed to conception, design, data collection and analysis and critically revised the manuscript.
Xingyu Chang and Yaqi Wang: contributed to the conception, design, data processing, statistical analysis graphics rendering and drafting the manuscript.
Yi Liu: contributed to the conception, design, revised the manuscript and drawing statistical chart.
Yanyu Shen and Jiaqing Feng: contributed to the design, data processing, statistical analysis and drafting the manuscript.
Chenjun Jiang: contributed to the conception, paper writing, article language modification.
Qianqian Liu and Jing Yu: contributed to the data processing, statistical analysis and Graphics rendering.
Xulei Tang, Gaojing Jing and Qianglong Niu: contributed to participation in data collection and critically revised the manuscript.
All authors agree to be accountable for all aspects of the work.
## Disclosure statement
No potential conflict of interest was reported by the author(s).
## Data availability statement
The data that support the findings of this study are available from the corresponding author.
## References
1. Jia W, Weng J, Zhu D. **Standards of medical care for type 2 diabetes in China 2019**. *Diabetes Metab Res Rev* (2019) **35** e3158. PMID: 30908791
2. Tabák AG, Herder C, Rathmann W. **Prediabetes: a high-risk state for diabetes development**. *Lancet* (2012) **379** 2279-2290. PMID: 22683128
3. Li Y, Teng D, Shi X. **Prevalence of diabetes recorded in mainland China using 2018 diagnostic criteria from the American diabetes association: national cross sectional study**. *BMJ* (2020) **369** m997. PMID: 32345662
4. Li Y, Teng D, Ba J. **Efficacy and safety of Long-Term universal salt iodization on thyroid disorders: epidemiological evidence from 31 provinces of mainland China**. *Thyroid* (2020) **30** 568-579. PMID: 32075540
5. Krysiak R, Marek B, Okopień B.. **Subkliniczna niedoczynność tarczycy [subclinical hypothyroidism]**. *Wiadomosci Lekarskie (Warsaw, Poland: 1960)* (2008) **61** 139-145. PMID: 18939365
6. Gierach M, Gierach J, Junik R.. **Insulin resistance and thyroid disorders**. *Endokrynol Pol* (2014) **65** 70-76. PMID: 24549605
7. Liu Q, Chang X, Wang Y. **Hypertension and subclinical hypothyroidism: a cross-sectional survey based on population in Gansu Province**. *Res Square* (2022)
8. **Guidelines for the diagnosis and treatment of adult hypothyroidism[J]**. *Chin J Endocrinol Metab* (2017) **33** 167-180
9. Zhai X, Zhang L, Chen L. **An Age-Specific serum thyrotropin reference range for the diagnosis of thyroid diseases in older adults: a Cross-Sectional survey in China**. *Thyroid* (2018) **28** 1571-1579. PMID: 30351201
10. Biondi B, Kahaly GJ, Robertson RP.. **Thyroid dysfunction and diabetes mellitus: two closely associated disorders**. *Endocr Rev* (2019) **40** 789-824. PMID: 30649221
11. Palma CC, Pavesi M, Nogueira VG. **Prevalence of thyroid dysfunction in patients with diabetes mellitus**. *Diabetol Metab Syndr* (2013) **5** 58. PMID: 24499529
12. Biondi B, Cappola AR, Cooper DS.. **Subclinical hypothyroidism: a review**. *JAMA* (2019) **322** 153-160. PMID: 31287527
13. Gussekloo J, van Exel E, de Craen AJ. **Thyroid status, disability and cognitive function, and survival in old age**. *JAMA* (2004) **292** 2591-2599. PMID: 15572717
14. El-Eshmawy MM, El-Hafez A, El Shabrawy HA. **Subclinical hypothyroidism is independently associated with microalbuminuria in a cohort of prediabetic egyptian adults**. *Diabetes Metab J* (2013) **37** 450-457. PMID: 24404516
15. Huang X, Zhang X, Zhou X. **Prevalence of thyroid dysfunction in a Chinese population with different glucose intolerance status: a Community-Based Cross-Sectional study**. *Diabetes Metab Syndr Obes Targets Therapy* (2020) **13** 4361-4368
16. Luo J, Wang X, Yuan L. **Association of thyroid disorders with gestational diabetes mellitus: a meta-analysis**. *Endocrine* (2021) **73** 550-560. PMID: 33983493
17. Minnetti M, Sada V, Feola T. **Selenium supplementation in pregnant women with autoimmune thyroiditis: a practical approach**. *Nutrients* (2022) **14** 2234. PMID: 35684035
18. Okten A, Akcay S, Cakir M. **Iodine status, thyroid function, thyroid volume and thyroid autoimmunity in patients with type 1 diabetes mellitus in an iodine-replete area**. *Diabetes Metab* (2006) **32** 323-329. PMID: 16977259
19. Pilhatsch M, Schlagenhauf F, Silverman D. **Antibodies in autoimmune thyroiditis affect glucose metabolism of anterior cingulate**. *Brain Behav Immun* (2014) **37** 73-77. PMID: 24365060
20. Garmendia Madariaga A, Santos Palacios S, Guillén-Grima F. **The incidence and prevalence of thyroid dysfunction in Europe: a meta-analysis**. *J Clin Endocrinol Metab* (2014) **99** 923-931. PMID: 24423323
|
---
title: PI3K/AKT/SERBP-1 pathway regulates Alisma orientalis beverage treatment of
atherosclerosis in APOE−/− high-fat diet mice
authors:
- Ruiyi Liu
- Yan Sun
- Dong Di
- Xiyuan Zhang
- Boran Zhu
- Haoxin Wu
journal: Pharmaceutical Biology
year: 2023
pmcid: PMC9970249
doi: 10.1080/13880209.2023.2168020
license: CC BY 4.0
---
# PI3K/AKT/SERBP-1 pathway regulates Alisma orientalis beverage treatment of atherosclerosis in APOE−/− high-fat diet mice
## Abstract
### Context
Previously, we found *Alisma orientalis* beverage (AOB), a classic traditional Chinese medicine (TCM) formulation, had the potential effect of treating atherosclerosis (AS). The underlying mechanism was still unclear.
### Objective
As an extention of our previous work, to investigate the underlying mechanism of action of AOB in the treatment for AS.
### Materials and methods
Network pharmacology was conducted using SwissTargetPrediction, GeneCards, DrugBank, Metascape, etc., to construct component-target-pathway networks. In vivo, AS models were induced by a high-fat diet (HFD) for 8 consecutive weeks in APOE−/− mice. After the administration of AOB (3.8 g/kg, i.g.) for 8 weeks, we assessed the aortic plaque, four indicators of blood lipids, and expression of the PI3K/AKT/SREBP-1 pathway in liver.
### Results
Network pharmacology showed that PI3K/AKT/SREBP-1 played a role in AOB’s treatment for AS (PI3K: degree = 18; AKT: degree = 17). Moreover, we found that the arterial plaque area and four indicators of blood lipids were all significantly reversed by AOB treatment in APOE−/− mice fed with HFD (plaque area reduced by about $37.75\%$). In addition, phosphorylated expression of PI3K/AKT and expression of SREBP-1 were obviously increased in APOE−/− mice fed with HFD, which were all improved by AOB (PI3K: $51.6\%$; AKT: $23.6\%$; SREBP-1: $40.0\%$).
### Conclusions
AOB had therapeutic effects for AS by improving blood lipids and inhibition of the PI3K/AKT/SERBP-1 pathway in the liver. This study provides new ideas for the treatment of AS, as well as new evidence for the clinical application of AOB.
## Introduction
The latest data from the Global Burden of Disease (GBD) showed that approximately 18.6 million people worldwide died of cardiovascular disease in 2019, which has surpassed infectious diseases as the leading cause of death and disability worldwide (Roth et al. 2020). Atherosclerosis (AS) is the main factor leading to the global epidemic of cardiovascular and cerebrovascular diseases, which is mainly characterized by lipid deposition and chronic inflammation in the arterial wall (Roth et al. 2020). Atherosclerosis (AS) is characterized by fibrofatty lesions formed on the inner wall of arteries and is the primary pathological basis of cardiovascular and cerebrovascular diseases (Kobiyama and Ley 2018). Increasing evidence has indicated that hypercholesterolemia-induced vascular inflammation and cholesterol deposition together constitute a risk factor for AS (Koeth et al. 2019). Statins, lipid-lowering drugs such as atorvastatin and rosuvastatin, are the first-line drugs for the treatment of AS in modern medicine and they have significant clinical efficacy in lowering blood lipids, but they can also cause adverse reactions such as liver damage and rhabdomyolysis (Soppert et al. 2020; Aryal et al. 2021). The latest study showed that a serine protease, PCSK9, actively targets LDL-R and causes its excessive accumulation, while PCKS9 inhibitors significantly reduce LDL-C levels and reverse plaque-like changes (Solanki et al. 2018). However, the high cost of the compound and lack of long-term safety and efficacy data limit its use in patients. Therefore, there is an urgent need to find drugs with safe effects and better efficacy.
Hypercholesterolemia is recognized as the main factor leading to AS; reducing blood cholesterol levels is an important way to prevent the development of AS (Francis 2010). Various studies suggested that lipid metabolism mechanisms play a key role in the pathophysiology of AS and that elevated LDL cholesterol leads to AS independent of inflammation, whereas residual cholesterol can drive the inflammatory component of AS (Geovanini and Libby 2018). However, this evidence was not capable of solving the root cause of treating AS. Alisma orientalis beverage (AOB) was first recorded in ‘Huangdi Neijing’, an ancient Chinese medical book, consisted of three herbs including *Alismatis rhizoma* (Sam.) Juzep. ( Alismataceae) (Zexie), *Atractylodis macrocephalae* rhizoma Koidz. ( Asteraceae) (Baizhu), and *Pyrolae calliantha* H. Andres (Pyrolaceae) (Luxiancao) based on the Chinese Pharmacopeia (2020 Edition). A previous study found that AOB can effectively inhibit the progression of atherosclerosis and improvement of blood lipid levels, and its mechanism of mitigating atherosclerosis may be related to gut microbiota and its metabolite (Zhu, Zhai, et al. 2020). However, how AOB influenced blood lipid levels was not identified. Due to the complex components of this formula, it is difficult to explore multiple targets in traditional Chinese medicine (TCM) formulation. Therefore, the underlying mechanism of AOB’s therapeutic actions have not been fully elucidated.
TCM has characteristics of multi-component, multi-target and integrity. Network pharmacology is based on theories of systems biology, genomics, proteomics and other disciplines, using high-throughput omics data analysis, computer simulation and network database retrieval (Hopkins 2008). The technology reveals the network relationship of drug-gene-target-disease interactions, predicts the mechanism of action of drugs through network relationships, evaluates drug efficacy, adverse reactions, etc., explores essential attributes of TCM by referring to the research ideas of network pharmacology, and has achieved good preliminary results in revealing the comprehensive overall effect of multiple pathways, multiple targets and multiple components of TCM (Zeng et al. 2019; Zhang et al. 2019; Zhu, Cai, et al. 2020). This study adopted network pharmacological results to pre-clinical experiments, starting from the material basis of AOB, analyzing and exploring the mechanism of action of AOB in the treatment for AS, and at the same time providing a certain theoretical basis for clinical application and follow-up research.
## Collection of chemical components for AOB and screening of active compounds
We used Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP, https://old.tcmsp-e.com/tcmsp.php) to screen the active ingredients of each herb (Ru et al. 2014). We then identified compounds with oral bioavailability (OB) ≥ $30\%$ (Xu, Zhang, et al. 2012) and drug-likeness (DL) ≥ 0.18 (Jia et al. 2020) in AOB as compounds with pharmacological activity, based on absorption, distribution, metabolism and excretion (ADME) characteristics of the drugs in the body. After the preliminary screening of compounds, the PubChem database (https://pubchem.ncbi.nlm.nih.gov) was used to confirm their molecular structure and name, to improve the credibility of screening results. The identified molecules were entered into the SwissTargetPrediction website (swisstargetprediction.ch) to find the protein targets of the active compounds, and related targets were added based on published literature reports. Then, the screened protein targets were unified in the Uniprot protein database (https://www.uniprot.org) for specification and protein-gene docking for further prediction and analysis.
## Prediction of potential targets of AOB for treatment
GeneCards is a searchable comprehensive database that automatically integrates gene-centric data from approximately 150 web sources, including genomics, transcriptomics, proteomics, genetics, clinical and functional information (Rebhan et al. 1997). With ‘Atherosclerosis’ as the keyword, relevant gene target information was searched in the GeneCards database (https://www.genecards.org) (Rebhan et al. 1997), and potential genes were supplemented using the TTD database (http://db.idrblab.net/ttd/) (Hamosh et al. 2005). When the number of targets is too large, the Score value in the Genecards database can be used for screening. The larger the score value, the closer the relationship between the target and the disease. The median of the Score value is used as the screening value. When there is too much data, multiple screening can be performed to obtain AS-related targets. The intersection of drug component-related targets and AS targets was operated by Venny2.1 (https://bioinfogp.cnb.csic.es/tools/venny/).
## Construction of an active compound-disease-target network
Upload the intersection of targets to the STRING11.0 database (https://string-db.org) (Szklarczyk et al. 2017) to construct a protein-protein interaction (PPI) network model, set the biological species to ‘Homo sapiens’, and set ‘highest confidence’ > 0.9. The PPI network was obtained by screening, and the PPI network was further clustered by Cytoscape_v3.8.2 (Shannon et al. 2003) to obtain potential protein functional modules. The core targets are selected according to the comprehensive ranking of node connectivity (degree), node closeness (closeness) and node betweenness (betweenness).
## Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis
The core target genes obtained in the above steps were uploaded to the Metascape platform (http://metascape.org) (Zhou et al. 2019), and a threshold of $p \leq 0.01$ was set. The main biological processes and metabolic pathways were analyzed, including the KEGG pathway, GO biological process, Cell composition, and molecular function enrichment analysis. Then the data were saved and the results were visualized. Finally, a visualization network of ‘prescriptions – traditional Chinese medicine – chemical components – core targets – key pathways’ was built.
## The preparation process of AOB
Alismatis rhizoma, *Atractylodis macrocephalae* rhizoma, and *Pyrolae herba* were purchased from Tong Ren Tang, Beijing. Alismatis rhizoma (100 ± 5 g), *Atractylodis macrocephalae* rhizoma (100 ± 5 g) and *Pyrolae herba* (50 ± 2 g) were placed into 5 L water to soak for 1 h. After soaking, we put the drug at 100 °C for condensation reflux extraction for 2 h and then recycled the medicine liquid and performed rotary evaporation and concentration at a constant temperature of 55 °C. After the concentrated medicinal liquid was recovered, it was freeze-dried at −60 °C and made into freeze-dried powder. The yield is $33.7\%$, which meets the requirements of drug preparation. The HPLC profile is shown in Figure 1. Both the peak area and concentration of Alisol A and Alisol B 23-monoacetate in AOB are presented in Table 1.
**Figure 1.:** *The HPLC fingerprint of AOB (A), benchmark sample (B).* TABLE_PLACEHOLDER:Table 1.
## Animals
Ten male C57BL/6J mice and 20 male APOE−/− mice were purchased from Changzhou Cavens Laboratory Animal Co., Ltd. (NO. SCXK (SU)-2016-0010). The weight of the mice was 22-24 g and the age of the mice was 4 weeks. All mice were acclimated to a standard rearing environment (Temperature: 18–22 °C, Humidity: 50–$60\%$) for 1 week with 5 mice per cage before experiments carried out. All animal experiments were approved by the Animal Ethics Committee of Nanjing University of Chinese Medicine (NO. 202106A024). All animal experiments complied with animal ethics and all experiments were double-blind.
## High-fat diet model (HFD), experimental design and AOB treatment
We designed three groups including control (CON), model (MOD), *Alisma orientalis* beverage (AOB). In the study, C57BL/6J mice were assigned to the control group, which were fed with a normal mouse diet; and APOE−/− were randomly assigned to the MOD group and AOB group. Mice in the MOD group were fed with a high-fat diet model for 8 weeks; AOB (3.8 g/kg, i.g.) was given daily while feeding with a high-fat diet for 8 weeks in the AOB group. All drugs followed the dosage of the Chinese Pharmacopoeia and did not impair liver and kidney function in mice after 8 weeks of AOB (Table 2).
**Table 2.**
| Unnamed: 0 | Reference range | Reference range.1 | Reference range.2 | Reference range.3 |
| --- | --- | --- | --- | --- |
| Groups | Liver function | Liver function | Renal function | Renal function |
| Groups | ALT(10.06–96.47 U/L) | AST(36.31–235.48 U/L) | CR(10.91–85.09 Umol/L) | BUN(10.81–34.74 mg/dL) |
| CON (Mean ± SD) | 43.549 ± 1.655 | 209.314 ± 4.336 | 26.314 ± 4.824 | 22.624 ± 1.106 |
| MOD (Mean ± SD) | 74.953 ± 15.255 | 171.605 ± 8.584 | 56.165 ± 4.455 | 15.499 ± 1.232 |
| AOB (Mean ± SD) | 59.124 ± 9.392 | 192.546 ± 9.641 | 45.406 ± 9.737 | 13.734 ± 2.006 |
| AOD (Mean ± SD) | 68.734 ± 16.558 | 213.935 ± 18.442 | 51.105 ± 6.759 | 14.733 ± 2.396 |
## Analysis of blood lipids
All mice were anesthetized by using pentobarbital sodium (45 mg/kg; i.p.). After anesthetization, blood was collected by enucleating the eyeball. The collected blood was centrifuged at low speed (3000 rpm) at 4 °C for 10 min, then collected supernatant and stored in −20 °C. High-density lipoprotein cholesterol (HDL-C) and triglycerides (TG), the serum levels of CHO and low-density lipoprotein cholesterol (LDL-C) were measured using a Chemray 240 automatic biochemical analyzer (Wuhan Servicebio Technology, Co., Ltd., China). All experiments were performed as described by the manufacturer.
## Aortic plaque analysis
The heart and the aortic arch were taken out at a low temperature of 4 °C, and the residual blood was washed with 0.01 M PBS. Tissues were placed in a cryostat (Leica, Germany) and serial sections (10 μm) from the aortic sinus to the aortic arch were made from the aortic root according to anatomical markers for histological examination of atherosclerotic aortic sinus lesions. Oil red O staining (ORO) and HE staining were subsequently performed. The plaque area was analyzed using Image Pro Plus 6.0 (*Image analysis* software, Media Cybernetics, Rockville, MD, USA).
## Western blot
The mouse liver tissue (100 mg/kg) was taken out at a low temperature and placed in a lysate for sufficient grinding to a particle-free state, followed by the addition of protease inhibitors. We then centrifuged for 10 min and took the supernatant for a protein concentration test (protein concentration by BCA method). Moreover, the samples were equalized according to the protein concentration and cooked at 100 °C for 5 min with Loading buffer added until the protein was stable. After the target protein was separated by gel electrophoresis (80 v, 90 min), which was transferred to the PVDF membrane under constant flow (300 mA, 60 min, 4 °C). After 18 h of primary antibody including pPI3K (1:1000), PI3K (1:1000), pAKT (1:1000), AKT (1:1000), SREBP-1 (1:1000), GAPDH (1:5000) incubation at 4 °C, we made 2 h of secondary antibody (IgG-Rabbit, 1:4000) incubation at room temperature (20–26 °C), ECL imaging was performed. Visualization of the blot was performed with the chemiluminescent substrate SuperSignal West Pico (Thermo Fisher Science Inc.) and displayed as density relative to GAPDH. Experiments were performed at least 3 times.
## Statistical analysis
All data were shown in the form of mean ± SEM. One-way ANOVA was used with the honestly important difference from Tukey or the post-hoc test from Dunnett. For all statistical tests, GraphPad Prism 8.0 was used, and one-way ANOVA was used in three groups. $p \leq 0.05$ was considered statistically significant.
## Identification of potential action targets of AOB
We collected 137 compounds in AOB from the TCMSP database, 46 of which belonged to Alismatis rhizoma, 55 belonged to *Atractylodis macrocephalae* rhizoma and 36 belonged to Pyrolae herba. After screening by ADME, a total of 7 active ingredients of Alismatis rhizoma, 4 from *Atractylodis macrocephalae* rhizoma, 5 from Pyrolae herba, and 1 common active ingredient from *Alismatis rhizoma* and *Pyrolae herba* were obtained, including Alisma alcohol, kaempferol, quercetin, gallic acid, atractylodes lactone, etc. ( Table 3). Furthermore, we collected the targets of 17 active compounds in AOB from the TCMSP. After the integration of UniProt database entries and the deletion of duplicates, 601 targets were obtained (Table 4).
We collected 4481 AS targets from the Genecards database. The median of the Score value was used as the screening value, so the target with a Score > 2.77 was set as the potential target of AS. We combined with OMIM, TTD, and DRUGBANK databases to supplement relevant targets, and deleted duplicate values after merging, and finally obtaines 1128 dyslipidemia-related targets. We took the intersection of the screened drug active ingredient targets and AS targets, and drew Venn diagrams through Venny2.1 to obtain 171 common targets of AOB and AS (Figure 2 and Table 5).
**Figure 2.:** *Targets screening involved in AOB for the treatment of AS. Venn diagram of disease targets.* TABLE_PLACEHOLDER:Table 5.
## The potential targets of AOB for the treatment of AS
To comprehensively elucidate the possible mechanism of AOB in the treatment of AS, 171 AOB anti-AS target gene names were imported into the STRING database to construct a PPI network. The required interaction score was 0.9 and the disconnected node network was hidden to draw a PPI network graph (Figure 3(A)). To achieve better visualization and identify core targets, we build a network using Cytoscape based on target degrees. With this network, core targets were obtained: PIK3R1, AKT1, PIK3CA, MAPK1, PTPN11, EGFR and MAPK4 (Figure 3(B) and Table 6). These targets may be considered as primary targets of action for AOB for AS treatment, and their identification suggests that AOB treats AS through multiple potential targets.
**Figure 3.:** *The potential targets of AOB for the treatment of AS. (A) The PPI network was constructed by the STRING database. (B) Drawing the PPI core network with Cytoscape 3.8.2 for visual display.* TABLE_PLACEHOLDER:Table 6.
## GO and KEGG enrichment analysis for identification of the pathway mechanisms of AOB
The *Metascape data* platform was used to analyze the signal pathway of the related targets in the regulation of AS by AOB. AOB was mainly involved in the biological processes including regulation of cell adhesion, wound healing, positive regulation of protein phosphorylation, positive regulation of cell migration, etc. The main cellular components involved include membrane raft, the extrinsic component of the membrane, phosphatidylinositol 3-kinase complex, focal adhesion, etc. GO molecular functions of AOB involved include phosphotransferase activity, alcohol group as acceptor, protein kinase activity, kinase activity, kinase binding, protein kinase binding, phosphatase binding, transmembrane receptor protein tyrosine kinase activity, protein tyrosine kinase activity, transmembrane receptor protein kinase activity, protein phosphatase binding. The pathways involved mainly include cancer pathways, PI3K-AKT signaling pathway, EGFR tyrosine kinase inhibitor resistance, fluid shear stress and atherosclerosis, AGE-RAGE signaling pathway in diabetic complications, hepatitis B, etc. ( Table 7). The top 10 significantly enriched ($p \leq 0.01$) terms in BP, CC and MF of GO analysis were selected (Figure 4(A)). The top 20 pathways with significant enrichment ($p \leq 0.01$) were selected (Figure 4(B)).
**Figure 4.:** *GO and KEGG enrichment analysis for identification of the pathway mechanisms of AOB. (A) The top 10 significantly enriched ($p \leq 0.01$) terms in BP, CC and MF of GO analysis were selected. (B) The top 20 pathways with significant enrichment ($p \leq 0.01$) were selected.* TABLE_PLACEHOLDER:Table 7.
## Network-based revelation of Compound-Disease-Pathway-Target network correlations
Combined with the above analysis results, the connection between traditional Chinese medicine, disease, pathways and targets was established. CytoScape3.8.2 was used to construct a Compound-Disease-Pathway-Target network (Figure 5). With the use of the built-in NetworkAnalyzer of CytoScape3.8.2, the network topology parameters of AOB treatment AS were analyzed, and the core components and core role targets were obtained. According to network analysis, 3 main components in the AOB treatment of AS: 16β-methoxyalisol B monoacetate (degree = 36), 3β-acetoxyatractylone (degree = 23), and 5, 2′-dihydroxy-6,7,8-trimethoxyflavone (degree = 19) (Table 8). Therefore, these compounds were regarded as the potential bioactive compounds of AOB against AS. Then, the top 10 core targets were selected according to the comprehensive ranking of degree, closeness and betweenness (Table 9). Interestingly, according to the network analysis results, the PIK3 family and AKT are the core targets of AOB in the treatment of AS, which is consistent with the previous PPI analysis results.
**Figure 5.:** *Compound-Disease-Pathway-Target Network. The orange hexagons represent AOB, the yellow hexagons represent AS, the green circles represent three traditional Chinese medicines, the blue diamonds around the green circles are the main components of the medicine, the red arrows represent the pathways, and the outermost blue circles are the targets. The darker the color, the more important the node is.* TABLE_PLACEHOLDER:Table 8. TABLE_PLACEHOLDER:Table 9.
## AOB reversed the aortic plaque area of as in APOE−/− mice stimulated with a high-fat diet
We performed an AS model with a high-fat diet (HFD). We found that C57BL/6J mice fed with a normal diet for 8 weeks showed no change in arterial plaque area and APOE−/− mice fed with HFD for 8 weeks showed a significant increase in arterial plaque area, which was compared with C57BL/6J mice fed with normal diet (Figure 5). However, after AOB treatment for 8 weeks, the arterial plaque area was significantly reversed both in ORO and HE staining (Figure 6; One-way ANOVA, ORO, F [2, 26] = 62.35, $p \leq 0.001$; HE, F [2, 26] = 86.91, $p \leq 0.001$). These data suggested that AOB had a potential effect to treat AS.
**Figure 6.:** *AOB reversed aortic plaque area of AS in APOE−/− mice. HE and ORO stained sections of aortic valve area in the control group, the model group and the AOB group. The atherosclerotic lesion area was quantitatively analyzed by Image J. Data show mean ± SEM values of 6 or more independent samples. # Represents comparison with the control group, ###represents p < 0.001; * represents comparison with the model group, *** represents p < 0.001.*
## AOB improved four indicators of blood lipids of as in APOE−/− mice stimulated with a high-fat diet
We then detected four indicators of blood lipids related to AS including TG, CHO, HDL and LDL. We found that TG, CHO, LDL were all increased after HFD feeding in APOE−/− mice for 8 weeks, which were all reversed obviously by AOB for 8 weeks [Figure 7, One-way ANOVA, TG, F [2, 17] = 85.17, $p \leq 0.001$; CHO, F [2, 17] = 59.30, $p \leq 0.001$; LDL, F [2, 17] = 19.20, $p \leq 0.001$]. Moreover, HDL in blood serum was decreased after HFD feeding in APOE−/− mice for 8 weeks, which was also reversed obviously by AOB for 8 weeks [Figure 7, One-way ANOVA, HDL, F [2, 17] = 76.44, $p \leq 0.001$].
**Figure 7.:** *AOB improved four indicators of blood lipid of AS in APOE−/− mice stimulated with high-fat diet. Data show mean ± SEM values of 6 independent samples. # Represents comparison with the control group, ### represents p < 0.001; * represents comparison with the model group, * represents p < 0.05, *** represents p < 0.001.*
## AOB alleviated as by regulating PI3K pathway
Based on our network pharmacological analysis, we chose the PI3K pathway to reveal the underlying mechanism of AOB treatment for AS. Then we measured the levels of PI3K/AKT/SREBP-1 in the liver. The results showed that HFD increased phosphorylated expressions of PI3K/AKT and expression of SREBP-1 in APOE−/− mice compared with C57BL/6J mice fed with a normal diet. Interestingly, AOB treatment for 8 weeks reversed all of them in the liver [Figure 8, One-way ANOVA, pPI3K/PI3K, F [2, 17] = 6.997, $$p \leq 0.0147$$; pAKT/AKT, F [2, 17] = 7.925, $$p \leq 0.0087$$; SREBP-1, F [2, 17] = 18.14, $p \leq 0.001$]. Although the expression of SREBP-2 in the liver was significantly increased by HFD, which was not altered after AOB treatment for 8 weeks [Figure 9, One-way ANOVA, F [2, 8] = 9.281 $$p \leq 0.0082$$].
**Figure 8.:** *AOB alleviated AS by regulating the PI3K pathway. The expression levels of the PI3K/AKT/SERBP-1 pathway proteins in each group were detected by western blots. The densitometric values of bands were quantitatively analyzed by Image J Densitometric values normalized to those in the model group and are presented as relative intensity. Data show mean ± SEM values of 6 independent samples. # Represents comparison with the control group, # represents p < 0.05, ## represents p < 0.01; * represents comparison with the model group, * represents p < 0.05, ** represents p < 0.01 *** represents p < 0.001.* **Figure 9.:** *The expression of SREBP-2 in liver after AOB treatment for 8 weeks. The densitometric values of bands were quantitatively analyzed by Image J Densitometric values normalized to those in the model group and are presented as relative intensity. Data show mean ± SEM values of 6 independent samples. # Represents comparison with the control group, ## represents p < 0.01.*
## Discussion
The relationship between the lipid metabolism pathway of AOB and AS was still unclear. In this study, we used network pharmacology combining with experiments to reveal the role of AOB in lipid metabolism, which played a major role in the treatment for AS. We found 3 main components in the AOB treatment of AS: 16β-methoxyalisol B monoacetate (degree = 36), 3β-acetoxyatractylone (degree = 23), and 5, 2′-dihydroxy-6,7,8-trimethoxyflavone (degree = 19) according to network analysis. Moreover, core targets were obtained: PIK3R1, AKT1, PIK3CA, MAPK1, PTPN11, EGFR and MAPK4, which might participate in the treatment of AS. In addition, the top 10 significantly enriched ($p \leq 0.01$) terms in BP, CC and MF of GO analysis were selected and the top 20 pathways with significantly enriched were selected, including the PI3K/AKT/SREBP-1 pathway. Based on the results, we used experiments to identified the therapeutic actions of AOB in AS via the PI3K/AKT/SREBP-1 pathway. The results showed that AOB reversed the aortic plaque area of AS, and improved main indicators of blood lipids related to AS and alleviated AS by regulating PI3K/AKT/SREBP-1 pathway in APOE−/− mice stimulated with HFD. Taken together, we firstly demonstrated that AOB was capable of ameliorating AS by regulating the PI3K/AKT/SREBP-1 pathway.
Although a previous study has identified the effects of AOB on the treatment for AS (Zhu, Zhai, et al. 2020), the underlying molecular mechanism was still unclear. The aortic plaque was significantly increased in AS-related diseases and had been shown to be closely related to high-fat diets (HFD) (Pan et al. 2022). In our study, we found a significant increase in the aortic plaque area in APOE−/− mice fed with HFD for 8 weeks, which was reversed by AOB after 8 weeks of continuous treatment. In addition, abnormal blood lipid indicators including TG, CHO, HDL and LDL, which contributed to AS (Jaquish et al. 1996; Barboza et al. 2016), were all relieved by AOB. To sum up, AOB has the effect of alleviating AS in APOE−/− mice stimulated with HFD.
Network pharmacological analysis showed that 3 main components, 16β-methoxyalisol B monoacetate, 3β-acetoxyatractylone, and 5,2′-dihydroxy-6,7,8-trimethoxyflavone, are associated with AOB treatment of AS. 16β-Methoxyalisol B monoacetate from *Alismatis rhizoma* has been identified to have an antibacterial effect (Jin et al. 2012) and the pathophysiology of bacterial is associated with the development of inflammation (Ge et al. 2022; Keir and Chalmers 2022), risk factors for atherosclerosis. Meanwhile, the component was proved to have an inhibitory effect on phosphorylation of the PI3K/Akt pathway (Xu, Zhao, et al. 2009). Moreover, 3β-acetoxyatractylone from *Atractylodis macrocephalae* rhizoma has been indicated to have treatment-related effects of AS (Chen et al. 2017; Li et al. 2018). In addition, 5,2′-dihydroxy-6,7,8-trimethoxyflavone from Pyrolae herba, a natural flavonoid, plays a role in lipid decreasing (Lin et al. 2022) and the progression of treatment of AS (Kimura et al. 2022; Liu et al. 2022), which also participate in anti-inflammation (Huang et al. 2022; Liu et al. 2022). Network pharmacological results were further identified in our experimental studies, which showed that AOB was capable of releiving AS by regulating the PI3K/AKT/SREBP-1 pathway.
Hypercholesterolemia is recognized as a major contributor to AS, and lowering blood cholesterol levels is an important means in the treatment of AS. Sterol-regulating element-binding proteins (SREBPs) are a family of transcription factors involved in the biosynthesis of cholesterol, fatty acids, and triglycerides (Moslehi and Hamidi-Zad 2018), consisting of SREBP-1 and SREBP-2. SREBP-1 is responsible for the synthesis of fatty acids and cholesterol, while SREBP-2 only regulates the synthesis of cholesterol (Jeon and Osborne 2012). Studies showed that SREBPs in the liver plays a catalytic role in AS by increasing lipid synthesis (Karasawa et al. 2011; Pérez-Belmonte et al. 2017), and inhibiting SREBP-1 led to lower serum cholesterol levels, further alleviating AS (Karasawa et al. 2011). PI3K/AKT is the upstream signaling pathway of SREBP-1, whose activation increased the expression of SREBP-1 (Jeon and Osborne 2012). Recent studies showed that PI3K/AKT signaling is significantly upregulated in patients with nonalcoholic fatty liver disease, one of the risk factors for AS, and inhibitors against PI3K and AKT have potential regulatory effects on lipid metabolism (Aljabban et al. 2022). Our results showed that phosphorylation of the PI3K/AKT signaling pathway in the liver of AS model mice is significantly activated, resulting in elevated SREBPs. After 8 weeks of AOB administration, phosphorylation of the PI3K/AKT signaling pathway in the liver is restored, further lowering SREBP-1 signaling in the liver instead of SREBP-2. These results indicate that AOB regulated the PI3K/AKT/SERBP-1 pathway leading to therapeutic actions in AS by reducing lipid levels.
## Conclusions
AOB has the therapeutic response of AS, which requires suppression of the PI3K/AKT/SERBP-1 pathway. These were our first findings on AOB’s treatment of AS and the underlying mechanism were associated with inhibition of the PI3K/AKT/SERBP-1 pathway, which suggests that traditional Chinese medicine has an obvious curative effect in the treatment of AS, and had a similar molecular mechanism as other western medicines (Mahtta et al. 2022), which provides strong evidence for our later development and extensive use of traditional Chinese medicine.
## Consent form
All authors have approved the manuscript and agree with its submission.
## Author contributions
Ruiyi Liu, Yan Sun, Boran Zhu and Haoxin Wu designed the study and wrote the manuscript. Ruiyi Liu and Yan Sun performed network pharmacology. Ruiyi Liu, Dong Di and Yan Sun performed the experiments. Ruiyi Liu, Boran Zhu and Yan Sun performed the analyzed the data. All data were generated inhouse, and no paper mill was used. All authors agree to be accountable for all aspects of work ensuring integrity and accuracy.
## Disclosure statement
No potential conflict of interest was reported by the author(s).
## Data availability statement
The raw data supporting the conclusions of this manuscript will be available from the corresponding author on reasonable request.
## References
1. Aljabban J, Rohr M, Syed S, Khorfan K, Borkowski V, Aljabban H, Segal M, Mukhtar M, Mohammed M, Panahiazar M. **Transcriptome changes in stages of non-alcoholic fatty liver disease**. *World J Hepatol* (2022) **14** 1382-1397. PMID: 36158924
2. Aryal A, Harmon AC, Dugas TR.. **Particulate matter air pollutants and cardiovascular disease: strategies for intervention**. *Pharmacol Ther* (2021) **223** 107890. PMID: 33992684
3. Barboza LN, Lívero FA, Prando TB, Ribeiro Rde C, Lourenço EL, Budel JM, de Souza LM, Acco A, Dalsenter PR, Gasparotto AJ.. **Atheroprotective effects of**. *J Ethnopharmacol* (2016) **187** 134-145. PMID: 27125593
4. Chen Y, Yang W, Guo L, Wu X, Zhang T, Liu J, Zhang J.. **Atractylodes lactone compounds inhibit platelet activation**. *Platelets* (2017) **28** 194-202. PMID: 27560602
5. Francis GA.. **The complexity of HDL**. *Biochim Biophys Acta* (2010) **1801** 1286-1293. PMID: 20736082
6. Ge L, Liu D, Mao X, Liu S, Guo J, Hou L, Chen X, Huang K.. **Low dose of deoxynivalenol aggravates intestinal inflammation and barrier dysfunction induced by enterotoxigenic**. *J Agric Food Chem* (2022) **70** 3009-3022. PMID: 35201764
7. Geovanini GR, Libby P.. **Atherosclerosis and inflammation: overview and updates**. *Clin Sci* (2018) **132** 1243-1252
8. Hamosh A, Scott AF, Amberger JS, Bocchini CA, McKusick VA.. **Online Mendelian Inheritance in Man (OMIM), a knowledgebase of human genes and genetic disorders**. *Nucleic Acids Res* (2005) **33** D514-517. PMID: 15608251
9. Hopkins AL.. **Network pharmacology: the next paradigm in drug discovery**. *Nat Chem Biol* (2008) **4** 682-690. PMID: 18936753
10. Huang H, He Y, Li Y, Gu M, Wu M, Ji L.. **Eriodictyol suppresses the malignant progression of colorectal cancer by downregulating tissue specific transplantation antigen P35B (TSTA3) expression to restrain fucosylation**. *Bioengineered* (2022) **13** 5551-5563. PMID: 35184647
11. Jaquish CE, Mahaney MC, Blangero J, Haffner SM, Stern MP, MacCluer JW.. **Genetic correlations between lipoprotein phenotypes and indicators of sex hormone levels in Mexican Americans**. *Atherosclerosis* (1996) **122** 117-125. PMID: 8724118
12. Jeon TI, Osborne TF.. **SREBPs: metabolic integrators in physiology and metabolism**. *Trends Endocrinol Metab* (2012) **23** 65-72. PMID: 22154484
13. Jia CY, Li JY, Hao GF, Yang GF.. **A drug-likeness toolbox facilitates ADMET study in drug discovery**. *Drug Discov Today* (2020) **25** 248-258. PMID: 31705979
14. Jin HG, Jin Q, Ryun Kim A, Choi H, Lee JH, Kim YS, Lee DG, Woo ER.. **A new triterpenoid from**. *Arch Pharm Res* (2012) **35** 1919-1926. PMID: 23212633
15. Karasawa T, Takahashi A, Saito R, Sekiya M, Igarashi M, Iwasaki H, Miyahara S, Koyasu S, Nakagawa Y, Ishii K. **Sterol regulatory element-binding protein-1 determines plasma remnant lipoproteins and accelerates atherosclerosis in low-density lipoprotein receptor-deficient mice**. *Arterioscler Thromb Vasc Biol* (2011) **31** 1788-1795. PMID: 21546605
16. Keir HR, Chalmers JD.. **Neutrophil extracellular traps in chronic lung disease: implications for pathogenesis and therapy**. *Eur Respir Rev* (2022) **31** 210241. PMID: 35197267
17. Kimura I, Kagawa S, Tsuneki H, Tanaka K, Nagashima F.. **Multitasking bamboo leaf-derived compounds in prevention of infectious, inflammatory, atherosclerotic, metabolic, and neuropsychiatric diseases**. *Pharmacol Ther* (2022) **235** 108159. PMID: 35183589
18. Kobiyama K, Ley K.. **Atherosclerosis**. *Circ Res* (2018) **123** 1118-1120. PMID: 30359201
19. Koeth RA, Lam-Galvez BR, Kirsop J, Wang Z, Levison BS, Gu X, Copeland MF, Bartlett D, Cody DB, Dai HJ. **l-Carnitine in omnivorous diets induces an atherogenic gut microbial pathway in humans**. *J Clin Invest* (2019) **129** 373-387. PMID: 30530985
20. Li W, Zhi W, Liu F, Zhao J, Yao Q, Niu X.. **Paeoniflorin inhibits VSMCs proliferation and migration by arresting cell cycle and activating HO-1 through MAPKs and NF-κB pathway**. *Int Immunopharmacol* (2018) **54** 103-111. PMID: 29121532
21. Lin YP, Fu SN, Li XP, Wang MM, Fang QL, Qiao X, Yan Q, Hua Y.. **Two novel flavonoids with lipid-lowering activity from Yi medicine Shekaqi**. *Chem Biodivers* (2022) **19** e202200363. PMID: 35934672
22. Liu J, Zhang W, Li Y, Li X, Li Y, Guo F.. **Flavonoids extract from the seeds of**. *Phytomedicine* (2022) **98** 153983. PMID: 35152088
23. Mahtta D, Lee MT, Ramsey DJ, Akeroyd JM, Krittanawong C, Khan SU, Sinh P, Alam M, Garratt KN, Schofield RS. **Significant facility-level variation in utilization of and adherence with secondary prevention therapies among patients with premature atherosclerotic cardiovascular disease: insights from the VITAL (Veterans wIth premaTure AtheroscLerosis) Registry7**. *Cardiovasc Drugs Ther* (2022) **36** 93-102. PMID: 33400053
24. Moslehi A, Hamidi-Zad Z.. **Role of SREBPs in liver diseases: a mini-review**. *J Clin Transl Hepatol* (2018) **6** 332-338. PMID: 30271747
25. Pan X, Wan R, Wang Y, Liu S, He Y, Deng B, Luo S, Chen Y, Wen L, Hong T. **Salvianolic acid B inhibiting chemically and mechanically activated Piezo1 channels as a mechanism for ameliorating atherosclerosis**. *Br J Pharmacol* (2022) **179** 3778-3814. PMID: 35194776
26. Pérez-Belmonte LM, Moreno-Santos I, Cabrera-Bueno F, Sánchez-Espín G, Castellano D, Such M, Crespo-Leiro MG, Carrasco-Chinchilla F, Alonso-Pulpón L, López-Garrido M. **Expression of sterol regulatory element-binding proteins in epicardial adipose tissue in patients with coronary artery disease and diabetes mellitus: preliminary study**. *Int J Med Sci* (2017) **14** 268-274. PMID: 28367087
27. Rebhan M, Chalifa-Caspi V, Prilusky J, Lancet D.. **GeneCards: integrating information about genes, proteins and diseases**. *Trends Genet* (1997) **13** 163. PMID: 9097728
28. Roth GA, Mensah GA, Johnson CO, Addolorato G, Ammirati E, Baddour LM, Barengo NC, Beaton AZ, Benjamin EJ, Benziger CP. **Global burden of cardiovascular diseases and risk factors, 1990–2019: update from the GBD 2019 study**. *J Am Coll Cardiol* (2020) **76** 2982-3021. PMID: 33309175
29. Ru J, Li P, Wang J, Zhou W, Li B, Huang C, Li P, Guo Z, Tao W, Yang Y. **TCMSP: a database of systems pharmacology for drug discovery from herbal medicines**. *J Cheminform* (2014) **6** 13. PMID: 24735618
30. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T.. **Cytoscape: a software environment for integrated models of biomolecular interaction networks**. *Genome Res* (2003) **13** 2498-2504. PMID: 14597658
31. Solanki A, Bhatt LK, Johnston TP.. **Evolving targets for the treatment of atherosclerosis**. *Pharmacol Ther* (2018) **187** 1-12. PMID: 29414673
32. Soppert J, Lehrke M, Marx N, Jankowski J, Noels H.. **Lipoproteins and lipids in cardiovascular disease: from mechanistic insights to therapeutic targeting**. *Adv Drug Deliv Rev* (2020) **159** 4-33. PMID: 32730849
33. Szklarczyk D, Morris JH, Cook H, Kuhn M, Wyder S, Simonovic M, Santos A, Doncheva NT, Roth A, Bork P. **The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible**. *Nucleic Acids Res* (2017) **45** D362-D368. PMID: 27924014
34. Xu X, Zhang W, Huang C, Li Y, Yu H, Wang Y, Duan J, Ling Y.. **A novel chemometric method for the prediction of human oral bioavailability**. *Int J Mol Sci* (2012) **13** 6964-6982. PMID: 22837674
35. Xu YH, Zhao LJ, Li Y.. **Alisol B acetate induces apoptosis of SGC7901 cells via mitochondrial and phosphatidylinositol 3-kinases/Akt signaling pathways**. *World J Gastroenterol* (2009) **15** 2870-2877. PMID: 19533808
36. Zeng Q, Li L, Siu W, Jin Y, Cao M, Li W, Chen J, Cong W, Ma M, Chen K. **A combined molecular biology and network pharmacology approach to investigate the multi-target mechanisms of Chaihu Shugan San on Alzheimer’s disease**. *Biomed Pharmacother* (2019) **120** 109370. PMID: 31563815
37. Zhang R, Zhu X, Bai H, Ning K.. **Network pharmacology databases for traditional Chinese medicine: review and assessment**. *Front Pharmacol* (2019) **10** 123. PMID: 30846939
38. Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, Benner C, Chanda SK.. **Metascape provides a biologist-oriented resource for the analysis of systems-level datasets**. *Nat Commun* (2019) **10** 1523. PMID: 30944313
39. Zhu C, Cai T, Jin Y, Chen J, Liu G, Xu N, Shen R, Chen Y, Han L, Wang S. **Artificial intelligence and network pharmacology based investigation of pharmacological mechanism and substance basis of Xiaokewan in treating diabetes**. *Pharmacol Res* (2020) **159** 104935. PMID: 32464328
40. Zhu B, Zhai Y, Ji M, Wei Y, Wu J, Xue W, Tao WW, Wu H.. *Front Pharmacol* (2020) **11** 570555. PMID: 33101028
|
---
title: Apelin-13 alleviates contrast-induced acute kidney injury by inhibiting endoplasmic
reticulum stress
authors:
- Qian Liu
- Shao-Bin Duan
- Lin Wang
- Xiao-Qin Luo
- Hong-Shen Wang
- Ying-Hao Deng
- Xi Wu
- Ting Wu
- Ping Yan and
- Yi-Xin Kang
journal: Renal Failure
year: 2023
pmcid: PMC9970253
doi: 10.1080/0886022X.2023.2179852
license: CC BY 4.0
---
# Apelin-13 alleviates contrast-induced acute kidney injury by inhibiting endoplasmic reticulum stress
## Abstract
Contrast-induced acute kidney injury (CI-AKI) is a severe complication associated with significant morbidity and mortality, and effective therapeutic strategies are still lacking. Apelin is an endogenous physiological regulator with antioxidative, anti-inflammatory and antiapoptotic properties. However, the role of apelin-13 in CI-AKI remains unclear. In our study, we found that the protein expression levels of apelin were significantly downregulated in rat kidney tissues and HK-2 cells during contrast media treatment. Moreover, we explored the protective effect of apelin-13 on renal tubule damage using in vitro and in vivo models of CI-AKI. Exogenous apelin-13 ameliorated endoplasmic reticulum stress, reactive oxygen species and apoptosis protein expression in contrast media-treated cells and rat kidney tissues. Mechanistically, the downregulation of endoplasmic reticulum stress contributed critically to the antiapoptotic effect of apelin-13. Collectively, our findings reveal the inherent mechanisms by which apelin-13 regulates CI-AKI and provide a prospective target for the prevention of CI-AKI.
## Graphical Abstract
## Introduction
Contrast-induced acute kidney injury (CI-AKI) is a frequent and serious complication of radiologic diagnosis and therapy using intravascular contrast media (CM) [1,2]. CI-AKI has a high incidence worldwide, and approximately $4.4\%$ to $22.1\%$ of patients develop acute kidney injury (AKI) after receiving CM injections [3,4]. The poor short-term and long-term health outcomes of CI-AKI are still far from satisfactory and include irreversible kidney injury, prolonged hospitalization, higher medical costs and increased mortality [5,6]. In clinical therapy, the intravenous hydration strategy is commonly used to prevent and treat CI-AKI, but it has limited efficacy [7]. Therefore, there is an urgent need to find new strategies for overcoming current challenges in CI-AKI treatment.
Previously, direct cytotoxicity associated with CM-induced apoptosis, the overproduction of reactive oxygen species (ROS) and renal hemodynamics were the potential pathogenic mechanisms of CI-AKI [8]. The use of drugs targeting apoptosis and oxidative stress regulation may be an effective therapeutic strategy [1,8]. Numerous studies have indicated that apoptosis is related to endoplasmic reticulum (ER) stress, which is a cellular stress response to the accumulation of unfolded or misfolded proteins in the ER lumen [9,10]. ER stress is thought to be an essential pathological process leading to tubular cell injury in AKI, the progression of chronic kidney disease (CKD) and the transition of AKI to CKD [9,11]. Recently, the involvement of ROS-mediated ER stress has been found in contrast-induced renal tubular cell apoptosis [12]. Therefore, the amelioration of ER stress is critical for the attenuation of CI-AKI.
Apelin-13 is the most widespread and bioactive subtype of the apelin family in humans [13]. As an endogenous physiological regulator with antioxidative, anti-inflammatory and antiapoptotic properties, apelin is becoming a therapeutic target for kidney diseases, including renal ischemia-reperfusion injury (IRI), diabetic nephropathy, and renal fibrosis [14–16]. Studies have also shown that the changes in serum apelin levels in patients with kidney diseases may be partly correlated with disease progression [17]. Interestingly, previous studies reported that ER stress was a critical regulator of apelin-mediated protective effects in ischemic stroke, diabetes and heart failure [15,18,19]. Thus, we proposed that exogenous apelin-13 has potential clinical applications for targeting ER stress, oxidative stress and apoptosis. In this study, we aimed to explore the protective role and mechanism of apelin-13 in CI-AKI through in vivo and in vitro models.
## Antibodies and specific reagents
Anti-PERK [3192], anti-p-PERK (3179S), anti-CHOP [2895], anti-caspase-4 [4450], anti-cleaved caspase-3 (Cleaved caspase-3) [9664] and anti-GAPDH [5174] antibodies were obtained from Cell Signaling (Danvers, USA). Anti-CHOP (15204-1-AP), anti-78-kDa glucose-regulated protein (GRP78) (66574-1-Ig), anti-Nuclear respiratory factor 2 (Nrf2) (16396-1-AP), anti-Keap1 (60027-1-Ig), anti-GAPDH (10494-1-AP) and anti-β-actin (20536-1-AP) were obtained from Proteintech (Chicago, USA). Anti-apelin (ab125213) and anti-caspase-12 (ab62484) were obtained from Abcam (Cambridge, UK). Anti-β-tubulin (GB11017) was obtained from Servicebio (Wuhan, China). All secondary antibodies were obtained from Abcam (Cambridge, UK). Apelin-13 (A6469, purity ≥ $95.0\%$) and 4-phenylbutyrate (4-PBA) (P21005) were purchased from Sigma-Aldrich (St. Louis, USA). GSK2656157 (5.04651) was purchased from EMD Millipore (Massachusetts, USA). Tunicamycin (TM) (ab120296) was purchased from Abcam (Cambridge, UK). Iohexol was purchased from GE Healthcare (Shanghai, China).
## Cell culture
Human proximal tubular cell lines (HK-2 cells) were cultured in DMEM/F12 medium containing $10\%$ fetal bovine serum and $1\%$ penicillin-streptomycin (Gibco, Waltham, USA) in a humidified incubator at 37 °C with an atmosphere containing $5\%$ CO2. CM injury was induced by adding nonionic low-osmolar contrast media iohexol. Briefly, after being washed with phosphate-buffered saline, HK-2 cells were incubated in the medium containing iohexol (200 mg iodine/mL) for 6 h. Additionally, the cells were treated according to the grouping requirements and collected for further analysis.
## Animals and surgical protocol
All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Central South University (NO. 100:2020sydw0899). Sprague-Dawley rats (male, 7 weeks, 220–240 g) were purchased and raised at the Department of Laboratory Animals of Central South University. All rats were housed in the pathogen-free animal facility with free access to food and water under a 12 h light-dark cycle and were acclimatized for 7 days before each animal experiment.
The rats were divided into five groups: control group ($$n = 5$$), 100 nM/kg apelin-13 group ($$n = 5$$), iohexol group ($$n = 5$$), iohexol + 10 nM/kg apelin-13 group ($$n = 5$$), and iohexol + 100 nM/kg apelin-13 group ($$n = 5$$). The model of rat CI-AKI was established as previously described [20]. Blood samples were collected from the orbital venous plexus through capillary glass tubes and used to measure serum creatinine (SCr) and blood urea nitrogen (BUN) before the rats were deprived of water. All rats that had been dehydrated for 48 h were injected intraperitoneally with furosemide (10 mL/kg) 30 min prior to the time point at which they were injected with iohexol (15 mL/kg). Apelin-13 (10 nM/kg, 100 nM/kg) was injected 10 min before the iohexol injection. Apelin-13 and iohexol were administered intravenously via rapid tail vein injection as previously described. The rats were sacrificed 24 h after the iohexol injection, and kidney tissues and blood samples were collected for further experiments (Figure S1).
**Figure 1.:** *Iohexol induces ER stress, oxidative stress and apoptosis in rat kidney tissues. The method for building the CI-AKI rat model has been described in our previous article. (A, B) Changes in the levels of Scr and BUN. (C) Renal GSH content. (D) Renal MDA activity. (E, F) Representative immunoblot analysis and semi-quantitative analysis of GRP78, CHOP, caspase-12 and Cleaved caspase-3, GAPDH was used as a loading control. Data are expressed as means ± SEMs. n = 4. *p < 0.05, **p < 0.01, significantly different from control group.*
## Cell viability assay
The viability of HK-2 cells was measured with a cell counting kit-8 (CCK-8) (Dojindo Molecular Technologies, Tokyo, Japan) according to the manufacturer’s instructions. Cells were seeded in 96-well plates at a density of 5000 cells/well, and eight replicate wells were used for each group. Then, the cells were incubated with iohexol or apelin-13 for the indicated times. Ten microliters of CCK-8 solution were added to each culture well, followed by incubation for 2 h at 37 °C. A microplate reader (MD SPECTRA M2, Molecular Devices, Sunnyvale, US) was used to spectrophotometrically measure the absorbance at 450 nm. HK-2 cells cultured in DMEM/F12 medium were used as a negative control. Culture media without cells served as blank controls. Cell viability was calculated and expressed as a percentage of the absorbance of the treated group to that of the control group.
## Immunoblot analysis
Immunoblot analysis was performed according to standard procedures. A BCA Protein Detection Kit (Beyotime Institute of Biotechnology, Shanghai, China) was used to determine the protein concentrations. The protein samples were electrophoresed on a polyacrylamide gel at suitable concentrations and then transferred to polyvinylidene fluoride membranes. The polyvinylidene fluoride membrane was then blocked with $5\%$ bovine serum albumin or $5\%$ fat-free milk for 1 h, and then incubated overnight with specific primary antibodies diluted according to the manufacturer’s recommendations, followed by incubation with horseradish peroxidase-conjugated secondary antibodies. Antigen-antibody complexes on the membranes were detected using an enhanced chemiluminescence kit (Thermo Fisher Scientific, Rockford, USA). Images were obtained by Tanon 5200 Multi Image Analysis software 1.0 (Tanon, Shanghai, China). Finally, the band intensity was evaluated by ImageJ software.
## Transmission electron microscope analysis (TEM)
Fresh samples were fixed in $2.5\%$ glutaraldehyde solution, followed by conventional dehydration, osmosis, embedding, sectioning, and staining as previously described [21]. Typical images were acquired with a Hitachi H7700 electron microscope (Hitachi, Tokyo, Japan).
## Determination of blood parameters
SCr and BUN were measured using the automatic biochemical analyzer Hitachi 7170A (Hitachi, Tokyo, Japan) in the laboratory department of Second Xiangya Hospital of Central South University.
## Hematoxylin-eosin (HE) staining
Renal tissues were fixed in $4\%$ paraformaldehyde solution, dehydrated, and paraffin-embedded. Paraffin-embedded kidney tissues were sectioned at a thickness of 4 μm and then stained with hematoxylin-eosin for histopathological analysis. Ten high-magnification (× 200) fields in the cortex and outer medulla of HE-stained kidney sections were randomly selected for semiquantitative analysis of morphological alterations. The specimens were scored based on the percentage of damaged renal tubules as previously described: 0, no damage; 1, < $25\%$ damage; 2, $25\%$ − $50\%$ damage; 3, $50\%$ − $75\%$ damage; and 4, > $75\%$ damage.
## Renal immunohistochemistry (IHC)
Paraffin-embedded kidney tissues were sectioned at a thickness of 4 μm for IHC analysis. After dewaxing, rehydration, blocking, and antigen retrieval, the kidney tissue sections were exposed to anti-apelin antibodies (1:100) at 4 °C overnight. Then the sections were incubated with biotinylated goat anti-rabbit secondary antibodies (PV-9000, Zhongshan Jinqiao Biotechnology, Beijing, China) for 30 min at room temperature. A DAB kit (ZLI-9018, Zhongshan Jinqiao Biotechnology, Beijing, China) was used to detect the signal of the antigen-antibody complexes. Finally, the slides were counterstained in hematoxylin.
## Immunofluorescence (IF) staining
Paraffin-embedded kidney sections were also used for IF studies. The sections were deparaffinized and sequentially incubated with 0.1 M sodium citrate for antigen retrieval, and $3\%$ H2O2 to block endogenous peroxidase activity. Then $2\%$ normal goat serum blocking buffer was used to reduce nonspecific binding. The sections were then incubated with anti-CHOP (Proteintech, 1:200) and anti-GRP78 antibodies (1:500) at 4 °C. After being incubated overnight incubation at 4 °C, the samples were incubated with secondary antibodies for 1 h at 37 °C in the dark. The sections were counterstained using the antifade mounting medium with DAPI (P0131, Beyotime, Shanghai, China) and then observed and photographed under a fluorescence microscope (BX51, Olympus, Tokyo, Japan).
## Malondialdehyde (MDA), glutathione (GSH) and ROS assays
Commercial kits were used to test MDA (A003-1, Jiancheng Bioengineering Institute, Nanjing, China) and GSH (S0052, Beyotime, Shanghai, China) concentrations according to the manufacturer’s instructions.
Dihydroethidium (DHE) staining was used to measure ROS levels in renal tissues as previously described [22]. The whole process can be divided into three parts. First, after the rat kidneys were collected, the kidney was rinsed in cold PBS and then placed in Tissue-Tek optimal cutting temperature compound, snap frozen in liquid nitrogen and stored at -80 °C. Second, freshly isolated 20 μm-thick kidney slices were incubated in 10 μm DHE (Thermo Fisher Scientific, Rockford, USA) for 30 min in the dark at 37 °C and then counterstained with DAPI (P0131, Beyotime, Shanghai, China) at room temperature. Finally, photos were taken with a fluorescence microscope (BX51, Olympus, Tokyo, Japan), and the fluorescence intensity in 10 random optical sections was determined with ImageJ software.
ROS levels in cells were tested by a ROS assay kit purchased from Beyotime Institute of Biotechnology according to the manufacturer’s instructions (S0033S, Shanghai, China). 2′,7′-Dichlorofluorescin diacetate (DCFH-DA) was prepared as a 10 mM solution using a serum-free medium. After being processed according to the grouping requirements, the cells were incubated in a medium containing 10 μM DCFH-DA at 37 °C for 20 min. Finally, fluorescence microscopy (BX51, Olympus, Tokyo, Japan) was used to observe the ROS production in HK-2 cells.
## Apoptosis assay
Apoptosis in renal tissues was detected using terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) reagent (12156792910, Roche Life Science, Basel, Switzerland) according to the manufacturer’s protocol. Briefly, the dewaxed kidney tissue sections were permeabilized with 0.1 M sodium citrate for 30 min. Then, the sections were incubated with TUNEL reaction mixture in the dark at 37 °C for 1 h. The number of TUNEL-positive cells and total cells in different tissue sections were counted in 10 representative fields per section by using fluorescence microscopy (BX51, Olympus, Tokyo, Japan). TUNEL-positive cells are represented as a percentage of total cells.
## Statistical analysis
All quantitative data are representative of at least 3 independent experiments. All statistical analyses were performed using SPSS 20.0 software. The results are expressed as the means ± SEMs. A significant difference between the two groups was evaluated using the Student’s t-test. A significant difference among three or more groups was determined by one-way analysis of variance (ANOVA) followed by the LSD test for post hoc comparisons. $p \leq 0.05$ was considered significantly different. All statistical results were graphed using GraphPad Prism 6.0 software.
## Iohexol induces ER stress, oxidative stress, and apoptosis in the CI-AKI model
We established a CI-AKI rat model as previously reported to determine whether iohexol induced ER stress in renal tissues in vivo. As shown in Figure 1(A–B), both SCr and BUN were significantly increased in rats after iohexol injection (CM group) compared to rats that were injected with saline (control group). The oxidative stress-related indicator GSH was decreased, while MDA was increased in the CM group compared with the control group (Figure 1(C–D)). The protein expression of ER stress indicators (GRP78), ER stress-induced apoptosis indicators (CHOP, caspase-12) and classical apoptosis indicators (Cleaved caspase-3) were all significantly increased compared with the control group (Figure 1(E–F)). These above results demonstrated that iohexol could indeed cause ER stress, oxidative stress and apoptosis in the kidney.
Then we administered iohexol (200 mg iodine/mL) to HK-2 human kidney proximal tubular cells for 0 h, 2 h, 4 h, 6 h, 8 h and 10 h. Iohexol caused a time-dependent decrease in cell viability (Figure S2(A)). Then, we further confirmed that iohexol increased the protein expression of ER stress indicators (GRP78, p-PERK), ER stress-induced apoptosis indicators (CHOP, caspase-4 (an alternative to caspase-4 in humans)) and classical apoptosis indicators (Cleaved caspase-3) in a time-dependent manner by in HK-2 cells immunoblotting (Figure S2(B–H)). TEM also revealed that the rough ER showed swelling, dilatation, and partial vesiculation with many shedding ribosomes in a time-dependent manner of the iohexol groups compared to the control group (Figure S2(I)). Taken together, these results indicate that iohexol triggers the induction of ER stress and apoptosis in renal tubular epithelial cells.
**Figure 2.:** *Iohexol intervention reduced apelin expression in tubular epithelial cells in vitro and in vivo. HK-2 cells were treated with iohexol (200 mg iodine/mL) at the indicated time points. The expression of apelin was detected by immunoblot analysis. (A, B) Representative immunoblot analysis and semi-quantitative analysis of apelin in rat kidney tissues, GAPDH was used as a loading control (n = 4). (C) Representative immunohistochemical staining of apelin. Scar bar, 50 μm. (D, E) Representative immunoblot analysis and semi-quantitative analysis of apelin in HK-2 cells. Tubulin was used as a loading control (n = 3). *p < 0.05, **p < 0.01, significantly different from the control group. Data are expressed as means ± SEMs.*
## Iohexol reduced apelin expression in rat kidney tissues and HK-2 cells
The effect of the iohexol on apelin expression was examined. In the CI-AKI rat model, immunoblot analysis showed a significant reduction in apelin in kidney tissues after iohexol intervention (Figure 2(A,B)). Immunohistochemical analysis also revealed lower apelin levels in renal tissues after iohexol intervention, and the reduction occurred mainly in renal tubular cells (Figure 2(C)). As shown in Figure 2(D–E), immunoblot analysis showed a sustained decrease in apelin expression with prolonged iohexol treatment in the in vitro model. Taken together, these results suggested that apelin was downregulated by iohexol, and that apelin might modulate renal tubular injury.
## Apelin-13 attenuated renal injury induced by iohexol in vivo
We further investigated the effect of apelin-13 on a rat CI-AKI model. Compared with the control group, renal function indicators (SCr and BUN) showed no significant difference in the 100 nM apelin-13 group and significant increases in the iohexol group. In contrast, there were significant improvement in SCr and BUN in the CM + 10 nM apelin-13 and CM + 100 nM apelin-13 groups (Figure 3(A,B)). HE staining was used to detect renal histopathology, and the renal tubular injury score of each group was calculated. The vacuolization of renal tubular epithelial cells is an indicator of drug toxicity and a histopathological feature of CI-AKI [8,23]. Almost normal renal tissue structure was observed in the control group and 100 nM apelin-13 groups. Serious tubular epithelial cell vacuolization and shedding, interstitial edema, brush border rarefaction, tubular dilation, and intratubular cast formation were observed in the iohexol group. Apelin-13 treatment significantly alleviated the development of these lesions and tissue damage. Quantitative analysis revealed that following iohexol intervention, apelin-13-treated rats had significantly lower tubular injury scores (∼2.7, ∼1.8) than rats without apelin-13 treatment (∼3.7) (Figure 3(C,D)). In addition, changes in renal tubular epithelial cells were observed by TEM. In the control group, the rough ER was flat-saccular and regularly arranged, with ribosomes attached outside the membrane. Mitochondria and other organelles had normal structures. In the iohexol group, the rough ER showed different degrees of swelling, dilatation, partial vesiculation and ribosomes attached to the ER membrane, and some mitochondria showed swelling, fragmentation, vacuoles, and the loss of cristae. These microstructural alterations in the renal tubular epithelial cells of apelin-13-treated rats were reduced to varying degrees (Figure 3(E)). Collectively, these findings support the therapeutic potential of apelin-13.
**Figure 3.:** *The protective effects of apelin-13 on renal function and pathological injury in rat kidneys. We observed the changes of SCr, BUN, and renal histopathology changes through HE staining in the control group, 100 nM apelin-13 group, iohexol group, iohexol + 10 nM apelin-13 group and iohexol + 100 nM apelin-13 group. (A) Changes of SCr level. (B) Changes of BUN level. (C) Representative images of HE staining. Scar bar, 100 μm. (D) Histopathological score of tubular damage. (E) Representative images of transmission electron microscope analysis (original magnification: ×5000, ×20000; scar bar: 1 μm; under a Hitachi H7700 electron microscope). *p < 0.05, **p < 0.01, significantly different from control group; #p < 0.05, ##p < 0.01, significantly different from the CM group. All quantitative data are expressed as means ± SEMs, n = 5.*
## The levels of oxidative stress and apoptosis decreased in the renal tissues of CI-AKI rat model after apelin-13 administration
In vivo, we measured MDA and GSH levels in renal tissues. Iohexol increased GSH expression and decreased MDA expression, while apelin-13 treatment ameliorated these changes (Figure 4(A,B)). Moreover, DHE assays revealed a significant increase in oxidative stress in the renal tubular cells of CI-AKI rats, whereas apelin-13 treatment significantly reversed these changes (Figure 4(C,D)). Moreover, since the Keap1-Nrf2 axis serves as a classical antioxidant pathway [21,24], we examined the Keap1 and Nrf2 expression levels by immunoblot analysis. Nrf2 expression was significantly decreased while Keap1 expression was increased in the renal tissues of the iohexol group compared with the control group. The level of Nrf2 activity in the apelin-13-treated group was significantly higher than that in the iohexol group, suggesting that apelin-13 treatment could reduce oxidative stress induced by iohexol in rat renal tissues. Moreover, the results of immunoblot analysis of caspase-12, and Cleaved caspase-3 revealed that apelin-13 reduced iohexol-induced apoptosis (Figure 4(E,F)). TUNEL staining also suggested a decrease in DNA double-strand breaks in the apelin-13-treated group, which was a positive indication of apoptosis (Figure 4(G,H)). These results suggested that apelin-13 plays a vital role in reducing renal tubular cell oxidative stress and apoptosis during CM treatment.
**Figure 4.:** *Apellin13 treatment reduced the oxidative stress and apoptosis in rat kidney tissues. Rats were treated as described above. The expression of Keap1, Nrf2 in renal tissues was assessed by immunoblot analysis. Oxidative stress in renal tubular cells was assessed using dihydroethidium (DHE). Apoptosis levels of renal tubular cells were assessed using TUNEL staining. (A) Renal MDA content. (B) Renal GSH activity. (C, D) Representative images of DHE staining and semi-quantitative analysis of DHE fluorescence intensity, Scar bar, 100 μm. (E, F) Representative immunoblot analysis and semi-quantitative analysis of Keap1, Nrf2, caspase-12 and Cleaved caspase-3 in rat kidney tissues, GAPDH was used as a loading control. (G) Representative images of TUNEL staining. The apoptotic cells were detected by TUNEL (red), and the nuclei were detected by DAPI (blue). Scar bar, 100 μm. (H) Quantitative analysis of TUNEL staining positive cells. *p < 0.05, **p < 0.01, significantly different from control group; #p < 0.05, ##p < 0.01, significantly different from the CM group. All quantitative data are expressed as means ± SEMs, n = 3.*
## ER stress induced by iohexol was downregulated in rats injected with apelin-13
To further confirm the renoprotective effect of apelin-13 on CI-AKI, we examined ER stress by evaluating the expression of ER stress-related proteins. Immunoblot analysis showed that GRP78 and CHOP expression levels and PERK phosphorylation levels were significantly increased in iohexol-treated rat kidney tissues, and these effects were inhibited by apelin-13 treatment (Figure 5(A,B)). The immunoblot analysis results were further validated by the immunohistochemical staining assay, which revealed that GRP78 expression was almost absent in the cortex in the control group, and high levels of GRP78 were observed in the cytoplasm of renal tubules in the iohexol group. Similarly, CHOP expression in the kidney cortex was low in the control group but was increased in the cytoplasm and nucleus after iohexol intervention. The shuttling of CHOP to the nucleus is an indication of the transcriptional activation of CHOP. Notably, the apelin-13 intervention alleviated the increase in the expression of GRP78 and CHOP, especially the accumulation of CHOP in the nucleus (Figure 5(C)). Taken together, these results suggest that apelin-13 may alleviate ER stress in the CI-AKI rat model, further supporting the therapeutic potential of apelin-13.
**Figure 5.:** *Apelin-13 attenuated iohexol-induced ER stress in rat tubular cells. Rats were treated as described above. The expression of GRP78, p-PERK, PERK and CHOP in kidney tissues was detected by immunoblot analysis. Immunofluorescence was used to detect the expression of GRP78 and CHOP in kidney tissues. (A, B) Representative immunoblot analysis and semi-quantitative analysis of GRP78, p-PERK, PERK, CHOP in rat kidney tissues, GAPDH was used as a loading control. (C) Representative immunofluorescence images of the GRP78 and CHOP. *p < 0.05, **p < 0.01, significantly different from control group; #p < 0.05, ##p < 0.01, significantly different from the CM group. All quantitative data are expressed as means ± SEMs, n = 3.*
## Apelin-13 protects HK-2 cells from iohexol-induced ER stress, oxidative stress and apoptosis in vitro
We also performed in vitro experiments to determine the effect of apelin-13 on ER stress, oxidative stress and apoptosis in HK-2 cells. We initially evaluated the effect of exogenous apelin-13 on the viability of HK-2 cells treated with iohexol (CM, 200 mg iodine/mL, 6 h) by CCK8 assay. As shown in Figure 6(A–B), within a specific range of concentrations and times, apelin-13 intervention dose-dependently improved cell activity. Moreover, the effect was better when apelin-13 was added prior to iohexol or simultaneously. To further confirm whether exogenous apelin-13 could protect renal tubular cells from ER stress through antioxidative and antiapoptotic effects, we added exogenous apelin-13 to HK-2 cells along with iohexol. The MDA and GSH levels in HK-2 cells were consistent with the results of the rat experiments (Figure 6(C,D)). As shown in Figure 6(E,G), Nrf2 was decreased in the iohexol group but increased after apelin-13 treatment. The change in Keap1 was opposite to that of Nrf2. Furthermore, exogenous apelin-13 decreased the expression of caspase-4, Cleaved caspase-3, CHOP and PERK phosphorylation induced by iohexol treatment (Figure 6(E–H)). Apparently, apelin-13 alleviated ER stress, oxidative stress and apoptosis induced by iohexol treatment in HK-2 cells in a dose-dependent manner. These results further validated the in vitro experimental results, which indicated the therapeutic potential of exogenous apelin-13 in CI-AKI.
**Figure 6.:** *Apelin-13 attenuates iohexol-induced ER stress, oxidative stress and apoptosis in HK-2 cells. (A, B) HK-2 cells were incubated in medium containing 200 mg iodine/mL iohexol and / or interfered with different concentrations of apelin-13 (0, 0.1, 1, 10, 100, 1000 nM) for 6 h. HK-2 cells were incubated in the medium containing 200 mg iodine/mL iohexol for 6 h and / or given the intervention of 100 nM apelin-13 at indicated time points (0 h, 3 h, 6 h, 9 h, 12 h) prior to cell collection. The cells cultured in a normal medium without iohexol were used as a control. Cell viability was detected with CCK-8 assay. Cell viability of the control group was set to 100 %, and other groups were normalized to indicate cell viability changes with the control group (n = 8). (C - H) HK 2 cells were incubated in medium containing 200 mg iodine/mL iohexol and apelin(10 nM, 1000 nM) for 6 h. Cells cultured in the normal medium without iohexol or apelin-13 were used as a control. (C) Cell GSH content. (D) Cell MDA activity. (E - H) Representative bands of p-PERK, CHOP, caspase-4, Cleaved caspase-3, Keap1, Nrf2 and semi-quantitative analysis of these protein expression. *p < 0.05, **p < 0.01, significantly different from control group; #p < 0.05, ##p < 0.01, significantly different from the CM group. Data are expressed as means ± SEMs.*
## ER stress plays a significant role in the antioxidative and antiapoptotic effects of apelin-13 in HK-2 cells treated with iohexol
To further determine the role of ER stress in iohexol-induced oxidative stress and apoptosis in HK-2 cells, we pretreated HK-2 cells with GSK2656157 and 4-PBA. GSK2656157 is a PERK-specific inhibitor, and 4-PBA is a chemical chaperone that can eliminate the accumulation of unfolded proteins in the ER and attenuate ER stress [25]. As shown in Figures S2 and S3, HK-2 cells were incubated in a medium with 1.0 μM GSK2656157 for 6.5 h, 2.0 mM 4-PBA for 6.5 h and/or 200 mg iodine/mL iohexol for 6 h. The effect of 4-PBA on cell death was evaluated by CCK8 and immunoblot analysis of Cleaved caspase-3. HK-2 cell death induced by iohexol treatment was significantly inhibited by 4-PBA (Figure S3(A)). In contrast to the effect of iohexol alone, the level of MDA was decreased, and the level of GSH was increased after treatment with 4-PBA (Figure S3(B,C)). In addition, 4-PBA reduced the intensity of red fluorescent ROS staining after iohexol treatment (Figure S3(D)). Moreover, immunoblot analysis of Cleaved caspase-3 also showed apoptosis in HK-2 cells induced by iohexol treatment was significantly inhibited by 4-PBA (Figure S3(E–F)). GSK2656157 exerted similar therapeutic effects as 4-PBA (Figure S4(A–F)). Taken together, these results suggest that inhibiting ER stress may provide a promising strategy for the prevention of CI-AKI. However, the effect and detailed mechanism of apelin-13 on ER stress in CI-AKI remain unclear.
Based on these results, we presumed that the antioxidative and antiapoptotic effects of apelin-13 on CI-AKI might be related to the inhibition of ER stress. To prove this hypothesis, we used TM, a classical ER stress inducer, to perform rescue experiments in vitro. As expected, TM partially reversed the protective effect of apelin-13 treatment (Figure 7(A)). As shown in Figure 7(B–F), TM partially blocked the antioxidative and antiapoptotic effects of apelin-13 in the model of CI-AKI, as shown by MDA, GSH, ROS and the expression of Nrf2, Keap1 and Cleaved caspase-3. As mentioned previously, the results showed that apelin-13 could partially protect HK-2 cells from iohexol-induced cytotoxicity mediated by relieving ER stress.
**Figure 7.:** *Activation of ER stress abolished the antioxidative and antiapoptotic effects of apelin-13 in HK-2 cells treated with iohexol. HK-2 cells were treated with iohexol (200 mg iodine/mL) for 6 h, apelin-13 (100 nM) for 6 h, and / or TM (100 nM) for 8 h. (A) Cell viability. (n = 8). (B) Cell MDA content. (C) Cell GSH activity. (D, E) Representative immunoblot analysis and semi quantitative analysis of Keap1, Nrf2 and Cleaved caspase-3 in HK-2 cells, GAPDH was used as a loading control. (F) Representative images of ROS staining, Scar bar, 100 μm. *p < 0.05, **p < 0.01, significantly different from CM group; #p < 0.05, ##p < 0.01, significantly different from the CM + Apelin-13 group. All quantitative data are expressed as means ± SEMs, n = 3.*
## Discussion
In the present study, we found that CM downregulated apelin-13 and demonstrated the protective effect of exogenous apelin-13 on CI-AKI rats and CM-treated HK-2 cells. Moreover, our results indicated that exogenous apelin-13 alleviated CI-AKI by reducing ER stress, oxidative stress and apoptosis. Mechanistically, apelin-13 exerted antioxidative and antiapoptotic effects, at least in part, by alleviating ER stress in renal tubular epithelial cells (Figure 8).
**Figure 8.:** *Central illustration - the mechanistic pathway of apelin-13 regulation in CI AKI. ER stress contributed to the oxidative stress and apoptosis of tubular epithelial cells induced by CM. Apelin-13 played a protective role by downregulating ER stress, oxidative stress and apoptosis. Apelin-13 may alleviate oxidative stress and apoptosis through modulating ER stress in tubular epithelial cells.*
CI-AKI is a common cause of AKI and iatrogenic nephropathy, and effective treatments are needed [26]. Impaired renal tubular cells are a hallmark of CI-AKI and are mainly caused by direct cytotoxicity leading to tubular epithelial cell apoptosis, the excessive production of ROS and intrarenal vasoconstriction. Our study and others have shown that renal oxidative stress induced by ROS overproduction results in apoptosis in CI-AKI [21,27], which can be an important pathway that influences kidney damage. In this study, we also examined the oxidative stress and apoptosis indices under the conditions of CI-AKI and showed that apoptosis plays an essential role in the initiation of CI-AKI. Apelin, which is the cognate ligand for the G-protein-coupled receptor APJ, is a member of the adipokine family and has been reported to be a broad regulator of physiology [14]. Apelin was demonstrated to exert protective effects in several renal injury models, such as IRI and diabetic kidney disease, by antioxidative, anti-inflammatory, and antiapoptotic effects [14,19]. However, the pathophysiological effects of apelin in CI-AKI remain elusive. In this study, we provided in vitro and in vivo evidence that the expression of apelin was reduced during CM intervention. The administration of a certain range of exogenous apelin-13 may alleviate renal tubular epithelial cell injury induced by CM. In addition, apelin-13 treatment attenuated changes in SCr and BUN levels and kidney histological changes in the rat CI-AKI model. More importantly, our study showed the oxidative stress and apoptosis indices under apelin-13 treatment in the rat CI-AKI model, such as the expression of Nrf2, Keap1 and Cleaved caspase-3 and the levels of GSH, MDA ROS and TUNEL, and the downregulation of oxidative stress and apoptosis suggested the antioxidative and antiapoptotic effects of apelin-13 in vivo and in vitro. Collectively, these findings provide further evidence supporting a potential therapeutic role for apelin-13 in the prevention of CI-AKI.
Apelin-13 has been shown to be beneficial in aortic calcification, ischemic stroke, and subarachnoid hemorrhage by alleviating ER stress [18,28,29]. The protective effect of apelin-13 against ER stress may be mediated by Gαi/Gαq-CK2 signaling [18]. Moreover, increasing evidence suggests that ER stress plays a vital role in renal tubular epithelial injury induced by AKI [9,30]. In terms of the three ER stress pathways, PERK initiates the immediate adaptive reaction to ER stress and crucially determines cell fate in response to ER stress, and so we chose the most classical pathway as the focus of our research. Previous studies confirmed that the PERK-CHOP pathway was altered, and ER stress had an essential effect on apoptosis induced by iodinated CM [12,31–33]. However, injury was mainly induced by high-osmolar contrast media (HOCM), which is currently less frequently used in clinical practice. The relationship between low-osmolar contrast media (LOCM)-induced ER stress and the activation of apoptosis has not been precisely studied. Consistent with LOCM, our study showed that LOCM induced ER stress in vivo and in vitro. The degree of LOCM-induced cell damage and apoptosis was positively correlated with ER stress levels in a time-dependent manner. By blocking ER stress, we further demonstrated that ER stress was directly involved in contrast-induced tubular epithelial apoptosis. Moreover, apelin-13 treatment reduced ER stress, while the antiapoptotic effect of apelin-13 was weakened after activating ER stress with TM. Therefore, it was proven that the antiapoptotic effect of apelin-13 in CI-AKI was related to inhibiting ER stress.
ER stress and oxidative stress are mutually causative. Low molecular-weight oxidants are produced and amassed in the ER lumen, which is highly oxidative [10,34]. PERK signaling has been reported to be involved in the crosstalk between ER stress and the oxidative stress signaling pathway [35,36]. The Nrf2 pathway is the dominant defense mechanism against oxidative stress [37]. Studies have shown that the Nrf2 pathway is activated in response to ER stress. Activation of the Nrf2 pathway has been shown to reduce the accumulation of misfolded proteins that can contribute to ER stress. In addition, activation of the Nrf2 pathway may also help to reduce inflammation and cell death that can result from ER stress [38,39]. Interestingly, as a regulatory transcription factor, Nrf2 has also been reported to be a direct substrate of PERK [40]. In this study, we provide evidence that the ER stress is associated with oxidative stress induced by CM via inhibiting ER stress and specifically inhibiting PERK in CI-AKI, which was shown as a notable change in the levels of MDA, GSH, ROS and Nrf2 expression. Notably, rescue experiments with an ER stress inducer indicated that moderate amounts of apelin-13 prevented iohexol-induced oxidative stress by antagonizing ER stress. Based on these observations, we concluded that apelin-13 maintained ER homeostasis to alleviate contrast-induced renal tubular epithelial injury.
This study has some limitations. First, apelin-13 was administered systemically without tissue or cell-type-specific genetic approaches. Further studies with larger sample sizes are needed to verify the conclusion in other CI-AKI models. Second, apelin-13 may affect renal tubular epithelial cells, renomedullary interstitial cells and renal hemodynamics, but our study focused on the effect of apelin-13 on renal tubular epithelial cell injury in vitro experiments. In addition, the critical molecular mechanisms of ER stress in response to apelin-13 treatment and its other precise regulatory mechanisms remain to be investigated in the future.
This study identified the novel beneficial effects of apelin-13 on renal damage in CI-AKI through in vitro and in vivo experiments. Functionally, treatment with apelin-13 markedly alleviated tubular epithelial damage by reducing ER stress, oxidative stress and apoptosis. Mechanistically, apelin-13 may alleviate oxidative stress and apoptosis by modulating ER stress in tubular epithelial cells. These results suggest that modulation of apelin levels could be regarded as a promising approach to the prevention of CI-AKI in the future.
## Author contributions
S.B.D. conception, design, supervision and financial support. Q.L. conception, design and implementation, collection of data, data analysis and interpretation, manuscript writing. L.W. and Y.H.D. collection of data and data analysis. X.W. and X.Q.L. data analysis and manuscript writing. H.S.W., T.W., Y.X.K. and P.Y. revision of the manuscript for important intellectual content. All authors read and approved the final version of the manuscript.
## Disclosure statement
The authors state no conflict of interest.
## Data availability statement
The data used to support the findings of this study are available from the corresponding authors upon request.
## References
1. Fahling M, Seeliger E, Patzak A. **Understanding and preventing contrast-induced acute kidney injury**. *Nat Rev Nephrol* (2017) **13** 169-180. PMID: 28138128
2. Davenport MS, Perazella MA, Yee J. **Use of intravenous iodinated contrast media in patients with kidney disease: consensus statements from the American college of radiology and the national kidney foundation**. *Radiology* (2020) **294** 660-668. PMID: 31961246
3. Nash K, Hafeez A, Hou S.. **Hospital-acquired renal insufficiency**. *Am J Kidney Dis* (2002) **39** 930-936. PMID: 11979336
4. Eng J, Wilson RF, Subramaniam RM. **Comparative effect of contrast media type on the incidence of contrast-induced nephropathy: a systematic review and meta-analysis**. *Ann Intern Med* (2016) **164** 417-424. PMID: 26830055
5. Weisbord SD, Palevsky PM, Kaufman JS. **Contrast-associated acute kidney injury and serious adverse outcomes following angiography**. *J Am Coll Cardiol* (2020) **75** 1311-1320. PMID: 32192658
6. Cheng W, Wu X, Liu Q. **Post-contrast acute kidney injury in a hospitalized population: short-, mid-, and long-term outcome and risk factors for adverse events**. *Eur Radiol* (2020) **30** 3516-3527. PMID: 32080754
7. Nijssen EC, Rennenberg RJ, Nelemans PJ. **Prophylactic hydration to protect renal function from intravascular iodinated contrast material in patients at high risk of contrast-induced nephropathy (AMACING): a prospective, randomised, phase 3, controlled, open-label, non-inferiority trial**. *Lancet* (2017) **389** 1312-1322. PMID: 28233565
8. Nijssen EC, Nelemans PJ, Wildberger JE. **Contrast-associated acute kidney injury**. *N Engl J Med* (2019) **381** 1295-1296
9. Cybulsky AV.. **Endoplasmic reticulum stress, the unfolded protein response and autophagy in kidney diseases**. *Nat Rev Nephrol* (2017) **13** 681-696. PMID: 28970584
10. Zhang Z, Zhang L, Zhou L. **Redox signaling and unfolded protein response coordinate cell fate decisions under ER stress**. *Redox Biol* (2019) **25** 101047. PMID: 30470534
11. Shu S, Zhu J, Liu Z. **Endoplasmic reticulum stress is activated in post-ischemic kidneys to promote chronic kidney disease**. *EBioMedicine* (2018) **37** 269-280. PMID: 30314894
12. Yang Y, Yang D, Yang D. **Role of reactive oxygen species-mediated endoplasmic reticulum stress in contrast-induced renal tubular cell apoptosis**. *Nephron Exp Nephrol* (2014) **128** 30-36. PMID: 25359386
13. Pitkin SL, Maguire JJ, Bonner TI. **International union of basic and clinical pharmacology**. *Pharmacol Rev* (2010) **62** 331-342. PMID: 20605969
14. Huang Z, Wu L, Chen L.. **Apelin/APJ system: a novel potential therapy target for kidney disease**. *J Cell Physiol* (2018) **233** 3892-3900. PMID: 28796300
15. Falcao-Pires I, Ladeiras-Lopes R, Leite-Moreira AF.. **The apelinergic system: a promising therapeutic target**. *Expert Opin Ther Targets* (2010) **14** 633-645. PMID: 20350050
16. Chapman FA, Nyimanu D, Maguire JJ. **The therapeutic potential of apelin in kidney disease**. *Nat Rev Nephrol* (2021) **17** 840-853. PMID: 34389827
17. Lu X, Liu S, Luan R. **Serum elabela and apelin levels during different stages of chronic kidney disease**. *Ren Fail* (2020) **42** 667-672. PMID: 32713238
18. Wu F, Qiu J, Fan Y. **Apelin-13 attenuates ER stress-mediated neuronal apoptosis by activating Galphai/Galphaq-CK2 signaling in ischemic stroke**. *Exp Neurol* (2018) **302** 136-144. PMID: 29337146
19. Muller T, Kalea AZ, Marquez A. **Apelinergic system in the kidney: implications for diabetic kidney disease**. *Physiol Rep* (2018) **6** e13939. PMID: 30548130
20. Cheng W, Li XW, Xiao YQ. **Non-coding RNA-associated ceRNA networks in a new contrast-induced acute kidney injury rat model**. *Mol Ther Nucleic Acids* (2019) **17** 102-112. PMID: 31234008
21. Zhao F, Feng LX, Liu Q. **Stanniocalcin-1 alleviates contrast-induced acute kidney injury by regulating mitochondrial quality control via the Nrf2 pathway**. *Oxid Med Cell Longev* (2020) **2020** 1898213. PMID: 32318235
22. Feng LX, Zhao F, Liu Q. **Role of Nrf2 in lipopolysaccharide-induced acute kidney injury: protection by human umbilical cord blood mononuclear cells**. *Oxid Med Cell Longev* (2020) **2020** 1-20
23. Kiss N, Hamar P.. **Histopathological evaluation of contrast-induced acute kidney injury rodent models**. *Biomed Res Int* (2016) **2016** 3763250. PMID: 27975052
24. Dodson M, Castro-Portuguez R, Zhang DD.. **NRF2 plays a critical role in mitigating lipid peroxidation and ferroptosis**. *Redox Biol* (2019) 23
25. Cao AL, Wang L, Chen X. **Ursodeoxycholic acid and 4-phenylbutyrate prevent endoplasmic reticulum stress-induced podocyte apoptosis in diabetic nephropathy**. *Lab Invest* (2016) **96** 610-622. PMID: 26999661
26. Yang L, Xing G, Wang L. **Acute kidney injury in China: a cross-sectional survey**. *Lancet* (2015) **386** 1465-1471. PMID: 26466051
27. Kusirisin P, Chattipakorn SC, Chattipakorn N.. **Contrast-induced nephropathy and oxidative stress: mechanistic insights for better interventional approaches**. *J Transl Med* (2020) **18** 400. PMID: 33081797
28. Li Y, Li Y, Li Y. **Inhibition of endoplasmic reticulum stress mediates the ameliorative effect of apelin on vascular calcification**. *J Mol Cell Cardiol* (2021) **152** 17-28. PMID: 33279504
29. Xu W, Li T, Gao L. **Apelin-13/APJ system attenuates early brain injury via suppression of endoplasmic reticulum stress-associated TXNIP/NLRP3 inflammasome activation and oxidative stress in a AMPK-dependent manner after subarachnoid hemorrhage in rats**. *J Neuroinflammation* (2019) **16** 247. PMID: 31791369
30. Yan M, Shu S, Guo C. **Endoplasmic reticulum stress in ischemic and nephrotoxic acute kidney injury**. *Ann Med* (2018) **50** 381-390. PMID: 29895209
31. Wu CT, Weng TI, Chen LP. **Involvement of caspase-12-dependent apoptotic pathway in ionic radiocontrast urografin-induced renal tubular cell injury**. *Toxicol Appl Pharmacol* (2013) **266** 167-175. PMID: 23116894
32. Peng P, Ma Q, Wang L. **Preconditioning with tauroursodeoxycholic acid protects against contrast-induced HK-2 cell apoptosis by inhibiting endoplasmic reticulum stress**. *Angiology* (2015) **66** 941-949. PMID: 25818106
33. Peng PA, Wang L, Ma Q. **Valsartan protects HK-2 cells from contrast media-induced apoptosis by inhibiting endoplasmic reticulum stress**. *Cell Biol Int* (2015) **39** 1408-1417. PMID: 26248489
34. Eletto D, Chevet E, Argon Y. **Redox controls UPR to control redox**. *J Cell Sci* (2014) **127** 3649-3658. PMID: 25107370
35. Cullinan SB, Diehl JA.. **Coordination of ER and oxidative stress signaling: the PERK/Nrf2 signaling pathway**. *Int J Biochem Cell Biol* (2006) **38** 317-332. PMID: 16290097
36. Verfaillie T, Rubio N, Garg AD. **PERK is required at the ER-mitochondrial contact sites to convey apoptosis after ROS-based ER stress**. *Cell Death Differ* (2012) **19** 1880-1891. PMID: 22705852
37. Wei W, Ma N, Fan X. **The role of Nrf2 in acute kidney injury: novel molecular mechanisms and therapeutic approaches**. *Free Radic Biol Med* (2020) **158** 1-12. PMID: 32663513
38. Xu B, Qin Y, Li D. **Inhibition of PDE4 protects neurons against oxygen-glucose deprivation-induced endoplasmic reticulum stress through activation of the Nrf-2/HO-1 pathway**. *Redox Biol* (2020) **28** 101342. PMID: 31639651
39. Zhang B, Wan S, Liu H. **Naringenin alleviates renal ischemia reperfusion injury by suppressing er stress-induced pyroptosis and apoptosis through activating Nrf2/HO-1 signaling pathway**. *Oxid Med Cell Longev* (2022) **2022** 1-24
40. Digaleh H, Kiaei M, Khodagholi F.. **Nrf2 and Nrf1 signaling and ER stress crosstalk: implication for proteasomal degradation and autophagy**. *Cell Mol Life Sci* (2013) **70** 4681-4694. PMID: 23800989
|
---
title: The association of plasma NT-proBNP level and progression of diabetic kidney
disease
authors:
- Yuancheng Zhao
- Lijun Zhao
- Yiting Wang
- Junlin Zhang
- Honghong Ren
- Rui Zhang
- Yucheng Wu
- Yutong Zou
- Nanwei Tong
- Fang Liu
journal: Renal Failure
year: 2023
pmcid: PMC9970255
doi: 10.1080/0886022X.2022.2158102
license: CC BY 4.0
---
# The association of plasma NT-proBNP level and progression of diabetic kidney disease
## Abstract
### Aims
Diabetic kidney disease (DKD) is the most common cause of end-stage kidney disease (ESKD). The identification of risk factors involved in the progression of DKD to ESKD is expected to result in early detection and appropriate intervention and improve prognosis. This study aimed to explore whether plasma N-terminal pro-B-type natriuretic peptide (NT-proBNP) was associated with kidney outcomes in patients with type 2 diabetes mellitus (T2DM) and biopsy-proven DKD.
### Methods
Patients with biopsy-proven DKD who were followed up at West China Hospital over 12 months were enrolled. The kidney outcome was defined as progression to ESKD. The cutoff value of plasma NT-proBNP concentration was calculated by using receiver operating characteristic (ROC) curve analysis. The influence of NT-proBNP levels on kidney outcome in patients with DKD was assessed using Cox regression analysis.
### Results
A total of 30 ($24.5\%$) patients reached ESKD during a median follow-up of 24.1 months. The baseline serum NT-proBNP level had a significant correlation with baseline proteinuria, kidney function, glomerular lesions, interstitial fibrosis tubular atrophy (IFTA), and arteriolar hyalinosis. Multivariate Cox regression analysis indicated that increased NT-proBNP level was significantly associated with a higher risk of progression to ESKD (HR 6.43; $95\%$ CI (1.65–25.10, $$p \leq 0.007$$), and each 1 SD increase in LG (NT-proBNP) was also associated with a higher risk (HR 2.43; $95\%$ CI 1.94–5.29, $$p \leq 0.047$$) of an adverse kidney outcome after adjusting for confounding factors.
### Conclusions
A higher level of plasma NT-proBNP predicts kidney prognosis in patients with biopsy-proven DKD. This warrants further investigation into the potential mechanisms.
## Introduction
Diabetic kidney disease (DKD) is one of the most serious microvascular complications in patients with diabetes. Centers for Disease Control and Prevention (CDC) showed that approximately $44\%$ of the patients initiating end-stage kidney disease (ESKD) treatment had diabetes ranked as the leading cause of ESKD in the United States in 2017 [1]. In addition, Afkarian et al. [ 2]. observed that DKD carries a 10-year mortality rate of up to $31\%$ in patients with type 2 diabetes mellitus (T2DM). Early detection and better management of DKD in patients with T2DM may delay the progression to ESKD and improve its complications and outcomes. Although renoprotective interventions have been universally implemented to improve glycemia, blood pressure, and serum lipid regulation over the last decades, the risk of ESKD and the health burden in DKD patients are still increasing [3]. Searching for further insight into the pathogenesis and risk factors for DKD development is extremely urgent and essential for the clinical management of DKD.
ProBNP (pro-B-type natriuretic peptide) is secreted by cardiomyocytes in response to stretching and is quickly cleaved into two circulating fragments—the biologically active 32-amino acid C-terminal BNP (B-type natriuretic peptide) and the inert 76-amino acid NT-proBNP—with a 1:1 molar ratio [4]. NT-proBNP is secreted from the ventricular myocardium in response to increased myocyte stress and volume overload. Volume overload is frequently observed in patients with T2DM at high cardio-kidney risk [5]. Recently, the ADVANCE Trial indicated that NT-proBNP may help to identify patients with T2DM who are at greatest risk of microvascular complications, particularly nephropathy [6], and the CRIC Study also showed that NT-proBNP is strongly associated with CKD progression among those with and without diabetes [7]. Diabetes with rapidly worsening kidney disease is often ‘clinically’ labeled as having diabetic kidney disease (DKD), whereas in many cases, they are developing nondiabetic kidney disease (NDKD) or mixed forms (DKD + NDKD). Nondiabetic kidney disease (NDKD) and mixed forms are common in patients with CKD and diabetes (27–$82.9\%$ and 4–$45.5\%$, respectively) [8].
Thus, this study was performed in patients with T2DM and biopsy-proven DKD 1) to explore the relationship between NT-proBNP concentration and pathological changes and 2) to explore whether NT-proBNP as a biomarker could predict kidney prognosis in patients with T2DM and biopsy-proven DKD.
## Patient selection and study design
To explore the relationship between plasma NT-proBNP concentration and the progression of DKD, this study included patients with T2DM and DKD who underwent kidney biopsy from 2010 to 2019 at the West China Hospital of Sichuan University. T2DM was diagnosed according to American Diabetes Association (ADA) criteria [9]. DKD was defined according to the standard published by An et al. [ 10] in 2015 and was diagnosed based on the Kidney Pathology Society (RPS) classification [11]. The indications for a kidney biopsy at our institution were T2DM and kidney damage, especially in T2DM patients without diabetic retinopathy, with sudden onset overt proteinuria, with obvious glomerular hematuria, or with rapidly declining kidney function [12]. Patients with coexisting nondiabetic kidney diseases (such as membranous and IgA nephropathy, systemic lupus erythematosus, ANCA-associated vasculitis), T1DM, acute cardiovascular events (such as hospitalization for heart failure, myocardial infarction, and stroke within three months before kidney biopsy) [13], acute pulmonary embolism [14], acute kidney injury [15], urosepsis [16], progression to ESKD before kidney biopsy, and patients without baseline serum NT-proBNP level information were excluded (Figure 1). All patients provided written informed consent, and this study was approved by the institutional review board at the West China Hospital of Sichuan University.
**Figure 1.:** *Flowchart of study participants.*
## Measurement of plasma NT-proBNP levels
Fasting blood samples were collected from a peripheral vein into tubes containing aprotinin and ethylenediaminetetraacetic acid at the time of kidney biopsy. Plasma samples were stored at -150 °C and thawed just before testing. The plasma NT-proBNP concentration was measured with an electrochemiluminescence immunoassay kit (Roche Diagnostics, Grenzach Wyhlen, Germany) at West China Hospital of Sichuan University [17].
## Clinical and laboratory data collection
Clinical and pathologic data, including age, sex, body mass index (BMI), systolic/diastolic blood pressure, and duration of diabetes, were obtained from electronic medical records at the time of kidney biopsy. Laboratory data at the time of biopsy were also obtained from the medical records. The Chronic Kidney Disease Epidemiology *Collaboration formula* was used to evaluate the estimated glomerular filtration rate (eGFR) [18]. Patient follow-up examinations were performed 2–4 times per year based on the patient’s condition.
## Pathological characteristics
Kidney biopsy samples were prepared for light microscopy, immunofluorescence, and electron microscopy using standard procedures at West China Hospital of Sichuan University. For light microscopy examination, hematoxylin and eosin, periodic acid–Schiff, Masson’s trichrome, and periodic acid–Schiff silver methenamine were used to stain the kidney specimens. The original findings of immunofluorescence microscopy and electron microscopy were used to confirm the diagnosis of DKD. RPS glomerular classifications, interstitial fibrosis tubular atrophy (IFTA), interstitial inflammation, arteriosclerosis, and arteriolar hyalinosis were assessed and scored according to the RPS classification [11]. The pathologists were blinded to the clinical data and kidney outcome. Finally, the overall pathological risk score, the diabetic pathological score (D-score) [19], which is considered a useful DKD pathological scoring system in terms of predicting kidney outcome, was introduced. The D-score was calculated by summing the scores of all kidney pathological characteristics.
## Kidney outcomes
The kidney outcome was defined by the progression to ESKD, which was defined as the need for chronic kidney replacement therapy or eGFR <15 mL/min/1.73 m2 over three months [20]. All patients enrolled in this study were followed until January 2020.
## Statistical analysis
Continuous variables that were normally distributed were expressed as the mean and standard deviation (SD). When continuous variables were not normally distributed, they were expressed as the median and interquartile range (IQR). Categorical variables were expressed as percentages and counts. Differences in continuous variables between patients with different groups of NT-proBNP were analyzed by ANOVA or the Kruskal–Wallis H test, as appropriate. Categorical variables were compared with the chi-squared test. Spearman correlation analysis was used to analyze the association between plasma NT-proBNP levels and clinical-pathological covariates.
Based on the receiver operating characteristic (ROC) curve analysis, the values of plasma NT-proBNP level that best characterized the patients were selected. Survival curves of plasma NT-proBNP levels were obtained by Kaplan–Meier methods with a log-rank test. Univariate and multivariable Cox proportional hazard models were used to estimate the hazard ratios (HRs) for ESKD. We applied multivariable Cox proportional hazard models, including clinicopathological parameters (age, sex, SBP, history of CVD, eGFR, proteinuria, RPS classification, and IFTA). Age and sex were chosen based on biological plausibility. The clinical and pathological covariates were selected as potential confounders because of their significance in univariate analysis or association with ESKD in previous studies [10,21]. Parameters with $p \leq 0.05$ in the adjusted model were considered to be significant predictors of prognosis. All statistical analyses were performed using IBM SPSS statistics (version 23, Chicago, IL, USA). Statistical significance was accepted at $p \leq 0.05.$
## Baseline clinical characteristics according to the plasma NT-proBNP concentrations
The baseline clinical and laboratory characteristics of the patients are displayed in Table 1. The mean age of the patients was 51.2 years, and 82 ($67.2\%$) patients were men. The mean BMI of the patients was 25.5 kg/m2, the mean duration of diabetes was 112.4 months, the mean SBP was 141.6 mmHg, the mean serum albumin was 36.9 g/L, and the mean hemoglobin was 125.3 g/L. The median baseline eGFR was 59.0 mL/min/1.73 m2, the median 24-h proteinuria was 3.0 g/d, and the median HbA1c was $7.3\%$. Thirteen ($10.6\%$) patients had a history of CVD at baseline.
**Table 1.**
| Parameters | All (n = 122) | Group 1 (n = 54) | Group 2 (n = 34) | Group 3 (n = 34) | p |
| --- | --- | --- | --- | --- | --- |
| Age (years) | 51.2 ± 10.6 | 47.9 ± 10.8 | 51.2 ± 8.9 | 56.6 ± 9.9 | 0.001 |
| Body mass index (kg/m2) | 25.5 ± 3.0 | 25.5 ± 2.6 | 25.7 ± 3.3 | 25.3 ± 3.1 | 0.873 |
| Gender (Male, %) | 82 (67.2) | 39 (72.2) | 21 (61.8) | 22 (64.7) | 0.557 |
| Duration of diabetes (months) | 102 (48–168) | 96 (45–159) | 102 (57.5–147.0) | 120 (36–180) | 0.909 |
| Hypertension (%) | 74 (60.6) | 28 (51.9) | 20 (58.8) | 26 (76.5) | 0.064 |
| SBP (mm Hg) | 141.6 ± 25.1 | 136.2 ± 24.8 | 140.8 ± 23.2 | 150.9 ± 25.5 | 0.026 |
| DBP (mm Hg) | 84 (74-92) | 84 (74.0-90.0) | 84 (71.5-93.5) | 84.5 (76.2-95.0) | 0.970 |
| 24-h proteinuria (g/d) | 3.3 (1.2–6.4) | 1.75 (0.5–3.6) | 3.3 (2.1–5.9) | 6.9 (5.1–9.7) | <0.001 |
| Serum creatinine (umol/L) | 112.0 (75.0–144.0) | 86.5 (65.5–125.5) | 102.7 (74.2–141.7) | 141.0 (114.0–164.0) | <0.001 |
| e-GFR (mL/min/1.73 m2) | 59.0 (45.5–94.2) | 83.3 (54.1–109.2) | 67.0 (51.4–94.2) | 43.0 (36.3–52.2) | <0.001 |
| BUN (mg/dl) | 7.4 (5.6–10.0) | 6.5 (5.2–9.4) | 7.3 (5.2–9.5) | 9.0 (7.4–11.9) | 0.002 |
| Serum albumin (g/L) | 36.9 ± 7.1 | 40.5 ± 5.6 | 36.5 ± 6.3 | 31.4 ± 6.3 | <0.001 |
| FBS (mmol/L) | 7.9 (6.2–10.4) | 8.3 (6.3–10.8) | 7.5 (5.5–9.5) | 7.9 (6.5–9.4) | 0.359 |
| HbA1c (%) | 7.3 (6.5–8.6) | 7.5 (6.7–8.2) | 6.9 (6.6–9.0) | 7.1 (6.4–9.4) | 0.588 |
| Triglyceride (mmol/L) | 1.9 (1.4–2.5) | 1.95 (1.37–2.80) | 1.90 (1.40–2.50) | 1.75 (1.27–2.22) | 0.553 |
| Total cholesterol (mmol/L) | 4.8 (4.1–5.9) | 4.60 (3.60–5.45) | 5.15 (4.17–5.95) | 5.2 (4.2–6.7) | 0.125 |
| Uric acid (mmol/L) | 390.4 ± 80.8 | 397.4 ± 92.6 | 400.9 ± 68.5 | 368.7 ± 69.5 | 0.182 |
| HDL cholesterol (mmol/L) | 1.1 (0.9–1.4) | 1.10 (0.90–1.30) | 1.10 (0.90–1.50) | 1.20 (1.10–1.40) | 0.744 |
| LDL cholesterol (mmol/L) | 2.9 ± 1.2 | 2.63 ± 1.05 | 2.83 ± 1.10 | 3.32 ± 1.32 | 0.025 |
| Hemoglobin (g/L) | 125.3 ± 27.1 | 135.8 ± 27.8 | 122.9 ± 25.5 | 110.8 ± 19.9 | <0.001 |
| History of CVD | 13 (10.7) | 1 (1.9) | 2 (5.9) | 10 (29.4) | <0.001 |
| Anti-hypertension drugs | Anti-hypertension drugs | Anti-hypertension drugs | Anti-hypertension drugs | Anti-hypertension drugs | Anti-hypertension drugs |
| α-blockers | 24 (19.7) | 4 (7.4) | 8 (23.5) | 12 (35.3) | 0.005 |
| β-blockers | 34 (27.9) | 6 (11.1) | 8 (23.5) | 20 (58.8) | <0.001 |
| CCB | 61 (50) | 15 (27.8) | 22 (64.7) | 24 (70.6） | <0.001 |
| ACEI | 18 (14.8) | 5 (9.3) | 8 (23.5) | 5 (14.7) | 0.185 |
| ARB | 95 (77.9) | 43 (79.6) | 29 (85.3) | 23 (67.6) | 0.197 |
| Diuretics | 22 (18.0) | 4 (7.4) | 5 (14.7) | 13 (38.2) | 0.001 |
| Lipid-lowering therapy | | | | | |
| Statin | 70 (57.9) | 27 (50) | 22 (64.7) | 21 (63.6) | 0.29 |
| Baseline glucose-lowering therapies | Baseline glucose-lowering therapies | Baseline glucose-lowering therapies | Baseline glucose-lowering therapies | Baseline glucose-lowering therapies | Baseline glucose-lowering therapies |
| Metformin | 47 (38.8) | 30 (55.6) | 11 (32.4) | 6 (18.2) | 0.002 |
| Sulfonylurea | 7 (5.7) | 6 (11.1) | 0 (0) | 1 (2.9) | 0.073 |
| Dipeptidyl peptidase-4 inhibitor | 42 (34.4) | 20 (37) | 12 (35.3) | 10 (29.4) | 0.758 |
| Insulin | 84 (68.9) | 32 (59.3) | 25 (73.5) | 27 (79.4) | 0.109 |
The area under the ROC curve for NT-proBNP at baseline for prediction of progression of diabetic kidney disease was 0.72 [$95\%$ CI (0.618,0.822), $p \leq 0.001$] (Figure 2). A cutoff value in the ROC curve analysis for baseline NT-proBNP was 416.2 pg/ml, which had a sensitivity of $60.0\%$ and a specificity of $81.5\%$ for the prediction of kidney outcome.
**Figure 2.:** *ROC curve analysis of the NT-proBNP level for prediction of progression of diabetic kidney disease.*
All patients in this study were divided into three groups according to the normal level and cutoff value of baseline plasma NT-proBNP concentration: Group 1 (normal level): ≤125 pg/mL ($$n = 54$$); Group 2: 125–416 pg/mL ($$n = 34$$); and Group 3: >416 pg/mL ($$n = 34$$). Compared with patients in group 1, patients in group 2 and group 3 had a lower eGFR slope (Supplement Table 1) and eGFR, hemoglobin, and serum albumin levels; however, they were older and had higher SBP, serum creatinine, BUN levels, and proteinuria. There were no differences in sex distribution, BMI, duration of diabetes, DBP, baseline fasting plasma glucose, HbA1c, serum total cholesterol, uric acid, high-density lipoprotein cholesterol, or low-density lipoprotein cholesterol concentrations among the three groups. The baseline clinical characteristics of the three groups divided by the tertile of plasma NT-proBNP concentrations are shown in Supplement Table 2.
## Baseline pathological characteristics according to the plasma NT-proBNP concentrations
According to the RPS classification, severe glomerular lesions (class III + IV) of group 1, group 2, and group 3 were observed in 18 ($33.3\%$), 23 ($67.7\%$), and 22 ($64.7\%$) patients, respectively. The severe interstitial fibrosis and tubular atrophy (IFTA) scores (scores 2 and 3) of groups 1, 2, and 3 were observed in 19 ($35.2\%$), 16 ($47.1\%$), and 23 ($67.7\%$) patients, respectively. The severe interstitial inflammation scores (score 2) of groups 1, 2, and 3 were observed in 8 ($14.8\%$), 7 ($20.6\%$), and 11 ($32.4\%$) patients, respectively. Among 122 patients, arteriolar hyalinosis was absent in 12 patients ($9.8\%$). At least one area of arteriolar hyalinosis (scored 1) was found in 77 ($63.2\%$) patients. More than one area of arteriolar hyalinosis (scored as 2) was found in 33 patients ($27.0\%$). Regarding arteriosclerosis changes, 16 ($13.1\%$) had no intimal thickening (scored as 0), 59 ($48.4\%$) had intimal thickening less than the thickness of the media (scored as 1), and 47 ($38.5\%$) had severe arteriosclerosis (scored as 2). Compared with patients in group 1, the patients in groups 2 and 3 had severe RPS glomerular classification, IFTA, interstitial inflammation, arteriosclerosis, or arteriolar hyalinosis and had high kidney pathological scores (Table 2).
**Table 2.**
| Pathological lesions | All (n = 122) | Group 1 (n = 54) | Group 2 (n = 34) | Group 3 (n = 34) | Unnamed: 5 | p value* | Correlation coefficient (r) | p value# |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| RPS classification | | | | | | 0.002 | 0.3 | 0.001 |
| I | 10 (8.2) | 10 (18.5) | 0 (0) | 0 (0) | 0.0 | | | |
| Iia | 32 (26.2) | 22 (40.7) | 7 (20.6) | 3 (8.8) | 3.0 | | | |
| Iib | 17 (13.9) | 12 (22.2) | 4 (11.8) | 1 (2.9) | 4.0 | | | |
| III | 47 (38.5) | 10 (18.5) | 19 (55.9) | 18 (52.9) | 6.0 | | | |
| IV | 16 (13.1) | 0 (11.1) | 4 (11.8) | 12 (35.3) | 6.0 | | | |
| IFTA | | | | | | 0.002 | 0.34 | <0.001 |
| 0 | 6 (4.9) | 6 (11.1) | 0 (0) | 0 (0) | 0.0 | | | |
| 1 | 58 (47.5) | 29 (53.7) | 18 (52.9) | 11 (32.4) | 7.0 | | | |
| 2 | 48 (39.3) | 18 (33.3) | 11 (32.4) | 19 (55.9) | 9.0 | | | |
| 3 | 10 (8.2) | 1 (1.9) | 5 (14.7) | 4 (11.8) | 11.0 | | | |
| Interstitial inflammation | | | | | | 0.248 | 0.23 | 0.012 |
| 0 | 3 (2.5) | 2 (3.7) | 1 (2.9) | 0 (0) | 0.0 | | | |
| 1 | 93 (76.2) | 44 (81.5) | 26 (76.5) | 23 (67.6) | 3.0 | | | |
| 2 | 26 (21.3) | 8 (14.8) | 7 (20.6) | 11 (32.4) | 4.0 | | | |
| Arteriolar hyalinosis | | | | | | 0.001 | 0.35 | <0.001 |
| 0 | 12 (9.8) | 11 (20.4) | 1 (2.9) | 0 (0) | 0.0 | | | |
| 1 | 77 (63.2) | 33 (61.1) | 24 (70.6) | 20 (58.8) | 0.0 | | | |
| 2 | 33 (27.0) | 10 (18.5) | 9 (26.5) | 14 (41.2) | 3.0 | | | |
| Arteriosclerosis | | | | | | 0.023 | 0.3 | 0.001 |
| 0 | 16 (13.1) | 13 (24.1) | 2 (5.9) | 1 (2.9) | 0.0 | | | |
| 1 | 59 (48.4) | 24 (44.4) | 18 (52.9) | 17 (50.0) | 0.0 | | | |
| 2 | 47 (38.5) | 17 (31.5) | 14 (41.2) | 16 (47.1) | 1.0 | | | |
| D-Score | | | | | | <0.001 | 0.42 | <0.001 |
| ≤14 | 31 (25.4) | 31 (57.4) | 0 (0) | 0 (0) | | | | |
| 15–18 | 48 (39.3) | 23 (42.6) | 25 (73.5) | 0 (0) | | | | |
| 19–21 | 30 (24.6) | 0 (0) | 9 (26.5) | 21 (61.8) | | | | |
| 22–25 | 13 (10.7) | 0 (0) | 0 (0) | 13 (38.2) | | | | |
## Associations between NT-proBNP concentration and clinical or pathological covariates
Spearman correlation analyses were performed to analyze the association between NT-proBNP and clinical characteristics (Table 3). The results showed that baseline NT-proBNP was significantly associated with 24-h proteinuria and kidney function and moderately associated with age, hypertension, systolic blood pressure, and BUN. Regarding the pathological findings, the baseline plasma NT-proBNP was significantly associated with RPS glomerular classifications, IFTA, and arteriolar hyalinosis and moderately associated with interstitial inflammation and arteriosclerosis.
**Table 3.**
| Parameters | Correlation coefficient (r) | p value |
| --- | --- | --- |
| Age (years) | 0.310 | 0.001 |
| SBP (mm Hg) | 0.264 | 0.003 |
| 24-h proteinuria (g/d) | 0.552 | <0.001 |
| e-GFR (mL/min/1.73 m2) | −0.516 | <0.001 |
| BUN (mmol/L) | 0.381 | <0.001 |
## Risk of progression to ESKD according to baseline plasma NT-proBNP concentrations
A total of 30 ($24.5\%$) patients reached ESKD during a median follow-up of 24.1 months. The percentages of patients in group 1 (normal level), group 2 (high), and group 3 (higher) who progressed to ESKD were 7 ($13\%$), 6 ($17.6\%$), and 17 ($50\%$), respectively. Figure 3 shows the survival curves of groups 1, 2, and 3 by Kaplan–Meier methods. The results showed that the baseline plasma NT-proBNP concentrations were significant for ESKD (log-rank test $p \leq 0.001$). The results of univariable and multivariable Cox proportional hazard analyses are presented in Table 4. Adjusted for sex, age, SBP, history of CVD, baseline eGFR, proteinuria, RPS classification, and IFTA (Model 3), the upper tertile NT-proBNP level group (Group 3) experienced a higher risk of ESKD (HR 6.43; $95\%$ CI 1.65–25.10, $$p \leq 0.007$$). Each 1 SD increase in LG (NT-proBNP) was also associated with a higher risk (HR 2.43; $95\%$ CI 1.94–5.29, $$p \leq 0.047$$) of adverse kidney outcomes after adjusting for the abovementioned confounding factors (Model 3). Supplement Table 3 shows the association among the kidney outcomes and plasma NT-proBNP concentrations that were divided into three groups by the tertile of NT-proBNP concentrations.
**Figure 3.:** *Kaplan–Meier curves of renal survival rate in patients with different plasma NT-proBNP concentrations. Log rank $$p \leq 0.001.$$* TABLE_PLACEHOLDER:Table 4.
## Discussion
In this longitudinal observational analysis of 122 patients with T2DM and biopsy-proven DKD, we found that the baseline plasma NT-proBNP levels were negatively correlated with eGFR and positively correlated with 24-h proteinuria and kidney pathological damage. Compared with patients with normal serum NT-proBNP levels (Group 1), patients in the upper tertile serum NT-proBNP level group (Group 3) had a significantly lower cumulative survival rate (Figure 2) and were associated with a higher risk of subsequent ESKD (HR = 6.43, $95\%$ [CI] 1.65–25.10, $$p \leq 0.007$$), even after adjusting for relevant confounding factors (Table 4). Our observations that plasma NT-proBNP levels are a noninvasive marker associated with adverse kidney outcomes may help nephrologists further manage DKD.
Previous studies [22,23] found a positive correlation between NT-proBNP concentrations and the severity of proteinuria in patients with T2DM. Other studies [24,25] showed that NT-proBNP concentrations were negatively correlated with baseline eGFR and positively correlated with kidney function decline in patients with CKD. This is consistent with the results of our study. A possible reason is that NT-proBNP is only excreted from the kidney [4]. When patients have high NT-proBNP concentrations, it is suggested that the kidney has been impaired, as evidenced by increased proteinuria and decreased eGFR. In addition, some studies [13,26] have shown that NT-proBNP is strongly associated with hypertension and cardiovascular events. In this study, patients in the high NT-proBNP concentration group (Group 3) had a higher baseline SBP and a history of cardiovascular events. Therefore, patients in group 3 would be more likely to use antihypertensive drugs such as alpha-blockers, beta-blockers, CCB, and diuretics.
This study found a correlation between plasma NT-proBNP and kidney pathology changes and severe kidney pathological lesions, including severe glomerular lesions, severe IFTA, severe interstitial inflammation, severe arteriolar hyalinosis, and arteriosclerosis, mainly occurring in patients with a high level of NT-proBNP. BNP and NT-proBNP, mainly secreted from cardiomyocytes at a 1:1 molar ratio, may not only participate in fluid balance by natriuresis and vasodilation but also play an important role against the overactivation of the renin-angiotensin-aldosterone system (RAAS) and sympathetic nervous system through the NPR-A/cGMP/PKG pathway [27]. Additionally, BNP was reported to promote natriuresis and diuresis by relaxing mesangial cells, modulating tubule-glomerular feedback, and water-channel protein aquaporin-2 (AQP2) translocation [27]. It has been considered that the overactivity of the RAAS and sodium and water retention have a central role in the pathogenesis and progression of diabetic kidney disease [28]. In the early stages of DKD progression, BNP can promote natriuresis and dieresis and inhibit the overactivity of the RAAS and central nervous system (SNS) to preserve kidney structure and function. With the further deterioration of kidney damage, however, more BNP would be secreted to counterinteract against the negative impact of overactivity of SNS and RAAS and sodium and water retention in the kidney. Therefore, a high level of BNP or NT-proBNP may indicate that there is severe kidney damage in patients with DKD. However, the underlying mechanism necessitates further research to clarify.
NT-proBNP has already been proven to be a prognostic marker for patients with CKD [29,30], which includes both diabetes-related CKD and nondiabetic CKD. Amanda et al. [ 7] reported that plasma NT-proBNP levels were a risk factor for adverse kidney outcomes (the composite of halving eGFR or initiating kidney replacement therapy) in nondiabetic CKD patients (HR ≥ 1.5) and in diabetic CKD patients (HR ≥ 2.0) among 3379 CKD participants ($47\%$ participants with concomitant diabetes). In addition, another study [30] found that higher levels of NT-proBNP were independently associated with higher rates of ESKD and death among patients with T2DM. After adjusting for baseline eGFR, proteinuria, and other known predictors of CKD progression, such as hemoglobin and albumin, the baseline NT-proBNP remained independently associated with ESKD. All of the above studies suggest that plasma NT-proBNP may be a prognostic marker for diabetes-related CKD patients. However, nondiabetic kidney disease (NDRD) is common (27–$82.9\%$) [8] among patients with diabetes and CKD undergoing kidney biopsy. Thus, our findings in patients with the biopsy-based diagnosis of DKD may be more justified.
In a previous study [31], a machine learning algorithm was used to develop and validate a predictive model for the risk of ESKD in patients with diabetic nephropathy, with a random forest algorithm identifying five major factors: cystatin-C, serum albumin, hemoglobin, 24-h urine urinary total protein, and eGFR (AUC 0.90 and ACC $82.65\%$). Zhao et al. [ 32] previously used machine learning algorithms to develop and validate a predictive model for the risk of kidney failure in patients with DKD, with a nomogram that included five factors: hemoglobin, NLR, serum cystatin C, eGFR, and 24-h urine protein (C-statistic 0.863). However, neither of these two studies had data on baseline plasma BNP or NT-proBNP concentrations. Thus, BNP or NT-proBNP was not available in the final model. In addition, Zhao et al. [ 33] also used machine learning algorithms to develop and validate a predictive model for the risk of ESKD in patients with DKD in another study, with a clinical-pathological model including cystatin C, eGFR, BNP, Log ACR, pathological grade and renin-angiotensin system (C-statistics 0.865). This suggested that BNP may predict the risk of ESKD in patients with DKD and improve the C-statistics of the predictive model. This is consistent with our findings and further shows the predictive role of NT-proBNP on DKD prognosis. This result may indicate that clinicians should take more aggressive and effective measures to prevent the progression of DKD with an increased level of NT-proBNP.
Recently, cardio-kidney syndrome, often defined as a bidirectional association between kidney disease and cardiovascular disease (CVD), has been broadly recognized [34]. There are several plausible mechanisms between cardiovascular and kidney disease involving neurohormonal activation, inflammation, oxidative stress, endothelial dysfunction, and anemia [35]. Given the role of NT-proBNP in terms of the heart and kidney, it may become a common marker for kidney and cardiovascular disease. This warrants further investigation into the role of NT-proBNP in cardio-kidney syndrome.
The KDIGO 2020 clinical practice guidelines for diabetic kidney disease [36] recommend that most patients with T2DM, CKD, and eGFR ≥ 30 mL/min per 1.73 m2 would benefit from treatment with both metformin and a sodium-glucose cotransporter-2 inhibitor (SGLT2i) to reduce the risk of cardiovascular events and kidney outcomes. Glucagon-like peptide-1 receptor agonist (GLP-1 RA) is generally preferred to manage glycemia when SGLT2i are contraindicated, especially in patients with eGFR < 30 mL/min per 1.73 m2. Therefore, in patients with diabetic kidney disease and high plasma NT-proBNP concentrations, especially those with NT-proBNP > 416 pg/mL and eGFR > 30 mL/min per 1.73 m2, almost all patients should be actively treated with SGLT2i to improve their cardio-kidney outcomes. GLP-1 RA should be considered when SGLT2i are contraindicated, especially in patients with NT-proBNP > 416 pg/mL and eGFR < 30 mL/min per 1.73 m2. Unfortunately, the 122 patients in this study were not prescribed SGLT2i and GLP-1 RA, as the drug was not available in China before the completion of this study.
This study, for the first time, revealed that the plasma NT-proBNP level was significantly associated with the progression of DKD. Of course, a few limitations in this study should be noted. First, it was a retrospective cohort study; therefore, selection bias was inevitable. Kidney biopsy is an invasive procedure, resulting in a proportion of patients being reluctant to undergo kidney biopsy, which contributes to the limited sample size and mismatched baseline data between groups in this study. However, the HRs were still significant after important confounding factors were adjusted in multivariable Cox analysis, which shows that the results in this study are still reliable. Second, the sample size was limited. Third, the severity of the patient’s condition varies when performing a kidney pathology biopsy. Fourth, the measurement of plasma NT-proBNP level was only measured once at baseline, and sequential measurements during the follow-up may help to further investigate its association with diabetic ESKD. Fifth, during follow-up, NT-pro-BNP levels did not correlate with HbA1c levels, insulin usage, or the fasting glucose levels of the patients with T2DM or DKD. Finally, we did not control therapeutic interventions (especially antidiabetic drugs with nephroprotective effects) during follow-up, which may be confounders to the results. In summary, our findings provide evidence that NT-proBNP levels can predict kidney prognosis in patients with T2DM and biopsy-proven DKD.
## Conclusion
In summary, our findings provide evidence that NT-ProBNP levels can predict kidney prognosis in patients with T2DM and biopsy-proven DKD.
## Ethics approval and consent to participate
The study protocol was approved by the Institutional Review Board at the West China Hospital of Sichuan University [number 2013R01], and written informed consent was obtained from all participants.
## Consent for publication
All patients provided informed consent.
All patients provided informed consent.
## Authors’ contributions
Fang Liu, Nanwei Tong, and Lijun Zhao designed the experiments. Yuancheng Zhao, Lijun Zhao, Yiting Wang, Junlin Zhang, Rui Zhang, Honghong Ren, Yucheng Wu, and Yutong Zou collected the data. Yuancheng Zhao and Lijun Zhao wrote the main manuscript. Yuancheng Zhao prepared Tables (1–4) and Figures (1–2).
## Disclosure statement
The authors have no conflicts of interest to disclose, and the results in this paper have not been published previously in whole or in part.
## Data availability statement
The datasets generated during and analyzed during the current study are available from the corresponding author upon reasonable request.
## References
1. Saran R, Robinson B, Abbott KC. **US kidney data system 2016 annual data report: epidemiology of kidney disease in the United States**. *Am J Kidney Dis* (2017) **69** A7-A8. PMID: 28236831
2. Afkarian M, Sachs MC, Kestenbaum B. **Kidney disease and increased mortality risk in type 2 diabetes**. *J Am Soc Nephrol* (2013) **24** 302-308. PMID: 23362314
3. Saran R, Li Y, Robinson B. **US kidney data system 2014 annual data report: epidemiology of kidney disease in the United States**. *Am J Kidney Dis* (2015) **66** S1-S305
4. Hall C.. **NT-proBNP: the mechanism behind the marker**. *J Card Fail* (2005) **11** S81-S83. PMID: 15948107
5. Idzerda NMA, Persson F, Pena MJ. **N-terminal pro-brain natriuretic peptide (NT-proBNP) predicts the cardio-kidney response to aliskiren in patients with type 2 diabetes at high kidney and cardiovascular risk**. *Diabetes Obes Metab* (2018) **20** 2899-2904. PMID: 29987919
6. Welsh P, Woodward M, Hillis GS. **Do cardiac biomarkers NT-proBNP and hsTnT predict microvascular events in patients with type 2 diabetes? Results from the ADVANCE trial**. *Diabetes Care* (2014) **37** 2202-2210. PMID: 24879844
7. Anderson AH, Xie D, Wang X. **Novel risk factors for progression of diabetic and nondiabetic CKD: findings from the chronic kidney insufficiency cohort (CRIC) study**. *Am J Kidney Dis* (2021) **77** 56-73. PMID: 32866540
8. Fiorentino M, Bolignano D, Tesar V. **Kidney biopsy in patients with diabetes: a pooled meta-analysis of 48 studies**. *Nephrol Dial Transplant* (2017) **32** 97-110. PMID: 27190327
9. Pinsker JE, Shank T, Dassau E. **Comment on American diabetes association. Approaches to glycemic treatment. Sec. 7. In standards of medical care in diabetes-2015**. *Diabetes Care* (2015) **38** S41-S48. PMID: 25537707
10. An Y, Xu F, Le W. **Kidney histologic changes and the outcome in patients with diabetic nephropathy**. *Nephrol Dial Transplant* (2015) **30** 257-266. PMID: 25063425
11. Tervaert TW, Mooyaart AL, Amann K. **Pathologic classification of diabetic nephropathy**. *J Am Soc Nephrol* (2010) **21** 556-563. PMID: 20167701
12. Li L, Zhang X, Li Z. **Kidney pathological implications in type 2 diabetes mellitus patients with kidney involvement**. *J Diabetes Complications* (2017) **31** 114-121. PMID: 27838100
13. Yancy CW, Jessup M, Bozkurt B. **ACC/AHA/HFSA focused update of the 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Failure Society of America**. *Circulation* (2017) **136** e137-e161. PMID: 28455343
14. Wang J, Wang L, Jin L. **Predictive value of MPV and plasma NT-proBNP combined with the simplified Geneva scale for the prognosis of acute pulmonary embolism**. *Evid Based Complement Alternat Med* (2021) **2021** 1292921. PMID: 34712339
15. Ostermann M, Bellomo R, Burdmann EA. **Controversies in acute kidney injury: conclusions from a kidney disease: improving global outcomes (KDIGO) conference**. *Kidney Int* (2020) **98** 294-309. PMID: 32709292
16. Hu W, Zhou PH, Wang W. **Prognostic value of adrenomedullin and natriuretic peptides in uroseptic patients induced by ureteroscopy**. *Mediators Inflamm* (2016) **2016** 9743198. PMID: 26880865
17. Chen F, Li JQ, Ou YW. **The impact of kidney function on the prognostic value of N-terminal pro-B-type natriuretic peptide in patients with coronary artery disease**. *Cardiol J* (2019) **26** 696-703. PMID: 29611168
18. Michels WM, Grootendorst DC, Verduijn M. **Performance of the Cockcroft-Gault, MDRD, and new CKD-EPI formulas in relation to GFR, age, and body size**. *Clin J Am Soc Nephrol* (2010) **5** 1003-1009. PMID: 20299365
19. Hoshino J, Mise K, Ueno T. **A pathological scoring system to predict kidney outcome in diabetic nephropathy**. *Am J Nephrol* (2015) **41** 337-344. PMID: 26067713
20. Mottl AK, Gasim A, Schober FP. **Segmental sclerosis and extracapillary hypercellularity predict diabetic ESKD**. *J Am Soc Nephrol* (2018) **29** 694-703. PMID: 29180393
21. Colhoun HM, Marcovecchio ML.. **Biomarkers of diabetic kidney disease**. *Diabetologia* (2018) **61** 996-1011. PMID: 29520581
22. Zhou L, Cai X, Li M. **Plasma NT-proBNP is independently associated with albuminuria in type 2 diabetes**. *J Diabetes Complications* (2016) **30** 669-674. PMID: 26922580
23. Seki N, Matsumoto T, Fukazawa M.. **Relationship between the brain natriuretic peptide (BNP) level and prognosis of diabetic nephropathy with microalbuminuria: a 7-year follow-up study**. *Horm Metab Res* (2018) **50** 389-396. PMID: 29723897
24. Saulnier PJ, Gand E, Velho G. **Association of circulating biomarkers (adrenomedullin, TNFR1, and NT-proBNP) with kidney function decline in patients with type 2 diabetes: a French prospective cohort**. *Diabetes Care* (2017) **40** 367-374. PMID: 27998909
25. Bansal N, Katz R, Dalrymple L. **NT-proBNP and troponin T and risk of rapid kidney function decline and incident CKD in elderly adults**. *Clin J Am Soc Nephrol* (2015) **10** 205-214. PMID: 25605700
26. Malachias MVB, Jhund PS, Claggett BL. **NT-proBNP by itself predicts death and cardiovascular events in high-risk patients with type 2 diabetes mellitus**. *J Am Heart Assoc* (2020) **9** e017462. PMID: 32964800
27. Zhang X, Zhou Y, Ma R.. **Potential effects and application prospect of angiotensin receptor-neprilysin inhibitor in diabetic kidney disease**. *J Diabetes Complications* (2022) **36** 108056. PMID: 34893426
28. Thomas MC, Brownlee M, Susztak K. **Diabetic kidney disease**. *Nat Rev Dis Primers* (2015) **1** 15018. PMID: 27188921
29. Tsai YC, Tsai HJ, Lee CS. **The interaction between N-terminal pro-brain natriuretic peptide and fluid status in adverse clinical outcomes of late stages of chronic kidney disease**. *PLOS One* (2018) **13** e0202733. PMID: 30133531
30. Desai AS, Toto R, Jarolim P. **Association between cardiac biomarkers and the development of ESKD in patients with type 2 diabetes mellitus, anemia, and CKD**. *Am J Kidney Dis* (2011) **58** 717-728. PMID: 21820220
31. Zou Y, Zhao L, Zhang J. **Development and internal validation of machine learning algorithms for end-stage kidney disease risk prediction model of people with type 2 diabetes mellitus and diabetic kidney disease**. *Ren Fail* (2022) **44** 562-570. PMID: 35373711
32. Cheng Y, Shang J, Liu D. **Development and validation of a predictive model for the progression of diabetic kidney disease to kidney failure**. *Ren Fail* (2020) **42** 550-559. PMID: 32524865
33. Sun L, Shang J, Xiao J. **Development and validation of a predictive model for end-stage kidney disease risk in patients with diabetic nephropathy confirmed by kidney biopsy**. *PeerJ* (2020) **8** e8499. PMID: 32095345
34. Schefold JC, Filippatos G, Hasenfuss G. **Heart failure and kidney dysfunction: epidemiology, mechanisms and management**. *Nat Rev Nephrol* (2016) **12** 610-623. PMID: 27573728
35. Szymanski MK, de Boer RA, Navis GJ. **Animal models of cardiokidney syndrome: a review**. *Heart Fail Rev* (2012) **17** 411-420. PMID: 21909828
36. **Clinical practice guideline for diabetes management in chronic kidney disease**. *Kidney Int* (2020) **98** S1-S115. PMID: 32998798
|
---
title: COVID-19 mRNA vaccine protects against SARS-CoV-2 Omicron BA.1 infection in
diet-induced obese mice through boosting host innate antiviral responses
authors:
- Yanxia Chen
- Wenchen Song
- Can Li
- Jiaxuan Wang
- Feifei Liu
- Zhanhong Ye
- Peidi Ren
- Yihan Tong
- Junhua Li
- Zhihua Ou
- Andrew Chak-Yiu Lee
- Jian-Piao Cai
- Bosco Ho-Yin Wong
- Jasper Fuk-Woo Chan
- Kwok-Yung Yuen
- Anna Jin-Xia Zhang
- Hin Chu
journal: eBioMedicine
year: 2023
pmcid: PMC9970285
doi: 10.1016/j.ebiom.2023.104485
license: CC BY 4.0
---
# COVID-19 mRNA vaccine protects against SARS-CoV-2 Omicron BA.1 infection in diet-induced obese mice through boosting host innate antiviral responses
## Body
Research in contextEvidence before this studyWe searched PubMed in May 2022, with no starting date limitations, using the terms “SARS-CoV-2” and “Omicron BA.1 or Alpha” and “Diet-induced obese mouse” and “COVID-19 mRNA vaccination” for articles in English. Our search did not reveal any report investigated in SARS-CoV-2 Omicron BA.1 and Alpha infection and COVID-19 mRNA vaccination of wild type diet-induced obese mouse for both in vitro and in vivo studies. Added value of this studyIn this study, we demonstrate that the pathogenicity of SARS-CoV-2 Omicron BA.1 is similar in DIO mice when compared with Alpha, despite results in milder diseases in lean mice. Next, we reveal that DIO mice are more susceptible to SARS-CoV-2 re-infection and are less efficiently protected by COVID-19 mRNA vaccine due to impaired adaptive immune response. However, after two doses of COVID-19 mRNA vaccination, the lower respiratory tract of vaccinated DIO mice is largely protected from Omicron BA.1 infection despite undetectable serum neutralizing antibody in DIO mice. Both in vivo and in vitro studies suggest the mRNA vaccine may protect DIO mice from SARS-CoV-2 infection by improving the host innate immune responses including the type I interferon signaling responses. Implications of all the available evidenceThese findings suggest that obesity increases the susceptibility of SARS-CoV-2 re-infection and vaccine breakthrough infections due to impaired adaptive immune responses. Nevertheless, COVID-19 mRNA vaccination offers partial protection in DIO mice by boosting the host innate immune responses.
## Summary
### Background
Obesity is a worldwide epidemic and is considered a risk factor of severe manifestation of Coronavirus Disease 2019 (COVID-19). The pathogenicity of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and host responses to infection, re-infection, and vaccination in individuals with obesity remain incompletely understood.
### Methods
Using the diet-induced obese (DIO) mouse model, we studied SARS-CoV-2 Alpha- and Omicron BA.1-induced disease manifestations and host immune responses to infection, re-infection, and COVID-19 mRNA vaccination.
### Findings
Unlike in lean mice, Omicron BA.1 and Alpha replicated to comparable levels in the lungs of DIO mice and resulted in similar degree of tissue damages. Importantly, both T cell and B cell mediated adaptive immune responses to SARS-CoV-2 infection or COVID-19 mRNA vaccination are impaired in DIO mice, leading to higher propensity of re-infection and lower vaccine efficacy. However, despite the absence of neutralizing antibody, vaccinated DIO mice are protected from lung damage upon Omicron challenge, accompanied with significantly more IFN-α and IFN-β production in the lung tissue. Lung RNAseq and subsequent experiments indicated that COVID-19 mRNA vaccination in DIO mice boosted antiviral innate immune response, including the expression of IFN-α, when compared to the nonvaccinated controls.
### Interpretation
Our findings suggested that COVID-19 mRNA vaccination enhances host innate antiviral responses in obesity which protect the DIO mice to a certain degree when adaptive immunity is suboptimal.
### Funding
A full list of funding bodies that contributed to this study can be found in the Acknowledgements section.
## Evidence before this study
We searched PubMed in May 2022, with no starting date limitations, using the terms “SARS-CoV-2” and “Omicron BA.1 or Alpha” and “Diet-induced obese mouse” and “COVID-19 mRNA vaccination” for articles in English. Our search did not reveal any report investigated in SARS-CoV-2 Omicron BA.1 and Alpha infection and COVID-19 mRNA vaccination of wild type diet-induced obese mouse for both in vitro and in vivo studies.
## Added value of this study
In this study, we demonstrate that the pathogenicity of SARS-CoV-2 Omicron BA.1 is similar in DIO mice when compared with Alpha, despite results in milder diseases in lean mice. Next, we reveal that DIO mice are more susceptible to SARS-CoV-2 re-infection and are less efficiently protected by COVID-19 mRNA vaccine due to impaired adaptive immune response. However, after two doses of COVID-19 mRNA vaccination, the lower respiratory tract of vaccinated DIO mice is largely protected from Omicron BA.1 infection despite undetectable serum neutralizing antibody in DIO mice. Both in vivo and in vitro studies suggest the mRNA vaccine may protect DIO mice from SARS-CoV-2 infection by improving the host innate immune responses including the type I interferon signaling responses.
## Implications of all the available evidence
These findings suggest that obesity increases the susceptibility of SARS-CoV-2 re-infection and vaccine breakthrough infections due to impaired adaptive immune responses. Nevertheless, COVID-19 mRNA vaccination offers partial protection in DIO mice by boosting the host innate immune responses.
## Introduction
Coronavirus Disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), which has resulted in more than 630 million infections with over 6.5 million deaths.1, 2, 3 Advanced age and comorbidities including hypertension, obesity, certain cancers, and cardiovascular diseases have been suggested as risk factors of developing severe COVID-19.4 Among these conditions, obesity is currently considered a worldwide epidemic and is associated with type II diabetes mellitus, cardiovascular disease, hypertension, and nonalcoholic fatty liver disease due to the state of chronic low-grade systemic inflammation.5 Previous studies reported that diet-induced obese (DIO) mice mounted dysregulated innate immune responses and developed severe lung damage after pandemic influenza H1N1 infection.6, 7, 8 Since the outbreak of COVID-19 pandemic, clinical reports suggested that people with obesity-related conditions are at higher risk of COVID-19 and are more prone to severe illness.9,10 Additionally, a recent study indicated that obesity is associated with delayed SARS-CoV-2 clearance and unfavorable prognosis in COVID-19 patients.11 However, the pathogenicity of SARS-CoV-2 and host responses to re-infection and vaccination in people with obesity remain incompletely understood.12 Since its emergence in 2019, SARS-CoV-2 continues to evolve and generate new SARS-CoV-2 variants of concerns (VOCs) including Alpha, Beta, Gamma, Delta, and Omicron. These VOCs carry key mutations in spike and other viral proteins that collectively modulate their pathogenesis, immune evasiveness, and transmissibility. In late November of 2021, SARS-CoV-2 Omicron BA.1 was first reported in South Africa, which rapidly spread and replaced Delta as the dominantly circulating SARS-CoV-2 variant in early 2022. Omicron BA.1 contains over 30 amino acid changes in spike in comparison to the ancestral SARS-CoV-2, which results in its unique virological features including reduced pathogenicity, elevated transmissibility13, 14, 15, 16, 17 and evasion of vaccine induced immunity.18, 19, 20 Since Omicron and its sublineages are now the predominant SARS-CoV-2 strains, it is important to understand the manifestation of Omicron infection in the obese population. Most importantly, the effectiveness of COVID-19 mRNA vaccination in obese state remains incompletely explored.
In this study, we used DIO mice as an animal model to simulate the manifestation of Alpha and Omicron BA.1 in the context of SARS-CoV-2 pathogenesis, re-infection, and COVID-19 mRNA vaccine-mediated protection in the obese individuals. First, we found that SARS-CoV-2 resulted in more severe disease outcomes in DIO mice than in lean mice. Meanwhile, we found that serum neutralizing antibody responses and viral specific T cell interferon-γ responses induced by either virus infection or COVID-19 mRNA vaccination were severely attenuated in DIO mice, which resulted in higher susceptibility to re-infection and less COVID-19 mRNA vaccine protection of DIO mice when compared with lean mice. Second, although Omicron BA.1 infection caused attenuated diseases in lean mice, it resulted in severe disease in DIO mice similar to that of Alpha. Third, we showed that two-dose of COVID-19 mRNA vaccinations failed to induce serum neutralizing antibody against Omicron BA.1 but was capable of ameliorating Omicron BA.1-induced lung damage in DIO mice. Transcriptomics studies of lung tissues demonstrated that the innate immune antiviral responses in DIO mice were significantly upregulated by vaccination, indicating that the COVID-19 mRNA vaccination offered protection to DIO mice through boosting host innate antiviral immunity. Altogether, our study revealed important knowledge of SARS-CoV-2 Alpha and Omicron BA.1 infection, re-infection, and COVID-19 mRNA vaccination in DIO mice.
## Ethics statement
All experiments involving live SARS-CoV-2 were performed in the Biosafety Level-3 facility at Department of Microbiology, the University of Hong Kong (HKU) according to the standard operating procedures. All the animal experimental procedures were approved by the Committee on the Use of Live Animals in Teaching and Research of the HKU under CULATR 5786-21 and compliance with animal use guidelines.
## Virus and biosafety
SARS-CoV-2 B.1.1.7/Alpha (GISAID: EPI_ISL_1273444) and B.1.1.529.1/Omicron BA.1 (GenBank:OM212469)18 were isolated from laboratory-confirmed COVID-19 patients in Hong Kong. Alpha and Omicron BA.1 were cultured and titrated by plaque assay in VeroE6-TMPRSS2 cells and stored at −80 °C before use.
## Animals
C57BL/6N mice were obtained from the Center of Comparative Medicine Research of HKU and kept in BSL-2 animal laboratory with a 12-h-interval day/night cycle.21 Obesity induction was performed as described previously.6 3 weeks-old newly weaned female mice were randomly divided into 2 groups, one group was fed with 45 Kcal% high-fat diet (D12451, Research Diet Inc, New Brunswick, N. J.) for 20 weeks to induce diet-induced obese (DIO) mice and the control group was fed with standard pellet food containing 13.2 Kcal% diet (PicoLab Rodent Diet 20, LabDiet Code 5053, PMI) as lean mice. Average body weight of DIO mouse used in this study was 40–50 g and the control lean mice group was 25–30 g.
## Cell culture and stimulation
Alveolar macrophages (AMs) were isolated by bronchoalveolar lavage (BAL). Briefly, After the mice were sacrificed by intra-peritoneal injection of pentobarbital, the lungs were flushed with 1 mL cool PBS for three times to obtain 3 mL bronchoalveolar lavage fluid (BALF). BALF from mice in the same group were pooled and centrifuged at 500×g for 10 min at 4 °C to collect precipitated cells, AMs were seeded with a density of 50,000 cells/well in 96-well plates cultured in RPMI 1640 plus $1\%$ penicillin-streptomycin at 37 °C in a moist atmosphere of $5\%$ CO2 for 2 h, then monolayers of AMs could adhere to culture plates, nonadherent cells were washed from plates with PBS and AMs were cultured in RPMI 1640 plus $1\%$ penicillin-streptomycin and $10\%$ fetal bovine serum (FBS) with or without stimulation of 1 μg/mL mRNA vaccine, 100 μg/mL Poly (I:C) or 100 ng/mL spike protein for 24 h.22 Supernatants were collected for ELISA test. Cells were washed thoroughly with PBS and harvested for RNA extraction.
## Alpha and Omicron BA.1 challenge in mice
DIO and lean mice under anesthesia by Ketamine (100 mg/kg) and Xylazine (10 mg/kg)23 were intranasal inoculation with 103 PFUs of Alpha or Omicron BA.1 diluted in 20 μl phosphate buffered saline (PBS), control mice were inoculated with the same volume of PBS. Body weight and symptoms of the infected mice were monitored for 14 days, the symptoms of disease including ruffled fur, hunched posture and labored breathing and one score was given to each sign. Mice were sacrificed by intra-peritoneal injection of pentobarbital at 2 and 4 days post infection (dpi), 3–6 mice in each group were euthanized to harvest blood samples, lung and nasal turbinate tissues for virological, histopathological and immunological assessment. Lung samples from each mouse were collected for the assessments. Each data dot in the figure panel represents result from one mouse.
## Vaccination procedure
Mice were randomly divided into groups and given a two-dose regimen of COVID-19 mRNA vaccination (BNT162b2, lot number 1B004A, BioNTech, Germany) at 14-day interval. The vaccine is based on a nucleoside-modified mRNA formulated in lipid nanoparticles that encodes the ancestral SARS-CoV-2 full-length. DIO and lean mice were intramuscularly injected with 50 μl (5 μg) of COVID-19 mRNA vaccine or normal saline as the control.24,25 At day 14 after the first dose of vaccination, we collected blood samples and then provided the second dose of mRNA vaccine. At 14 days after the second dose, blood samples were collected for assessment of antibody responses. Mice were subsequently intranasally inoculated with 103 PFU of Alpha or Omicron BA.1. At 2 dpi, mice were sacrificed by intra-peritoneal injection of pentobarbital and samples including blood samples, lung and nasal turbinate tissues were collected for virological, histological and immunological assessment. Lung tissues were separated into three parts, the left lung was harvested for histological assessment, the caudal lobe of the right lung was collected as lung homogenate and the rest of right lung for RNA extraction.
## Determination of SARS-CoV-2 gene copies and infectious viral titre in lung and nasal turbinate tissues
Total RNA from lung and nasal turbinate (NT) tissues of mice with Alpha and Omicron BA.1 infection were extracted by the MiniBEST Universal RNA extraction kit (9766, Takara Bio Inc. Shiga, Japan). SARS-CoV-2 RNA-dependent RNA polymerase (RdRp) gene copies were quantified by a QuantiNova Probe RT-PCR Kit (208354, Qiagen). qRT-PCR for detection of SARS-CoV-2 RdRp gene copies and the house-keeping gene β-actin for normalization were performed on a LightCycler 96 system (Roche Applied Sciences, Indianapolis, USA).21 *For infectious* virus titration, homogenates from lung (caudal lobe of the right lung) and NT tissues infected with Alpha and Omicron BA.1 were performed by a $50\%$ tissue culture infectious dose (TCID50) assay in VeroE6-TMPRSS2 cells.26 Homogenates were 10-fold serial dilution and incubated with VeroE6-TMPRSS2 cells in 96-well plate for 1 h at a 37 °C incubator, supernatants were discarded and cells were further incubated in 37 °C for 72 h. Cytopathic effect (CPE) was observed and $50\%$ tissue infectious titre were determined by the Reed & Munch endpoint calculation method.26
## Histopathology, immunohistochemistry, and immunofluorescence staining of lung and nasal turbinate tissue sections
Haematoxylin and Eosin (H&E) staining of formalin-fixed and paraffin-embedding tissue sections (4 μm each) were observed for histopathology changes. Severity of histopathology in lungs were given score under complete masking27 by assessment of pulmonary congestion, interstitial infiltration, alveolar infiltration, hemorrhage and scored 0–4 as described previously.26 The following criteria were used for scoring: 0, normal lung section; 1, blood vessel congestion, perivascular or peribronchiolar infiltration; 2, in addition to 1 with diffuse alveolar wall congestion and infiltration; 3, air space infiltration, exudation, hemorrhage of localized alveolitis; 4, diffuse alveolitis were observed. Immunohistochemistry staining was performed by a DAB (3,3′-diaminobenzidine) substrate kit (Vector Laboratories) as we previously described.28 Briefly, For ACE2 antigen detection, ACE2 recombinant rabbit monoclonal antibody (MA5-32307, Invitrogen) were used and followed with color development by using the DAB substrate kit. The ACE2 protein was detected by haematoxylin and then mounted the tissue sections with the VectaMount permanent mounting medium (Vector Laboratories). For SARS-CoV-2 antigen expression, slides of lung and NT tissues were stained with an in-house antibody of rabbit anti SARS-CoV-2 nucleocapsid protein (NP) followed by a secondary antibody of FITC–conjugated goat anti rabbit IgG (65-6111, Thermo Fisher Scientific, Waltham, MA, USA). The following criteria were used for NP scoring. Lung: “score 0”- no fluorescence staining signal; “score 1”- only in 1–3 bronchiolar epithelium with N antigen positive cells; “score 2”- more than 3 bronchiolar epithelium with N antigen positive cells; “score 3”- *Bronchiolar epithelium* with a few positive cells in nearby alveolar; “score 4”- multiple foci or large area of alveoli with N antigen positive cells. NT: “score 0”- no fluorescence staining signal; “score 1”- a few N antigen positive cells scattered in the epithelium; “score 2”- epithelium showing continually positive N antigen focus in adjacent cells; “score 3”- more N antigen positive of epithelial foci distributed in different area. Images were captured by using microscope of Olympus BX53 semi-motorized fluorescence or bright-field with OLYMPUS CellSense Standard Software.
## RNA isolation and real-time reverse-transcription polymerase chain reaction
Total RNA was extracted from tissue homogenates (the cranial, middle and accessory lobes were harvested for lungs) and reverse-transcription for cDNA was performed by MiniBEST Universal RNA Extraction Kit and RT Reagent Kit (RR036A Takara Bio Inc.) following the manufacturer's instruction. Expression level of cytokines, chemokines, and interferons were detected by qRT-PCR with specific primers (Supplementary Table S1) using a SYBR Premix Ex Taq II Kit (RR820A, Takara Bio Inc.). Values of each gene were normalized with house-keeping gene β-actin and presented as 2−ΔCt as we previously described.28,29
## Enzyme-linked immunosorbent assay (ELISA)
SARS-CoV-2 nucleoprotein (N), spike protein receptor-binding domain (RBD) and inactivated SARS-CoV-2 were coated in 96-well immunoplates (Nunc-Immuno Modules; Nunc A/S, Roskilde, Denmark) in 0.05 M NaHCO3 and incubated at 4 °C, overnight. Serum samples were 2-fold serially diluted and added to the coated plate, incubated at 4 °C for 1 h followed by horseradish peroxidase (HRP)-conjugated secondary antibodies (Rabbit anti-mouse IgG, Goat anti-mouse IgG1, IgG2a, IgG2b, ab6728, ab98693, M32307, M32507, Abcam and Invitrogen) at 37 °C for 1 h. Color development was performed with 3,3′,5,5′-tetramethylbenzidine solution (#N301, Thermo Fisher Scientific) at 37 °C for 15 min and stop with H2SO4. The optical density (OD) values were read at 450 nm. Antibody titres were determined by a cut off OD value which was set at the mean OD of uninfected serum at all dilutions plus 3 standard deviations, and the highest dilution which produces an OD value above the cut off was determined as the antibody titre of serum.30 IFN-α, IFN-β, Albumin and hemoglobin concentrations were determined using a mouse IFN-α (Invitrogen, USA), IFN-β (R&D systems, USA), albumin and hemoglobin (Abcam, Cambridge, UK) ELISA kit following the manufacturer's instructions.
## Microneutralization (MNT) assay
Serum samples were serially diluted 2-fold starting from 1:10 in PBS and mixed with 100 TCID 50 of SARS-CoV-2 for 1 h at 37 °C, the mixture was added into pre-seeded VeroE6-TMPRSS2 cells in 96-well plate at 37 °C for 72 h. CPE was observed and neutralizing antibody titres were determined as the highest dilution of serum that completely inhibited the cytopathic effect.
## Enzyme-linked immunospot (ELISpot) assay
Virus-specific IgG producing cells were detected by seeding single cell (2.5 × 105 cells/per well) suspension of lung and spleen tissues into ELISpot plate with inactivated SARS-CoV-2 (5 μg/mL) at 37 °C for 48 h. IgG-producing cells were determined with alkaline phosphatase (AP) conjugated-goat anti mouse IgG antibody (62-6522, Invitrogen).31 For Virus-specific IFN-γ producing cells detection, 2.5 × 105 cells/per well single cell suspension of lung and spleen tissues were incubated in IFN-γ ELISpot plate stimulating with SARS-CoV-2 RBD peptide pool and N protein at 37 °C for 48 h, IFN-γ producing cells were determined using a mouse IFN-γ ELISpot BASIC kit (3321-2A, Mabtech, Inc., Stockholm, Sweden) following the manufacturer's instructions.32
## RNA sequencing and data analysis
Total RNA from lung tissue cells of DIO and lean mice (n ≥ 3 per group) was isolated using NucleoSpin RNA Kit (740955.250, MACHEREY-NAGEL, Duren, Germany). DNA-depleted and purified RNA was used to construct double-stranded (ds) cDNA library using MGIEasy RNA Library preparation reagent set (MGI, Shenzhen, China) following the standard protocol. Sequencing data were filtered with fastp v0.20.1 to remove adapter and low quality reads.33 Ribosomal RNA (rRNA) reads was filtered with URMAP v1.0.1480 (Edgar, 2020),34 HISAT2 v2.2.0 was used to map the reads against the mouse reference genome (GRCm38/ENSEMBL 84).35 The alignment file was used for assembling transcripts, estimating their abundances, and detecting differential expression of genes, the gene expression levels were quantified by StringTie v2.1.5.36 Principal components analysis (PCA) was conducted with R v4.0. Differentially expressed genes (DEGs) and determined based on gene counts with DESeq2 v3.15.37 DEGs between different treatment groups were identified by clusterProfile with the threshold of |log2FC| > 1 and FDR value < 0.05, and used for enrichment analysis involving Gene Ontology (GO), the Kyoto Encyclopedia of Genes and Genomes Pathway (KEGG).38 *All* genes in the mouse genome were used as the enrichment background. ImmuCellAI_Mouse was used to determine the immune cell abundance based on the RNAseq data.39
## Statistical analysis
Data represented means and standard deviations. Statistical differences between two groups were evaluated with Student's t-test using GraphPad Prism 8. Statistical differences between three or more groups were evaluated with one-way or two-way ANOVA using GraphPad Prism 9. Differences were considered statistically significant when $p \leq 0.05.$ The figures and graphs in the manuscript were prepared with GraphPad Prism 8, Adobe Illustrator, or BioRender.com.
## Reagent validation
Antibodies used in the study have been validated by the commercial source that they are purchased from. Detailed information of reagents used in the study can be found in the Reagent Validation file of Supplemental Data (Supplementary Table S2).
## Role of funders
The funding sources had no role in study design, data collection, analysis or interpretation or writing of the report.
## SARS-CoV-2 Alpha and Omicron BA.1 result in more severe diseases in DIO mice than in lean mice
To understand the pathogenicity of SARS-CoV-2 in the context of obesity, DIO and lean mice were inoculated with 103 PFU of Omicron BA.1 (B.1.1.529.1) or Alpha (B.1.1.7) via the intranasal route. Omicron BA.1 and Alpha carry the N501Y substitution in spike that allow them to infect wild-type mice.40,41 *In a* 14-day disease course, we observed a slow but constant decline of body weight in Alpha-infected lean mice with a mean body weight loss of $5\%$ at 14 dpi (Fig. 1a). In contrast, we observed a mean body weight increase of $1.5\%$ in Omicron BA.1-infected lean mice, indicating that the pathogenicity of Omicron BA.1 in lean mice was significantly attenuated (Fig. 1a), which is in keeping with recent studies from us and others.14,42 However, infection of both Alpha and Omicron BA.1 resulted in more severe body weight loss in diet-induced obese (DIO) mice, which was approximately $12\%$ at 9 dpi when compared with their original body weight ($100\%$) at 0 dpi (Fig. 1a). Interestingly, while there was a clear difference in the mean body weight change between Alpha- and Omicron BA.1-infected lean mice (14 dpi: −$5.0\%$ vs +$1.5\%$), the body weight loss of Alpha- and Omicron BA.1-infected DIO mice were largely the same (14 dpi: −$11.1\%$ vs −$11.4\%$) (Fig. 1a). Consistent with the body weight measurements, Alpha and Omicron BA.1 infection resulted in comparable clinical symptoms in DIO mice, including ruffled fur, hunched back, and labored breathing that peaked at 4 dpi (Fig. 1b), while we did not observe any sign of disease in lean mice upon Alpha or Omicron BA.1 infection. Next, we assessed the histological changes in upper and lower respiratory tissues. Mock-infected mouse nasal turbinate (NT) and lung tissues were shown as control (Supplementary Fig. S1). We observed mild virus-induced epithelium destruction and inflammatory infiltration in NT of both Alpha- and Omicron BA.1-infected lean mice at 2 dpi. The epithelial desquamation and immune infiltration were more dramatic in the NT sections of DIO mice infected with Alpha or Omicron BA.1 at 2 dpi when compared to lean mice (Fig. 1c). At 4 dpi, mild NT epithelium destruction with a few luminal cell debris was detected in Alpha-infected lean mice, while NT of Omicron BA.1-infected lean mice appeared relatively intact. However, we continued to detect severe epithelial destruction, luminal debris and submucosal immune cells infiltration in the NT sections of DIO mice at day 4 after Alpha or Omicron BA.1 infection (Fig. 1c). In the lung tissues of lean mice, Alpha infection resulted in localized interstitial inflammation and mild alveolar capillary congestion at 2 and 4 dpi, while Omicron BA.1 infection resulted in lung interstitial inflammation at 2 dpi but was largely resolved at 4 dpi (Fig. 1d). In sharp contrast, more severe histological damages of alveoli were observed in the lung tissues of DIO mice after both Alpha or Omicron BA.1 infection, which manifested as severe pulmonary blood vessel congestion at 2 dpi. Increased peribronchiolar and peri-vascular immune cell infiltration, as well as immune cells and fluid exudates were observed in alveolar sacs at 4 dpi (Fig. 1d). The concentration of albumin in the bronchoalveolar lavage fluid was significantly higher in DIO mice than in lean mice at day 4 after Alpha infection, suggesting increased alveolar capillary permeability in SARS-CoV-2-infected DIO mice (Supplementary Fig. S2). Consistent with these findings, semi-quantitative histology assessment of lung tissues indicated that infection of both Omicron BA.1 and Alpha resulted in more severe lung histopathology in DIO mice than that in lean mice (Fig. 1e). Overall, these results indicate that SARS-CoV-2 infection results in more severe disease manifestations in DIO mice than in lean mice. Importantly, while Omicron BA.1 is less pathogenic than Alpha in lean mice, the two SARS-CoV-2 variants are more pathogenic and cause similar diseases in DIO mice. Fig. 1SARS-CoV-2 Alpha and Omicron BA.1 cause more severe diseases in DIO mice than in lean mice. Diet-induced obese (DIO) and lean (Ln) mice were intranasally inoculated with 103 PFU of Alpha and Omicron BA.1. Body weight and signs of disease of the infected mice were monitored for 14 days. Lung and nasal turbinate (NT) tissues were collected at day 2 and day 4 post infection ($$n = 6$$ in each group). a Body weight changes in DIO and Ln mice infected with Alpha or Omicron BA.1. b Clinical scores of disease signs upon virus infection. A score of 1 was given to each disease sign (ruffled fur, hunched back and laboured breathing). Highest total score = 3 for each mouse. c and d Representative images of hematoxylin and eosin (H&E) staining for NT sections (c) and lung sections (d) of mice infected with Alpha or Omicron BA.1 at 2 and 4 dpi. The arrows indicated nasal epithelium destruction and detachment into the nasal cavity. e Quantification of histopathological damage in lung sections at 2 and 4 dpi. Images in (c and d) are representative images from 6 mice, $$n = 6$$ in each group. dpi, days post infection; NT, nasal turbinate. Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with one way-ANOVA (e), two-way ANOVA (a and b). ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$ Scale bar in (c and d) represented 100 μm.
## SARS-CoV-2 replicates more efficiently in DIO mice than in lean mice
We next asked whether Omicron BA.1 and Alpha replicate more effectively in the respiratory tissues of DIO mice in comparison to that of lean mice. We found that both Alpha and Omicron BA.1 replicated to higher levels in the lung tissues of DIO mice than in lean mice at 2 and 4 dpi (Fig. 2a and b). In keeping with previous reports, Omicron BA.1 replicated to lower levels than Alpha in the lung tissues of lean mice. In particular, the infectious titre of Omicron BA.1 was 26.1- and 3.8-folds lower than that of Alpha at 2 and 4 dpi, respectively. In contrast, Omicron BA.1 and Alpha replicated to similar levels in the lung tissues of DIO mice by measuring RdRp gene copies and infectious titres (Fig. 2a and b). In addition, we also detected a similar pattern of viral RdRp gene and infectious titre in the NT tissues, where Omicron BA.1 and Alpha replicated to comparable levels in DIO mice but not in lean mice (Fig. 2c and d). By immunofluorescence staining, we detected more prominent viral nucleocapsid (N) expression in the lung and NT tissues of DIO mice infected with Omicron BA.1 or Alpha when compared to lean mice at both 2 and 4 dpi (Fig. 2e and f and Supplementary Fig. S3). In parallel, we quantified the expression of key interferons (IFNs) and pro-inflammatory cytokines in the lung tissues. Our results demonstrated that Omicron BA.1 and Alpha triggered significantly lower levels of IFN-β in DIO mice when compared to their lean counterparts (Fig. 2g and Supplementary Fig. S4). In contrast, pro-inflammatory cytokines including IL-6, and IP-10 were triggered at higher levels in DIO mice by Omicron BA.1 and Alpha infection when compared to lean mice (Fig. 2h). In addition, we evaluated the baseline expression of ACE2 and TMPRSS2 in uninfected lean and DIO mice and found that ACE2 and TMPRSS2 in lean and DIO mice were expressed at comparable levels. However, immunohistochemistry staining performed on ACE2 antigen expression detected higher intensity of ACE2 in the bronchiolar and alveolar epithelium of uninfected DIO mice when compared to lean mice (Supplementary Fig. S5). Together, these findings suggest that the impaired IFN-β expression and higher ACE2 expression in DIO mice may contribute to the higher virus replication and increased pro-inflammatory response, leading to more severe tissue damage in the DIO mice. Fig. 2SARS-CoV-2 replicates more efficiently in DIO mice than in lean mice. DIO and Ln mice were intranasally inoculated with 103 PFU Alpha or Omicron BA.1. Lung and NT tissues were collected at 2 and 4 dpi for virological analysis and qRT-PCR. a–d RNA-dependent RNA polymerase (RdRp) gene copies were quantified by RT-qPCR in lung (a) and nasal turbinate (NT) tissues (c) at 2 or 4 days post infection (dpi) ($$n = 6$$). Dashed lines represent detection limit of the assays. Infectious virus titre in lung (b) and NT tissues (d) were quantified by $50\%$ tissue culture infection dose (TCID50) on VeroE6-TMPRSS2 cells ($$n = 6$$). e and f Representative images of immunofluorescence staining of nucleocapsid (N) protein in lung (e) and NT tissues (f) infected with Alpha or Omicron BA.1 at 2 dpi and 4 dpi. SARS-CoV-2 N was stained in green color and indicated with white arrows. Cell nuclei were stained in blue color with 4′, 6-diamidino-2phenylindole (DAPI). Images in (e and f) were representative images from 6 mice. g and h Inflammatory cytokines, chemokines and interferons in lung homogenates of Ln and DIO mice infected with Alpha or Omicron BA.1 at 2 dpi and 4 dpi were quantified by qRT-PCR ($$n = 6$$). Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with one way-ANOVA (g and h), two-way ANOVA (a–d). ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$ Scale bar in (f) represented 100 μm and (e) represented 200 μm.
## Adaptive immunity acquired from previous SARS-CoV-2 infection inefficiently protects DIO mice
Next, to understand the protective efficiency of adaptive immunity against re-infection in the obese state, convalescent DIO mice and lean mice were re-challenged with 103 PFU of Alpha at day 28 after primary infection and samples were harvested at 2 days post re-infection (dpr) (Fig. 3a). We analyzed the IgG antibody against SARS-CoV-2 antigen in the serum samples taken at day 14 after primary Alpha infection and found that the titre of IgG against SARS-CoV-2 N and spike receptor-binding domain (RBD) were significantly lower in DIO mice than that of lean mice by 1.9- ($$p \leq 0.0128$$) and 3.4-folds ($$p \leq 0.0362$$), respectively (Fig. 3b). In addition, the titre of viral binding total IgG and subtype IgG1, IgG2a, and IgG2b were all significantly lower in DIO mice when compared with lean mice (Fig. 3b). Serum neutralization titre of 1:10–1:20 were detected from 5 out of 6 lean mice but was not detected from any of the 6 DIO mice (Fig. 3b), which suggested an impaired adaptive immune response in DIO mice upon SARS-CoV-2 infection. At 2dpr, viral RdRp gene copy was readily detected from all NT tissues and $\frac{3}{6}$ ($50\%$) of lung tissues of re-challenged DIO mice (Fig. 3c), whereas no viral RdRp gene copy was retrieved from the NT and lung tissues of re-challenged lean mice. Consistently, we frequently detected viral N protein in the epithelium of NT and lung tissues in DIO mice but not lean mice (Fig. 3d). Histological examinations revealed epithelial destruction in NT tissues as well as alveolar wall congestion, peribronchiolar infiltration, and localized alveolar hemorrhage in lung tissues of DIO mice but not lean mice (Fig. 3e). Next, we evaluated the recall of immune memory responses upon re-infection in lean and DIO mice, we found that virus-specific IFN-γ and IgG producing cells in both lung and spleen tissues of DIO mice were significantly lower than that of lean mice at 2dpr (Fig. 3f and g). Importantly, we were not able to detect any neutralizing antibody titre against Alpha in the serum of DIO mice at day 2 upon re-infection of Alpha, while $\frac{4}{6}$ ($66.7\%$) of re-infected lean mice had a neutralizing titre of 1:10 (Fig. 3h). Thus, our findings suggest that DIO mice are more susceptible to re-infections due to the insufficiently mounted adaptive B cell and T cell immune responses upon SARS-CoV-2 infection. Fig. 3Adaptive immunity acquired from previous SARS-CoV-2 infection inefficiently protects DIO mice. a Schematic of primary infection and re-infection of Alpha. After primary infection with 103 PFU of Alpha, mice were re-challenged with the same dose of Alpha at 28 dpi. Blood samples were collected for antibody detection at 14 dpi. Blood samples, spleen, lung and NT tissues at 2 days post re-infection (dpr) were harvested for virological, histological and immunological analysis. b *Mouse serum* of IgG against RBD and N, total IgG, viral binding IgG, and IgG subtype (IgG1, IgG1a, IgG1b) at 14 dpi were determined by ELISA. Neutralizing antibody was detected by microneutralization assay ($$n = 6$$). c RdRp gene copies of lung and NT tissues at 2 dpr were quantified by probe-specific RT-qPCR ($$n = 6$$). Dashed lines represent detection limit of the assays. d Representative images of immunofluorescence staining of nucleocapsid protein in NT (left panel) and lung (right panel) tissues at 2 dpr. Squared area were magnified. SARS-CoV-2 N was stained in green color, cell nuclei were stained in blue color with 4′, 6-diamidino-2 phenylindole (DAPI) e Representative images of H&E staining for detection of pathological damage of NT (left panel) and lung (right panel) tissues at 2 dpr. Images in (d and e) are representative images from 6 mice. f Virus-specific IFN-γ producing cells in lung and spleen tissues at 2 dpr detected by ELISpot assay ($$n = 3$$). g Virus-specific IgG-producing cells were detected by ELISpot assay ($$n = 3$$). h Serum neutralizing antibody at 2 dpr was detected by microneutralization assay ($$n = 6$$). Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with two-tail Student's t-test (b, f and g) and two-way ANOVA (c). ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$ Scale bar in (d and e) represented 100 μm.
## COVID-19 mRNA vaccination offers less protection against Alpha infection in DIO mice due to attenuated adaptive immune response
To explore the efficiency of protection for DIO mice after immunization with mRNA vaccines, we intramuscularly injected DIO and lean mice with a two-dose COVID-19 mRNA vaccination regimen followed by Alpha challenge at 14 days post boost as illustrated in Fig. 4a. For NT tissues, we detected viral RdRp gene in $\frac{3}{6}$ ($50\%$) and infectious titre in $\frac{2}{6}$ ($33.3\%$) of vaccinated lean mice, while we detected viral RdRp gene and infectious titre in all vaccinated DIO mice (Fig. 4b), suggesting that breakthrough infection post-vaccination occurred significantly more readily in DIO mice in comparison to the lean mice. In keeping with previous reports that suggested vaccinations offer better protection of the lower respiratory tract when compared to the upper respiratory tract in human and animal models,43,44 we did not detect viral gene copy or infectious virus from any of the lung tissues of vaccinated lean mice (Fig. 4c). In contrast, we detected viral RdRp gene in $\frac{3}{6}$ ($50\%$) and infectious titre in $\frac{4}{6}$ ($66.7\%$) of lung tissues of DIO mice. Histologically, we observed extensive epithelium destruction in NT tissues of vaccinated DIO mice after challenge by Alpha virus, while a lesser degree of NT tissues damage was observed in vaccinated lean mice (Fig. 4d). In the lung tissues of vaccinated DIO mice, we detected congestion and infiltration in alveolar space accompanied by foci of alveolar hemorrhage upon Alpha infection, while these histopathological damages were largely absent in lung tissues of vaccinated lean mice (Fig. 4e). In line with these findings, COVID-19 mRNA vaccination reduced SARS-CoV-2 N expression in both lean and DIO mice in the NT tissues (Fig. 4f). In lung tissues, viral N signals were only detected in vaccinated DIO mice but not vaccinated lean mice (Fig. 4g). Next, we found that the virus-specific IFN-γ and IgG producing cells in the spleen tissues of vaccinated DIO mice were significantly lower when compared with that in vaccinated lean mice at day 2 after Alpha infection (Fig. 4h). Importantly, microneutralization assay only detected neutralizing antibody against Alpha in the serum of DIO mice at 14 days after second dose of vaccination and 2 days after virus challenging, but the level was 2.8-folds ($p \leq 0.0001$) and 21.8-folds ($$p \leq 0.0053$$) lower than that of the lean mice, respectively (Fig. 4i). Collectively, these results indicate that adaptive antibody responses to COVID-19 mRNA vaccination are impaired in DIO mice. Fig. 4COVID-19 mRNA vaccination offers less protection against Alpha infection in DIO mice due to attenuated adaptive immune response. a Schematic of vaccination schedule and virus infection of Alpha. DIO and Ln mice were intramuscularly vaccinated with two doses of COVID-19 mRNA vaccine (5 μg of antigen per mouse) or normal saline as control at a 14-day interval. Blood was collected at 14 and 28 days after primary vaccination. Mice were intranasally inoculated with 103 PFU of Alpha at 14 days post secondary vaccination. Blood samples, lung, NT and spleen tissues were harvested at 2 dpi for immunological, virological and histological analysis. b and c RdRp gene copies of NT (b) and lung (c) tissues at 2 dpi were quantified by RT-qPCR and infectious titre of NT (b) and lung (c) tissues were determined by TCID50 assay ($$n = 6$$). Dashed lines represent detection limit of the assays. d and e Representative images of H&E staining of NT (d) and lung (e) tissues at 2 dpi. f and g Representative images of immunofluorescence staining of N protein in NT (f) and Lung (g) tissues at 2 dpi. Images were representative images from 6 mice. SARS-CoV-2 N was stained in green color and indicated with white arrows, cell nuclei were stained in blue color with 4′, 6-diamidino-2 phenylindole (DAPI). h Virus-specific IFN-γ (left panel) and virus-specific IgG (right panel) producing cells of spleen tissues were detected by ELISpot assay ($$n = 3$$). i Serum neutralizing antibody responses against Alpha at 14, 28 days after primary vaccination and 2 dpi were determined by microneutralization assay. ( $$n = 12$$ for 14, 28 days after primary vaccination, $$n = 6$$ for 2 dpi). Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with two-tail Student's t-test (i) or one-way ANOVA (b, c, and h). ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$ Scale bar in (d and f) represented 100 μm, in (e and g) represented 200 μm.
## COVID-19 mRNA vaccination ameliorates lung damage caused by Omicron BA.1 challenge in DIO mice in the absence of detectable neutralizing antibody
Next, we asked to what extend COVID-19 mRNA vaccination protects DIO mice against SARS-CoV-2 Omicron BA.1. To this end, we first tested serum neutralizing antibody titre against Omicron BA.1, and detected low titre in lean mice at 14 days post boost, while no neutralizing antibody titre was detected in DIO mice even after the two doses of vaccination COVID-19 mRNA boost or 2 days after Omicron BA.1 infection (Fig. 5a and b), suggesting vaccination in DIO mice induced little cross antibody against Omicron BA.1. Next, DIO and lean mice were challenged with 103 PFU of Omicron BA.1 via the intranasal route at 14 days post boost vaccination (Fig. 5a). In the NT tissues, COVID-19 mRNA vaccination modestly reduced Omicron BA.1 replication and infectious titre in lean mice but had little effect in DIO mice comparing to unvaccinated controls (Fig. 5c). Interestingly, COVID-19 mRNA vaccination significantly reduced Omicron BA.1 replication in the lung tissues of DIO mice comparing to their unvaccinated controls, though not as effective as the complete inhibition of Omicron BA.1 in the lung tissues of vaccinated lean mice (Fig. 5d). Specifically, the Omicron BA.1 RdRp gene copy was reduced by 867-folds ($$p \leq 0.0087$$) and infectious titre by 7979.9-folds ($p \leq 0.0001$) in vaccinated DIO mice. Consistent with the above virological findings, COVID-19 mRNA vaccination was less effective in reducing Omicron BA.1 antigen expression in NT tissues of lean and DIO mice but reduced Omicron BA.1 antigen expression in lung tissues of lean and DIO mice (Fig. 5e). At 2 dpi, Omicron BA.1 infection resulted in certain degree of congestion and infiltration in alveolar space accompanied with pulmonary hemorrhage and epithelium damage in unvaccinated DIO mice, which were reduced by COVID-19 mRNA vaccination (Fig. 5f). Importantly, we found that the concentration of IFN-α and IFN-β in the lung tissues of vaccinated DIO mice were significantly augmented comparing to the unvaccinated mice at 2 dpi after Omicron BA.1 challenge, while serum neutralizing antibody remained undetectable from the vaccinated DIO mice (Fig. 5g). The overall disease severity by body weight changes showed that COVID-19 mRNA vaccination reduced the maximum body weight loss in DIO mice to $9\%$ when compared with $13\%$ in unvaccinated DIO mice (Supplementary Fig. S6). Taken together, these results indicate that despite triggering an undetectable level of antibody response, COVID-19 mRNA vaccination may enhance the innate antiviral responses to ameliorates Omicron BA.1-induced lung tissues damage in DIO mice. Fig. 5COVID-19 mRNA vaccine ameliorates lung damage in the absence of neutralizing antibody response in Omicron BA.1-infected DIO mice.a Schematic of vaccine schedule and virus infection of Omicron BA.1. DIO and Ln mice were intramuscularly vaccinated with two doses of COVID-19 mRNA vaccine (5 μg of antigen per mouse) or normal saline as control at a 14-day interval. Blood samples were collected at day 14 and day 28 after primary vaccination. Mice were intranasally inoculated with 103 PFU of Omicron BA.1 at day 14 post secondary vaccination. Blood samples, lung, NT tissues were harvested at 2 dpi for immunological, virological and histological analysis. b Serum neutralizing antibody responses of DIO and Ln mice against Omicron BA.1 at 14, 28 days after primary vaccination were determined by microneutralization assay ($$n = 12$$ for 14 and 28 days after primary vaccination). c and d RdRp gene copies of NT (c) and lung (d) tissues at 2 dpi were quantified by RT-qPCR and infectious titre of NT (c) and lung (d) tissues were determined by TCID50 assay ($$n = 6$$). Dashed lines represent detection limit of the assays. e Representative images of immunofluorescence staining of N protein in NT and Lung tissues at 2 dpi. SARS-CoV-2 N was stained in green color and indicated with white arrows, cell nuclei were stained in blue color with 4′, 6-diamidino-2 phenylindole (DAPI). f Representative images of H&E staining of NT and lung tissues at 2 dpi. Images in e and f were representative images from 6 mice. g Protein concentrations of IFN-α and IFN-β in lung homogenate at 2 dpi were determined by ELISA. Serum neutralizing antibody responses against Omicron BA.1 of vaccinated DIO and Ln mice at 2 dpi were determined by microneutralization assay ($$n = 6$$). Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with two-tail Student's t-test (b and g) and one-way ANOVA (c and d). ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$ Scale bar in (e and f of NT) represented 100 μm, in (e and f of Lung) represented 200 μm.
## COVID-19 mRNA vaccination upregulates antiviral responses in lungs of DIO mice
To better understand the immune responses in the lung tissues of DIO mice with COVID-19 mRNA vaccination, we harvested mice lung samples and explored their transcriptome profiles at day 2 upon Omicron BA.1 infection (Fig. 6a and Supplementary Fig. S7). Intriguingly, we observed a notable enrichment of genes involved in pathways related to innate antiviral responses, including “cellular response to IFN-α and IFN-β” and “positive regulation of IFN-α and IFN-β production” in the lung tissues of vaccinated DIO mice (Fig. 6b). Analysis of immune gene sets further revealed that the vaccinated DIO mice had the highest abundance of M1 macrophages among all evaluated groups (Fig. 6c).Fig. 6COVID-19 mRNA vaccination upregulates antiviral responses in lungs of DIO mice. a Schematic of RNA sequencing experiment design: DIO and Ln mice were intramuscularly vaccinated with two doses of COVID-19 mRNA vaccine (5 μg of antigen per mouse) or normal saline as control at a 14-day interval. After two doses of vaccination, mice were intranasally inoculated with 103 PFU of Omicron BA.1 and lung tissues were harvested from unvaccinated or vaccinated DIO and Ln mice at 2 dpi. b Differentially expressed genes (DEGs) in pathways related to the cellular responses to IFN-β/α (GO:0035458 and GO:0035457) and positive regulation of IFN-β/α production (GO:0032728 and GO:0032727) identified by R packages clusterProfile and used for enrichment analysis involving Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes Pathway (KEGG). c Estimation of Immune cell type enrichment by ImmuCellAI Mouse ($$n = 5$$ in control unvaccinated group and $$n = 6$$ in vaccinated group). Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with two-way ANOVA. ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$
To evaluate if COVID-19 mRNA vaccine boosted antiviral interferon responses in alveolar macrophage, we isolated alveolar macrophages from lean mice and DIO mice and stimulated with mRNA vaccine (Fig. 7a). We found that mRNA vaccine in vitro stimulation of mouse alveolar macrophages induced significant upregulation of RIG-I, MDA5, STAT1, STAT2, ISG15, IFIT3, and OAS3, while only the alveolar macrophages from DIO mice showed significant enhancement of IFN-α production on mRNA and protein levels (Fig. 7b and c). These findings suggested a differential response profile of alveolar macrophage from DIO mice and lean mice. Next, we isolated mouse alveolar macrophages from vaccinated or unvaccinated lean and DIO mice, followed by stimulating the cells with poly (I:C) or SARS-CoV-2 spike protein (Fig. 7d). In vitro stimulation with poly (I:C) further increased the expression of innate immune response related genes in alveolar macrophages from vaccinated lean and DIO mice. Interestingly, the interferon-stimulated genes including ISG15, IFIT3, and OAS3 were more highly upregulated in vaccine-primed alveolar macrophages from DIO mice when compared to those from lean mice (Fig. 7e). In parallel, recombinant SARS-CoV-2 spike protein stimulation increased the expression of RIG-I, TLR3, IL-6, TNF-α, IFN-α, IFN-β, IRF7, and STAT2 mRNA expression in vaccine-primed alveolar macrophages of DIO mice but not lean mice (Fig. 7g). Importantly, IFN-α production was dramatically increased in alveolar macrophages of vaccinated DIO mice in response to poly (I:C) or S protein stimulation (Fig. 7f and h). Taken together, these results indicate that COVID-19 mRNA vaccination may restore the innate antiviral response in DIO mice lung through modulating the expression of type-I interferon related genes in alveolar macrophages. Fig. 7Alveolar macrophages (AMs) of DIO mice contribute to the upregulated antiviral responses. a Schematic of AMs from clean Ln and DIO mice stimulation with or without mRNA vaccine. Bronchoalveolar lavage fluid (BALF) from clean Ln and DIO mice were collected by intratracheal instillation with cool PBS, adherent AMs were stimulated with or without mRNA vaccine. b Gene expression levels of AMs from clean Ln and DIO mice stimulated with 1 μg/mL mRNA vaccine were quantified by qRT-PCR ($$n = 3$$). c Protein concentrations of IFN-α in AMs (from clean Ln and DIO mice) supernatants stimulated with 1 μg/mL mRNA vaccine were determined by ELISA ($$n = 3$$). d Schematic of AMs from unvaccinated or vaccinated Ln and DIO mice stimulation with poly (I:C) or SARS-CoV-2 spike protein. After two doses of COVID-19 mRNA vaccine or PBS, BALF from Ln and DIO mice were collected by intratracheal instillation with cool PBS, adherent AMs were stimulated with poly (I:C) or SARS-CoV-2 spike protein. e Gene expression levels of AMs from vaccinated or unvaccinated Ln and DIO mice stimulated with 100 μg/mL Poly (I:C) were quantified by qRT-PCR ($$n = 3$$). f Protein concentrations of IFN-α in AMs (from vaccinated or unvaccinated Ln and DIO mice) supernatant stimulated with 100 μg/mL Poly (I:C) were determined by ELISA ($$n = 3$$). g Gene expression levels of AMs from vaccinated or unvaccinated Ln and DIO mice stimulated with 100 ng/mL spike protein were quantified by qRT-PCR ($$n = 3$$). h Protein concentrations of IFN-α in AMs (from vaccinated or unvaccinated Ln and DIO mice) supernatant stimulated with 100 ng/mL spike protein were determined by ELISA ($$n = 3$$). Data represented means and standard deviations from the indicated number of biological repeats. Statistical significance between groups was determined with one-way ANOVA. ∗ represented $p \leq 0.05$, ∗∗ represented $p \leq 0.01$, ∗∗∗ represented $p \leq 0.001$, ∗∗∗∗ represented $p \leq 0.0001.$
## Discussion
Obesity is associated with exacerbated inflammation responses and disruption of immune system that consequently results in a high risk of severity and hospital admission for COVID-19 patients.45 It remains unclear how these chronic low-grade inflammation and dysfunction of immune responses modulate the course of SARS-CoV-2 infection. To explore the pathogenesis of SARS-CoV-2 and assess the efficiency of COVID-19 mRNA vaccination in obese individuals, appropriate small animal models are urgently needed to mimic the clinical features of COVID-19 patients. However, ancestral SARS-CoV-2 does not infect wild-type laboratory mouse since RBD of ancestral spike does not efficiently bind to mouse ACE2.21,40,41 Interestingly, recently emerged SARS-CoV-2 variants, including Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), and Omicron BA.1 (B.1.1.529.1), have acquired the N501Y mutation in spike, which enables cross-species transmission of SARS-CoV-2 to wild-type murines.41,46, 47, 48, 49, 50 *In this* study, we use a diet-induced obese wild-type mouse model to investigate SARS-CoV-2 Alpha and Omicron BA.1 pathogenesis, re-infection, and vaccination. Our results demonstrated that Omicron BA.1 infection is less severe than that of Alpha in lean mice according to the bodyweight change, viral load, virus titre, lung pro-inflammatory markers, and histopathology findings. These results are in accordance with recent reports that observed a lower level of infection by Omicron BA.1 compared with previous SARS-CoV-2 variants in animal models.14,42 The increased severity of SARS-CoV-2 infection in DIO mice than in lean mice is in line with results from recent studies performed in diet-induced obese mice and hamsters.51, 52, 53 Interestingly, we revealed that Omicron BA.1 replicated to similar level compared to Alpha in DIO mice, resulting in comparable pathogenicity of Omicron BA.1 and Alpha in DIO mice. In particular, representative pro-inflammatory cytokines including IP-10 and IL-6 were robustly triggered by Omicron BA.1 infection in DIO mice, indicating increased inflammatory damage and morbidity.54 Meanwhile, production of antiviral IFN in lung tissues of DIO mice was significantly lower than that of lean mice upon both Alpha and Omicron BA.1 infection. These results suggest that Omicron BA.1 infection in the DIO mice can cause severe disease similar to that of Alpha. It is currently incompletely understood how obesity contributes to severe diseases in COVID-19. The increased leptin level is a hallmark of obesity that regulates metabolism, Jak/STAT and Akt signaling pathways, and modulates T cell functions, which may contribute to the severe outcome.55,56 Antibody-mediated humoral immunity is essential for host defenses,57, 58, 59 while optimal virus clearance in SARS-CoV-2-infected mice also requires CD4+ and CD8+ T cell responses.60 *In this* study, we detected significantly lower levels of B cell and T cell responses from DIO mice when compared with lean mice, which may contribute to the increased susceptibility of re-infection in DIO mice. The underlying mechanism of the reduced B cell and T cell response in DIO mice remains incompletely explored, but may be associated with impairment of dendritic cell functionality as revealed in previous influenza virus studies.61, 62, 63 Despite the success of COVID-19 mRNA vaccination with high efficacy to protect against SARS-CoV-2 severe diseases, the emergence of immune evasiveness in SARS-CoV-2 variants may jeopardize vaccine efficacy.43 In our DIO mouse model with a two-dose immunization regimen, the neutralizing antibody in lean mice were readily detected and robustly boosted by the second dose of vaccination against Alpha and Omicron BA.1. In contrast, the two-dose immunization regimen of COVID-19 mRNA vaccination resulted in undetectable serum neutralizing antibody against both Alpha and Omicron BA.1 in DIO mice, and no neutralizing antibody was detected against Omicron BA.1 even at day 2 upon Omicron BA.1 challenge. These findings indicate that the adaptive B cell responses to mRNA vaccine are severely weakened in DIO mice, which may be due to impairments in B cell development, activation, and functions.64, 65, 66 Interestingly, vaccination offered a certain degree of protection to Omicron BA.1-challenged DIO mice in the lower respiratory but not in the nasal turbinates of infected animals. The insufficient protection in the nasal turbinate tissues may be due to insufficient mucosal immune responses upon intra-muscular vaccination that failed to evoke sufficient protective immunity at the mucosal sites.67,68 Through transcriptomic analysis, we revealed that the COVID-19 mRNA vaccination upregulated antiviral responses in the lungs of DIO mice. Subsequent experiments revealed that IFN-α production in responses to poly (I:C) or SARS-CoV-2 spike protein stimulation was augmented in the alveolar macrophages of vaccinated DIO mice. Together with the reduced viral load and histopathological changes, and increased concentration of IFN-α/β in the lungs of Omicron BA.1-challenged vaccinated DIO mice, our findings suggest that mRNA vaccination boosts the host innate immunity in the obese animals that contributes to the observed protection. Overall, our study reveals important knowledge of SARS-CoV-2 Alpha and Omicron BA.1 infection, re-infection, and vaccination in diet-induced obese mice, which provides insights into the management, treatment, and vaccination strategies for the obese populations.
Our study has a number of limitations. First, we performed re-challenge at 28 days post primary infection. In real-life scenarios, memory longevity and protection of re-infection in patients should be monitored over a longer time frame. Second, female mice were used in this study since females are known to mount stronger innate and adaptive immune responses than males.69 The differential immune responses against SARS-CoV-2 between sexes and their roles in disease development of COVID-19 should be further explored. Third, we only used the wild-type mice model for this study, additional evaluation in humans should be performed in future clinical studies to confirm and further explore our findings on whether the COVID-19 mRNA vaccination improves the innate immune responses upon SARS-CoV-2 infection in the vaccinated obese human patients. Nevertheless, our study provides important insights that suggest obese patients may develop more severe clinical diseases upon SARS-CoV-2 infection and are more susceptible to re-infections or vaccine-breakthrough infections. Importantly, our data suggests COVID-19 mRNA vaccines may remain effective in the obese individuals despite low or absence of antibody response, further emphasizing the importance of vaccination in these populations.
## Contributors
AJZ and HC had roles in the study design, data collection, data analysis, data interpretation, writing of the manuscript, supervision, and funding. YXC, WCS, CL, JXW, FFL, ZHY, PDR, YHT, JHL, ZHO, ACYL, PJC, and BHYW had roles in the experiments, data collection, and/or data analysis, JFWC and KYY had roles in experimental protocols and revision of the manuscript. All authors read and approved the final version of the manuscript. YXC, AJZ, and HC have verified the underlying data.
## Materials and correspondence
Correspondence and material requests should be addressed to Dr. Hin Chu or Dr. Anna Jin-Xia Zhang.
## Data sharing statement
The data that support the findings of this study are available from the corresponding authors upon reasonable request. Transcriptomes data is available in CNSA (CNGB Sequence Archive) of CNGBdb under project number CNP0003180 (https://db.cngb.org/cnsa/).
## Declaration of interests
J.F.-W.C has received travel grants from Pfizer Corporation Hong Kong and Astellas Pharma Hong Kong Corporation Limited, and was an invited speaker for Gilead Sciences Hong Kong Limited and Luminex Corporation. K.Y.Y. is the inventor of an intranasal influenza virus-vectored vaccine for SARS-CoV-2.
All other authors declare no competing interests.
## Supplementary data
Supplementary Tables and Figures
## References
1. Zhou P., Yang X.L., Wang X.G.. **A pneumonia outbreak associated with a new coronavirus of probable bat origin**. *Nature* (2020) **579** 270-273. PMID: 32015507
2. Chu H., Chan J.F., Yuen T.T.. **Comparative tropism, replication kinetics, and cell damage profiling of SARS-CoV-2 and SARS-CoV with implications for clinical manifestations, transmissibility, and laboratory studies of COVID-19: an observational study**. *Lancet Microbe* (2020) **1** e14-e23. PMID: 32835326
3. Chan J.F., Kok K.H., Zhu Z.. **Genomic characterization of the 2019 novel human-pathogenic coronavirus isolated from a patient with atypical pneumonia after visiting Wuhan**. *Emerg Microbes Infect* (2020) **9** 221-236. PMID: 31987001
4. Williamson E.J., Walker A.J., Bhaskaran K.. **Factors associated with COVID-19-related death using OpenSAFELY**. *Nature* (2020) **584** 430-436. PMID: 32640463
5. Guh D.P., Zhang W., Bansback N., Amarsi Z., Birmingham C.L., Anis A.H.. **The incidence of co-morbidities related to obesity and overweight: a systematic review and meta-analysis**. *BMC Public Health* (2009) **9** 88. PMID: 19320986
6. Zhang A.J.X., Zhu H., Chen Y.. **Prostaglandin E2-mediated impairment of innate immune response to A(H1N1)pdm09 infection in diet-induced obese mice could be restored by paracetamol**. *J Infect Dis* (2019) **219** 795-807. PMID: 30202973
7. Smith A.G., Sheridan P.A., Harp J.B., Beck M.A.. **Diet-induced obese mice have increased mortality and altered immune responses when infected with influenza virus**. *J Nutr* (2007) **137** 1236-1243. PMID: 17449587
8. Hulme K.D., Noye E.C., Short K.R., Labzin L.I.. **Dysregulated inflammation during obesity: driving disease severity in influenza virus and SARS-CoV-2 infections**. *Front Immunol* (2021) **12**
9. Sanchis-Gomar F., Lavie C.J., Mehra M.R., Henry B.M., Lippi G.. **Obesity and outcomes in COVID-19: when an epidemic and pandemic collide**. *Mayo Clin Proc* (2020) **95** 1445-1453. PMID: 32622449
10. Kruglikov I.L., Shah M., Scherer P.E.. **Obesity and diabetes as comorbidities for COVID-19: underlying mechanisms and the role of viral-bacterial interactions**. *Elife* (2020) **9**
11. Chen X., Hu W., Ling J.. **Hypertension and diabetes delay the viral clearance in COVID-19 patients**. *medRxiv* (2020). DOI: 10.1101/2020.03.22.20040774
12. Nasr M.C., Geerling E., Pinto A.K.. **Impact of obesity on vaccination to SARS-CoV-2**. *Front Endocrinol (Lausanne)* (2022) **13**
13. Viana R., Moyo S., Amoako D.G.. **Rapid epidemic expansion of the SARS-CoV-2 Omicron variant in Southern Africa**. *Nature* (2022) **603** 679-686. PMID: 35042229
14. Shuai H., Chan J.F., Hu B.. **Attenuated replication and pathogenicity of SARS-CoV-2 B.1.1.529 Omicron**. *Nature* (2022) **603** 693-699. PMID: 35062016
15. Hu B., Chan J.F., Liu H.. **Spike mutations contributing to the altered entry preference of SARS-CoV-2 Omicron BA.1 and BA.2**. *Emerg Microbes Infect* (2022) **11** 1-31. PMID: 34890524
16. Chan J.F., Hu B., Chai Y.. **Virological features and pathogenicity of SARS-CoV-2 Omicron BA.2**. *Cell Rep Med* (2022) **3**
17. Yuan S., Ye Z.W., Liang R.. **Pathogenicity, transmissibility, and fitness of SARS-CoV-2 Omicron in Syrian hamsters**. *Science* (2022) **377** 428-433. PMID: 35737809
18. Lu L., Mok B.W., Chen L.L.. **Neutralization of SARS-CoV-2 Omicron variant by sera from BNT162b2 or Coronavac vaccine recipients**. *Clin Infect Dis* (2022) **75** e822-e826. DOI: 10.1093/cid/ciab1041
19. Liu L., Iketani S., Guo Y.. **Striking antibody evasion manifested by the Omicron variant of SARS-CoV-2**. *Nature* (2022) **602** 676-681. PMID: 35016198
20. Iketani S., Liu L., Guo Y.. **Antibody evasion properties of SARS-CoV-2 Omicron sublineages**. *Nature* (2022) **604** 553-556. PMID: 35240676
21. Chen Y., Li C., Liu F.. **Age-associated SARS-CoV-2 breakthrough infection and changes in immune response in a mouse model**. *Emerg Microbes Infect* (2022) **11** 368-383. PMID: 34989330
22. Zhang X., Goncalves R., Mosser D.M.. **The isolation and characterization of murine macrophages**. *Curr Protoc Immunol* (2008) **83**
23. Li C., Chen Y., Zhao Y.. **Intravenous injection of coronavirus disease 2019 (COVID-19) mRNA vaccine can induce acute myopericarditis in mouse model**. *Clin Infect Dis* (2022) **74** 1933-1950. PMID: 34406358
24. Li C., Chen Y.X., Liu F.F.. **Absence of vaccine-enhanced disease with unexpected positive protection against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) by inactivated vaccine given within 3 days of virus challenge in Syrian hamster model**. *Clin Infect Dis* (2021) **73** e719-e734. PMID: 33515458
25. Vogel A.B., Kanevsky I., Che Y.. **BNT162b vaccines protect rhesus macaques from SARS-CoV-2**. *Nature* (2021) **592** 283-289. PMID: 33524990
26. Lee A.C., Zhang A.J., Chan J.F.. **Oral SARS-CoV-2 inoculation establishes subclinical respiratory infection with virus shedding in golden Syrian hamsters**. *Cell Rep Med* (2020) **1**
27. Meyerholz D.K., Beck A.P.. **Principles and approaches for reproducible scoring of tissue stains in research**. *Lab Invest* (2018) **98** 844-855. PMID: 29849125
28. Chu H., Shuai H., Hou Y.. **Targeting highly pathogenic coronavirus-induced apoptosis reduces viral pathogenesis and disease severity**. *Sci Adv* (2021) **7**
29. Chu H., Hou Y., Yang D.. **Coronaviruses exploit a host cysteine-aspartic protease for replication**. *Nature* (2022) **609** 785-792. PMID: 35922005
30. Zhang A.J., Li C., To K.K.. **Toll-like receptor 7 agonist imiquimod in combination with influenza vaccine expedites and augments humoral immune responses against influenza A(H1N1)pdm09 virus infection in BALB/c mice**. *Clin Vaccine Immunol* (2014) **21** 570-579. PMID: 24521786
31. Li C., To K.K.W., Zhang A.J.X.. **Co-Stimulation with TLR7 agonist imiquimod and inactivated influenza virus particles promotes mouse B cell activation, differentiation, and accelerated antigen specific antibody production**. *Front Immunol* (2018) **9** 2370. PMID: 30369932
32. Zhou R., To K.K., Wong Y.C.. **Acute SARS-CoV-2 infection impairs dendritic cell and T cell responses**. *Immunity* (2020) **53** 864-877.e5. PMID: 32791036
33. Chen S., Zhou Y., Chen Y., Gu J.. **fastp: an ultra-fast all-in-one FASTQ preprocessor**. *Bioinformatics* (2018) **34** i884-i890. PMID: 30423086
34. Edgar R.. **URMAP, an ultra-fast read mapper**. *PeerJ* (2020) **8** e9338. PMID: 32612885
35. Kim D., Paggi J.M., Park C., Bennett C., Salzberg S.L.. **Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype**. *Nat Biotechnol* (2019) **37** 907-915. PMID: 31375807
36. Pertea M., Pertea G.M., Antonescu C.M., Chang T.C., Mendell J.T., Salzberg S.L.. **StringTie enables improved reconstruction of a transcriptome from RNA-seq reads**. *Nat Biotechnol* (2015) **33** 290-295. PMID: 25690850
37. Love M.I., Huber W., Anders S.. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol* (2014) **15** 550. PMID: 25516281
38. Yu G., Wang L.G., Han Y., He Q.Y.. **clusterProfiler: an R package for comparing biological themes among gene clusters**. *OMICS* (2012) **16** 284-287. PMID: 22455463
39. Miao Y.R., Xia M., Luo M., Luo T., Yang M., Guo A.Y.. **ImmuCellAI-mouse: a tool for comprehensive prediction of mouse immune cell abundance and immune microenvironment depiction**. *Bioinformatics* (2021) **38** 785
40. Shuai H., Chan J.F., Yuen T.T.. **Emerging SARS-CoV-2 variants expand species tropism to murines**. *eBioMedicine* (2021) **73**
41. Gu H., Chen Q., Yang G.. **Adaptation of SARS-CoV-2 in BALB/c mice for testing vaccine efficacy**. *Science* (2020) **369** 1603-1607. PMID: 32732280
42. Halfmann P.J., Iida S., Iwatsuki-Horimoto K.. **SARS-CoV-2 Omicron virus causes attenuated disease in mice and hamsters**. *Nature* (2022) **603** 687-692. PMID: 35062015
43. Krause P.R., Fleming T.R., Longini I.M.. **SARS-CoV-2 variants and vaccines**. *N Engl J Med* (2021) **385** 179-186. PMID: 34161052
44. Corbett K.S., Nason M.C., Flach B.. **Immune correlates of protection by mRNA-1273 vaccine against SARS-CoV-2 in nonhuman primates**. *Science* (2021) **373**
45. Hamer M., Gale C.R., Kivimäki M., Batty G.D.. **Overweight, obesity, and risk of hospitalization for COVID-19: a community-based cohort study of adults in the United Kingdom**. *Proc Natl Acad Sci U S A* (2020) **117** 21011-21013. PMID: 32788355
46. Mannar D., Saville J.W., Zhu X.. **SARS-CoV-2 Omicron variant: antibody evasion and cryo-EM structure of spike protein-ACE2 complex**. *Science* (2022) **375** 760-764. PMID: 35050643
47. Zhang Y., Huang K., Wang T.. **SARS-CoV-2 rapidly adapts in aged BALB/c mice and induces typical pneumonia**. *J Virol* (2021) **95**
48. Huang H., Zhu Y., Niu Z., Zhou L., Sun Q.. **SARS-CoV-2 N501Y variants of concern and their potential transmission by mouse**. *Cell Death Differ* (2021) **28** 2840-2842. PMID: 34389814
49. Niu Z., Zhang Z., Gao X.. **N501Y mutation imparts cross-species transmission of SARS-CoV-2 to mice by enhancing receptor binding**. *Signal Transduct Target Ther* (2021) **6** 284. PMID: 34315847
50. Chu H., Chan J.F., Yuen K.Y.. **Animal models in SARS-CoV-2 research**. *Nat Methods* (2022) **19** 392-394. PMID: 35396468
51. Briand F., Sencio V., Robil C.. **Diet-induced obesity and NASH impair disease recovery in SARS-CoV-2-infected golden hamsters**. *Viruses* (2022) **14** 2067. PMID: 36146875
52. Port J.R., Adney D.R., Schwarz B.. **High-fat high-sugar diet-induced changes in the lipid metabolism are associated with mildly increased COVID-19 severity and delayed recovery in the Syrian hamster**. *Viruses* (2021) **13** 2506. PMID: 34960775
53. Zhang Y.N., Zhang Z.R., Zhang H.Q.. **Increased morbidity of obese mice infected with mouse-adapted SARS-CoV-2**. *Cell Discov* (2021) **7** 74. PMID: 34433811
54. Akbar A.N., Gilroy D.W.. **Aging immunity may exacerbate COVID-19**. *Science* (2020) **369** 256-257. PMID: 32675364
55. Rebello C.J., Kirwan J.P., Greenway F.L.. **Obesity, the most common comorbidity in SARS-CoV-2: is leptin the link?**. *Int J Obes* (2020) **44** 1810-1817
56. Mullen M., Gonzalez-Perez R.R.. **Leptin-induced JAK/STAT signaling and cancer growth**. *Vaccines (Basel)* (2016) **4** 26. PMID: 27472371
57. Zhu A., Chen Z., Wang Y.. **Immune responses to SARS-CoV-2 infection in humans and ACE2 humanized mice**. *Fundam Res* (2021) **1** 124-130
58. Achkar J.M., Casadevall A.. **Antibody-mediated immunity against tuberculosis: implications for vaccine development**. *Cell Host Microbe* (2013) **13** 250-262. PMID: 23498951
59. Casadevall A., Pirofski L.A.. **A new synthesis for antibody-mediated immunity**. *Nat Immunol* (2011) **13** 21-28. PMID: 22179281
60. Israelow B., Mao T., Klein J.. **Adaptive immune determinants of viral clearance and protection in mouse models of SARS-CoV-2**. *Sci Immunol* (2021) **6**
61. Smith A.G., Sheridan P.A., Tseng R.J., Sheridan J.F., Beck M.A.. **Selective impairment in dendritic cell function and altered antigen-specific CD8+ T-cell responses in diet-induced obese mice infected with influenza virus**. *Immunology* (2009) **126** 268-279. PMID: 18754811
62. Macia L., Delacre M., Abboud G.. **Impairment of dendritic cell functionality and steady-state number in obese mice**. *J Immunol* (2006) **177** 5997-6006. PMID: 17056524
63. Honce R., Schultz-Cherry S.. **Impact of obesity on influenza A virus pathogenesis, immune response, and evolution**. *Front Immunol* (2019) **10** 1071. PMID: 31134099
64. Frasca D., Blomberg B.B.. **Obesity accelerates age defects in mouse and human B cells**. *Front Immunol* (2020) **11** 2060. PMID: 32983154
65. Kennedy D.E., Knight K.L.. **Inhibition of B Lymphopoiesis by adipocytes and IL-1-producing myeloid-derived suppressor cells**. *J Immunol* (2015) **195** 2666-2674. PMID: 26268654
66. Pham T.D., Chng M.H.Y., Roskin K.M.. **High-fat diet induces systemic B-cell repertoire changes associated with insulin resistance**. *Mucosal Immunol* (2017) **10** 1468-1479. PMID: 28422186
67. Alu A., Chen L., Lei H., Wei Y., Tian X., Wei X.. **Intranasal COVID-19 vaccines: from bench to bed**. *EBioMedicine* (2022) **76**
68. Tiboni M., Casettari L., Illum L.. **Nasal vaccination against SARS-CoV-2: synergistic or alternative to intramuscular vaccines?**. *Int J Pharm* (2021) **603**
69. Klein S.L., Flanagan K.L.. **Sex differences in immune responses**. *Nat Rev Immunol* (2016) **16** 626-638. PMID: 27546235
|
---
title: 'Knowledge, attitudes, perceptions, and practice toward seasonal influenza
and its vaccine: A cross-sectional study from a country of conflict'
authors:
- Wesam S. Ahmed
- Rana Abu Farha
- Abdulsalam M. Halboup
- Arwa Alshargabi
- Ahmed Al-mohamadi
- Eman Y. Abu-rish
- Mohammed Zawiah
- Yousf K. Al-Ashbat
- Sayida Al-Jamei
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9970292
doi: 10.3389/fpubh.2023.1030391
license: CC BY 4.0
---
# Knowledge, attitudes, perceptions, and practice toward seasonal influenza and its vaccine: A cross-sectional study from a country of conflict
## Abstract
### Background
The seasonal influenza vaccine is an important preventive measure against influenza and its associated complications. In Yemen, there is no seasonal influenza vaccination policy, and the influenza vaccine is excluded from the national immunization program. Data on vaccination coverage remain scarce with no previous surveillance programs or awareness campaigns implemented in the country. The current study aims to assess the awareness, knowledge, and attitudes of the public in Yemen toward seasonal influenza and their motivators and perceived barriers to receiving its vaccine.
### Methods
A cross-sectional survey was carried out using a self-administered questionnaire that was distributed to eligible participants using convenience sampling.
### Results
A total of 1,396 participants completed the questionnaire. The respondents showed a median knowledge score of influenza of $\frac{11.0}{15.0}$, and most of them ($70\%$) were able to recognize its modes of transmission. However, only $11.3\%$ of the participants reported receiving the seasonal influenza vaccine. Physicians were the respondents' most preferred information source for influenza ($35.2\%$), and their recommendation ($44.3\%$) was the most cited reason for taking its vaccine. On the contrary, not knowing about the vaccine's availability ($50.1\%$), concerns regarding the safety of the vaccine ($17\%$), and not considering influenza as a threat ($15.9\%$) were the main reported barriers to getting vaccinated.
### Conclusion
The current study showed a low uptake of influenza vaccines in Yemen. The physician's role in promoting influenza vaccination seems to be essential. Extensive and sustained awareness campaigns would likely increase the awareness of influenza and remove misconceptions and negative attitudes toward its vaccine. Equitable access to the vaccine can be promoted by providing it free of charge to the public.
## 1. Introduction
Seasonal influenza is an acute infection of the respiratory tract caused by influenza virus type A, B, or C [1]. Influenza virus types A, B, and C are known to infect humans, while type D is believed to infect cattle [2, 3]. Type A influenza virus is subdivided into several serotypes based on the viral hemagglutinin (H) and neuraminidase (N) surface proteins. Some of these serotypes were responsible for outbreaks throughout recent history [4]. Most notoriously are the 1918 Spanish flu (H1N1), the 1857 Asian flu (H2N2), the 1068 Hong Kong flu (H3N2) [5], and the 2009 Swine flu outbreak (H1N1) [6]. Other viral respiratory pandemics in the last couple of decades include the 2002–2004 severe acute respiratory syndrome (SARS) outbreak caused by the newly identified—at the time—SARS coronavirus (SARS-CoV) [7], and the ongoing coronavirus 2019 (COVID-19) pandemic caused by a related coronavirus strain, SARS-CoV-2 (8–10).
Uncomplicated seasonal influenza is manifested by a combination of symptoms, most commonly headache, cough, sore throat, runny nose, fatigue, muscle pain, and fever [11, 12]. Complications of influenza include pneumonia, sinus and ear infections, and worsening of existing chronic medical conditions such as asthma, diabetes, and heart failure [13]. According to the World Health Organization (WHO), seasonal influenza afflicts one billion individuals worldwide annually resulting in 3–5 million severe illness cases and between 290,000 and 650,000 influenza-associated respiratory mortality [14]. Moreover, the disease poses an additional load on the healthcare system and increases the economic burden as a result of work absenteeism and loss of productivity [15]. High-risk groups of influenza complications include immunocompromised individuals, patients with chronic conditions, pregnant females, adults older than 65 years of age, and children younger than 5 years old with those younger than 2 years of age being at even higher risk of influenza complications [16, 17]. Healthcare workers (HCWs) are also considered an at-risk group [12].
The seasonal influenza vaccine (flu vaccine) is an important preventive measure against influenza and its associated complications [18]. The vaccine protects against 3 or 4 influenza viruses that are expected to circulate in the upcoming flu season [18]. The directors of the WHO collaborating centers, laboratories, and academies recommend the composition of the flu vaccine based on surveillance and clinical studies [19]. The Centers for Disease Control and Prevention (CDC) recommend all individuals aged ≥6 months take the vaccine [18, 20, 21] by the end of October each year [21]. Priority is given to healthcare workers and other high-risk groups [16, 22, 23]. The flu vaccine provides several benefits including protection against flu infection (24–26), severity [27], and hospitalization [28], especially in high-risk groups (29–33).
In Yemen, little is known about the public's knowledge of influenza, and their attitudes and practice toward its vaccine. According to the WHO Regional Office of the Eastern Mediterranean region, the country has neither a seasonal influenza vaccination policy for the general public or subgroups nor does it include the influenza vaccine in its national immunization program [34]. Despite increased recommendations on the value of influenza vaccination, data on the burden of influenza and vaccination coverage in the country remain scarce [34]. In addition to COVID-19 [35], the country is simultaneously struck by three other fever-causing infectious diseases: dengue fever, chikungunya fever, and malaria (36–38). Thus, it is imperative to address these severe infections and complement this with the implementation of equitable influenza vaccine programs to reduce the overall disease burden and exhaustion of medical resources which are already scarce in the country as a result of the ongoing war. No previous surveillance programs or awareness campaigns that cover influenza and its vaccine have been implemented in the country. Therefore, the aim of the current study is to assess knowledge, attitudes, and practices toward influenza and its vaccine among the public in Yemen and to understand the main determinants of vaccine acceptance which could be a critical step toward future planning of national influenza campaigns and in implementing national influenza vaccination policy to improve vaccination coverage in the country. To the best of our knowledge, this is the first study to address this topic in the country.
## 2.1. Study design and participants
A questionnaire survey was distributed in public places in Sana'a city to eligible participants over the period of March 2019 to February 2020 using a convenience sampling approach. Potential participants were approached on the streets by experienced interviewers and were invited to participate in the survey. The face-to-face sampling was then discontinued due to the COVID-19 outbreak. This was followed by an online recruitment phase where the questionnaire survey was deployed online using social media platforms (WhatsApp and Facebook) between the 5th of July and the 25th of October 2020. The online sampling included participants from different cities. The survey was directed to the public in Yemen who are aged 18 years or older, are competent, and can read and understand the Arabic language. Consent was obtained orally for the face-to-face sampling. For online recruitment, a consent statement was included at the beginning of the questionnaire as well as in the recruitment invite that was shared through social media platforms.
The questionnaire was adopted from a previously validated questionnaire [39] with few modifications to be applicable to the public in Yemen. The questionnaire was distributed in Arabic since it is the native language of the country. Several response formats were utilized in the questionnaire, including multiple choice, “Yes,” “No,” or “I am not sure,” multiple check box, the Likert scale, and open-ended items. The survey consisted of four sections. The first section solicited sociodemographic information from respondents, including age, gender, marital status, education level, and whether they have ever taken the flu vaccine before. Demographic questions also focused on whether the respondents are working in the medical field, have any chronic medical conditions, or are medically insured. The following section of the questionnaire was designed to assess participants' knowledge of influenza, its modes of transmission, and its preventive measures. The third section assessed the source of knowledge participants used to gain information about influenza and its vaccine. The last section of the questionnaire was designed to determine participants' motivating factors and barriers toward taking the influenza vaccine. The study was verified by the ethics committee of the Scientific Research Center of Yemen University (Ref #: ERC/$\frac{2018}{123}$).
## 2.2. Data analysis and figure preparation
Data were analyzed using the statistical package for social science (SPSS) version 22 (SPSS Inc., Chicago, IL, USA). The descriptive analysis was executed using the median and interquartile range for continuous variables and frequency (percentage) for qualitative variables. Checking for data normality was carried out using the Shapiro–Wilk test (with $P \leq 0.05$ indicating a normally distributed continuous variable). To assess the respondents' level of knowledge of influenza, a score of 1 was given to each correct answer to the 15 questions exploring general knowledge, mode of transmission, and preventive measures. A score of 0 was given for wrong answers. The total score in this construct ranged from 0 to 15. Screening for factors affecting participants' previous uptake of the seasonal influenza vaccine was carried out using univariate and multivariate logistic regression. Following univariate logistic regression analysis, any variable found to be significant on the single predictor level (P-value < 0.25) was entered into the multivariate logistic regression analysis to explore the factors that were significantly and independently associated with participants' previous uptake of the seasonal influenza vaccine. Odds ratios were calculated to measure the effect of each predictor on the practice of seasonal influenza vaccine uptake. Variables were selected after checking their multicollinearity, where tolerance values >0.1 and variance inflation factor (VIF) values < 10 were checked to indicate the absence of multicollinearity between the independent variables in regression analysis. A p-value of ≤0.05 was considered to be statistically significant. Figures were prepared using Microsoft Excel 13.
## 3.1. Demographics
The demographic analysis is reported in Table 1. A total of 1,396 participants completed the questionnaire (472 on-site and 924 online). The response rate for the on-site sampling was $61\%$. The completion rate for the on-site and online sampling was $72\%$. Approximately $41\%$ ($$n = 575$$) were aged between 18 and 24 years of age, and men were overrepresented ($63.0\%$, $$n = 879$$). Most participants had a diploma or a higher education degree ($80.5\%$, $$n = 1$$,124), and approximately one-third were from the medical field ($36.9\%$, $$n = 515$$). Many participants were from big cities such as Sana'a ($34\%$, $$n = 476$$), Taiz ($23.4\%$, $$n = 326$$), and Ibb ($12\%$, $$n = 172$$). A minority were smokers ($12.2\%$, $$n = 172$$), and $12.2\%$ ($$n = 170$$) reported having a chronic medical condition. Most participants were not medically insured ($75.4\%$, $$n = 1$$,053). In addition, $11.3\%$ ($$n = 158$$) stated that they had received the seasonal influenza vaccine before (Table 1).
**Table 1**
| Parameter | n (%) |
| --- | --- |
| Age (years) | Age (years) |
| 18–24 | 575 (41.2) |
| 25–35 | 496 (35.5) |
| ≥35 | 325 (23.3) |
| Gender | Gender |
| Males | 879 (63.0) |
| Females | 517 (37.0) |
| Marital status | Marital status |
| Married | 630 (45.1) |
| Others | 766 (54.9) |
| Education level | Education level |
| School level | 255 (18.3) |
| Diploma | 118 (8.5) |
| Undergraduate degree | 874 (62.6) |
| Post-graduate degree | 132 (9.5) |
| Missing data | 17 (1.2) |
| Governorate | Governorate |
| Sana'a | 476 (34.1) |
| Taiz | 326 (23.4) |
| Ibb | 172 (12.3) |
| Hodidah | 58 (4.2) |
| Hajah | 58 (4.2) |
| Others | 306 (21.9) |
| Are you from the medical field? | Are you from the medical field? |
| No | 881 (63.1) |
| Yes | 515 (36.9) |
| Do you have any chronic medical conditions? | Do you have any chronic medical conditions? |
| No | 1,225 (87.8) |
| Yes | 171 (12.2) |
| Do you have medical insurance? | Do you have medical insurance? |
| No | 1,053 (75.4) |
| Yes | 343 (24.6) |
| Smoking status | Smoking status |
| Non-smoker/ex-smoker | 1,225 (87.8) |
| Current smoker | 171 (12.2) |
| Have you ever had the seasonal influenza vaccine? | Have you ever had the seasonal influenza vaccine? |
| No | 1,233 (88.3) |
| Yes | 158 (11.3) |
| Missing data | 5 (0.4) |
## 3.2. Participants' knowledge
Participants' knowledge of influenza was assessed and is reported in Table 2. Participants showed a median knowledge score of 11.0 out of 15.0. Most participants correctly identified the definition of influenza ($82.7\%$, $$n = 1$$,155), its risk factors such as comorbid chronic diseases ($75.8\%$, $$n = 1$$,058) and age ≥65 years and ≤5 years ($70.9\%$, $$n = 990$$). However, only $21.5\%$ ($$n = 300$$) were aware that not every H1N1-infected person will experience complications that need hospitalization and $16.8\%$ ($$n = 234$$) thought that cures are available to treat complicated cases of influenza. Considering modes of transmission and prevention, more than $77\%$ were able to recognize the different modes of influenza virus transmission, and the majority believed that avoiding crowded places helps prevent the transmission ($90.0\%$, $$n = 1$$,257), although influenza vaccine was the least recognized preventive measure ($62.5\%$, $$n = 872$$). Sources of information are illustrated in Figure 1.
## 3.3. Motivators, barriers, and factors influencing vaccine uptake
Participants who reported receiving the flu vaccine cited the motivating factors that contributed to their vaccine acceptance (Table 3). Compliance with the physician's recommendation ($44.3\%$, $$n = 70$$) was the most cited motivator for taking the flu vaccine, followed by fear of catching H1N1 influenza ($29.7\%$, $$n = 47$$) and preventing disease transmission to family members ($16.5\%$, $$n = 26$$).
**Table 3**
| Factor | Participants agreed, n (%)† |
| --- | --- |
| Reasons for getting vaccinated [reported by participants who had ever received the vaccine (n = 158)] | Reasons for getting vaccinated [reported by participants who had ever received the vaccine (n = 158)] |
| Compliance with physician's recommendation | 70 (44.3) |
| Fear from catching H1N1 influenza | 47 (29.7) |
| Worries about becoming severely ill following influenza infection | 18 (11.4) |
| To prevent disease transmission to family members | 26 (16.5) |
| Having a chronic medical condition | 4 (2.5) |
| Reasons for not getting vaccinated [reported by participants who had never received the vaccine (n = 1,233)] | Reasons for not getting vaccinated [reported by participants who had never received the vaccine (n = 1,233)] |
| Not considering influenza as a threat | 196 (15.9) |
| Doubts regarding the vaccine's efficacy | 121 (9.8) |
| Doubts regarding the vaccine's safety | 210 (17.0) |
| Time constraints | 55 (4.5) |
| Unaware of vaccine availability | 618 (50.1) |
| Cost of the vaccine | 50 (4.1) |
On the contrary, participants who had never been vaccinated reported being unaware of vaccine availability ($50.1\%$, $$n = 618$$) as the most common barrier for not getting vaccinated, followed by safety concerns regarding the vaccine ($17.0\%$, $$n = 210$$) and not considering influenza as a threat ($15.9\%$, $$n = 196$$) (Table 3).
In addition, participants reported factors that would encourage them to get vaccinated in the future (Table 4). Approximately two-thirds agreed or strongly agreed to take the vaccine if it was recommended by their physician ($67.4\%$, $$n = 941$$) or if the vaccine was better validated for safety and efficacy ($66.4\%$, $$n = 918$$).
**Table 4**
| Factor | Strongly agreed/agreed (%)† | Missing data |
| --- | --- | --- |
| The vaccine is recommended by the physician | 941 (67.4) | 5 (0.4) |
| The vaccine is more validated for safety and efficacy | 918 (66.4) | 5 (0.4) |
| Vaccine uptake is encouraged by the government | 835 (59.8) | 5 (0.4) |
| The vaccine is offered free of charge by the government | 891(63.8) | 5 (0.4) |
## 3.4. Predictors for low vaccination uptake
The results from univariate logistic regression on a single predictor level (P-value < 0.25) revealed that the practice toward seasonal influenza uptake was significantly less frequent among participants who are women, have a diploma or higher education degree, work in the medical field, and are older in age. These factors were further analyzed through multivariate linear regression analysis (backward method) to explore the factors that were significantly and independently associated with low prior vaccine uptake (P ≤ 0.05 with OR < 1). Two factor fulfilled the criteria, these are older age (OR = 0.967, $$P \leq 0.040$$) and being from the medical field (OR = 0.686, $$P \leq 0.044$$) (Table 5). The model fit was found to be significant with χ2 (df = 4) = 17.491 at $$P \leq 0.002$$, which indicated that our full model predicts significantly better or more accurately than the null model.
**Table 5**
| Parameter | Previous vaccine uptake [0: No, 1: Yes] | Previous vaccine uptake [0: No, 1: Yes].1 | Previous vaccine uptake [0: No, 1: Yes].2 | Previous vaccine uptake [0: No, 1: Yes].3 |
| --- | --- | --- | --- | --- |
| | OR | P -value † | OR | P -value ‡ |
| Age (years) | 0.967 | 0.005§ | 0.976 | 0.040* |
| Gender | Gender | Gender | Gender | Gender |
| Male | Reference | | | |
| Female | 0.697 | 0.049§ | 0.709 | 0.064 |
| Marital status | Marital status | Marital status | Marital status | Marital status |
| Married | Reference | | | |
| None-married (single, widowed, or divorced) | 1.134 | 0.462 | – | – |
| Educational level | Educational level | Educational level | Educational level | Educational level |
| School or lower | Reference | | | |
| Diploma of higher | 0.768 | 0.199§ | 0.693 | 0.104 |
| Are you from the medical field? | Are you from the medical field? | Are you from the medical field? | Are you from the medical field? | Are you from the medical field? |
| No | Reference | | | |
| Yes | 0.665 | 0.017§ | 0.686 | 0.044* |
| Do you have any chronic medications? | Do you have any chronic medications? | Do you have any chronic medications? | Do you have any chronic medications? | Do you have any chronic medications? |
| No | Reference | | | |
| Yes | 0.938 | 0.726 | – | – |
| Do you have medical insurance? | Do you have medical insurance? | Do you have medical insurance? | Do you have medical insurance? | Do you have medical insurance? |
| No | Reference | | | |
| Yes | 1.006 | 0.976 | – | – |
| Smoking status | Smoking status | Smoking status | Smoking status | Smoking status |
| Non-smoker/ex-smoker | Reference | | | |
| Current smoker | 0.908 | 0.714 | – | – |
| Knowledge score | 1.038 | 0.270 | – | – |
## 4. Discussion
Influenza is an overlooked contributor to morbidity and mortality (40–42). Several studies assessed knowledge about influenza and attitudes toward its vaccine among healthcare professionals (43–50). However, studies to assess the same among the public are generally scarce (46, 51–53), especially in a low-income, war-torn developing country such as Yemen. This is the first large-scale study in the country to assess the public's knowledge, attitudes, and perceptions of influenza and its vaccine, and their motivating factors and perceived barriers to vaccine acceptance. Findings from this study will inform future national influenza vaccination regimes and assess improving vaccination coverage in the country.
Overall, the results from the current study showed an acceptable median score of knowledge of influenza among the participants, but some major gaps in knowledge were identified. Physicians were the main source that participants sought to gain information about influenza. Only a minority of participants had ever received the influenza vaccine. Lack of awareness of vaccine availability in the country was identified as the main barrier toward vaccine uptake, while physicians' recommendation to take the vaccine was cited as the prime motivator behind receiving the vaccine. Older age and being a healthcare worker were identified as predictors for low prior vaccine acceptance.
Assessing knowledge about influenza, its modes of transmission, and its preventive measures revealed an acceptable median knowledge score but with critical knowledge gaps that were mostly related to H1N1-associated infection. The median knowledge score would seem to reflect the education level of the participants as most participants reported receiving a higher education degree. The exaggeration of the H1N1 fatality risk reported by the participants in the current study is in disagreement with a former study from China where only a minority believed that H1N1 has a high fatality rate [54]. On the contrary, although the majority correctly identified all influenza preventive measures, the influenza vaccine was the least recognized preventive strategy. This is similar to other studies from Italy and Jordan, where most participants did not recognize the vaccine as a major preventive measure to control influenza transmission [39, 46]. These findings highlight the need for educational campaigns in the country to raise public awareness of influenza, especially H1N1, and the role of the flu vaccine in preventing the spread of infection.
Assessments of major sources of information that participants used to gain medical information about influenza revealed that physicians were the main source followed by television and newspapers. As a result, these information sources can be utilized in the future to enhance awareness of influenza and its vaccine among the public in Yemen. This is different from a study in Saudi Arabia where less than one-fourth of participants reported receiving information from a healthcare provider [55]. In comparison, in a study from Jordan, a country that has the highest literacy rate in the Arab region [56, 57], newspapers were the major source of influenza information [39]. In the United Kingdom, television and the Internet were the leading sources of knowledge about influenza [52]. As such, it seems that education level and cultural differences between countries play a role in selecting a reliable source of information about influenza and its vaccine. Importantly, these findings suggest a higher trust in physicians' knowledge among the public in Yemen.
Regarding influenza vaccine uptake, although the respondents had a median knowledge score of $\frac{11}{15}$, and most ($62.5\%$) believed that the flu vaccine is an important preventive measure, the practice toward influenza vaccine uptake was relatively poor. Only a minority ($11\%$) of respondents reported receiving the seasonal influenza vaccine previously. This low vaccine uptake may as well be an overestimation of the vaccination coverage for any given year, which has not been investigated in the current study. This is in agreement with the vaccination status reported in the MENA region according to the 7th Middle East and North Africa Influenza Stakeholder Network (MENA-ISN) report [58]. As such, having knowledge about influenza and its vaccine does not seem to predict vaccination uptake. More evidence of this is that healthcare workers were less likely to receive the vaccine compared to those who were not enrolled in the medical field. Previous studies from the Arab MENA region such as Lebanon [44], Jordan [39], Saudi Arabia [45], Kuwait, Oman, and the United Arab Emirates [43] showed that individuals enrolled in a non-health related field had higher vaccination acceptance rate compared to those working in a medical field. This observation is not confined to the Arab MENA region but also extends globally [46, 51, 59]. In a study from France, general practitioners who perceived the risk of influenza illness to outweigh the risk of its vaccination had a higher vaccination acceptance rate compared to those who did not perceive the same [51]. Therefore, underestimating the risks and complications of influenza and overestimating the risks associated with its vaccine among individuals working in the medical field appears to be a major barrier to vaccination acceptance in this group. In addition to being from the medical field, the current study also identified older age as a predictor of low vaccine acceptance. Although older adults are at higher risk of influenza complications, this group is more hesitant about vaccine uptake. Similar findings were reported in other countries (60–64).
In addition to respondents' beliefs and knowledge, the current study assessed factors that contributed to prior vaccine acceptance or rejection. In agreement with some previous studies [39, 65, 66], physician recommendation to receive the vaccine was the main motivating factor for vaccine uptake. This, again, indicates trust in physicians' knowledge and recommendations among the respondents. Fear of catching an H1N1 infection was also reported as a major motivator to receiving the vaccine. On the contrary, not knowing about the availability of the vaccine in the country was regarded as the main barrier to getting vaccinated. Surprisingly, although *Yemen is* a low-income developing country, the cost of the vaccine was the least cited barrier toward vaccine uptake. This suggests the willingness of the public to take the vaccine if they were aware of its availability. In addition, it stresses the need for governmental efforts to educate the public about the availability of the vaccine in the country. In addition to unawareness of vaccine availability, doubts regarding the safety of the vaccine and not considering influenza as a threat were other highly cited reasons for abstinence from vaccination. In agreement with these results, similar studies from Jordan and the United States showed that low perceived risk of influenza and concerns regarding the safety and efficacy of the vaccine are leading factors for rejecting vaccination [30, 39]. Additional challenges in Yemen such as political conflict, personal safety, food security, weak infrastructure, collapsed healthcare system, and lack of awareness campaigns on the importance and availability of influenza vaccine are all potential barriers to vaccination [67, 68].
When assessing factors that will encourage the participants to receive the influenza vaccine in the future, most participants expressed their willingness to take the vaccine if it was recommended by the physicians, was sufficiently validated for safety and efficacy, and was advocated and offered free of charge by the government. Similarly, in a recent study from Jordan, most participants were willing to get vaccinated if the influenza vaccine was recommended by physicians, was safe and effective, and was provided free by the government [39].
Overall, the study at hand is an important first step to inform researchers and decision-makers in the country of the current public awareness toward influenza and its vaccine, including the low vaccination coverage. Intervention studies, utilizing motivators and barriers toward vaccine uptake reported in the study, are likely to follow. Implementing equitable access to information about influenza and its vaccine in the country is possible using the available multimedia in the country, for instance, utilizing short SMS mobile phone messages to inform the public of information about influenza and the availability of the vaccine. On the other hand, equitable access to influenza vaccine would seem more challenging given the current conditions in the country. The latest World *Bank data* show that most of the Yemeni population (~$75\%$) live below the poverty line. Therefore, one important step toward implementing equitable vaccine access would be to offer it free of charge, for instance as part of the national immunization program.
## 5. Limitations
The current study comes with some limitations that can be mitigated through future research. One such limitation is that the current study inquired about “ever” receiving the flu vaccine. Future research can explore vaccine uptake in each year/duration, which will provide valuable insight into the effect of the ongoing conflict on vaccine uptake and the change in the annual vaccination coverage over time. Moreover, although our study recruited participants from different areas of different cities, our sampling method, by definition, is convenience sampling and the ability to generalize from convenience sampling remains limited compared to random sampling. Randomly sampling the population would have been really challenging given the current state of war in the country. In addition, even though our study did not assess the economic status of the participants, it still can be assumed from the health insurance statistics that most participants are from the low-to-middle class as only $25\%$ of participants reported having health insurance, which is not mandatory in the country and is not covered by government or private employers, and therefore only $25\%$ would seem to afford to have health insurance and, probably, the flu vaccine as well. However, our univariate and multivariate analyses did not reveal having health insurance as a predictor for vaccine uptake. As such, the low vaccine uptake reported in the current study would not seem to be attributed to the less wealthy participants but rather to other factors such as the lack of awareness of the availability of the vaccine which has been cited as the main barrier to vaccine uptake by the participants. Another limitation is that the assessment of influenza vaccine uptake was based on the self-reporting recall of the participants rather than reviewing medical records.
## 6. Conclusion
Critical gaps in knowledge of influenza were identified among the public in Yemen. The study revealed a low vaccine uptake in the country and identified major determinants of vaccine acceptance and rejection. Optimizing vaccine acceptance and coverage can be achieved by collaboration between the healthcare sector and governmental authorities. Efforts ensuring the free-of-charge provision of the vaccine will assess in establishing equitable vaccine access. In addition, implementing education programs utilizing different audiovisual platforms is recommended to enhance positive attitudes toward influenza vaccine, raise awareness toward vaccine availability, consolidate the public's trust in the safety of the vaccine, and promote the vaccine among high-risk groups in the community who are in critical need of the vaccine.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Yemen University (Ref #: ERC/$\frac{2018}{123}$). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
WA: conceptualization, study design, data interpretation, original draft preparation, project administration, reviewing, and editing. RA: data analysis, data interpretation, original draft preparation, reviewing, and editing. AH: data acquisition, data interpretation, original draft preparation, reviewing, and editing. AA and YA-A: data acquisition, reviewing, and editing. EA-r: study design, reviewing, and editing. MZ: data interpretation, reviewing, and editing. SA-J: conceptualization, study design, data acquisition, data interpretation, original draft preparation, and project administration. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Killingley B, Nguyen-Van-Tam J. **Routes of influenza transmission**. *Influenza Other Respir Viruses* (2013.0) **7** 42. DOI: 10.1111/irv.12080
2. Ducatez MF, Pelletier C, Meyer G. **Influenza D virus in cattle, France, 2011–2014**. *Emerg Infect Dis.* (2015.0) **21** 368. DOI: 10.3201/eid2102.141449
3. He J, Hou S, Chen Y, Yu J-L, Chen Q-Q, He L. **The epidemiological pattern and co-infection of influenza A and B by surveillance network from 2009 to 2014 in Anhui Province, China**. *Front Public Health* (2022.0) **10** 825645. DOI: 10.3389/fpubh.2022.825645
4. Meng X, Zhao H, Ou R, Zeng Q, Lv H, Zhu H. **Epidemiological and clinical characteristics of influenza outbreaks among children in Chongqing, China**. *Front Public Health.* (2022.0) **10** 760746. DOI: 10.3389/fpubh.2022.760746
5. Kilbourne ED. **Influenza pandemics of the 20th century**. *Emerg Infect Dis.* (2006.0) **12** 9. DOI: 10.3201/eid1201.051254
6. Presanis AM, De Angelis D, New York City Swine Flu Investigation Team, Hagy A, Reed C, Riley S. **The severity of pandemic H1N1 influenza in the United States, from April to July 2009: a Bayesian analysis**. *PLoS Med.* (2009.0) **6** e1000207. DOI: 10.1371/journal.pmed.1000207
7. Low D, McGeer A. **SARS–one year later**. *N Engl J Med.* (2003.0) **349** 2381. DOI: 10.1056/NEJMp038203
8. Wu D, Wu T, Liu Q, Yang Z. **The SARS-CoV-2 outbreak: What we know**. *Int J Infect Dis.* (2020.0) **94** 44-8. DOI: 10.1016/j.ijid.2020.03.004
9. Zhang J, Huang X, Tao Z. **Correlation of clinical characteristics in seasonal influenza patients with those in wild-type or delta variant SARS-CoV-2 infected patients**. *Front Public Health*. DOI: 10.3389/fpubh.2022.981233
10. Guan Z, Chen C, Li Y, Yan D, Zhang X, Jiang D. **Impact of coinfection with SARS-CoV-2 and influenza on disease severity: a systematic review and meta-analysis**. *Front Public Health* (2021.0) **9** 773130. DOI: 10.3389/fpubh.2021.773130
11. Hart AM. **Influenza: a clinical update following a century of influenza science**. *J Nurse Pract.* (2019.0) **15** 429-33. DOI: 10.1016/j.nurpra.2018.12.026
12. Maita H, Kobayashi T, Akimoto T, Matsuoka F, Osawa H, Kato H. **Factors associated with seasonal influenza self-diagnosis: a prospective observational study in Japan**. *NPJ Prim Care Respir Med.* (2020.0) **30** 9. DOI: 10.1038/s41533-020-0165-3
13. 13.CDC. Flu Symptoms & Complications. (2019). Available online at: https://www.cdc.gov/flu/symptoms/symptoms.htm (accessed 28 May, 2020).. *Flu Symptoms & Complications* (2019.0)
14. 14.WHO. Global Influenza Strategy 2019–2030. (2019). Available online at: https://www.who.int/influenza/Global_Influenza_Strategy_2019_2030_Summary_English.pdf?ua=1 (accessed 26 May, 2020).. *Global Influenza Strategy 2019–2030* (2019.0)
15. 15.CDC. Flu Vaccination Ceoverage, United States, 2014–15 Influenza Season. (2015). Available online at: https://www.cdc.gov/flu/pdf/fluvaxview/nfid-coverage-2014-15.pdf (accessed 28 May, 2020).. *Flu Vaccination Ceoverage, United States, 2014–15 Influenza Season* (2015.0)
16. 16.CDC. People at High Risk For Flu Complications. (2018). Available online at: https://www.cdc.gov/flu/highrisk/index.htm (accessed 28 May, 2020).. *People at High Risk For Flu Complications* (2018.0)
17. 17.ECDC. Seasonal Influenza Vaccination in Europe – Vaccination Recommendations and Coverage Rates for Eight Influenza Seasons (2007–2008 to 2014–2015). (2017). Available online at: https://www.ecdc.europa.eu/en/publications-data/seasonal-influenza-vaccination-europe-vaccination-recommendations-and-coverage-2007-2015 (accessed 28 May, 2200).. *Seasonal Influenza Vaccination in Europe – Vaccination Recommendations and Coverage Rates for Eight Influenza Seasons (2007–2008 to 2014–2015)* (2017.0)
18. 18.CDC. Preventive Steps. (2019). Available online at: https://www.cdc.gov/flu/prevent/prevention.htm (accessed 28 May, 2020).. *Preventive Steps* (2019.0)
19. 19.CDC. Selecting Viruses for the Seasonal Influenza Vaccine (2018). Available online at: https://www.cdc.gov/flu/prevent/vaccine-selection.htm (accessed 28 May, 2020).. *Selecting Viruses for the Seasonal Influenza Vaccine* (2018.0)
20. Grohskopf LA, Alyanak E, Broder KR, Walter EB, Fry AM, Jernigan DB. **Prevention and control of seasonal influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices—United States, 2019–20 influenza season**. *MMWR Recommend Rep.* (2019.0) **68** 1. DOI: 10.15585/mmwr.rr6803a1
21. 21.CDC. Who Needs a Flu Vaccine and When. (2019). Available online at: https://www.cdc.gov/flu/prevent/vaccinations.htm#when-vaccinate (accessed 28 May, 2020).. *Who Needs a Flu Vaccine and When* (2019.0)
22. 22.WHO. Recommendations on Influenza Vaccination During the 2019–2020 Winter Season. (2019). Available online at: http://www.euro.who.int/__data/assets/pdf_file/0017/413270/Influenza-vaccine-recommendations-2019–2020_en.pdf?ua=1 (accessed 28 May, 2020).. *Recommendations on Influenza Vaccination During the 2019–2020 Winter Season* (2019.0)
23. **Meeting of the Strategic Advisory Group of Experts on immunization, April 2012-conclusions and recommendations**. *Wkly Epidemiol Rec.* (2012.0) **87** 201-16. PMID: 24340402
24. Dawood F, Chung J, Kim S, Zimmerman R, Nowalk M, Jackson M. **Interim Estimates of 2019–20 Seasonal Influenza Vaccine Effectiveness-United States, February 2020**. *MMWMorbid R Mortality Wkly Rep.* (2020.0) **69** 177. DOI: 10.15585/mmwr.mm6907a1
25. Balasubramani GK, Nowalk MP, Sax TM, Suyama J, Bobyock E, Rinaldo CR Jr. **Influenza vaccine effectiveness among outpatients in the US Influenza Vaccine Effectiveness Network by study site 2011–2016**. *Influenza Other Respir Viruses* (2020.0) **14** 380-90. DOI: 10.1111/irv.12741
26. Fu C, He Q, Li Z, Xu J, Li Y, Lu J. **Seasonal influenza vaccine effectiveness among children, 2010–2012**. *Influenza Other Respir Viruses* (2013.0) **7** 1168-74. DOI: 10.1111/irv.12157
27. Arriola C, Garg S, Anderson EJ, Ryan PA, George A, Zansky SM. **Influenza vaccination modifies disease severity among community-dwelling adults hospitalized with influenza**. *Clin Infect Dis.* (2017.0) **65** 1289. DOI: 10.1093/cid/cix468
28. Ferdinands J, Gaglani M, Martin E, Middleton D, Monto A, Murthy K. **Prevention of influenza hospitalization among adults in the United States, 2015–2016: results from the US Hospitalized Adult Influenza Vaccine Effectiveness Network (HAIVEN)**. *J Infect Dis.* (2019.0) **220** 1265. DOI: 10.1093/infdis/jiy723
29. 29.CDC. CDC Study Finds Flu Vaccine Saves Children's Lives. (2017). Available online at: https://www.cdc.gov/media/releases/2017/p0403-flu-vaccine.html (accessed 28 May, 2020).. *CDC Study Finds Flu Vaccine Saves Children's Lives* (2017.0)
30. Ferdinands J, Olsho L, Agan A, Bhat N, Sullivan R, Hall M. **Effectiveness of influenza vaccine against life-threatening RT-PCR-confirmed influenza illness in US children, 2010–2012**. *J Infect Dis.* (2014.0) **210** 674. DOI: 10.1093/infdis/jiu185
31. Rondy M, El Omeiri N, Thompson MG, Levêque A, Moren A, Sullivan SG. **Effectiveness of influenza vaccines in preventing severe influenza illness among adults: a systematic review and meta-analysis of test-negative design case-control studies**. *J Infect.* (2017.0) **75** 381. DOI: 10.1016/j.jinf.2017.09.010
32. Thompson M, Kwong J, Regan A, Katz M, Drews S, Azziz-Baumgartner E. **Influenza vaccine effectiveness in preventing influenza-associated hospitalizations during pregnancy: a multi-country retrospective test negative design study, 2010–2016**. *Clin Infect Dis.* (2019.0) **68** 1444. DOI: 10.1093/cid/ciy737
33. Vamos EP, Pape UJ, Curcin V, Harris MJ, Valabhji J, Majeed A. **Effectiveness of the influenza vaccine in preventing admission to hospital and death in people with type 2 diabetes**. *CMAJ.* (2016.0) **188** E342. DOI: 10.1503/cmaj.151059
34. Abubakar A, Melhem N, Malik M, Dbaibo G, Khan W, Zaraket H. **Seasonal influenza vaccination policies in the Eastern Mediterranean Region: current status and the way forward**. *Vaccine* (2019.0) **37** 1601. DOI: 10.1016/j.vaccine.2019.02.001
35. AlTaher N. *Lack of testing leads to undercounting of coronavirus cases in Aden, Yemen* (2020.0)
36. Yildiz H. *Mysterious disease in Yemen claiming lives; Deaths from unknown disease reported in Yemen's temporary capital of Aden* (2020.0)
37. 37.IFRC. Yemen: Dengue Fever - Outbreak Operation Update Report (DREF n° MDRYE008) (2020). Available online at: https://reliefweb.int/report/yemen/yemen-dengue-fever-outbreak-operation-update-report-dref-n-mdrye008 (accessed May, 2020).. *Yemen: Dengue Fever - Outbreak Operation Update Report (DREF n° MDRYE008)* (2020.0)
38. 38.OCHA. Yemen Situation Report, 6 May 2020 (2020). Available online at: https://reliefweb.int/report/yemen/yemen-situation-report-6-may-2020 (accessed 28 May, 2020).. *Yemen Situation Report, 6 May 2020* (2020.0)
39. Abu-Rish EY, Elayeh ER, Mousa LA, Butanji YK, Albsoul-Younes AM. **Knowledge, awareness and practices towards seasonal influenza and its vaccine: implications for future vaccination campaigns in Jordan**. *Fam Pract.* (2016.0) **33** 690-7. DOI: 10.1093/fampra/cmw086
40. Macias AE, McElhaney JE, Chaves SS, Nealon J, Nunes MC, Samson SI. **The disease burden of influenza beyond respiratory illness**. *Vaccine* (2021.0) **39** A6-14. DOI: 10.1016/j.vaccine.2020.09.048
41. Molinari N-AM, Ortega-Sanchez IR, Messonnier ML, Thompson WW, Wortley PM, Weintraub E. **The annual impact of seasonal influenza in the US: measuring disease burden and costs**. *Vaccine* (2007.0) **25** 5086-96. DOI: 10.1016/j.vaccine.2007.03.046
42. Iuliano AD, Roguski KM, Chang HH, Muscatello DJ, Palekar R, Tempia S. **Estimates of global seasonal influenza-associated respiratory mortality: a modelling study**. *Lancet* (2018.0) **391** 1285-300. DOI: 10.1016/S0140-6736(17)33293-2
43. Abu-Gharbieh E, Fahmy S, Rasool BA, Khan S. **Influenza vaccination: healthcare workers attitude in three Middle East countries**. *Int J Med Sci.* (2010.0) **7** 319. DOI: 10.7150/ijms.7.319
44. Romani M, Musharrafieh U, Lakkis N, Hamadeh G. **Family physicians beliefs and attitudes regarding adult pneumococcal and influenza immunization in Lebanon**. *Fam Pract.* (2011.0) **28** 632. DOI: 10.1093/fampra/cmr038
45. Alshammari T, AlFehaid L, AlFraih J, Aljadhey H. **Health care professionals' awareness of, knowledge about and attitude to influenza vaccination**. *Vaccine* (2014.0) **32** 5957. DOI: 10.1016/j.vaccine.2014.08.061
46. Lino M, Di Giuseppe G, Albano L, Angelillo I. **Parental knowledge, attitudes and behaviours towards influenza A/H1N1 in Italy**. *Eur J Public Health* (2012.0) **22** 568. DOI: 10.1093/eurpub/ckr115
47. Lehmann B, Ruiter R, Chapman G, Kok G. **The intention to get vaccinated against influenza and actual vaccination uptake of Dutch healthcare personnel**. *Vaccine* (2014.0) **32** 6986. DOI: 10.1016/j.vaccine.2014.10.034
48. Wesley MG, Soto G, Arriola CS, Gonzales M, Newes-Adeyi G, Romero C. **Prospective cohort study of influenza vaccine effectiveness among healthcare personnel in Lima, Peru: Estudio Vacuna de Influenza Peru, 2016–2018**. *Influenza Other Respir Viruses.* (2020.0) **14** 391-402. DOI: 10.1111/irv.12737
49. Hussain H, McGeer A, McNeil S, Katz K, Loeb M, Simor A. **Factors associated with influenza vaccination among healthcare workers in acute care hospitals in Canada**. *Influenza Other Respir Viruses* (2018.0) **12** 319-25. DOI: 10.1111/irv.12545
50. Chen S, Jiang Y, Tang X, Gan L, Xiong Y, Chen T. **Research on knowledge, attitudes, and practices of influenza vaccination among healthcare workers in Chongqing, China-based on structural equation model**. *Front Public Health* (2022.0) **10** 853041. DOI: 10.3389/fpubh.2022.853041
51. Massin S, Ventelou B, Nebout A, Verger P, Pulcini C. **Cross-sectional survey: risk-averse French general practitioners are more favorable toward influenza vaccination**. *Vaccine* (2015.0) **33** 610. DOI: 10.1016/j.vaccine.2014.12.038
52. Janks M, Cooke S, Odedra A, Kang H, Bellman M, Jordan R. **Factors affecting acceptance and intention to receive pandemic influenza A H1N1 vaccine among primary school children: a cross-sectional study in Birmingham, UK**. *Influenza Res Treat.* (2012.0) **2012** 182565. DOI: 10.1155/2012/182565
53. Latiff LA, Parhizkar S, Zainuddin H, Chun GM, Rahiman MA, Liyana ARN. **Pandemic influenza a (H1N1) and its prevention: a cross sectional study on patients' knowledge, attitude and practice among patients attending primary health Care Clinic in Kuala Lumpur, Malaysia**. *Glob J Health Sci.* (2012.0) **4** 95. DOI: 10.5539/gjhs.v4n2p95
54. Lau JT, Griffiths S, Choi KC, Tsui HY. **Widespread public misconception in the early phase of the H1N1 influenza epidemic**. *J Infect.* (2009.0) **59** 122-7. DOI: 10.1016/j.jinf.2009.06.004
55. Balkhy HH, Abolfotouh MA, Al-Hathlool RH, Al-Jumah MA. **Awareness, attitudes, and practices related to the swine influenza pandemic among the Saudi public**. *BMC Infect Dis.* (2010.0) **10** 42. DOI: 10.1186/1471-2334-10-42
56. Amr M. **Teacher education for inclusive education in the Arab world: the case of Jordan**. *Prospects* (2011.0) **41** 399-413. DOI: 10.1007/s11125-011-9203-9
57. Al-Shami KM, Ahmed WS, Alzoubi KH. **Motivators and barriers towards clinical research participation: a population-based survey from an Arab MENA country**. *PLoS ONE* (2022.0) **17** e0270300. DOI: 10.1371/journal.pone.0270300
58. Al Awaidi S, Abusrewil S, AbuHasan M, Akcay M, Aksakal F, Bashir U. **Influenza vaccination situation in Middle-East and North Africa countries: report of the 7th MENA Influenza Stakeholders Network (MENA-ISN)**. *J Infect Public Health* (2018.0) **11** 845. DOI: 10.1016/j.jiph.2018.07.003
59. Albano L, Matuozzo A, Marinelli P, Di Giuseppe G. **Knowledge, attitudes and behaviour of hospital health-care workers regarding influenza A/H1N1: a cross sectional survey**. *BMC Infect Dis.* (2014.0) **14** 1-7. DOI: 10.1186/1471-2334-14-208
60. Troiano G, Nardi A. **Vaccine hesitancy in the era of COVID-19**. *Public Health* (2021.0) **194** 245. DOI: 10.1016/j.puhe.2021.02.025
61. Bödeker B, Remschmidt C, Schmich P, Wichmann O. **Why are older adults and individuals with underlying chronic diseases in Germany not vaccinated against flu? A population-based study**. *BMC Public Health* (2015.0) **15** 618. DOI: 10.1186/s12889-015-1970-4
62. Nicholls LAB, Gallant AJ, Cogan N, Rasmussen S, Young D, Williams L. **Older adults' vaccine hesitancy: psychosocial factors associated with influenza, pneumococcal, and shingles vaccine uptake**. *Vaccine* (2021.0) **39** 3520-7. DOI: 10.1016/j.vaccine.2021.04.062
63. Ziello A, Scavone C, Di Battista ME, Salvatore S, Cesare DDG, Moreggia O. **Influenza vaccine hesitancy in patients with multiple sclerosis: a monocentric observational study**. *Brain Scie.* (2021.0) **11** 890. DOI: 10.3390/brainsci11070890
64. González-Block MÁ, Pelcastre-Villafuerte BE, Knauth DR, Fachel-Leal A, Comes Y, Crocco P. **Influenza vaccination hesitancy in large urban centers in South America. Qualitative analysis of confidence, complacency and convenience across risk groups**. *PLoS ONE* (2021.0) **16** e0256040. DOI: 10.1371/journal.pone.0256040
65. Malosh R, Ohmit SE, Petrie JG, Thompson MG, Aiello AE, Monto AS. **Factors associated with influenza vaccine receipt in community dwelling adults and their children**. *Vaccine* (2014.0) **32** 1841-7. DOI: 10.1016/j.vaccine.2014.01.075
66. Flicoteaux R, Pulcini C, Carrieri P, Schwarzinger M, Leport C, Verger P. **Correlates of general practitioners' recommendations to patients regarding vaccination for the 2009–2010 pandemic influenza (A/H1N1) in France: implications for future vaccination campaigns**. *Vaccine* (2014.0) **32** 2281-7. DOI: 10.1016/j.vaccine.2014.02.074
67. Halboup A, Abdi A, Ahmed M, Al-Qadasi F, Othman G. **Access to antibiotics without prescription in community pharmacies in Yemen during the political conflict**. *Public Health* (2020.0) **183** 30-5. DOI: 10.1016/j.puhe.2020.03.003
68. Torbosh A, Al Amad MA, Al Serouri A, Khader Y. **The impact of war in Yemen on immunization coverage of children under one year of age: descriptive study**. *JMIR Public Health Surveill.* (2019.0) **5** e14461. DOI: 10.2196/14461
|
---
title: Differential expression profile of urinary exosomal microRNAs in patients with
mesangial proliferative glomerulonephritis
authors:
- Rong Dai
- Lei Zhang
- Hua Jin
- Dong Wang
- Meng Cheng
- Yunhui Xu
- Haiyin Zhang
- Yiping Wang
journal: Aging (Albany NY)
year: 2023
pmcid: PMC9970301
doi: 10.18632/aging.204527
license: CC BY 3.0
---
# Differential expression profile of urinary exosomal microRNAs in patients with mesangial proliferative glomerulonephritis
## Abstract
Objective: To investigate the differential expression profile of urinary exosomal microRNA (miRNA) in patients with mesangial proliferative glomerulonephritis (MsPGN) and healthy controls and their potential role in the pathogenesis of MsPGN.
Methods: Urine specimens were collected from five MsPGN patients and five healthy controls, and differentially expressed miRNAs were screened using high-throughput sequencing technology. The sequenced urinary exosomal miRNAs were further investigated by quantitative real-time polymerase chain reaction (qRT-PCR) in a validation cohort (16 MsPGN patients and 16 healthy controls). Correlation and receiver operating characteristic (ROC) curve analyses were used to determine the association between clinical features and miRNA expression in MsPGN. Finally, fluorescence in situ hybridization was performed to detect miRNA expression in the renal tissues of MsPGN patients.
Results: Five differentially expressed miRNAs (miR-125b-2-3p, miR-205-5p, let-7b-3p, miR-1262, and miR-548o-3p) were identified by qRT-PCR. The expression of these miRNAs correlated with ACR, 24hUpro, mAlb, UA, and combined yielded a ROC curve area of 0.916 in discriminating MsPGN patients from the controls. In addition, the expression of miR-205-5p, let-7b-3p, miR-1262, and miR-548o-3p was elevated in the MsPGN patient group, and miR-125b-2-3p was decreased in the MsPGN patient group.
Conclusions: Differential expression of urinary exosomal miRNAs may pose a risk of MsPGN and help distinguish MsPGN patients from controls. Certain miRNA expressions may be associated with disease progression, contributing to the epigenetic understanding of the pathophysiology of MsPGN.
## INTRODUCTION
Mesangial proliferative glomerulonephritis (MsPGN), one of the most important kidney diseases, is characterized by apoptosis, proliferation, and extracellular matrix (ECM) secretion of glomerular mesangial cells (GMCs) and is representative of nephritis leading to chronic kidney disease (CKD) [1, 2]. Furthermore, MsPGN accounts for approximately $10.5\%$ of primary glomerulonephritis diseases in China and can develop into glomerulosclerosis, interstitial fibrosis, and end-stage renal disease [3]. The etiology of MsPGN is unclear, and some studies have reported that it is a severe immune-mediated inflammatory disease caused by the deposition of immune circulating complexes in the glomerular thylakoid membrane through some antigens, which stimulate the body to produce antibodies [4]. To date, the pathogenesis of MsPGN has not been well elucidated.
Epigenetic mechanisms can affect gene expression and function without altering the underlying DNA sequence and mediate crosstalk between genes and the environment [5]. Unlike dichotomous genetic variation, they provide continuous regulation of the genetic effects of phenotypic changes. Epigenetic regulation includes non-coding RNAs, DNA hypo- or hypermethylation, histone modifications, and heterochromatin [6]. MicroRNAs (miRNAs) are single-stranded non-coding RNAs consisting of 18–25 nucleotides that can repress specific target messenger RNAs (mRNAs) by cleavage or translation and post-transcriptionally regulate gene expression variously, with significant biological functions in the physiology and pathology of various diseases [7]. Exosomes are extracellular vesicles with a diameter of 40–160 nm (average 100 nm), comprising lipids, proteins, and nucleic acids, including mRNA, miRNA, or long-stranded non-coding RNA [8]. They are present in almost all biological fluids, including urine [9].
Currently, renal biopsy is the gold standard for diagnosing MsPGN; however, since it is an invasive method, repeated renal biopsies are not significantly effective in assessing disease severity and progression. It has been shown that a few miRNAs are highly expressed in the plasma of MsPGN patients compared with that of healthy controls, highlighting the possible presence of altered miRNA levels in the disease that could serve as a potential novel diagnostic biomarker for MsPGN [10]. Urinary exosomes are composed of proteins, mRNA and miRNA produced by glomerular cells (podocytes, endothelial cells, and GMCs), and renal tubular cells [11]. Thus, urinary exosomes may provide sensitive and accurate biomarkers of renal dysfunction and structural damage [12]. Urine is a suitable sample source for isolating and extracting exosomes, as it is easily accessible.
Numerous studies have revealed different urinary exosome miRNA expression patterns in patients with kidney disease. Urinary exosomes miR-29c, miR-146a, and miR-205 may serve as IgAN biomarkers [13], while miR-3135b, miR-654-5p, and miR-146a-5p are candidate biomarkers for combined cellular crescent in type IV lupus nephritis [14], and miR-146a may be a potential indicator of early renal injury in hypertension [15]. However, the MsPGN urinary exosome miRNA expression profile is yet to be elucidated. Therefore, it is crucial to explore new noninvasive diagnostic biomarkers. MiRNA expression is differentially expressed in circulating exosomes in MsPGN, which could result in aberrant regulatory information being transmitted to the target organs and functional abnormalities of associated genes and pathways. Considering the underlying mechanisms of multisystem diseases, studying exosomal miRNAs may reveal new targets and aberrant epigenetic regulation contributing to the risk of MsPGN. Therefore, in this study, we determined the expression profile of exosomal miRNAs in the urine of MsPGN patients to partially reveal the pathophysiology of the disease, which could be used to design prospective molecular targets for MsPGN diagnosis.
## Clinical characteristics
The MsPGN patient and control groups participating in the sequencing and validation study and their clinical characteristics are summarized in Table 1. No statistically significant differences were observed between the MsPGN patient and control groups in age, sex, blood pressure, Alanine transaminase (ALT), Aspartate transaminase (AST), *Blood urea* nitrogen (BUN), Serum creatinine (Scr), urinary β2-microglobulin (β2-MG), urine red blood cell (URBC), Immunoglobulin G (IgG), and complement C3 (C3). However, there were significant differences in Total Protein (TP), albumin (ALB), Urine Albumin to. Creatinine Ratio (ACR), Urine Total Protein to. Creatinine Ratio (TCR), 24 hours urine protein (24hUpro), microalbuminuria (mAlb), uric acid (UA), Immunoglobulin A (IgA), Immunoglobulin M (IgM), and cystatin C (CysC) between the MsPGN patient and control groups.
**Table 1**
| Parameters | MsPGN | Control | t/χ2 | P | Difference and 95% confidence interval |
| --- | --- | --- | --- | --- | --- |
| Age | 38.57±15.36 | 36.14±3.39 | 0.708 | 0.483 | 2.49 (-4.51–9.36) |
| Sex [M/F, n(%) | 9(42.86)/12(57.14) | 10(47.62)/11(52.38) | | | |
| Systolic blood pressure (mmHg) | 121.57±14.37 | 120.19±1.81 | 0.437 | 0.664 | 1.38 (-5.01–7.76) |
| Diastolic blood pressure (mmHg) | 79.86±16.38 | 80.29±1.90 | -0.119 | 0.906 | -0.43 (-7.70–6.85) |
| ALT (U/L) | 24.10±14.65 | 23.76±6.70 | 0.095 | 0.925 | 0.33 (-6.77–7.43) |
| AST (U/L) | 22.33±10.49 | 17.76±3.13 | 1.914 | 0.063 | 4.57 (-0.25–9.39) |
| TP (g/dL) | 55.52±12.03 | 67.46±6.08 | 4.061 | 0.001 | -11.94 (-17.88–5.99) |
| ALB (g/dL) | 32.41±7.71 | 40.14±3.05 | 4.273 | 0.001 | -7.72 (-11.38–4.07) |
| BUN | 6.21±2.15 | 5.29±1.37 | 1.659 | 0.105 | 0.92 (-0.21–2.05) |
| Scr | 118.35±152.67 | 50.03±4.26 | 2.05 | 0.054 | 68.32 (0.96–135.68) |
| ACR (mg/mmol) | 1,557.92±2,216.29 | 12.40±4.07 | 3.196 | 0.005 | 1,545.51 (568.05–2,522.98) |
| TCR (mg/mmol) | 2.57±3.39 | 0.09±0.04 | 3.35 | 0.003 | 2.48 (0.98–3.98) |
| 24hUpro (g/24 h) | 2.73±3.25 | 0.09±0.03 | 3.717 | 0.001 | 2.64 (1.21–4.07) |
| mAlb (mg/L) | 1,617.42±1,704.14 | 9.73±1.06 | 4.323 | 0.001 | 1,607.69 (856.10–2,359.27) |
| β2-MG (mg/L) | 0.60±1.11 | 0.009±0.007 | 1.627 | 0.114 | 0.60 (0.11–1.09) |
| UA (μmol/L) | 424.57±96.51 | 217.37±52.33 | 8.606 | 0.001 | 206.20 (157.77–254.61) |
| URBC | 347.07±992.84 | 1.96±0.09 | 1.593 | 0.127 | 345.10 (-92.77–782.99) |
| IgA (g/L) | 2.45±0.91 | 0.98±0.56 | 6.317 | 0.001 | 1.48 (1.00–1.95) |
| IgG (g/L) | 7.41±2.77 | 7.02±0.86 | 0.612 | 0.544 | 0.38 (-0.89–1.67) |
| IgM (g/L) | 0.99±0.34 | 0.80±0.08 | 2.53 | 0.019 | 0.19 (-0.04–0.35) |
| C3 (mg/dL) | 1.13±0.23 | 1.19±0.03 | 1.308 | 0.198 | -0.06 (-0.17–0.04) |
| CysC (mg/L) | 1.09±0.43 | 0.71±0.32 | 3.205 | 0.003 | 0.37 (0.13–0.61) |
## Identification of isolated exosomes in urine
Exosomes from human urine are characterized by their morphology, diameter, concentration, and the presence of exosome-rich protein markers such as CD9, CD63, and TSG101. Transmission electron microscopy revealed exosomal bilayers of membranous structure and oval shape, with a typical cup-shaped structures, ranging in size from 30–100 nm. The results of particle size analysis of urinary exosomes in normal humans showed that the average particle size of urinary exosomes was 75.97 nm after 100-fold dilution, and a total of 5,985 particles were detected, with $99.78\%$ of the particles being 30–150 nm and were at an average concentration of 2.64 × 1010 particles/mL. In patients with MsPGN, the mean particle size of urinary exosomes was 76.89 nm, and a total of 4,462 particles were detected, with $99.89\%$ being 30–150-nm particles at a mean concentration of 3.93 × 1011 particles/mL. In addition, western blotting revealed that these exosomes were positive for CD9, CD63, and TSG101 proteins (Figure 1). CD9 and CD63 are exosome-enriched proteins of the tetraspanin family, whereas TSG101 is a protein essential for multivesicular body biogenesis.
**Figure 1:** *(A) Electron microscopic images of extracted exosomes revealed cup-shaped structures with a diameter of about 30–160 nm. Scale bar: 100 nm. (B) Nanoparticle tracking analysis (NTA) revealed the diameter of isolated extracellular vesicles (EVs) is consistent with that of exosomes. (C) Western blotting revealed CD9, CD63 and TSG101 proteins in exosome samples.*
By comparing the TargetScan, miRDB, and miRWalk databases, 334, 445, and 279 miRNAs were found to be expressed in the normal control group, MsPGN patients, and co-expressed in both groups, respectively. The 50 significantly differentially expressed miRNAs (Table 2) were derived based on the log2 (fold change) absolute value ≥ 1, q-value < 0.05 screening criteria, among which 24 and 26 miRNAs were upregulated and downregulated, respectively (Figure 2A, 2B).
## Validation of differentially expressed miRNAs
The validation cohort comprised a sample size of 32 (16 MsPGN patients and 16 healthy controls). To ensure the accuracy of the sequencing results, miRNAs for the subsequent PCR validation were screened according to the following criteria: [1] the mean expression (TPM) within at least one of the two groups of samples was > 2; [2] the absolute value of log2 (fold change) was ≥ 2; [3] at least one sample in the MsPGN group had a TPM > 0. The final selection of upregulated miR-125b -2-3p, miR-1262, miR-508-3p, let-7b-3p, miR-548o-3p, and miR-193b-3p and down-regulated miR-205-5p and miR-891a-5p were used for PCR validation. After statistical analysis, there were significant differences in the expression of miR-125b-2-3p, miR-1262, let-7b-3p, miR-548o-3p, miR-205-5p, and miR-193b-3p in the MsPGN group compared with that in the normal control group ($P \leq 0.01$); let-7b-3p expression was generally significant ($P \leq 0.05$), whereas miR-508-3p and miR-891a-5p expression differences were not statistically significant ($P \leq 0.05$). Among them, the miR-193b-3p expression differences were inconsistent with the direction of sequencing results, although they were significant. Therefore, miR-125b-2-3p, miR-1262, let-7b-3p, miR-548o-3p, and miR-205-5p were finally selected for the next step of target gene prediction and KEGG pathway and GO enrichment analyses. The relative expression levels of the five differentially expressed miRNAs are shown in Figure 3C.
**Figure 3:** *(A) Top 20 statistically significantly enriched pathways in Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. (B) Gene Ontology (GO) analysis for the target genes of five differentially expressed miRNAs. (C) The relative expression level of five differentially expressed miRNAs.*
## KEGG and GO analyses of differentially expressed miRNAs
Based on RNA sequencing results and subsequent validation, we revealed that five miRNAs (miR-125b-2-3p, miR-1262, let-7b-3p, miR-548o-3p, and miR-205-5p) were differentially expressed between the two groups. The target genes of the five miRNAs were predicted using TargetScan, miRDB, and miRWalk databases. The main biological processes, molecular functions, and cellular components of the target genes of the five miRNAs were identified by GO analysis. A total of 246 meaningful GO functions were obtained, including 47 molecular functions (mainly ubiquitin-protein ligase, protein serine/threonine kinase, DNA-binding transcription factor, and protein kinase activities), 44 cellular components (mainly nucleoplasm, cytoplasm, and post-emphasis density), and 155 biological processes (mainly positive and negative regulation of RNA polymerase II promoter transcription, positive and negative regulation of gene expression, cell migration). The top 30 GO functions identified by GO annotation are shown in Figure 3B. To better understand the biological functions of the predicted target genes, KEGG analysis was performed to reveal the main pathways in which the candidate target genes might be involved. The top 20 significantly enriched pathways were identified, including the PI3K/Akt, MAPK, Rap1, and mTOR signaling pathways (Figure 3A), and classified into environmental information processing (six), molecular processes (three), organismal systems (four), and human diseases (seven).
## Correlation between clinical characteristics and urinary exosome miRNA in MsPGN patients
Spearman order correlation analysis was used to analyze the correlations between the five exosomal differential miRNAs (miR-125b-2-3p, miR-1262, let-7b-3p, miR-548o-3p, and miR-205-5p) and several clinical indicators in MsPGN patients. A P-value < 0.05 indicated correlation, and the larger the r-value, the greater the correlation. Both ACR and 24hUpro were inversely correlated with miR-125b-2-3p expression. Additionally, there was a positive correlation between uric acid levels and miR-548o-3p. miR-205-5p was positively correlated with urinary microalbumin. TP, ALB, TCR, IgA, and CysC were not correlated with any of the differentially expressed miRNAs (Table 3 and Figure 4). Meanwhile, fluorescence in situ hybridization was performed to detect miRNAs expression in renal tissues of MsPGN patients. The expression of miR-205-5p, let-7b-3p, miR-1262, and miR-548o-3p was elevated in the MsPGN patient group, and miR-125b-2-3p was decreased in the MsPGN patient group (Figure 5).
## DISCUSSION
The current study identified the urinary exosomal differential miRNA profile of MsPGN patients. The results revealed statistically significantly increased miR-125b-2-3p, miR-1262, let-7b-3p, and miR-548o-3p and decreased miR-205-5p. Differential miRNAs target genes regulate genes involved in cancer and PI3K/Akt, MAPK, Rap1, and mTOR signaling pathways. The expression levels of the five miRNAs were mainly correlated with 24-h urine protein quantification and ACR, mAlb, and UA levels.
Among these differential miRNAs, miR-205-5p has been previously associated with renal diseases and is involved in angiogenesis and regulates related cellular signaling pathways such as cell migration, proliferation, and apoptosis [16]. Jie et al. found that circ-ACTR2 levels were upregulated in diabetic kidney disease and human renal mesangial cells (HRMCs) treated with high glucose. Silencing circ-ACTR2 expression partially abrogated high-glucose-induced cell proliferation, inflammation, ECM accumulation, and oxidative stress in HRMCs [17]. In addition, miR-205-5p directly targets the high mobility group AT-hook2 (HMGA2), and HMGA2 downregulation protects against high-glucose-treated HRMCs injury. Circ-ACTR2 mediates high-glucose-induced HRMC injury through the miR-205-5p/HMGA2 axis. In addition, miR-205-5p is reportedly involved in renal clear cell carcinoma and colorectal cancer cell proliferation, invasion, and migration [18, 19]. In the present study, urinary exosomal miR-205-5p was downregulated in MsPGN patients, consistent with its expression trend in diabetic kidney disease. Furthermore, we found that miR-205-5p is associated with urinary microalbumin; therefore, whether miR-205-5p mediates downstream signaling pathways and thus affects thylakoid region proliferation deserves further investigation.
miR-125b, miR-1262, let-7b-3p, and miR-548o-3p were significantly upregulated. miR-125b-2-3p, miR-1262, and let-7b-3p have been primarily studied in cancer and are involved in cancer development through their target genes and downstream signaling pathways. However, miR-548o-3p has not been studied. Zeng et al. found that miR-125b-2-3p expression was significantly reduced in colorectal cancer tissues and cell lines. The high miR-125b-2-3p expression was associated with relatively lower proliferation rates and less metastasis, and functional experiments showed that miR-125b-2-3p overexpression attenuated tumor cell proliferation and epithelial-mesenchymal transition, and downregulation of its expression was associated with lower proliferation and metastasis. Moreover, miR-125b-2-3p expression promotes cell development and metastasis in vitro and in vivo [20]. In addition, Meng et al. proposed that miR-125b-2-3p expression was elevated in renal clear cell carcinoma, accelerated the migration of renal clear cell carcinoma cells, and promoted tumor metastasis by downregulating its target gene EGR1. The role of miR-125b-2-3p in different cancers is related to disease type, pathogenesis, and disease progression is a complex process regulated by multiple targets and pathways [21]. Ling et al. found that LRP8 is highly expressed in breast cancer tissues and cell lines compared with that in normal human breast tissues. The poor prognosis of breast cancer patients has been associated with the upregulation of LRP8 and negatively correlated with miR-1262 overexpression. Functionally, deletion of LRP8 in breast cancer cells impairs cell proliferation, clone formation, invasion, and migration abilities, consistent with an upregulated miR-1262 effect. Bioinformatics predictions and luciferase reporter analysis confirmed that miR-1262 is an LRP8 upstream factor and negatively regulates LRP8 expression. Therefore, miR-1262-regulated LRP8 can regulate cell proliferation and migration and other processes to promote breast cancer development [22]. Sun et al. reported that has-miR-1262 expression was significantly increased in exosomes from septic patients and regulated apoptosis and glycolysis in human cardiomyocytes via the has-miR-1262/SLC2A1 signaling pathway [23]. Li et al. found that Let-7b-3p is downregulated in lung adenocarcinoma cells and tissue samples, and low Let-7b-3p expression is associated with poor prognosis in lung adenocarcinoma patients. Furthermore, Let-7b-3p inhibits proliferation and metastasis of lung adenocarcinoma cells in vivo and in vitro by directly targeting the BRF2-mediated MAPK/ERK pathway [24]. In summary, current studies on miR-125b-2-3p, miR-1262, and let-7b-3p are mainly involved in cancer development by interfering with downstream signaling pathways to regulate biological processes such as cell proliferation and migration and apoptosis. In the present study, miR-125b-2-3p, miR-1262, and let-7b-3p were significantly highly expressed in urinary exosomes of MsPGN patients. Whether these miRNAs mediate pathological MsPGN damage by interfering with downstream-related signaling pathways to regulate cell proliferation and migration, apoptosis, and other biological processes warrants further study.
To verify the function of these differential miRNAs in MsPGN patients, we performed GO and KEGG analyses. The biological processes primarily involved the negative regulation of RNA polymerase II promoter transcription, positive regulation of DNA template transcription, and cell migration. Moreover, several enriched pathways were indicated, among which PI3K/Akt, MAPK, Rap1, and mTOR signaling pathways were listed as the top 4. The proliferation of GMCs in MsPGN appears to be critical for the subsequent increase in the ECM and glomerulosclerosis development [25, 26]. Activation of the PI3K/Akt signaling pathway has been shown to be associated with cell proliferation and ECM synthesis [27–29]. Regarding PI3K/Akt signaling pathway, numerous studies have demonstrated its important regulatory role in GMCs proliferation, and Liu et al. showed that paeoniflorin could effectively reduce 24-h urinary protein in MsPGN rats. The protective effect of paeoniflorin was accompanied by strong inhibition of the PI3K/AKT/GSK-3β pathway. Paeoniflorin enhanced the inhibitory effect of the PI3K inhibitor LY294002 and inhibited the activation of the PI3K/AKT/GSK-3β pathway by the PI3K agonist insulin-like growth factor 1 (IGF-1), thus downregulating the PI3K/AKT/GSK-3β pathway to inhibit thylakoid cell proliferation and the inflammatory response and ameliorate pathological injury in MsPGN [30]. Li et al. found that PAQR3 expression is significantly upregulated in human thylakoid cells under a high glucose environment [31]. PAQR3 knockdown effectively inhibits proliferation and ECM production and significantly reverses PI3K/AKT pathway activation of HRMCs stimulated by high glucose. The important role of MAPK in mediating apoptosis by activating apoptosis-related genes is supported by extensive evidence [32, 33]. Sublytic C5b-9 induces apoptosis via MEKK2-p38 MAPK-IRF-1-TRADD-Caspase 8 in rat Thy-1 nephritis in GMCs [34]. The inflammatory response is thought to be an important factor in the MsPGN development [35, 36]. C5α is a potent pro-inflammatory mediator that correlates with the severity of various renal diseases and induces the synthesis of IL-6 and TNF-α in rat GMCs through MAPK signaling pathway activation [37]. Rap1 (also known as Krev-1) is a small GTPase belonging to the Ras superfamily that is involved in the formation and stabilization of E-cadherin-based cell–cell adhesion in epithelial cells [38–40]. Thus, Rap1 is a key mediator of integrin-mediated cell–ECM adhesion [41].
In addition, we performed a clinical analysis using Spearman correlation and simple linear regression analyses to elucidate the relationship between each differentially expressed miRNA and various MsPGN clinical features. According to our findings, both ACR and 24hUpro showed significant negative correlations with miR-125b-2-3p. Moreover, there was a positive correlation between uric acid levels and miR-548o-3p, and miR-205-5p was positively correlated with urinary microalbumin. Since miRNA expression correlated with different clinical features of MsPGN, we combined these five miRNAs in ROC analysis with an area under the curve of 0.916, suggesting high diagnostic efficacy.
## CONCLUSIONS
In this study, we measured the urinary exosomal miRNA profile of MsPGN. Exosomal miRNAs can be fully carried and released into target cells to regulate subsequent transcriptional processes. This cell–cell-mediated and epigenetic regulation could better explain the phenotype of MsPGN rather than genetic variation. Moreover, this study is the first report on the differential expression profile of urinary exosomal miRNAs in MsPGN patients. Considering miRNA-driven disease regulation, future translational research in this field may provide new diagnostic and therapeutic approaches to treat MsPGN. However, some limitations remain. The cohort in the present study is relatively small; thus, the results need to be validated in a larger population with other specific phenotypes. Nevertheless, the present study extends our understanding of the urinary exosomal miRNA expression profile in MsPGN patients, and the aberrant epigenetic regulation increases the risk of the syndrome and may lead to complications, with specific mechanisms and target cells and tissues to be elucidated by further studies.
## Study participants
All MsPGN patients were from the wards and outpatient clinics of the Department of Nephrology, the First Affiliated Hospital of Anhui University of Chinese Medicine (October 2019–October 2021), and baseline demographic and clinical data were recorded at the time of renal biopsy. The control group was derived from sex- and age-matched healthy subjects (from healthy volunteers at the medical screening center). Patients were all aged 18–70 years; five were MsPGN patients, and five were controls enrolled as the discovery cohort for exosomal miRNA sequencing. An additional 16 participants were enrolled in each group, increasing the validation cohort to 21 MsPGN patients vs. 21 controls. Data for all participants were obtained with their informed consent, and the study was approved by the ethical review committee of the Anhui University of Chinese Medicine.
## Exosome isolation and RNA extraction
The morning urine of patients with MsPGN and that of normal healthy subjects was collected in 50-mL centrifuge tubes and sent to the laboratory within 1 h for preparation for centrifugation. Subsequently, the urine samples were centrifuged at 500 × g at 4° C for 10 min, after which the sediment was removed and the remaining supernatant transferred to new 50-mL sterile centrifuge tubes. The supernatant from the previous step was then centrifuged at 2,500 × g at 4° C for 30 min. After centrifugation, cells and debris were removed, the sediment discarded, and the supernatant transferred to new 50-mL Beckman special centrifuge tubes (Beckman Coulter, Brea, CA, USA). These were then centrifuged at 17,000 × g at 4° C for 30 min to remove large vesicles, after which the sediment was discarded. The resulting supernatant was transferred to new Beckman centrifuge tubes and centrifuged at 110,000 × g for 80 min at 4° C; the granular precipitate obtained at the bottom of the tube was the exosome. After resuspending the precipitate from the previous step with PBS, it was centrifuged at 110,000 × g for 80 min at 4° C, and the pellet-like precipitate at the bottom of the tube was the purified exosome. This was carefully blown with 100 μL of PBS for resuspension, after which the exosome suspension was aspirated into an EP tube and stored at -80° C in a refrigerator.
Total RNA was extracted using a mirVana miRNA isolation kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol and quantified using a Nanodrop2000 Spectrophotometer (Thermo Fisher Scientific). RNA integrity was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA).
## MicroRNA library construction and sequencing
One microgram of total RNA per sample was used for small RNA library construction using a TruSeq small RNA sample preparation kit (Cat. No. RS-200-0012; Illumina, San Diego, CA, USA). Briefly, total RNA was ligated to the adapters at both ends, after which the splice ligated RNA was reverse transcribed to cDNA and PCR amplified. Thereafter, the small RNA libraries were isolated and purified from the 140–160-bp PCR products. Library quality was assessed using an Agilent Bioanalyzer 2100 system (Agilent Technologies) with a DNA high sensitivity chip. The libraries were finally sequenced using an Illumina HiSeq X Ten platform (Illumina), and 150-bp paired-end reads were generated. The sequencing of small RNA library construction and data analysis were partially conducted by OE Biotech Co., Ltd. (Shanghai, China).
## Western blot
Gel configuration was conducted according to the Takara catalog (Takara Bio Inc., Shiga, Japan). The sample stock solution was diluted to the desired loading volume, and then 5× SDS-PAGE protein loading buffer was added to a sample dilution of 1:4. Next, the proteins were fully denatured in a boiling water bath for 15 min. After cooling to room temperature, an appropriate amount (approximately 100 μg) was subjected to SDS-PAGE electrophoresis and transferred to a polyvinylidene difluoride (PVDF) membrane. Subsequently, the membrane was blocked with $5\%$ skimmed milk. Thereafter, primary antibodies were added and diluted with primary antibody diluent at the appropriate ratios [CD9 (#Ab92726, 1:500; Abcam), CD63 (#Ab216130, 1:500; Abcam), and TSG101 (#Ab125011, 1:3,000; Abcam)] and incubated overnight at 4° C with slow shaking. Secondary antibodies included HRP-labeled Goat Anti-Mouse IgG(H+L) antibody (A0216, 1:1,000; Beyotime) and HRP-labeled Goat Anti-Rabbit IgG(H+L) antibody (A0208, 1:1,000; Beyotime). Finally, the PVDF membrane was immersed in the developing solution to develop the exposure, and the images were acquired using a UVP gel imaging analyzer (US, GelDoc-It Ts3 Imaging System).
## Quantitative real-time PCR
Total RNA was extracted from urinary exosomes using TRIzol reagent (Life Technologies, USA) and reverse transcribed to cDNA using a PrimeScript™ RT Reagent Kit with a gDNA Eraser (RR047A, AL21115A; Takara Bio Inc.) for miRNA expression verification. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the internal reference. All the primer sequences were designed and synthesized by Sangon Biotech Company (Shanghai, China) and data were analyzed according to the 2-ΔΔCt method. Primer design as shown in Table 4.
**Table 4**
| Gene | Forward primer (5'→3') | Reverse primer (5'→3') |
| --- | --- | --- |
| U6 | CTCGCTTCGGCAGCACA | AACGCTTCACGAATTTGCGT |
| hsa-let-7b-3p | ACACTCCAGCTGGGCTATACAACCTACTGC | TGGTGTCGTGGAGTCG |
| hsa-miR-125b-2-3p | ACACTCCAGCTGGGTCACAAGTCAGGCTCT | TGGTGTCGTGGAGTCG |
| hsa-miR-1262 | ACACTCCAGCTGGGATGGGTGAATTTGTAG | TGGTGTCGTGGAGTCG |
| hsa-miR-205-5p | ACACTCCAGCTGGGTCCTTCATTCCACCGG | TGGTGTCGTGGAGTCG |
| hsa-miR-548o-3p | ACACTCCAGCTGGGCCAAAACTGCAGTTAC | TGGTGTCGTGGAGTCG |
## Transmission electron microscopy
The exosome samples were removed from the -80-° C refrigerator, placed in an ice box, dissolved, and centrifuged, after which 15 μL were pipetted onto a copper grid for 1 min. Next, the exosome samples were blotted dry on the copper grid using filter paper, and then 15 μL of $2\%$ UO2 acetate staining solution were pipetted onto the copper grid for 1 min at room temperature. Finally, the finished samples were baked under a lamp for 10 min, and the final images were observed using a transmission electron microscope at 80 kV (Tecnai, G2 spirit; FEI Company, Hillsboro, OR, USA), photographed, and saved.
## Nanoparticle tracking analysis
The sample cells were washed with ultrapure water, and the instrument (NanoFCM, N30E) was calibrated using polystyrene microspheres (100 nm). Next, the exosomes were removed and diluted to the appropriate magnification. The instrument should be tested with the standard before the sample is loaded, and the sample should be diluted in a gradient to avoid blocking the injection needle. The particle size and concentration information of exosomes can then be obtained after the sample is tested.
## Fluorescence in situ hybridization
The expression levels and localization of 5 miRNAs in renal tissues were detected by fluorescence in situ hybridization (FISH). Five probes for Cy3-labeled miRNAs were designed (hsa-miR-125b-2-3p, 5'-GTCCCAAGAGCC+TGACT+TGTGA-3'; hsa-miR-1262, 5'-ATCCT+ TCTACAAAT+TCACCCAT-3'; hsa-let-7b-3p, 5'-GGGAAGGCAG+TAGGT+TGTATAG-3'; hsa-miR-548o-3p, 5'-GCAAAAGTAAC+TGCAGT+TTTGG-3'; hsa-miR-205-5p, 5'-CAGAC+TCCGG+TGGAATGAAGGA-3', purchased from GenePharma (Shanghai, China). Each probe was hybridized overnight according to the manufacturer's instructions.
## Bioinformatics analysis
To predict the target genes of differential miRNAs and take their intersection as candidate target genes, TargetScan Human version 7.2 (http://www.targetscan.org/vert_72/), MicroRNA Target Prediction Database (miRDB; http://mirdb.org/), and miRWalk (http://mirwalk.umm.uni-heidelberg.de/) were used. Gene ontology (GO; http://www.geneontology.org) analysis was then performed to analyze the main functions of the putative target genes, and Kyoto Encyclopedia of Genes and Genomes (KEGG; https://www.genome.jp/kegg) analysis was used to identify potentially altered molecular pathways.
## Statistical analyses
Statistical analyses were performed using SPSS (version 25.0; SPSS Inc., Chicago, IL, USA). Continuous variables were tested for abnormality using the Kolmogorov–Smirnov test. To determine statistical significance, normally distributed variables were analyzed using Student’s t-test, and nonparametric data were assessed using the Wilcoxon–Mann–Whitney test. Receiver operating characteristic (ROC) curve analysis was used to assess the diagnostic potential of plasma exosomal miRNAs to differentiate MsPGN samples from controls. Logistic regression analysis was used to assess the diagnostic efficacy of miRNA combinations. Spearman correlation analysis was performed to assess the correlation between miRNA expression and MsPGN phenotype. $P \leq 0.05$ was considered statistically significant.
## References
1. Geng W, Wei R, Liu S, Tang L, Zhu H, Chen P, Wu J, Zhang X, Zhu F, Yin Z, Chen X. **Shenhua Tablet inhibits mesangial cell proliferation in rats with chronic anti-Thy-1 nephritis.**. *Biol Res* (2016) **49** 17. DOI: 10.1186/s40659-016-0078-3
2. Jin M, Yin Z, Wei K, Xie Y, Bai X, Fu B, Feng Z, Li Q, Chen X. **Metanephric mesenchyme-derived Foxd1+ mesangial precursor cells alleviate mesangial proliferative glomerulonephritis.**. *J Mol Med (Berl)* (2019) **97** 553-61. DOI: 10.1007/s00109-019-01749-1
3. Yang Y, Zhang Z, Zhuo L, Chen DP, Li WG. **The Spectrum of Biopsy-Proven Glomerular Disease in China: A Systematic Review.**. *Chin Med J (Engl)* (2018) **131** 731-5. DOI: 10.4103/0366-6999.226906
4. Han X, Zheng X, Wang Y, Sun X, Xiao Y, Tang Y, Qin W. **Random forest can accurately predict the development of end-stage renal disease in immunoglobulin a nephropathy patients.**. *Ann Transl Med* (2019) **7** 234. DOI: 10.21037/atm.2018.12.11
5. Fitz-James MH, Cavalli G. **Molecular mechanisms of transgenerational epigenetic inheritance.**. *Nat Rev Genet* (2022) **23** 325-41. DOI: 10.1038/s41576-021-00438-5
6. Zhang L, Lu Q, Chang C. **Epigenetics in Health and Disease.**. *Adv Exp Med Biol* (2020) **1253** 3-55. DOI: 10.1007/978-981-15-3449-2_1
7. Yao Q, Chen Y, Zhou X. **The roles of microRNAs in epigenetic regulation.**. *Curr Opin Chem Biol* (2019) **51** 11-7. DOI: 10.1016/j.cbpa.2019.01.024
8. Kalluri R, LeBleu VS. **The biology, function, and biomedical applications of exosomes.**. *Science* (2020) **367** eaau6977. DOI: 10.1126/science.aau6977
9. Cheng L, Sun X, Scicluna BJ, Coleman BM, Hill AF. **Characterization and deep sequencing analysis of exosomal and non-exosomal miRNA in human urine.**. *Kidney Int* (2014) **86** 433-44. DOI: 10.1038/ki.2013.502
10. Wang L, Lin J, Yu T, Zuo Q, Shen B, Zhang H, Liu B, Cai D, Mao H, Zhao H, Zou Q, Xiao B. **Identification of plasma miR-106a-5p and miR-30a-5p as potential biomarkers for mesangial proliferative glomerulonephritis.**. *Clin Biochem* (2020) **84** 79-86. DOI: 10.1016/j.clinbiochem.2020.07.001
11. Vitorino R, Ferreira R, Guedes S, Amado F, Thongboonkerd V. **What can urinary exosomes tell us?**. *Cell Mol Life Sci* (2021) **78** 3265-83. DOI: 10.1007/s00018-020-03739-w
12. Sinha N, Kumar V, Puri V, Nada R, Rastogi A, Jha V, Puri S. **Urinary exosomes: Potential biomarkers for diabetic nephropathy.**. *Nephrology (Carlton)* (2020) **25** 881-7. DOI: 10.1111/nep.13720
13. Min QH, Chen XM, Zou YQ, Zhang J, Li J, Wang Y, Li SQ, Gao QF, Sun F, Liu J, Xu YM, Lin J, Huang LF. **Differential expression of urinary exosomal microRNAs in IgA nephropathy.**. *J Clin Lab Anal* (2018) **32** e22226. DOI: 10.1002/jcla.22226
14. Li Y, Xu X, Tang X, Bian X, Shen B, Zhao H, Luo S, Chen Z, Zhang K. **MicroRNA expression profile of urinary exosomes in Type IV lupus nephritis complicated by cellular crescent.**. *J Biol Res (Thessalon)* (2018) **25** 16. DOI: 10.1186/s40709-018-0088-0
15. Perez-Hernandez J, Olivares D, Forner MJ, Ortega A, Solaz E, Martinez F, Chaves FJ, Redon J, Cortes R. **Urinary exosome miR-146a is a potential marker of albuminuria in essential hypertension.**. *J Transl Med* (2018) **16** 228. DOI: 10.1186/s12967-018-1604-6
16. Oltra M, Vidal-Gil L, Maisto R, Sancho-Pelluz J, Barcia JM. **Oxidative stress-induced angiogenesis is mediated by miR-205-5p.**. *J Cell Mol Med* (2020) **24** 1428-36. DOI: 10.1111/jcmm.14822
17. Yun J, Ren J, Liu Y, Dai L, Song L, Ma X, Luo S, Song Y. **Circ-ACTR2 aggravates the high glucose-induced cell dysfunction of human renal mesangial cells through mediating the miR-205-5p/HMGA2 axis in diabetic nephropathy.**. *Diabetol Metab Syndr* (2021) **13** 72. DOI: 10.1186/s13098-021-00692-x
18. Huang J, Wang X, Wen G, Ren Y. **miRNA-205-5p functions as a tumor suppressor by negatively regulating VEGFA and PI3K/Akt/mTOR signaling in renal carcinoma cells.**. *Oncol Rep* (2019) **42** 1677-88. DOI: 10.3892/or.2019.7307
19. Chen S, Wang Y, Su Y, Zhang L, Zhang M, Li X, Wang J, Zhang X. **miR-205-5p/PTK7 axis is involved in the proliferation, migration and invasion of colorectal cancer cells.**. *Mol Med Rep* (2018) **17** 6253-60. DOI: 10.3892/mmr.2018.8650
20. Zeng ZL, Lu JH, Wang Y, Sheng H, Wang YN, Chen ZH, Wu QN, Zheng JB, Chen YX, Yang DD, Yu K, Mo HY, Hu JJ. **The lncRNA XIST/miR-125b-2-3p axis modulates cell proliferation and chemotherapeutic sensitivity via targeting Wee1 in colorectal cancer.**. *Cancer Med* (2021) **10** 2423-41. DOI: 10.1002/cam4.3777
21. Meng X, Liu K, Xiang Z, Yu X, Wang P, Ma Q. **MiR-125b-2-3p associates with prognosis of ccRCC through promoting tumor metastasis via targeting EGR1.**. *Am J Transl Res* (2020) **12** 5575-85. PMID: 33042439
22. Li L, Qu WH, Ma HP, Wang LL, Zhang YB, Ma Y. **LRP8, modulated by miR-1262, promotes tumour progression and forecasts the prognosis of patients in breast cancer.**. *Arch Physiol Biochem* (2022) **128** 657-65. DOI: 10.1080/13813455.2020.1716019
23. Sun F, Geng H, Sun Y, Feng W, Tian T, Ye L, Lei M. **Exosomes derived from the blood of patients with sepsis regulate apoptosis and aerobic glycolysis in human myocardial cells via the hsa-miR-1262/SLC2A1 signaling pathway.**. *Mol Med Rep* (2022) **25** 119. DOI: 10.3892/mmr.2022.12635
24. Li Y, Dong R, Lu M, Cheng C, Feng Z, Zhao R, Liang J, Han J, Jiang J, Xu-Welliver M, Renaud S, Tian H. **Let-7b-3p inhibits tumor growth and metastasis by targeting the BRF2-mediated MAPK/ERK pathway in human lung adenocarcinoma.**. *Transl Lung Cancer Res* (2021) **10** 1841-56. DOI: 10.21037/tlcr-21-299
25. Liu L, Ge W, Zhang Z, Li Y, Xie M, Zhao C, Yao C, Luo C, Wu Z, Wang W, Zhao D, Zhang J, Qiu W, Wang Y. **Sublytic C5b-9 triggers glomerular mesangial cell proliferation via enhancing FGF1 and PDGFα gene transcription mediated by GCN5-dependent SOX9 acetylation in rat Thy-1 nephritis.**. *FASEB J* (2021) **35** e21751. DOI: 10.1096/fj.202002814RR
26. Liu L, Li F, Zou H, He W, Zhang B, Zhang J. **A case of proliferative glomerulosclerosis with compound heterozygous TTC21B mutations.**. *Clin Chim Acta* (2022) **529** 17-20. DOI: 10.1016/j.cca.2022.01.020
27. Zhan B, Zhang Z, Piao C, Dong X, Du Y, Kong C, Jiang Y. **The Sigma-2 Receptor/TMEM97 Agonist PB28 Suppresses Cell Proliferation and Invasion by Regulating the PI3K-AKT-mTOR Signalling Pathway in Renal Cancer.**. *J Cell Mol Med* (2021) **25** 11244-56. DOI: 10.1111/jcmm.17047
28. Abe-Fukasawa N, Watanabe R, Gen Y, Nishino T, Itasaki N. **A liquid culture cancer spheroid model reveals low PI3K/Akt pathway activity and low adhesiveness to the extracellular matrix.**. *FEBS J* (2021) **288** 5650-67. DOI: 10.1111/febs.15867
29. Wei D, Rui B, Qingquan F, Chen C, Ping HY, Xiaoling S, Hao W, Jun G. **KIF11 promotes cell proliferation via ERBB2/PI3K/AKT signaling pathway in gallbladder cancer.**. *Int J Biol Sci* (2021) **17** 514-26. DOI: 10.7150/ijbs.54074
30. Liu B, Lin J, Bai L, Zhou Y, Lu R, Zhang P, Chen D, Li H, Song J, Liu X, Wu Y, Wu J, Liang C, Zhou J. **Paeoniflorin Inhibits Mesangial Cell Proliferation and Inflammatory Response in Rats With Mesangial Proliferative Glomerulonephritis Through PI3K/AKT/GSK-3β Pathway.**. *Front Pharmacol* (2019) **10** 978. DOI: 10.3389/fphar.2019.00978
31. Li H, Wang Y, Chen B, Shi J. **Silencing of PAQR3 suppresses extracellular matrix accumulation in high glucose-stimulated human glomerular mesangial cells via PI3K/AKT signaling pathway.**. *Eur J Pharmacol* (2018) **832** 50-5. DOI: 10.1016/j.ejphar.2018.05.032
32. Yue J, López JM. **Understanding MAPK Signaling Pathways in Apoptosis.**. *Int J Mol Sci* (2020) **21** 2346. DOI: 10.3390/ijms21072346
33. Guo Y, Liu C, Zhang J, Tian BB, Wang L, Liu GK, Liu Y. **A relationship between MAPK/ERK pathway expression and neuronal apoptosis in rats with white matter lesions.**. *Eur Rev Med Pharmacol Sci* (2020) **24** 4412-9. DOI: 10.26355/eurrev_202004_21023
34. Zhu G, Qiu W, Li Y, Zhao C, He F, Zhou M, Wang L, Zhao D, Lu Y, Zhang J, Liu Y, Yu T, Wang Y. **Sublytic C5b-9 Induces Glomerular Mesangial Cell Apoptosis through the Cascade Pathway of MEKK2-p38 MAPK-IRF-1-TRADD-Caspase 8 in Rat Thy-1 Nephritis.**. *J Immunol* (2017) **198** 1104-18. DOI: 10.4049/jimmunol.1600403
35. Lu Y, Mei Y, Chen L, Wu L, Wang X, Zhang Y, Fu B, Chen X, Xie Y, Cai G, Bai X, Li Q, Chen X. **The role of transcriptional factor D-site-binding protein in circadian CCL2 gene expression in anti-Thy1 nephritis.**. *Cell Mol Immunol* (2019) **16** 735-45. DOI: 10.1038/s41423-018-0020-4
36. Xia L, Liu Y, Zhang Z, Gong Y, Yu T, Zhao D, Qiu W, Wang Y, Zhang J. **Modulation of IL-6 Expression by KLF4-Mediated Transactivation and PCAF-Mediated Acetylation in Sublytic C5b-9-Induced Rat Glomerular Mesangial Cells.**. *Front Immunol* (2022) **12** 779667. DOI: 10.3389/fimmu.2021.779667
37. Ji M, Lu Y, Zhao C, Gao W, He F, Zhang J, Zhao D, Qiu W, Wang Y. **C5a Induces the Synthesis of IL-6 and TNF-α in Rat Glomerular Mesangial Cells through MAPK Signaling Pathways.**. *PLoS One* (2016) **11** e0161867. DOI: 10.1371/journal.pone.0161867
38. Kitayama H, Sugimoto Y, Matsuzaki T, Ikawa Y, Noda M. **A ras-related gene with transformation suppressor activity.**. *Cell* (1989) **56** 77-84. DOI: 10.1016/0092-8674(89)90985-9
39. Hogan C, Serpente N, Cogram P, Hosking CR, Bialucha CU, Feller SM, Braga VM, Birchmeier W, Fujita Y. **Rap1 regulates the formation of E-cadherin-based cell-cell contacts.**. *Mol Cell Biol* (2004) **24** 6690-700. DOI: 10.1128/MCB.24.15.6690-6700.2004
40. Price LS, Hajdo-Milasinovic A, Zhao J, Zwartkruis FJ, Collard JG, Bos JL. **Rap1 regulates E-cadherin-mediated cell-cell adhesion.**. *J Biol Chem* (2004) **279** 35127-32. DOI: 10.1074/jbc.M404917200
41. Yamamoto K, Takagi Y, Ando K, Fukuhara S. **Rap1 Small GTPase Regulates Vascular Endothelial-Cadherin-Mediated Endothelial Cell-Cell Junctions and Vascular Permeability.**. *Biol Pharm Bull* (2021) **44** 1371-9. DOI: 10.1248/bpb.b21-00504
|
---
title: 'Metformin use history and genome-wide DNA methylation profile: potential molecular
mechanism for aging and longevity'
authors:
- Pedro S. Marra
- Takehiko Yamanashi
- Kaitlyn J. Crutchley
- Nadia E. Wahba
- Zoe-Ella M. Anderson
- Manisha Modukuri
- Gloria Chang
- Tammy Tran
- Masaaki Iwata
- Hyunkeun Ryan Cho
- Gen Shinozaki
journal: Aging (Albany NY)
year: 2023
pmcid: PMC9970305
doi: 10.18632/aging.204498
license: CC BY 3.0
---
# Metformin use history and genome-wide DNA methylation profile: potential molecular mechanism for aging and longevity
## Abstract
Background: Metformin, a commonly prescribed anti-diabetic medication, has repeatedly been shown to hinder aging in pre-clinical models and to be associated with lower mortality for humans. It is, however, not well understood how metformin can potentially prolong lifespan from a biological standpoint. We hypothesized that metformin’s potential mechanism of action for longevity is through its epigenetic modifications.
Methods: To test our hypothesis, we conducted a post-hoc analysis of available genome-wide DNA methylation (DNAm) data obtained from whole blood collected from inpatients with and without a history of metformin use. We assessed the methylation profile of 171 patients (first run) and only among 63 diabetic patients (second run) and compared the DNAm rates between metformin users and nonusers.
Results: *Enrichment analysis* from the Kyoto Encyclopedia of Genes and Genome (KEGG) showed pathways relevant to metformin’s mechanism of action, such as longevity, AMPK, and inflammatory pathways. We also identified several pathways related to delirium whose risk factor is aging. Moreover, top hits from the Gene Ontology (GO) included HIF-1α pathways. However, no individual CpG site showed genome-wide statistical significance ($p \leq 5$E-08).
Conclusion: This study may elucidate metformin’s potential role in longevity through epigenetic modifications and other possible mechanisms of action.
## INTRODUCTION
We live in an aging society. According to the U.S. Census Bureau’s 2017 National Population Projections, 1 in every 5 residents will be in retirement age by 2030 [1]. Subsequently, a more significant percentage of the population will endure the challenges of age-related diseases than ever before. Treatments targeting these diseases, such as dementia or cancer, at most “delay” the disease process but have a limited ability to “cure.” Therefore, there are growing interests in treating aging itself as a disease [2].
Considerable evidence from basic and pre-clinical models shows that several interventions, such as exercise, intermittent fasting, and even ingestion of certain compounds can prolong lifespan. These promising compounds include rapamycin [3, 4], resveratrol [5–7], NAD [8], and metformin [9–11]. Our group also confirmed that inpatients using metformin had improved three-year survival rates compared to non-metformin users [12]. In addition, our data also showed that prevalence of delirium was lower among those who were on metformin compared to those without [12].
The mechanism (or mechanisms) of action that rationalizes how these interventions prolong lifespan, or potentially delay aging, has been investigated heavily. Nevertheless, no exact process is well understood, especially for metformin. It is believed that epigenetics is one of the most important molecular mechanisms of aging in animals and plants; thus, it is plausible that the “life-prolonging” effects of many interventions are through modification of epigenetic processes. For example, several reports show epigenetic changes from exercise [13], fasting [14], rapamycin [3], resveratrol [5], and NAD [8]. However, there are only a few studies investigating the direct influence of metformin on epigenetic changes [15–17], suggesting that information about the influence of metformin on the epigenetic profile in humans is currently limited.
To fill such gap of knowledge, we investigated the potential influence of metformin on the epigenetic profile by testing genome-wide DNA methylation (DNAm) in whole blood samples obtained from inpatients with and without a history of metformin use.
## Demographics
173 subjects were enrolled in this study, but only 171 were included in downstream data analysis. The average patient age was 74.4 (SD = 9.8). 58 ($33.9\%$) subjects were females while almost all the subjects were white per self-report ($$n = 167$$; $97.7\%$). 108 patients were non-diabetic (non-DM) while 63 were diabetic (DM). Among the DM group, 37 had diabetes with a history of metformin prescription DM(+)Met and 26 had diabetes without a history of metformin prescription DM(−)Met. Additionally, 43 ($68.3\%$) diabetic subjects had a history of insulin use. Charlson Comorbidity Index (CCI) and body mass index (BMI) information are also included in Table 1. No variable revealed statistically significant differences between the DM(−)Met and DM(+)Met. However CCI, BMI, and insulin use were significantly higher among the DM group compared to the non-DM group, as expected.
**Table 1**
| Classification | All Subjects | Diabetes | Diabetes.1 | p | Statistical test | DM subjects | DM subjects.1 | p.1 | Statistical test.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Classification | All Subjects | non-DM | DM | p | Statistical test | DM(−)Met | DM(+)Met | p | Statistical test |
| N | 171 | 108 | 63 | | | 26 | 37 | | |
| Age - yr | 74.4 | 74.6 | 74.1 | 0.77 | t = 1.98 | 73.8 | 74.3 | 0.833 | t = 2.01 |
| SD | 9.8 | 9.7 | 10.0 | | | 10.6 | 9.7 | | |
| Female sex (n) | 58 | 36 | 22 | 0.81 | χ2 = 0.10 | 11 | 11 | 0.303 | χ2 = 1.06 |
| % | 33.9 | 33.6 | 34.9 | | | 42.3 | 29.7 | | |
| Race, White (n) | 167 | 105 | 62 | 0.63 | χ2 = 0.23 | 25 | 37 | 0.229 | χ2 = 1.45 |
| % | 97.7 | 97.2 | 98.4 | | | 96.2 | 100 | | |
| CCI | 3.8 | 3.1 | 4.9 | 7.5E-06* | t = 1.98 | 4.8 | 5.0 | 0.756 | t = 2.00 |
| SD | 2.7 | 2.7 | 2.4 | | | 2.4 | 2.5 | | |
| BMI | 29.7 | 28.3 | 32.2 | 0.002* | t = 1.98 | 30.0 | 33.8 | 0.64 | t = 2.00 |
| SD | 7.6 | 6.3 | 8.8 | | | 5.0 | 10.5 | | |
| Insulin use history | 43 | 0 | 43 | 3.3E-23* | χ2 = 98.48 | 15 | 28 | 0.131 | χ2 = 2.28 |
| % | 25.1 | 0 | 68.3 | | | 57.7 | 75.7 | | |
## Met vs. non-Met (including all patients regardless of diabetes status): top hits, KEGG, GO
Table 2 shows the most significant genes that differed in methylation rates between patients with and without metformin use history regardless of diabetes status (171 subjects). None of the sites met the criteria for genome-wide statistical significance ($p \leq 5$E-8).
**Table 2**
| Gene name | CpG site | Chromosome | non-Met (%) | Met (%) | % mean difference (Δβ) | p-value |
| --- | --- | --- | --- | --- | --- | --- |
| PSME3 | cg22769787 | chr17 | 15.6% | 14.3% | 1.3% | 3.37e-07 |
| EPHA8 | cg27136384 | chr1 | 83.2% | −2.7% | −2.7% | 4.84e-07 |
| | cg22163972 | chr17 | 92.1% | 4.2% | 4.2% | 4.89e-07 |
| | cg23047680 | chr3 | 0.8% | −0.2% | −0.2% | 9.08e-07 |
| NEDD4 | cg11341892 | chr15 | 4.7% | 0.6% | 0.6% | 2.82e-06 |
| PRKCG | cg11293016 | chr19 | 52.9% | 4.0% | 4.0% | 4.68e-06 |
| SRSF11 | cg12923877 | chr1 | 97.5% | −0.3% | −0.3% | 4.94e-06 |
| RRP15 | cg24353272 | chr1 | 95.3% | −0.8% | −0.8% | 5.16e-06 |
| KIAA1688 | cg07969649 | chr8 | 91.1% | −1.6% | −1.6% | 5.22e-06 |
| TRIM27 | cg02525926 | chr6 | 97.4% | 0.8% | 0.8% | 6.98e-06 |
| | cg23067796 | chr12 | 93.7% | 1.7% | 1.7% | 7.29e-06 |
| RYR2 | cg04573831 | chr1 | 96.6% | −0.6% | −0.6% | 8.11e-06 |
| | cg15180899 | chr18 | 93.9% | 1.7% | 1.7% | 8.67e-06 |
| | cg12222244 | chr3 | 94.1% | 2.1% | 2.1% | 1.27e-05 |
| C1orf125 | cg20746459 | chr1 | 90.6% | 3.5% | 3.5% | 1.52e-05 |
| SERPINH1 | cg19586851 | chr11 | 97.2% | −0.5% | −0.5% | 1.55e-05 |
| PPL | cg12991522 | chr16 | 1.8% | −0.5% | −0.5% | 1.55e-05 |
| ACO1 | cg13567378 | chr9 | 89.0% | −1.3% | −1.3% | 1.71e-05 |
| | cg24525630 | chr17 | 1.6% | −0.3% | −0.3% | 1.72e-05 |
| TCF7L1 | cg20116596 | chr2 | 95.7% | −0.5% | −0.5% | 1.76e-05 |
Next, we conducted enrichment analysis using the top 330 CpG sites based on the absolute difference in methylation level (beta value) between metformin users (Met) and nonusers (non-Met) greater than $4\%$ and the p-value less than 0.01. Enrichment analysis from the KEGG top signals showed relevant pathways to metformin’s possible roles, such as “longevity regulating pathway”, “longevity regulating pathway – multiple species”, and “AMPK signaling pathway” (Table 3). In addition, other pathways, such as “mTOR signaling pathway”, “insulin secretion”, “glutamatergic synapse”, and “circadian entrainment” were discovered (Table 3). There were also relevant pathways revealed in the GO analysis, such as “regulation of hypoxia-inducible factor-1alpha signaling pathway”, “positive regulation of hypoxia-inducible factor-1alpha signaling pathway”, and “canonical Wnt signal pathway” (Table 4), although none of the pathways in either KEGG or GO reached the False Discovery Rate (FDR) significance level (FDR <0.05) (Tables 3 and 4).
## Met vs. non-Met (including only patients with type 2 diabetes mellitus): top hits, KEGG, GO
Table 5 shows the most significant genes that differed in methylation rate between metformin users and nonusers among the diabetes group (63 subjects). Similar to the previous analysis, no gene reached genome-wide statistical significance ($p \leq 5$E-8).
**Table 5**
| Gene name | CpG site | Chromosome | non-Met (%) | Met (%) | Mean difference (Δβ) | p-value |
| --- | --- | --- | --- | --- | --- | --- |
| | cg19873536 | chr10 | 78.3% | 67.9% | 10.4% | 1.28e-06 |
| | cg13596208 | chr9 | 1.9% | 2.7% | −0.9% | 2.29e-06 |
| HBA1 | cg01704105 | chr16 | 40.5% | 33.7% | 6.8% | 5.42e-06 |
| DUOX2 | cg02550961 | chr15 | 1.5% | 1.9% | −0.4% | 6.1e-06 |
| NEO1 | cg12516231 | chr15 | 2.2% | 3.2% | −0.9% | 6.97e-06 |
| C7orf46 | cg06685724 | chr7 | 2.1% | 2.9% | −0.8% | 1.28e-05 |
| NAT15 | cg00484396 | chr16 | 9.8% | 4.9% | 4.8% | 1.56e-05 |
| | cg14685975 | chr5 | 89.9% | 92.1% | −2.2% | 1.64e-05 |
| CTSL | cg02104500 | chr9 | 3.6% | 4.9% | −1.4% | 1.66e-05 |
| | cg12584257 | chr9 | 67.6% | 77.2% | −9.6% | 1.69e-05 |
| NAT15 | cg22508957 | chr16 | 10.9% | 6.3% | 4.6% | 1.84e-05 |
| AREL1 | cg11034672 | chr14 | 11.6% | 15.0% | −3.3% | 1.86e-05 |
| | cg24651265 | chr10 | 1.1% | 1.7% | −0.5% | 2.12e-05 |
| CMBL | cg17467873 | chr5 | 1.7% | 2.1% | −0.4% | 2.21e-05 |
| EBF4 | cg05857996 | chr20 | 77.6% | 63.6% | 13.9% | 2.23e-05 |
| | cg18482666 | chr2 | 95.8% | 94.8% | 1.0% | 2.39e-05 |
| HRASLS5 | cg00489394 | chr11 | 6.6% | 7.1% | −0.5% | 2.4e-05 |
| AKAP13 | cg21530087 | chr15 | 2.2% | 2.6% | −0.4% | 2.59e-05 |
| | cg15864571 | chr3 | 93.4% | 95.0% | −1.6% | 2.67e-05 |
| FLJ35024 | cg15981195 | chr9 | 2.3% | 3.5% | −1.1% | 2.91e-05 |
The enrichment analysis was generated using consistent parameters in methylation level differences (beta >$4\%$) and p-value (<0.01). This current analysis, however, included 1283 CpGs. KEGG showed many of the same signals discovered from the previous analysis, including “longevity regulating pathway”, “glutamatergic synapse”, “insulin secretion”, “circadian entrainment”, and “cholinergic synapse” (Table 6). GO also showed overlapping pathways compared to the first analysis, including “hypoxia-inducible factor-1alpha signaling pathway”, but also new pathways, such as “interleukin-8-mediated signaling pathway”, “negative regulation of leukocyte apoptotic process”, “neutrophil homeostasis”, and “neuron projection”, although these pathways did not reach the FDR significance level (FDR <0.05) (Table 7).
## DNA methylation age acceleration
Among the diabetes group, metformin nonusers had a mean age acceleration of −8.07 compared to a mean age acceleration of −4.47 for metformin users ($$p \leq 0.11$$) (Figure 1). This difference was smaller among all the subjects included regardless of diabetes status (−5.92 for metformin nonusers vs. −4.47 for metformin users; $$p \leq 0.34$$) (Figure 2). Both analyses did not reach statistical significance.
**Figure 1:** *Age acceleration between metformin users and nonusers among the diabetes group. Age acceleration was calculated using the Horvath epigenetic clock as DNAm age - chronological age. Metformin = 0: without history of metformin use, Metformin = 1: with history of metformin use. p = 0.11.* **Figure 2:** *Age acceleration between metformin users and nonusers. Age acceleration was calculated using the Horvath epigenetic clock as DNAm age - chronological age. Metformin = 0: without history of metformin use, Metformin = 1: with history of metformin use. p = 0.34.*
## DISCUSSION
In this study, we compared genome-wide DNA methylation rates among metformin users and nonusers to investigate the potential epigenetic effects of metformin exposure. Enrichment analysis was employed to elucidate the possible mechanisms of action induced by metformin. Our KEGG analysis revealed evidence of differences in epigenetic profiles involved in “longevity” such as “longevity regulating pathway” and “longevity regulating pathway – multiple species” (Tables 3 and 6). Although it was not statistically significant, the appearance of these pathways among top signals in the KEGG analysis demonstrates the potential role of the epigenetic processes manifesting the effect of metformin on longevity. The same KEGG analysis also showed “AMPK signaling pathway” (Table 3). AMP-activated protein kinase (AMPK), an energy sensor that regulates metabolism, is commonly referred to as one of the targets of metformin’s hypothetical mechanisms of action [18, 19], although there is also evidence that metformin’s effects are in part AMPK-independent [20]. Furthermore, AMPK activation is related to subsequent activation of hypoxia-inducible factors [21] which also appeared in our GO analyses as “regulation of hypoxia-inducible factor-1alpha signaling pathway” and “positive regulation of hypoxia-inducible factor-1alpha signaling pathway” (Table 4), as well as “hypoxia-inducible factor-1alpha signaling pathway” (Table 7). Hypoxia-inducible factor-1alpha (HIF-1α) is a transcription factor expressed in nucleated cells and mediated by oxygen levels. HIF-1α has been implicated in age-related diseases, endothelial senescence progression, AMPK, and many other pathways [22]. Beyond metformin’s potential epigenetic medication related to longevity, several pathways related to delirium, such as “circadian entrainment”, “cholinergic synapse”, and “glutamatergic synapse”, were identified (Tables 3 and 6). These pathways are intriguing from metformin’s possible “anti-aging” standpoint as age is a major risk factor of delirium.
The beneficial effects of metformin on lifespan have been widely studied. Previous studies reported that metformin increased median lifespan of C. elegans co-cultured with E.coli by more than $35\%$ [9, 23], and prolonged the lifespan of mice [10]. Patients with age-related diseases such as cardiovascular diseases and cancer who take metformin also had lower rates of mortality [24, 25]. Our recent study using a cohort of over 1,400 inpatients also revealed that diabetic patients with a history of metformin use have a significantly lower 3-year mortality than diabetic patients who have not taken metformin [12]. There are, however, conflicting reports as well. For example, the same effect was not observed in Drosophila [26]. Also, age-dependent, dose-dependent, and gender-dependent variable effects on lifespan were reported in mice [27, 28]. Although these previous studies’ results are not consistent, our cohort mentioned above (from which the present data are an analysis of its subgroup) clearly showed a positive influence of metformin use on survival among diabetic inpatients [12].
Our epigenetics data presented herein support metformin’s broad range of potential effects as indicated by the pathways identified through the enrichment analysis. The KEGG analysis (Table 7) showed several signals related to inflammation and the immune system, such as “interleukin-8 receptor activity” and “negative regulation of leukocyte apoptotic process.” The appearance of inflammation-related pathways is intriguing considering strong evidence showing that elderly people present with low-grade, chronic inflammation [29]. These signals identified in our study may support our hypothesis that metformin can modify the inflammatory process through epigenetic modification and influence the likelihood of survival. Consistent with our data, Barath et al. also reported that metformin inhibited cytokine production from Th17 by correcting age-related changes in autophagy and mitochondrial bioenergetics, indicating its potential for the medication to promote healthy aging [30]. Among the literature supporting metformin’s role in suppressing inflammation, clinical trials including the Diabetes Prevention Program (DPP) [31] and Bypass Angioplasty Revascularization Investigation 2 Diabetes (BARI 2D) [32] have provided further evidence of metformin’s role in changing inflammatory biomarker levels among diabetic patients, while other clinical trials, such as the Lantus for C-reactive Protein Reduction in Early Treatment of Type 2 Diabetes (LANCET) [33], have found opposing evidence. Although several studies mentioned here have investigated the relationship between metformin and its potential anti-inflammation, a clinical trial aimed to confirm metformin’s role in aging is yet to be seen [2, 34]. It is worth mentioning, nonetheless, a small clinical study that demonstrated the regression of epigenetic age of patients through the administration of recombinant human growth hormone (rhGH), dehydroepiandrosterone (DHEA), and metformin [15]. As the study team administered three medications to their subjects at the same time, it is impossible to distinguish epigenetic changes caused only by metformin. It is also worth mentioning the unexpected results from the Horvath epigenetic clock since subjects with history of metformin use had relatively higher age acceleration than subjects without history of metformin. Still, neither reached statistical significance ($p \leq 0.05$). Future prospective studies comparing epigenetics marks before and after metformin use would be needed to better understand the direct effect of the medication.
In DM-only subjects, A-kinase anchoring protein 13 (AKAP13) gene was found (Table 5). A recent study showed that AKAP13 inhibits mammalian target of rapamycin complex 1 (mTORC1), which was present in our enrichment analysis as “mTOR signaling pathway” (Table 3). Furthermore, the degree of AKAP13 expression in lung adenocarcinoma cell lines correlates with mTORC1 activity [35]. Metformin’s anti-inflammatory effect has been shown to occur through eventual AMPK activation, which also inhibits the mTOR signaling pathway [18]. Metformin’s connection to AKAP13, which has yet been fully understood, deserves further investigation.
To the best of our knowledge, our study is the largest of its kind. A smaller, previous study also investigated metformin’s effect on genome-wide DNA methylation in human peripheral blood, although their study power was limited to a sample size of 32 male subjects [36]. Enrichment analysis in the present study revealing the longevity pathway from a hypothesis-free approach further strengthens our hypothesis that metformin exhibits its potential benefit for longevity through epigenetic processes. We also identified other relevant pathways associated with metformin’s mechanisms of action, such as the AMPK signaling pathway and HIF-1α signaling pathway [37].
Our study has several limitations. Although 171 subjects were analyzed retrospectively in this study, a controlled prospective study with a larger sample size would provide a better picture of the epigenetic mechanism of metformin on longevity. In addition, none of the individual CpG sites reached genome-wide significance ($p \leq 5$E-08). Thus, our findings should be interpreted as exploratory and hypothesis-generating. However, the fact that we found their biological relevance to metformin’s roles is still worth noting. As diabetes and metformin use status of the subjects was determined based on a retrospective chart review of electronic medical records, there are possibilities for misclassification, although we were still able to find multiple relevant pathways and genes of interest related to metformin’s action. Moreover, the duration of metformin use was not precisely assessed, making our definition of “metformin history use” broad since it might have included patients who took metformin for only a few months and patients who took metformin for years, for instance. Also, other types of diabetic medications were not investigated, such as sulfonylureas and glinide drugs as we used an already completed study dataset from our previous work. The rationale for us not investigating the influence of other diabetic medications was based on past literature showing that those diabetic medications other than metformin did not show benefits for survival. In fact, sometimes they were associated with worse mortality [38–40].
In summary, the data presented here support our hypothesis that epigenetics, especially DNA methylation, may be altered by metformin use and that such epigenetic processes potentially contribute to molecular mechanisms leading to longevity. Further careful investigation with a larger sample size would be warranted.
## Study participants and recruitment
We have previously recruited patients at the University of Iowa Hospital and Clinics (UIHC) for a separate study related to delirium from January 2016 to March 2020 [41–44]. Among them, we used data from a subgroup of patients recruited from November 2017 to March 2020 who had blood samples collected and processed for the epigenetics analysis [45–47]. Patients 18 years or older, who were admitted to the emergency department, orthopedics floor, general medicine floor, or intensive care unit were approached. Only those who consented, or whose legally authorized representative consented, were enlisted in the study. Written informed consent was obtained from all participants. Exclusion criteria included subjects whose goals of care were comfort measures only, those who were prisoners, or individuals with droplet/contact precautions. Further details of the study subjects and enrollment process are described previously [41–44].
We tested 173 subjects for genome-wide DNA methylation (DNAm) status, then conducted a post-hoc analysis of the available data to assess the influence of metformin. This study was approved by the University of Iowa Hospital and Clinics Institutional Review Board, and all procedures were compliant with the Declaration of Helsinki.
## Clinical information
Clinical variables were gathered through electronic medical chart review, patient interviews, and collateral information from family members [41–44]. Metformin use, insulin use, and type 2 diabetes mellitus (DM) history were obtained by using the search terms “metformin”, “insulin”, and “DM” or “diabetes”, respectively [12]. Only type 2 diabetes mellitus (DM) was included, excluding type 1 diabetes mellitus or gestational diabetes. If there was a history of metformin prescription before the study enrollment, patients were categorized as metformin users (Met). Those who were prescribed metformin after participation were not categorized as metformin users (non-Met) since the blood was obtained prior to such prescription.
## Sample collection
Blood samples were collected in EDTA tubes during patients’ hospital stay. Samples were shipped to the research laboratory and stored at −80°C until downstream analysis as a batch.
## Sample analysis
DNA was extracted from whole blood following the MasterPure™ DNA Purification kit (Epicentre, MCD 85201). DNA passing quality control based on NanoDrop spectrometry and in sufficient amount through the Qubit dsDNA Broad Range Assay Kit (ThermoFischer Scientific, Q32850) was selected for analysis for genome-wide DNAm status. 500 ng of genomic DNA from each sample was bisulfite-converted with the EZ DNA Methylation™ Kit (Zymo Research, D5002) and analyzed using Infinium HumanMethylationEPICBeadChip™ Kit (Illumina, WG-317-1002). The Illumina iScan platform scanned the arrays.
## Statistics and bioinformatics analysis
All analyses were conducted using R. The R packages ChAMP [48] and minfi [49] were used to process the data. Data from a total of 175 samples from 173 subjects were included for the statistical and bioinformatic analysis. DNAm levels for each CpG site were first compared between those with and without a history of metformin prescription (first run; Supplementary Table 1). Then, comparison limited among only DM patients between those with and without a history of metformin prescription was conducted to avoid potential influence of DM on DNAm status (second run; Supplementary Table 2).
During quality control processes, 2 samples from the first run and no samples from the second run were excluded based on the density analysis plots as a part of our quality control pipeline. 2 samples were also excluded because two patients had their blood collected twice. The first collected samples were included for further analysis while the second samples were excluded to maintain consistency between samples from all subjects. Therefore, 171 subjects from the first run and 63 subjects from the second run remained for the analysis. Furthermore, during the data loading process, probes were filtered out if they (i) had a detection p-value >0.01, (ii) had <3 beads in at least $5\%$ of samples per probe, (iii) were non-CpG, SNP-related, or multi-hit probes, or (iv) were located on chromosome X or Y. Beta mixture quantile dilation [50] was used to normalize samples, while the combat normalization method was used to correct for batch effect in the first run [51, 52]. The second run, which only included diabetic patients, was not corrected for batch effect because there were individual patients who were not part of any batches.
Top hits based on each CpG site difference were obtained through the RnBeads package using the limma method [53, 54] and accounting for age, sex, insulin use, BMI and cell type proportions (CD8 T cells, CD4 T cells, natural killer cells, B cells, and monocytes) as covariates. DNAm Age Calculator available online [55] calculated the cell type proportions through the method reported previously [56].
After obtaining the top CpG sites, enrichment analysis followed using missMethyl [57] and unbalanced numbers of CpG sites on each gene were controlled using the EPIC array. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genome (KEGG) [58] analysis was conducted. The number of CpG sites included in the analysis was determined by the combination of p-value and beta value cutoffs of the methylation rates of each CpG site ($p \leq .01$ and beta >0.04). Genome-wide significance was set at a p-value of less than = 5.0E-08.
The chi-square test compared the categorical data (sex, race, and insulin use) between two groups, while the Welch’s t-test compared the numerical data (age, BMI, and CCI) between two groups.
## DNA methylation aging clock analysis
To investigate whether subjects with history of metformin use had slower “age acceleration” than subjects without history of metformin use, we submitted the raw DNA methylation beta values to a publicly available tool, which includes the Horvath [55] method. The calculated output was the difference between the DNA methylation age and the chronological age.
## Availability of data materials
The datasets analyzed during the current study are available from the corresponding author upon reasonable request.
## References
1. 1Older People Projected to Outnumber Children for First Time in U.S. History. United States Census Bureau. 2018. https://www.census.gov/newsroom/press-releases/2018/cb18-41-population-projections.html (11/7/2021).. *United States Census Bureau* (2018.0)
2. Barzilai N, Crandall JP, Kritchevsky SB, Espeland MA. **Metformin as a Tool to Target Aging.**. *Cell Metab* (2016.0) **23** 1060-5. DOI: 10.1016/j.cmet.2016.05.011
3. Bjedov I, Toivonen JM, Kerr F, Slack C, Jacobson J, Foley A, Partridge L. **Mechanisms of life span extension by rapamycin in the fruit fly Drosophila melanogaster.**. *Cell Metab* (2010.0) **11** 35-46. DOI: 10.1016/j.cmet.2009.11.010
4. Bitto A, Ito TK, Pineda VV, LeTexier NJ, Huang HZ, Sutlief E, Tung H, Vizzini N, Chen B, Smith K, Meza D, Yajima M, Beyer RP. **Transient rapamycin treatment can increase lifespan and healthspan in middle-aged mice.**. *Elife* (2016.0) **5** e16351. DOI: 10.7554/eLife.16351
5. Howitz KT, Bitterman KJ, Cohen HY, Lamming DW, Lavu S, Wood JG, Zipkin RE, Chung P, Kisielewski A, Zhang LL, Scherer B, Sinclair DA. **Small molecule activators of sirtuins extend Saccharomyces cerevisiae lifespan.**. *Nature* (2003.0) **425** 191-6. DOI: 10.1038/nature01960
6. Baur JA, Sinclair DA. **Therapeutic potential of resveratrol: the in vivo evidence.**. *Nat Rev Drug Discov* (2006.0) **5** 493-506. DOI: 10.1038/nrd2060
7. Baur JA, Pearson KJ, Price NL, Jamieson HA, Lerin C, Kalra A, Prabhu VV, Allard JS, Lopez-Lluch G, Lewis K, Pistell PJ, Poosala S, Becker KG. **Resveratrol improves health and survival of mice on a high-calorie diet.**. *Nature* (2006.0) **444** 337-42. DOI: 10.1038/nature05354
8. Rajman L, Chwalek K, Sinclair DA. **Therapeutic Potential of NAD-Boosting Molecules: The In Vivo Evidence.**. *Cell Metab* (2018.0) **27** 529-47. DOI: 10.1016/j.cmet.2018.02.011
9. Onken B, Driscoll M. **Metformin induces a dietary restriction-like state and the oxidative stress response to extend C. elegans Healthspan via AMPK, LKB1, and SKN-1.**. *PLoS One* (2010.0) **5** e8758. DOI: 10.1371/journal.pone.0008758
10. Martin-Montalvo A, Mercken EM, Mitchell SJ, Palacios HH, Mote PL, Scheibye-Knudsen M, Gomes AP, Ward TM, Minor RK, Blouin MJ, Schwab M, Pollak M, Zhang Y. **Metformin improves healthspan and lifespan in mice.**. *Nat Commun* (2013.0) **4** 2192. DOI: 10.1038/ncomms3192
11. de Kreutzenberg SV, Ceolotto G, Cattelan A, Pagnin E, Mazzucato M, Garagnani P, Borelli V, Bacalini MG, Franceschi C, Fadini GP, Avogaro A. **Metformin improves putative longevity effectors in peripheral mononuclear cells from subjects with prediabetes. A randomized controlled trial.**. *Nutr Metab Cardiovasc Dis* (2015.0) **25** 686-93. DOI: 10.1016/j.numecd.2015.03.007
12. Yamanashi T, Anderson ZE, Modukuri M, Chang G, Tran T, Marra PS, Wahba NE, Crutchley KJ, Sullivan EJ, Jellison SS, Comp KR, Akers CC, Meyer AA. **The potential benefit of metformin to reduce delirium risk and mortality: a retrospective cohort study.**. *Aging (Albany NY)* (2022.0) **14** 8927-43. DOI: 10.18632/aging.204393
13. Ngwa JS, Nwulia E, Ntekim O, Bedada FB, Kwabi-Addo B, Nadarajah S, Johnson S, Southerland WM, Kwagyan J, Obisesan TO. **Aerobic Exercise Training-Induced Changes on DNA Methylation in Mild Cognitively Impaired Elderly African Americans: Gene, Exercise, and Memory Study - GEMS-I.**. *Front Mol Neurosci* (2022.0) **14** 752403. DOI: 10.3389/fnmol.2021.752403
14. Ng GY, Sheng DPL, Bae HG, Kang SW, Fann DY, Park J, Kim J, Alli-Shaik A, Lee J, Kim E, Park S, Han JW, Karamyan V. **Integrative epigenomic and transcriptomic analyses reveal metabolic switching by intermittent fasting in brain.**. *Geroscience* (2022.0) **44** 2171-94. DOI: 10.1007/s11357-022-00537-z
15. Fahy GM, Brooke RT, Watson JP, Good Z, Vasanawala SS, Maecker H, Leipold MD, Lin DTS, Kobor MS, Horvath S. **Reversal of epigenetic aging and immunosenescent trends in humans.**. *Aging Cell* (2019.0) **18** e13028. DOI: 10.1111/acel.13028
16. Kim D, Kim Y, Lee BB, Cho EY, Han J, Shim YM, Kim DH. **Metformin Reduces Histone H3K4me3 at the Promoter Regions of Positive Cell Cycle Regulatory Genes in Lung Cancer Cells.**. *Cancers (Basel)* (2021.0) **13** 739. DOI: 10.3390/cancers13040739
17. Nwanaji-Enwerem JC, Chung FF, Van der Laan L, Novoloaca A, Cuenin C, Johansson H, Bonanni B, Hubbard AE, Smith MT, Hartman SJ, Cardenas A, Sears DD, Herceg Z. **An epigenetic aging analysis of randomized metformin and weight loss interventions in overweight postmenopausal breast cancer survivors.**. *Clin Epigenetics* (2021.0) **13** 224. DOI: 10.1186/s13148-021-01218-y
18. Burkewitz K, Weir HJ, Mair WB. **AMPK as a Pro-longevity Target.**. *Exp Suppl* (2016.0) **107** 227-56. DOI: 10.1007/978-3-319-43589-3_10
19. Ma T, Tian X, Zhang B, Li M, Wang Y, Yang C, Wu J, Wei X, Qu Q, Yu Y, Long S, Feng JW, Li C. **Low-dose metformin targets the lysosomal AMPK pathway through PEN2.**. *Nature* (2022.0) **603** 159-65. DOI: 10.1038/s41586-022-04431-8
20. Foretz M, Hébrard S, Leclerc J, Zarrinpashneh E, Soty M, Mithieux G, Sakamoto K, Andreelli F, Viollet B. **Metformin inhibits hepatic gluconeogenesis in mice independently of the LKB1/AMPK pathway via a decrease in hepatic energy state.**. *J Clin Invest* (2010.0) **120** 2355-69. DOI: 10.1172/JCI40671
21. Salminen A, Kaarniranta K, Kauppinen A. **AMPK and HIF signaling pathways regulate both longevity and cancer growth: the good news and the bad news about survival mechanisms.**. *Biogerontology* (2016.0) **17** 655-80. DOI: 10.1007/s10522-016-9655-7
22. Alique M, Sánchez-López E, Bodega G, Giannarelli C, Carracedo J, Ramírez R. **Hypoxia-Inducible Factor-1α: The Master Regulator of Endothelial Cell Senescence in Vascular Aging.**. *Cells* (2020.0) **9** 195. DOI: 10.3390/cells9010195
23. Cabreiro F, Au C, Leung KY, Vergara-Irigaray N, Cochemé HM, Noori T, Weinkove D, Schuster E, Greene ND, Gems D. **Metformin retards aging in C. elegans by altering microbial folate and methionine metabolism.**. *Cell* (2013.0) **153** 228-39. DOI: 10.1016/j.cell.2013.02.035
24. Morales DR, Morris AD. **Metformin in cancer treatment and prevention.**. *Annu Rev Med* (2015.0) **66** 17-29. DOI: 10.1146/annurev-med-062613-093128
25. Luo F, Das A, Chen J, Wu P, Li X, Fang Z. **Metformin in patients with and without diabetes: a paradigm shift in cardiovascular disease management.**. *Cardiovasc Diabetol* (2019.0) **18** 54. DOI: 10.1186/s12933-019-0860-y
26. Slack C, Foley A, Partridge L. **Activation of AMPK by the putative dietary restriction mimetic metformin is insufficient to extend lifespan in Drosophila.**. *PLoS One* (2012.0) **7** e47699. DOI: 10.1371/journal.pone.0047699
27. Anisimov VN, Piskunova TS, Popovich IG, Zabezhinski MA, Tyndyk ML, Egormin PA, Yurova MV, Rosenfeld SV, Semenchenko AV, Kovalenko IG, Poroshina TE, Berstein LM. **Gender differences in metformin effect on aging, life span and spontaneous tumorigenesis in 129/Sv mice.**. *Aging (Albany NY)* (2010.0) **2** 945-58. DOI: 10.18632/aging.100245
28. Novelle MG, Ali A, Diéguez C, Bernier M, de Cabo R. **Metformin: A Hopeful Promise in Aging Research.**. *Cold Spring Harb Perspect Med* (2016.0) **6** a025932. DOI: 10.1101/cshperspect.a025932
29. Calçada D, Vianello D, Giampieri E, Sala C, Castellani G, de Graaf A, Kremer B, van Ommen B, Feskens E, Santoro A, Franceschi C, Bouwman J. **The role of low-grade inflammation and metabolic flexibility in aging and nutritional modulation thereof: a systems biology approach.**. *Mech Ageing Dev* (2014.0) **136-137** 138-47. DOI: 10.1016/j.mad.2014.01.004
30. Bharath LP, Agrawal M, McCambridge G, Nicholas DA, Hasturk H, Liu J, Jiang K, Liu R, Guo Z, Deeney J, Apovian CM, Snyder-Cappione J, Hawk GS. **Metformin Enhances Autophagy and Normalizes Mitochondrial Function to Alleviate Aging-Associated Inflammation.**. *Cell Metab* (2020.0) **32** 44-55.e6. DOI: 10.1016/j.cmet.2020.04.015
31. Haffner S, Temprosa M, Crandall J, Fowler S, Goldberg R, Horton E, Marcovina S, Mather K, Orchard T, Ratner R, Barrett-Connor E. **Intensive lifestyle intervention or metformin on inflammation and coagulation in participants with impaired glucose tolerance.**. *Diabetes* (2005.0) **54** 1566-72. DOI: 10.2337/diabetes.54.5.1566
32. Sobel BE, Hardison RM, Genuth S, Brooks MM, McBane RD, Schneider DJ, Pratley RE, Huber K, Wolk R, Krishnaswami A, Frye RL. **Profibrinolytic, antithrombotic, and antiinflammatory effects of an insulin-sensitizing strategy in patients in the Bypass Angioplasty Revascularization Investigation 2 Diabetes (BARI 2D) trial.**. *Circulation* (2011.0) **124** 695-703. DOI: 10.1161/CIRCULATIONAHA.110.014860
33. Pradhan AD, Everett BM, Cook NR, Rifai N, Ridker PM. **Effects of initiating insulin and metformin on glycemic control and inflammatory biomarkers among patients with type 2 diabetes: the LANCET randomized trial.**. *JAMA* (2009.0) **302** 1186-94. DOI: 10.1001/jama.2009.1347
34. Kulkarni AS, Gubbi S, Barzilai N. **Benefits of Metformin in Attenuating the Hallmarks of Aging.**. *Cell Metab* (2020.0) **32** 15-30. DOI: 10.1016/j.cmet.2020.04.001
35. Zhang S, Wang H, Melick CH, Jeong MH, Curukovic A, Tiwary S, Lama-Sherpa TD, Meng D, Servage KA, James NG, Jewell JL. **AKAP13 couples GPCR signaling to mTORC1 inhibition.**. *PLoS Genet* (2021.0) **17** e1009832. DOI: 10.1371/journal.pgen.1009832
36. Li M, Bao L, Zhu P, Wang S. **Effect of metformin on the epigenetic age of peripheral blood in patients with diabetes mellitus.**. *Front Genet* (2022.0) **13** 955835. DOI: 10.3389/fgene.2022.955835
37. Song H, Zhang X, Zhai R, Liang H, Song G, Yuan Y, Xu Y, Yan Y, Qiu L, Sun T. **Metformin attenuated sepsis-associated liver injury and inflammatory response in aged mice.**. *Bioengineered* (2022.0) **13** 4598-609. DOI: 10.1080/21655979.2022.2036305
38. **Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). UK Prospective Diabetes Study (UKPDS) Group.**. *Lancet* (1998.0) **352** 854-65. PMID: 9742977
39. Johnson JA, Majumdar SR, Simpson SH, Toth EL. **Decreased mortality associated with the use of metformin compared with sulfonylurea monotherapy in type 2 diabetes.**. *Diabetes Care* (2002.0) **25** 2244-8. DOI: 10.2337/diacare.25.12.2244
40. Cho EH, Han K, Kim B, Lee DH. **Gliclazide monotherapy increases risks of all-cause mortality and has similar risk of acute myocardial infarction and stroke with glimepiride monotherapy in Korean type 2 diabetes mellitus.**. *Medicine (Baltimore)* (2020.0) **99** e21236. DOI: 10.1097/MD.0000000000021236
41. Shinozaki G, Chan AC, Sparr NA, Zarei K, Gaul LN, Heinzman JT, Robles J, Yuki K, Chronis TJ, Ando T, Wong T, Sabbagh S, Weckmann MT. **Delirium detection by a novel bispectral electroencephalography device in general hospital.**. *Psychiatry Clin Neurosci* (2018.0) **72** 856-63. DOI: 10.1111/pcn.12783
42. Shinozaki G, Bormann NL, Chan AC, Zarei K, Sparr NA, Klisares MJ, Jellison SS, Heinzman JT, Dahlstrom EB, Duncan GN, Gaul LN, Wanzek RJ, Cramer EM. **Identification of Patients With High Mortality Risk and Prediction of Outcomes in Delirium by Bispectral EEG.**. *J Clin Psychiatry* (2019.0) **80** 19m12749. DOI: 10.4088/JCP.19m12749
43. Saito T, Malicoat JR, Leyden LR, Williams JC, Jellison SS, Long H, Hellman MM, Crutchley KJ, Anderson ZEM, Lo D, Modukuri MV, Schacher CJ, Yoshino A. **Mortality prediction by bispectral electroencephalography among 502 patients: its role in dementia.**. *Brain Commun* (2021.0) **3** fcab037. DOI: 10.1093/braincomms/fcab037
44. Yamanashi T, Crutchley KJ, Wahba NE, Sullivan EJ, Comp KR, Kajitani M, Tran T, Modukuri MV, Marra PS, Herrmann FM, Chang G, Anderson ZM, Iwata M. **Evaluation of point-of-care thumb-size bispectral electroencephalography device to quantify delirium severity and predict mortality.**. DOI: 10.1192/bjp.2021.101
45. Saito T, Toda H, Duncan GN, Jellison SS, Yu T, Klisares MJ, Daniel S, Andreasen AJ, Leyden LR, Hellman MM, Shinozaki E, Lee S, Yoshino A. **Epigenetics of neuroinflammation: Immune response, inflammatory response and cholinergic synaptic involvement evidenced by genome-wide DNA methylation analysis of delirious inpatients.**. *J Psychiatr Res* (2020.0) **129** 61-5. DOI: 10.1016/j.jpsychires.2020.06.005
46. Saito T, Braun PR, Daniel S, Jellison SS, Hellman M, Shinozaki E, Lee S, Cho HR, Yoshino A, Toda H, Shinozaki G. **The relationship between DNA methylation in neurotrophic genes and age as evidenced from three independent cohorts: differences by delirium status.**. *Neurobiol Aging* (2020.0) **94** 227-35. DOI: 10.1016/j.neurobiolaging.2020.06.003
47. Yamanashi T, Saito T, Yu T, Alario A, Comp K, Crutchley KJ, Sullivan EJ, Anderson ZM, Marra PS, Chang G, Wahba NE, Jellison SS, Meyer AA. **DNA methylation in the TNF-alpha gene decreases along with aging among delirium inpatients.**. *Neurobiol Aging* (2021.0) **105** 310-7. DOI: 10.1016/j.neurobiolaging.2021.05.005
48. Morris TJ, Butcher LM, Feber A, Teschendorff AE, Chakravarthy AR, Wojdacz TK, Beck S. **ChAMP: 450k Chip Analysis Methylation Pipeline.**. *Bioinformatics* (2014.0) **30** 428-30. DOI: 10.1093/bioinformatics/btt684
49. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, Irizarry RA. **Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays.**. *Bioinformatics* (2014.0) **30** 1363-9. DOI: 10.1093/bioinformatics/btu049
50. Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, Beck S. **A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data.**. *Bioinformatics* (2013.0) **29** 189-96. DOI: 10.1093/bioinformatics/bts680
51. Johnson WE, Li C, Rabinovic A. **Adjusting batch effects in microarray expression data using empirical Bayes methods.**. *Biostatistics* (2007.0) **8** 118-27. DOI: 10.1093/biostatistics/kxj037
52. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD, Zhang Y, Torres LC. *sva: Surrogate Variable Analysis. R package version 3.42.0* (2021.0)
53. Assenov Y, Müller F, Lutsik P, Walter J, Lengauer T, Bock C. **Comprehensive analysis of DNA methylation data with RnBeads.**. *Nat Methods* (2014.0) **11** 1138-40. DOI: 10.1038/nmeth.3115
54. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. **limma powers differential expression analyses for RNA-sequencing and microarray studies.**. *Nucleic Acids Res* (2015.0) **43** e47. DOI: 10.1093/nar/gkv007
55. Horvath S. **DNA methylation age of human tissues and cell types.**. *Genome Biol* (2013.0) **14** R115. DOI: 10.1186/gb-2013-14-10-r115
56. Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, Nelson HH, Wiencke JK, Kelsey KT. **DNA methylation arrays as surrogate measures of cell mixture distribution.**. *BMC Bioinformatics* (2012.0) **13** 86. DOI: 10.1186/1471-2105-13-86
57. Phipson B, Maksimovic J, Oshlack A. **missMethyl: an R package for analyzing data from Illumina's HumanMethylation450 platform.**. *Bioinformatics* (2016.0) **32** 286-8. DOI: 10.1093/bioinformatics/btv560
58. Kanehisa M, Goto S. **KEGG: kyoto encyclopedia of genes and genomes.**. *Nucleic Acids Res* (2000.0) **28** 27-30. DOI: 10.1093/nar/28.1.27
|
---
title: 'Intrinsic capacity differs from functional ability in predicting 10-year mortality
and biological features in healthy aging: results from the I-Lan longitudinal aging
study'
authors:
- Wei-Ju Lee
- Li-Ning Peng
- Ming-Hsien Lin
- Ching-Hui Loh
- Fei-Yuan Hsiao
- Liang-Kung Chen
journal: Aging (Albany NY)
year: 2023
pmcid: PMC9970311
doi: 10.18632/aging.204508
license: CC BY 3.0
---
# Intrinsic capacity differs from functional ability in predicting 10-year mortality and biological features in healthy aging: results from the I-Lan longitudinal aging study
## Abstract
This study aimed to explore the biological features and mortality risk of intrinsic capacity (IC) and functional ability (FA). Based on data from 1839 participants from the I-Lan Longitudinal Aging Study, multivariable Cox proportional hazard models were used to evaluate the predictive ability of IC (range 0–100) and FA (range 0–100) on 10-year mortality. Of 2038 repeated measurements for IC within a 7-year observational period, multivariable logistic regression was used to compare biological features of participants with maintained, improved and rapidly deteriorated IC. A 1-point increased IC value was associated with a $5\%$ (HR 0.95, $95\%$ CI 0.93–0.97, $p \leq 0.001$) decrease in mortality risk. Low IC (HR 1.94, $95\%$ CI 1.39–2.70, $p \leq 0.001$) was associated with higher mortality risk. Hyperglycemia (OR 1.40, $95\%$ CI 1.09–1.81, $$p \leq 0.010$$), low serum levels of DHEA-S (OR 3.33, $95\%$ CI 1.32–8.41, $$p \leq 0.011$$), and high serum levels of C-reactive protein (OR 1.45, $95\%$ CI 1.05–2.00, $$p \leq 0.023$$) were associated with low IC at baseline. Low serum levels of DHEA-S (middle tertile OR 1.84, $95\%$ CI 1.15–2.95, $$p \leq 0.012$$; lowest tertile OR 2.25, $95\%$ CI 1.34–3.77, $$p \leq 0.002$$) and vitamin D deficiency (OR 1.82, $95\%$ CI 1.02–3.27, $$p \leq 0.044$$) were associated with rapid deterioration of IC. IC and FA predicted 10-year mortality, whereas chronic inflammation, hyperglycemia, and low DHEA-S were associated with low IC status. Low DHEA-S and vitamin D deficiency may be considered as potential biomarkers of rapid IC declines, which implies underlying biological mechanisms of healthy aging.
## INTRODUCTION
The World Health Organization (WHO) published the World Report on Aging and Health in 2015 and the Integrated Care for Older People (ICOPE) in 2017, which has transformed health services from the traditional disease-focused approach into a function-centered one in the scheme of healthy aging. Healthy aging is defined as the process of developing and maintaining functional ability to ensure wellbeing in later life; IC and FA are proposed to describe and estimate the state of healthy aging [1, 2]. IC, defined as a composite measure for all physical and mental capacities of an individual, is conceptualized as a dynamic construct to serve as the potential functional reserve in the aging process [3]. Interacting with environmental factors, such as facilitators or barriers, IC may be a proxy to estimate the FA of an individual and the status of healthy aging [1]. Notably, as conceptualized, IC declines occur earlier before clinical manifestations of FA declines, so it is critical to capture IC declines in the life course for healthy longevity [4].
In the WHO Integrated Care for Older People, IC consists of five elements, i.e., locomotion, sensory, vitality, psychological, and cognition, which represent the physiological competence of individuals to support their FA [1]. Although there are strong theoretical and consensual bases to support this multidomain conceptualization of IC [5–8], most empirical studies have focused on the associations between IC and disability [4, 9], falls [10], quality of life [11], and mortality [12–14]. Unique diet, lifestyle, culture and policy may interact with IC, and contribute to FA, particularly in Asia [8]. In this context, functioning was used as a target to build an IC model instead of exploring the role of FA on health outcomes. Nevertheless, questions remained as there might be a gap between IC and FA against mortality due to their different conceptualization regarding a person’s competency and ability. To the best of our knowledge, no study has examined the impacts of these two distinct constructs on mortality in parallel to distinguish their potential impacts on clinical outcomes. A systemic review exploring possible biomarkers related to aging showed that lipids, glucose, inflammatory biomarkers, dehydroepiandrosterone sulfate (DHEA-S), growth hormone and insulin-like growth Factor 1 (IGF-1) were candidate biomarkers [15, 16]. Other studies indicated that elevated systemic inflammatory biomarkers, such as high-sensitivity C-reactive protein (hsCRP) and homocysteine, were associated with slow gait speed, weak muscle strength, and low IC [17, 18]. However, conflicting results have been reported regarding the associations between specific biomarkers and IC or FA. Although the underlying biological mechanisms of healthy aging remain unclear, identifying the biological features of healthy aging helps to capture the heterogenicity of aging over time [19]. Moreover, assessing the biological features of IC enables researchers and clinicians to understand potential pathophysiological mechanism, and to design personalized intervention programs to promote healthy aging.
To address the abovementioned knowledge gaps, the study aimed 1) to examine the association between declines in IC or FA and 10-year mortality risk and 2) to further explore the biological features of IC based on its longitudinal changes.
## Demographic characteristics of participants stratified by IC and FA status
The demographic characteristics of the participants (mean age 63.9 ± 9.3 years, $47.5\%$ men) are shown in Supplementary Table 1. Compared to the high IC group ($$n = 1190$$), those with low IC ($$n = 649$$) were older (70.0 ± 8.8 vs. 60.6 ± 7.7 years old), mainly women ($63.0\%$ vs. $46.8\%$), had fewer years of education (2.5 ± 3.4 vs. 8.2 ± 4.5 years), were less likely to consume alcohol ($23.0\%$ vs. $38.5\%$), and had a higher disease burden (CCI 1.6 ± 1.4 vs. 0.7 ± 1.1). Compared to the high FA group ($$n = 1769$$), those with low FA ($$n = 70$$) were older (76.0 ± 9.9 vs. 63.5 ± 8.9 years old), had fewer years of education (2.9 ± 4.1 vs. 6.4 ± 4.9 years), were less likely to consume alcohol ($15.7\%$ vs. $33.7\%$) and had a higher disease burden (CCI 2.3 ± 1.6 vs. 1.0 ± 1.2). Comparisons of subcategories of IC were shown in Supplementary Table 2. The distributions of IC and FA were left-skewed (Supplementary Figure 1). Overall, IC declined progressively from 86.0 ± 5.1 to 80.5 ± 5.1 points, and FA declined from 99.9 ± 1.3 to 99.3 ± 4.1 points in a mean follow-up period of 6.5 ± 0.8 years.
## Survival analysis for IC and FA status
There were 238 deaths in a mean follow-up period of 8.5 ± 1.5 years. Kaplan–*Meier analysis* showed that low FA (Figure 1A, log-rank test, $p \leq 0.001$) and low IC (Figure 1B, log-rank test, $p \leq 0.001$) were significantly associated with mortality. A one-point (percent) increase in IC score decreased the odds of mortality by $5\%$ (HR 0.95, $95\%$ CI 0.93–0.97, $p \leq 0.001$), and those with low IC had a greater risk of mortality (HR 1.94, $95\%$ CI 1.39–2.70, $p \leq 0.001$). The associations between FA and mortality were attenuated after adjusting for relevant confounders (Table 1).
**Figure 1:** *Kaplan–Meier survival plots for (A) functional ability and (B) intrinsic capacity.* TABLE_PLACEHOLDER:Table 1
## Biomarkers associated with low IC
The results of adjusted logistic regression analysis for the associations between biomarkers and low IC are shown in Figure 2. Participants with higher levels of fasting glucose had higher odds of having low IC (OR 1.40, $95\%$ CI 1.09–1.81, $$p \leq 0.010$$), but those with higher levels of cholesterol (OR 0.60, $95\%$ CI 0.47–0.79, $p \leq 0.001$) and LDL-C (OR 0.72, $95\%$ CI 0.55–0.94, $$p \leq 0.015$$) were less likely to have low IC. Those in the lowest tertile of DHEA-S had higher odds for low IC (OR 3.33, $95\%$ CI 1.32–8.41, $$p \leq 0.011$$). Of the inflammatory biomarkers, higher levels of hsCRP (OR 1.45, $95\%$ CI 1.05–2.00, $$p \leq 0.023$$) and neutrophil-to-lymphocyte ratio (NLR, highest tertile OR 1.77, $95\%$ CI 1.29–2.43, $p \leq 0.001$) were significantly associated with low IC (Figure 2).
**Figure 2:** *Logistic regression to explore biomarkers associated with low intrinsic capacity at baseline.*
## Biomarkers associated with longitudinal preservation or deterioration of IC
Adjusted logistic regressions showed that predictors for rapid deterioration of IC were lower levels of dehydroepiandrosterone sulfate (DHEA-S, middle tertile OR 1.84, $95\%$ CI 1.15–2.95, $$p \leq 0.012$$; lowest tertile OR 2.25, $95\%$ CI 1.34–3.77, $$p \leq 0.002$$) and vitamin D deficiency (OR 1.82, $95\%$ CI 1.02–3.27, $$p \leq 0.044$$) (Figure 3). Inverse probability weighting regressions were used to reduce potential selection bias from excluded participants, and findings were similar (OR 1.41, $95\%$ CI 1.09–1.83, $$p \leq 0.008$$ for lowest tertile of DHEA-S; OR 1.06, $95\%$ CI 1.01–1.12, $$p \leq 0.030$$ for vitamin D deficiency). Supplementary Figure 2 shows adjusted logistic regression to explore potential biomarkers for maintained or improved IC. All biomarkers showed insignificant associations.
**Figure 3:** *Logistic regression to explore biomarkers associated with rapidly deteriorated intrinsic capacity at a 7-year follow-up period.*
## DISCUSSION
The WHO conceptualized the scheme of healthy aging and addressed the importance of IC and FA in the life course that shifted health care services from disease-centric models to function-centric ones, as well as the focus on positive attributes of health. This study used a longitudinal cohort to compare the 10-year mortality risk between IC and FA. Unlike IC, the mortality risk of FA attenuated after adjustment for smoking, drinking, and disease burden, which supports the idea that environmental factors modify health outcomes in older age. However, the constant mortality risk of IC clearly reflects its adverse impacts on health outcomes in the aging process. In addition, this study found that hyperglycemia, proinflammatory status of hsCRP and NLR, and low serum levels of DHEA-S were associated with low IC. In particular, lower serum levels of DHEA-S and vitamin D deficiency were associated with rapid deterioration of IC. These findings suggest the potential roles of inflammation and endocrine and musculoskeletal systems in healthy aging through their influences on IC.
Based on the conceptual framework of healthy aging, the impacts of environmental or modifiable factors, including interventions on diseases and healthy behaviors, were emphasized to constitute FA instead of focusing on age-related declines in IC. The results of the current study completely support the conceptual framework of healthy aging. In contrast to frailty characterized by increased vulnerability to stressors from disrupted homeostasis of multiple physical systems, IC aimed to capture the nature of physiological reserves and residual capacities in aging, which is particularly suitable for longitudinal assessments [3]. Although frailty and IC are two interrelated but distinct constructs, both of them focus on function-centric health services with comprehensive assessment and are provided in a multidisciplinary fashion. Our previous study indicated that multidomain interventions combined with integrated primary care significantly improve physical function, cognitive performance, and quality of life among older adults with multimorbidity, which supported the core concepts of healthy aging [20, 21]. Nevertheless, the importance of FA should be addressed as well because of the significant modifying effects of environmental factors in older age.
By examining associations between IC and biomarkers, this study disclosed the multidimensional nature of biological mechanisms in the process of healthy aging. In this study, low IC status was associated with glucose metabolism, chronic inflammation, and neuroendocrine diseases. Previous studies have suggested associations between hsCRP, NLR, poor locomotion and vitality [17, 22]. This study further extended these associations from two specific domains (locomotion and vitality) to the whole composite IC, which strengthens the roles of inflammation in the aging process. On the other hand, cholesterol and LDL-C are well-known cardiovascular risk factors, but the associations are attenuated in late life [23]. A systemic review concluded that lower serum cholesterol levels were associated with greater mortality risk in older adults [24], which is in line with our study that lower cholesterol and low-density lipoprotein cholesterol (LDL-C) increased the risk of poor IC. These data justified the debates on statin treatment in older adults that routine statin use was not encouraged for those aged 75 years old and older. In addition, deprescription of statin therapy in those who have developed frailty is usually recommended [25]. Hyperglycemia is also a factor contributing to incident frailty and disability [26], which explained its association with low IC in our study.
DHEA-S is being widely used in the general public to prevent age-related disease. However, oral supplementation with DHEA-S failed to demonstrate a positive impact on the improvement of cognitive performance in healthy older people [27], whereas it preserved bone health [28]. Vitamin D is crucial to musculoskeletal health [29], but a systemic review of 81 randomized controlled trials reported disappointing results on falls, fracture preventions, or meaningful effects on bone mineral density [30]. Previous studies have suggested potential associations between DHEA-S and inflammatory [31] as well as the benefits for DHEA-S supplement administration on libido improvement and bone turnover biomarkers improvement [32]. In this study, low DHEA-S and vitamin D deficiency were both associated with rapid deterioration of IC but were not associated with maintained or improved IC in a follow-up period of up to 7 years. These findings highlight the importance of musculoskeletal health in preventing IC decline but also reflect the fact that neither monodimensional factor could counter the effects of age-related IC decline stemming from deficits in multiple physiological systems. Further interventional studies to explore the effects of oral supplementation with DHEA-S and vitamin D on IC are needed to further guide the optimal use of these supplements.
Despite all efforts made in this study, there are some limitations to note. First, environmental factors such as access to health facilities or means of transportation, etc., could not be adjusted for FA due to the limitation of data availability. Nevertheless, FA measured by the Functional Autonomy Measurement System (SMAF) considered available resources and environmental factors to compensate for functional impairment. Second, information regarding the cause of death could not be available in this study, which precludes cause-specific analysis. Further biological marker analysis could provide information on IC decline and potential mechanisms. Third, those excluded from the second part of this study might introduce selection bias. However, we have conducted an inverse probability weighting analysis to adjust for such potential selection bias and yields similar results. Fourth, as the participants from ILAS were community-dwelling adults, they are relative healthy. That is why the cutoff value of functional ability was quite high. Last, subdomains of vision and hearing were obtained from self-report questionnaires, which may probably underestimate the prevalence of vision and hearing impairment.
In conclusion, IC and FA predicted 10-year mortality, whereas the association for FA diminished after adjusting for smoking, drinking, and disease burden. Chronic inflammatory markers of hsCRP and NLR, hyperglycemia, and low DHEA-S were associated with low IC status. DHEA-S and vitamin D deficiency aggravated IC deterioration, which implies the importance of musculoskeletal health in healthy aging.
## Participants and study design
Data from the first and third waves of the I-Lan Longitudinal Aging Study (ILAS) were collected for the baseline cross-sectional (wave 1) and longitudinal cohort analyses (wave 1 and 3). ILAS was a prospective cohort study focused on interactions between sarcopenia, frailty, and cognitive functioning throughout the aging process. The details of the study design, participant recruitment, and data collection of ILAS have been reported previously [33]. Briefly, ILAS recruited community-dwelling adults aged ≥50 years without severe disability, dementia or communication difficulties, limited life expectancy due to major illness or being institutionalized. The first part of this study was a survival analysis of a 10-year follow-up, and the second part was a 7-year longitudinal study to capture biological features based on the status of IC changes (maintained, improved and rapidly deteriorated). In the first part of the study, data from 1,839 participants from the ILAS wave 1 survey in 2011 were used to evaluate the 10-year mortality risk of IC and FA (Figure 4A). In the second part of the study, 1019 participants who recruited in ILAS wave 1 and completed the ILAS wave 3 survey since 2018 (216 participants died before wave 3 survey and 604 participants declined to participate in wave 3 survey) were enrolled (Figure 4B). Compared to those enrolled in the second part of this study, those excluded were older (67.5 ± 10.0 vs. 61.1 ± 7.5 years, $p \leq 0.001$), less educated (4.5 ± 4.5 vs. 7.6 ± 4.8 years, $p \leq 0.001$), higher disease burden of Charlson comorbidity index (1.4 ± 1.4 vs. 0.7 ± 1.1, $p \leq 0.001$), but similar sex proportions.
**Figure 4:** *Study flowcharts of (A) survival analysis and (B) longitudinal study for biological features of intrinsic capacity. Abbreviations: IC: denotes intrinsic capacity; FA: denotes functional ability.*
This study was designed and conducted in accordance with the principles of the 1964 Declaration of Helsinki and later amendments. The institutional review board of National Yang Ming University (YM103008) and Taipei Veterans General Hospital (2018-05-003B) approved the study protocol, and written informed consent was obtained from each individual before inclusion. The observational design and reporting format followed the STROBE guidelines [34].
## Intrinsic capacity (IC)
The five IC elements, i.e., cognition, locomotion, vitality, psychology, and sensory, were selected based on the principle proposed by WHO ICOPE [5].
Cognition was evaluated by the Chinese version of the Mini-Mental State Examination (MMSE), with scores ranging from 0 to 30, and a higher score on the MMSE indicated better cognition [35]. Locomotion was assessed by a timed 6-meter gait speed at the usual pace in meters/second based on a consensus recommended by the Asian Working Group for Sarcopenia (AWGS) [36]. Vitality was assessed by the Mini Nutritional Assessment (MNA), with scores ranging from 0 to 30, and a higher score indicated better nutrition [37]. Psychology was measured with the Center for Epidemiologic Studies—Depression scale (CESD) with scores ranging from 0 to 60, and a higher score denoted a greater level of depressive symptoms [38]. Psychological scores were obtained by multiplying the original CESD scores by −1 in this study. The reason why we would like to adopt this approach is mainly because IC was constructed based on positive capacity for physical and mental reserves. However, we used CESD to measure mental reserve, but a higher CESD score indicates the higher level of depression. We thus rescale our CESD by multiplying the original CESD by −1. Sensory impairment was assessed through a self-reported score comprising visual and hearing impairments based on a previous study [12]. A sensory score was built from two questions of vision and hearing: score from 0 (independent), −1 (need supervision), −2 (need help), and −3 (dependent), yielding a total score ranging from −6 to 0. We further used the percent of maximum possible method - calculated by [100 × (observed − minimum)/(maximum − minimum)]- to rescale all individual variables with a ranging from 0 (minimum possible) to 100 (maximum possible) [39] to make the composite values of IC comparable longitudinally [40]. Psychological scores were obtained by 100 minus the rescaled CESD scores. For each study participant, we calculated their IC scores as the mean of the sum of subscores obtained in each of five elements.
## Functional ability (FA)
Functional ability (FA) was measured by SMAF through the percent of maximum possible method to rescale the original score with the range from 0 to 100. SMAF is a 29-item four-level measurement scale based on WHO’s classification of impairments, disabilities, and handicap with consideration of available resources and environmental factors to compensate for functional impairment [41]. The SMAF assessed functional ability in 5 areas: activities of daily living (ADL) [7 items], mobility [6 items], communication [3 items], mental functions [5 items] and instrumental activities of daily living (IADL) [8 items]. For each item, the disability was scored on a 4-level scale: 0 (independent), −0.5 (with difficulty), −1 (needs supervision), −2 (needs help), −3 (dependent). Resources available to compensate for the disability were also evaluated, and a handicap score was deducted. The stability of the resources was also assessed. A disability score (on −87) can be calculated, together with subscores for each dimension. Given the involvement of environmental factors, the SMAF evaluated a person’s FA rather than their IC alone [42].
## Acquisition of mortality
All participants in ILAS received a phone call by research nurses every 3 months to check their health conditions and survival status. These survival data were calculated from the index interview day until the last phone contact before 31 March 2022.
## Other variables
Potential confounding variables were identified from previous literature, which included age, sex, education years, smoking and alcohol consumption in the past six months (yes versus no). The burden of disease was assessed by Charlson’s comorbidity index (CCI) [43]. Venous blood samples were collected from all ILAS participants after a 10-hour overnight fast. Biomarkers related to cardiometabolic health, hormones and biochemistry were tested for all participants, including fasting glucose, total cholesterol, triglycerides, LDL, insulin level, and homeostasis model assessment-insulin resistance (HOMA-IR)-insulin resistance. Inflammatory biomarkers (NLR, platelet-to-lymphocyte ratio, homocysteine, and hsCRP), age-related hormones (growth hormone, IGF-1, DHEA-S, testosterone, sex hormone binding globulin (SHBG), intact parathyroid hormone, and micronutrients (vitamin B12, folic acid and vitamin D, measured by 25-OH vitamin D) were also tested for all participants. Details of the machine, limit of detection, and intra- and inter-assay coefficients of variation for serum biomarkers are shown in the appendix (Supplementary Table 3). The free androgen index (FAI) was calculated from testosterone and SHBG [44].
## Statistical analysis
Numerical variables were expressed as the mean ± standard deviation, and categorical variables were expressed as numbers with percentages. Comparisons of baseline characteristics between different IC or FA groups were performed by Student’s t test for continuous variables and chi-square tests or Fisher’s exact test for categorical variables. In the first part of this study, IC (≥82.7 vs. <82.7) and FA (≥98.9 vs. <98.9) at baseline were assigned to high IC/FA or low IC/FA groups according to the abovementioned cutoff values. Receiver operating curve (ROC) analysis and Youden’s index maxima were used to determine the optimal cutoff values that achieved optimal discrimination (Supplementary Table 4). Subcategories of IC as mobility (≥39.8 vs. <39.8), cognition (≥86.7 vs. <86.7), psychological (≥96.7 vs. <96.7), vitality (≥88.3 vs. <88.3), and sensory (≥83.3 vs. <83.3) were categorized into high and low for comparisons of baseline characteristics. Multivariable logistic regression was used to assess the associations between identified biomarkers and low IC. Cox proportional hazard regression was used to evaluate the association between IC, FA, and mortality. Schoenfeld residuals were used to test proportionality assumptions in Cox proportional hazard models. Kaplan–Meier survival plots and log-rank tests were applied to investigate the association between IC, FA, and mortality. In the second part of this study, multivariable logistic regressions were used to identify biomarkers associated with maintained or improved IC or rapidly deteriorated IC. The rapidly deteriorated IC group was defined as those with a decline in IC score >$10\%$ (equal to the 1 standard deviation (SD) lower than the group mean of decline in IC score) between wave 1 and 3 survey.
Biomarkers were dichotomized based on definitions used in previous literature: HOMA-IR (≥2 vs. <2) [45], high cholesterol (≥200 or taking lipid lower drugs vs. <200 mg/dL) [46], high triglycerides (≥150 or taking lipid lower drugs vs. <150 mg/dL) [46], high LDL-C (≥130 or taking lipid lower drugs vs. <130 mg/dL) [46], high fasting blood glucose (≥100 or taking oral anti-diabetic drugs vs. <100 mg/dL) [46], hyperhomocysteinemia (>15 vs. ≤15 μmol/L), and hsCRP (≥0.3 vs. <0.3 mg/dL) [47]. All other biomarkers, including the FAI, were tertilized [44]. Vitamin D insufficiency and deficiency were defined by 25-OH vitamin D <20 and <10 ng/mL, respectively [29].
A p value from two-sided tests <0.05 and $95\%$ CIs not spanning the null hypothesis values were considered to be statistically significant. All analyses were performed using the SAS statistical package version 9.4 (SAS Institute, Inc., Cary, NC, USA).
## References
1. 1World Health Organization. World report on ageing and health: World Health Organization. 2015. https://apps.who.int/iris/handle/10665/186463.. *World report on ageing and health: World Health Organization* (2015)
2. Beard JR, Officer A, de Carvalho IA, Sadana R, Pot AM, Michel JP, Lloyd-Sherlock P, Epping-Jordan JE, Peeters GME, Mahanani WR, Thiyagarajan JA, Chatterji S. **The World report on ageing and health: a policy framework for healthy ageing.**. *Lancet* (2016) **387** 2145-54. DOI: 10.1016/S0140-6736(15)00516-4
3. Belloni G, Cesari M. **Frailty and Intrinsic Capacity: Two Distinct but Related Constructs.**. *Front Med (Lausanne)* (2019) **6** 133. DOI: 10.3389/fmed.2019.00133
4. Stolz E, Mayerl H, Freidl W, Roller-Wirnsberger R, Gill TM. **Intrinsic Capacity Predicts Negative Health Outcomes in Older Adults.**. *J Gerontol A Biol Sci Med Sci* (2022) **77** 101-5. DOI: 10.1093/gerona/glab279
5. Cesari M, Araujo de Carvalho I, Amuthavalli Thiyagarajan J, Cooper C, Martin FC, Reginster JY, Vellas B, Beard JR. **Evidence for the Domains Supporting the Construct of Intrinsic Capacity.**. *J Gerontol A Biol Sci Med Sci* (2018) **73** 1653-60. DOI: 10.1093/gerona/gly011
6. George PP, Lun P, Ong SP, Lim WS. **A Rapid Review of the Measurement of Intrinsic Capacity in Older Adults.**. *J Nutr Health Aging* (2021) **25** 774-82. DOI: 10.1007/s12603-021-1622-6
7. Gonzalez-Bautista E, Andrieu S, Gutiérrez-Robledo LM, García-Chanes RE, de Souto Barreto P. **In the quest of a Standard Index of Intrinsic Capacity. A Critical Literature Review.**. *J Nutr Health Aging* (2020) **24** 959-65. DOI: 10.1007/s12603-020-1394-4
8. Fang EF, Xie C, Schenkel JA, Wu C, Long Q, Cui H, Aman Y, Frank J, Liao J, Zou H, Wang NY, Wu J, Liu X. **A research agenda for ageing in China in the 21st century (2nd edition): Focusing on basic and translational research, long-term care, policy and social networks.**. *Ageing Res Rev* (2020) **64** 101174. DOI: 10.1016/j.arr.2020.101174
9. Yu R, Leung G, Leung J, Cheng C, Kong S, Tam LY, Woo J. **Prevalence and Distribution of Intrinsic Capacity and Its Associations with Health Outcomes in Older People: The Jockey Club Community eHealth Care Project in Hong Kong.**. *J Frailty Aging* (2022) **11** 302-8. DOI: 10.14283/jfa.2022.19
10. Liu S, Yu X, Wang X, Li J, Jiang S, Kang L, Liu X. **Intrinsic Capacity predicts adverse outcomes using Integrated Care for Older People screening tool in a senior community in Beijing.**. *Arch Gerontol Geriatr* (2021) **94** 104358. DOI: 10.1016/j.archger.2021.104358
11. Salinas-Rodríguez A, González-Bautista E, Rivera-Almaraz A, Manrique-Espinoza B. **Longitudinal trajectories of intrinsic capacity and their association with quality of life and disability.**. *Maturitas* (2022) **161** 49-54. DOI: 10.1016/j.maturitas.2022.02.005
12. Sánchez-Sánchez JL, Rolland Y, Cesari M, de Souto Barreto P. **Associations Between Intrinsic Capacity and Adverse Events Among Nursing Home Residents: The INCUR Study.**. *J Am Med Dir Assoc* (2022) **23** 872-76.e4. DOI: 10.1016/j.jamda.2021.08.035
13. Meng LC, Huang ST, Peng LN, Chen LK, Hsiao FY. **Biological Features of the Outcome-Based Intrinsic Capacity Composite Scores From a Population-Based Cohort Study: Pas de Deux of Biological and Functional Aging.**. *Front Med (Lausanne)* (2022) **9** 851882. DOI: 10.3389/fmed.2022.851882
14. Yu R, Lai ETC, Leung G, Ho SC, Woo J. **Intrinsic capacity and 10-year mortality: Findings from a cohort of older people.**. *Exp Gerontol* (2022) **167** 111926. DOI: 10.1016/j.exger.2022.111926
15. Engelfriet PM, Jansen EH, Picavet HS, Dollé ME. **Biochemical markers of aging for longitudinal studies in humans.**. *Epidemiol Rev* (2013) **35** 132-51. DOI: 10.1093/epirev/mxs011
16. Belsky DW, Caspi A, Houts R, Cohen HJ, Corcoran DL, Danese A, Harrington H, Israel S, Levine ME, Schaefer JD, Sugden K, Williams B, Yashin AI. **Quantification of biological aging in young adults.**. *Proc Natl Acad Sci U S A* (2015) **112** E4104-10. DOI: 10.1073/pnas.1506264112
17. Lee WJ, Peng LN, Loh CH, Chen LK. **Sex-different associations between serum homocysteine, high-sensitivity C-reactive protein and sarcopenia: Results from I-Lan Longitudinal Aging Study.**. *Exp Gerontol* (2020) **132** 110832. DOI: 10.1016/j.exger.2020.110832
18. Giudici KV, de Souto Barreto P, Guerville F, Beard J, Araujo de Carvalho I, Andrieu S, Rolland Y, Vellas B. **Associations of C-reactive protein and homocysteine concentrations with the impairment of intrinsic capacity domains over a 5-year follow-up among community-dwelling older adults at risk of cognitive decline (MAPT Study).**. *Exp Gerontol* (2019) **127** 110716. DOI: 10.1016/j.exger.2019.110716
19. Beard JR, Jotheeswaran AT, Cesari M, Araujo de Carvalho I. **The structure and predictive value of intrinsic capacity in a longitudinal study of ageing.**. *BMJ Open* (2019) **9** e026119. DOI: 10.1136/bmjopen-2018-026119
20. Lee WJ, Peng LN, Lin CH, Chen RC, Lin SZ, Loh CH, Kao SL, Hung TS, Chang CY, Huang CF, Tang TC, Huang ST, Wen YW. **Effects of incorporating multidomain interventions into integrated primary care on quality of life: a randomised controlled trial.**. *Lancet Healthy Longev* (2021) **2** e712-23. DOI: 10.1016/S2666-7568(21)00248-8
21. Lee WJ, Peng LN, Lin MH, Lin CH, Chen LK. **Clinical Efficacy of Multidomain Interventions among Multimorbid Older People Stratified by the Status of Physio-Cognitive Declines: A Secondary Analysis from the Randomized Controlled Trial for Healthy Aging.**. *J Nutr Health Aging* (2022) **26** 909-17. DOI: 10.1007/s12603-022-1843-3
22. Öztürk ZA, Kul S, Türkbeyler İH, Sayıner ZA, Abiyev A. **Is increased neutrophil lymphocyte ratio remarking the inflammation in sarcopenia?**. *Exp Gerontol* (2018) **110** 223-9. DOI: 10.1016/j.exger.2018.06.013
23. Schatz IJ, Masaki K, Yano K, Chen R, Rodriguez BL, Curb JD. **Cholesterol and all-cause mortality in elderly people from the Honolulu Heart Program: a cohort study.**. *Lancet* (2001) **358** 351-5. DOI: 10.1016/S0140-6736(01)05553-2
24. Ravnskov U, Diamond DM, Hama R, Hamazaki T, Hammarskjöld B, Hynes N, Kendrick M, Langsjoen PH, Malhotra A, Mascitelli L, McCully KS, Ogushi Y, Okuyama H. **Lack of an association or an inverse association between low-density-lipoprotein cholesterol and mortality in the elderly: a systematic review.**. *BMJ Open* (2016) **6** e010401. DOI: 10.1136/bmjopen-2015-010401
25. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, Braun LT, de Ferranti S, Faiella-Tommasino J, Forman DE, Goldberg R, Heidenreich PA, Hlatky MA. **2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines.**. *Circulation* (2019) **139** e1082-143. DOI: 10.1161/CIR.0000000000000625
26. Kalyani RR, Tian J, Xue QL, Walston J, Cappola AR, Fried LP, Brancati FL, Blaum CS. **Hyperglycemia and incidence of frailty and lower extremity mobility limitations in older women.**. *J Am Geriatr Soc* (2012) **60** 1701-7. DOI: 10.1111/j.1532-5415.2012.04099.x
27. Grimley Evans J, Malouf R, Huppert F, van Niekerk JK. **Dehydroepiandrosterone (DHEA) supplementation for cognitive function in healthy elderly people.**. *Cochrane Database Syst Rev* (2006) **2006** CD006221. DOI: 10.1002/14651858.CD006221
28. von Mühlen D, Laughlin GA, Kritz-Silverstein D, Bergstrom J, Bettencourt R. **Effect of dehydroepiandrosterone supplementation on bone mineral density, bone markers, and body composition in older adults: the DAWN trial.**. *Osteoporos Int* (2008) **19** 699-707. DOI: 10.1007/s00198-007-0520-z
29. Chen CH, Liu LK, Chen MJ, Lee WJ, Lin MH, Peng LN, Chen LK. **Associations between vitamin D deficiency, musculoskeletal health, and cardiometabolic risk among community-living people in Taiwan: Age and sex-specific relationship.**. *Medicine (Baltimore)* (2018) **97** e13886. DOI: 10.1097/MD.0000000000013886
30. Bolland MJ, Grey A, Avenell A. **Effects of vitamin D supplementation on musculoskeletal health: a systematic review, meta-analysis, and trial sequential analysis.**. *Lancet Diabetes Endocrinol* (2018) **6** 847-58. DOI: 10.1016/S2213-8587(18)30265-1
31. Young DG, Skibinski G, Mason JI, James K. **The influence of age and gender on serum dehydroepiandrosterone sulphate (DHEA-S), IL-6, IL-6 soluble receptor (IL-6 sR) and transforming growth factor beta 1 (TGF-beta1) levels in normal healthy blood donors.**. *Clin Exp Immunol* (1999) **117** 476-81. DOI: 10.1046/j.1365-2249.1999.01003.x
32. Baulieu EE, Thomas G, Legrain S, Lahlou N, Roger M, Debuire B, Faucounau V, Girard L, Hervy MP, Latour F, Leaud MC, Mokrane A, Pitti-Ferrandi H. **Dehydroepiandrosterone (DHEA), DHEA sulfate, and aging: contribution of the DHEAge Study to a sociobiomedical issue.**. *Proc Natl Acad Sci U S A* (2000) **97** 4279-84. DOI: 10.1073/pnas.97.8.4279
33. Liu LK, Lee WJ, Chen LY, Hwang AC, Lin MH, Peng LN, Chen LK. **Sarcopenia, and its association with cardiometabolic and functional characteristics in Taiwan: results from I-Lan Longitudinal Aging Study.**. *Geriatr Gerontol Int* (2014) **14** 36-45. DOI: 10.1111/ggi.12208
34. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. **The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.**. *Prev Med* (2007) **45** 247-51. DOI: 10.1016/j.ypmed.2007.08.012
35. Guo NW, Liu HC, Wong PF, Liao KK, Yan SH, Lin KP, Chang CY, Hsu TC. **Chinese Version and Norms of the Mini-Mental State Examination.**. *J Phys Med Rehabil* (1988) 52-9
36. Chen LK, Woo J, Assantachai P, Auyeung TW, Chou MY, Iijima K, Jang HC, Kang L, Kim M, Kim S, Kojima T, Kuzuya M, Lee JSW. **Asian Working Group for Sarcopenia: 2019 Consensus Update on Sarcopenia Diagnosis and Treatment.**. *J Am Med Dir Assoc* (2020) **21** 300-7.e2. DOI: 10.1016/j.jamda.2019.12.012
37. Vellas B, Villars H, Abellan G, Soto ME, Rolland Y, Guigoz Y, Morley JE, Chumlea W, Salva A, Rubenstein LZ, Garry P. **Overview of the MNA--Its history and challenges.**. *J Nutr Health Aging* (2006) **10** 456-63. PMID: 17183418
38. Lewinsohn PM, Seeley JR, Roberts RE, Allen NB. **Center for Epidemiologic Studies Depression Scale (CES-D) as a screening instrument for depression among community-residing older adults.**. *Psychol Aging* (1997) **12** 277-87. DOI: 10.1037/0882-7974.12.2.277
39. Cohen P, Cohen J, Aiken LS, West SG. **The Problem of Units and the Circumstance for POMP.**. *Multivariate Behav Res* (1999) **34** 315-46. DOI: 10.1207/S15327906MBR3403_2
40. Moeller J. **A word on standardization in longitudinal studies: don't.**. *Front Psychol* (2015) **6** 1389. DOI: 10.3389/fpsyg.2015.01389
41. Hebert R, Carrier R, Bilodeau A. **The Functional Autonomy Measurement System (SMAF): description and validation of an instrument for the measurement of handicaps.**. *Age Ageing* (1988) **17** 293-302. DOI: 10.1093/ageing/17.5.293
42. Moreno-Agostino D, Prina M, Chua KC, Jotheeswaran AT, Sadana R, Officer A, Kamenov K, Cieza A. **Measuring functional ability in healthy ageing: a nationwide cross-sectional survey in the Philippine older population.**. *BMJ Open* (2021) **11** e050827. DOI: 10.1136/bmjopen-2021-050827
43. Charlson M, Szatrowski TP, Peterson J, Gold J. **Validation of a combined comorbidity index.**. *J Clin Epidemiol* (1994) **47** 1245-51. DOI: 10.1016/0895-4356(94)90129-5
44. Brannian JD, Long P, Kreger DO. **Is the free androgen index a useful clinical marker in male patients?**. *S D J Med* (1998) **51** 449-51. PMID: 9871405
45. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. **Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man.**. *Diabetologia* (1985) **28** 412-9. DOI: 10.1007/BF00280883
46. Bozkurt B, Aguilar D, Deswal A, Dunbar SB, Francis GS, Horwich T, Jessup M, Kosiborod M, Pritchett AM, Ramasubbu K, Rosendorff C, Yancy C. **Contributory Risk and Management of Comorbidities of Hypertension, Obesity, Diabetes Mellitus, Hyperlipidemia, and Metabolic Syndrome in Chronic Heart Failure: A Scientific Statement From the American Heart Association.**. *Circulation* (2016) **134** e535-78. DOI: 10.1161/CIR.0000000000000450
47. Casas JP, Shah T, Hingorani AD, Danesh J, Pepys MB. **C-reactive protein and coronary heart disease: a critical review.**. *J Intern Med* (2008) **264** 295-314. DOI: 10.1111/j.1365-2796.2008.02015.x
|
---
title: Hair follicle mesenchymal stem cell exosomal lncRNA H19 inhibited NLRP3 pyroptosis
to promote diabetic mouse skin wound healing
authors:
- Hongliang Yang
- Yan Zhang
- Zhenwu Du
- Tengfei Wu
- Chun Yang
journal: Aging (Albany NY)
year: 2023
pmcid: PMC9970314
doi: 10.18632/aging.204513
license: CC BY 3.0
---
# Hair follicle mesenchymal stem cell exosomal lncRNA H19 inhibited NLRP3 pyroptosis to promote diabetic mouse skin wound healing
## Abstract
Skin wounds caused by diabetes are a major medical problem. Mesenchymal stem cell-derived exosomes hold promise to quicken wound healing due to their ability to transfer certain molecules to target cells, including mRNAs, microRNAs, lncRNAs, and proteins. Nonetheless, the specific mechanisms underlying this impact are not elucidated. Therefore, this research aimed to investigate the effect of MSC-derived exosomes comprising long non-coding RNA (lncRNA) H19 on diabetic skin wound healing. Hair follicle mesenchymal stem cells (HF-MSCs) were effectively isolated and detected, and exosomes (Exo) were also isolated smoothly. Pretreatment with 30 mM glucose for 24 h (HG) could efficiently induce pyroptosis in HaCaT cells. Exosomal H19 enhanced HaCaT proliferation and migration and inhibited pyroptosis by reversing the stimulation of the NLRP3 inflammasome. Injection of exosomes overexpressing lncRNA H19 to diabetic skin wound promoted sustained skin wound healing, whereas sh-H19 exosomes did not have this effect. In conclusion, Exosomes overexpressing H19 promoted HaCaT proliferation, migration and suppressed pyroptosis both in vitro and in vivo. Therefore, HFMSC-derived exosomes that overexpress H19 may be included in strategies for healing diabetic skin wounds.
## INTRODUCTION
Chronic wounds that take a long time to heal owing to a variety of pathological reasons lead to financial and social problems as well as psychological stress [1]. With the aging of the population, the incidence of chronic wounds increases sharply owing to numerous age-associated illnesses, including diabetes. Chronic diabetic wounds are among the deadliest diabetic-related complications, affecting 15 percent of diabetic patients and contributing to a greater risk of amputation [2]. Under conditions of high glucose, inflammatory cells, fibroblasts, epidermal cells, and endothelial cells are dysfunctional, thus causing a delay in the wound healing process [3]. At present, it is still challenging to develop therapeutic strategies to improve wound healing in individuals with diabetes.
Mesenchymal stem cells (MSCs), including umbilical cord, bone marrow, and hair follicle stem cells, have aroused extensive research interest in the realm of regenerative medicine [4]. However, the mechanisms through which stem cells play a therapeutic role in the process of transplantation are still not well understood. Currently, a large amount of data confirm that transplanted stem cells play a positive role in tissue repair through a paracrine mechanism [5]. Among them, the cellular components called exosomes, produced during the paracrine process by stem cells, have attracted extensive attention. Exosomes are nanovesicles, and their biggest advantage is that they have almost no immunogenicity compared with the source cells themselves [6]. They are cup-shaped, with a density of approximately 1.13-1.19 g/ml and a diameter of approximately 40–100 nm. They contain lipid envelope and cytoplasmic components derived from cells. They are rich in mRNAs, microRNAs, lncRNAs, and proteins [7]. At present, they have been studied in a variety of disease models and have been shown to have certain effects, including renal damage repair enhancement, myocardial infarction area decrease, immunological response modulation, and skin wound healing. Stem cell-derived exosomes are expected to become a new promising direction of stem cell therapy [8].
Long non-coding RNAs (lncRNAs) have recently been demonstrated to exist in exosomes. It has been reported that exosomal lncRNAs can affect many diseases, such as diabetes, renal injury, and chronic skin wounds [9]. This suggests that lncRNAs may be transferred to cells through exosomes in the process of cell-to-cell communication and modulate gene expression [10]. We discovered a high expression of lncRNA H19 in hair follicle mesenchymal stem cell-derived exosomes (HF-MSCs-Exo). The lncRNA H19 has been depicted to perform a significant function in fibroblasts, enhancing proliferation and migration, and preventing apoptosis [11]. In this research, we also discovered that HF-MSCs-Exo can boost diabetic skin wound healing. However, the specific mechanism of exosomes in skin repair remains unclear. Excessive cell inflammation is a cause of chronic wounds [12]. Pyroptosis has recently been recognized as a type of programmed cell death. The activation of caspases, including Caspase-1, is mediated by inflammatory corpuscles, resulting in shear and polymerization of Gasdermin family members, including GSDMD, resulting in cell perforation and cell death [13]. Research reports have found that high glucose (HG)-mediated priming signal-induced NLRP3 (Nod-like receptor family pyrin domain comprising 3 mRNA) expression is reduced by exosomes from stem cells, indicating that exosomes perform instrumental functions in the modulation of NLRP3 gene expression of skin relative cells [14].
In this research, we postulated that lncRNA H19 delivered via HF-MSCs-Exo can regulate HG-induced human fibroblast (HDF) pyroptosis and enhance diabetic mouse skin wound healing. This research proposes a new mechanism through which HF-MSC-Exo promotes chronic skin wound repair, indicating that lncRNA H19 and exosomes could be a therapeutic strategy for diabetes focal skin defects.
## Isolation of exosomes from HF-MSCs
Fibroblast-like cells tended to migrate outward from the HFs ten days following the onset of HF culture. In tissue culture plates, HF-MSCs had a polygon-like and elongated shape. ( Figure 1A). Flow cytometry assays with immunofluorescence staining revealed that fibroblast-like cells presented MSC surface markers (negative for CD34, CD31, and CD45 and positive for CD44, CD90, CD73, CD105) (Figure 1B). In adipogenic differentiation culture conditions, the cells illustrated the formation of lipid droplets in the cytoplasm after Oil Red O staining. Under an osteogenic growth environment, the cells’ shape transformed from fibroblast-like to osteoblast-like, and they demonstrated significant levels of alkaline phosphatase activity. Alcian blue staining revealed the presence of glycosaminoglycan on the 21st day of chondrogenesis, indicating the differentiation ability into chondrogenesis (Figure 1C). Therefore, hair follicle-derived fibroblast-like cells exhibit mesenchymal stem cell surface markers and displayed trilineage differentiation potential toward osteoblasts, adipocytes and chondrogenesis. Consequently, these cells were labeled as HF-MSCs.
**Figure 1:** *Classification of hair follicle mesenchymal stem cells (HF-MSCs) and HFMSC-derived exosomes (HF-MSC-Exo). (A) Morphological observation of MSCs (200×). (B) MSCs surface marker molecules identified by flow cytometry. HF-MSCs were positive for CD90, CD105, CD73, CD44 and negative for CD34, CD45 and CD31. (C) Cell lineage-induced chondrogenic, osteogenic, and adipogenic differentiation were evaluated by toluidine blue staining, alizarin red staining, and Oil-Red-O staining. Scale bar=100 μm. (D) Micrographs of transmission electron microscopy of purified HF-MSC-Exo, showing a spheroid shape. Scale bar=50 nm. (E) The size distribution of the HF-MSC-s was assessed utilizing dynamic light scattering. (F) Western blotting results indicated the positive expression of Alix, CD63, and Tsg101 protein in HF-MSC-Exo.*
Exosomes were isolated by ultracentrifugation from HF-MSC cells using $2\%$ (w/v) exosome-deprived FBS. TEM illustrated that HF-MSCs produced vesicles with a distinct cup-shaped morphology (Figure 1D). According to nanoparticle tracking analyses, the particle size distribution of the vesicles was about 85 percent, ranging between 20 and 200 nm (Figure 1E). Western blotting revealed the existence of the exosomal marker proteins Alix, CD63, and Tsg101 (Figure 1F).
## Establishment of chronic hyperglycemic cell model
To determine the mechanism of the effects of exosomes on diabetic wounds, HaCaT cells were used to make the model chronic. HaCaT cells were treated with LG (5.5 mM) or HG (25 or 30 mM)12h to 96h. The CCK8 assay (Figure 2A) showed that HG stimulation resulted in a remarkable decrease in cell viability. Annexin V-PI staining in combination with flow cytometry assay (Figure 2B, 2C) illustrated that the proportion of PI-positive cells and annexin V-positive cells increased with time and concentration, being considerably higher compared with that in the control cohort ($P \leq 0.01$). The ROS detection (Figure 2D, 2E) showed that the fluorescence intensity of ROS dramatically increased as early as 12 h after treatment with HG, being substantially higher compared with that of the control cohort ($P \leq 0.05$); it continuously increased with time, reaching a maximum at 24 h.
**Figure 2:** *Glucose effects on HaCaT cells by CCK-8 assay, AnexinV-PI, and ROS staining. (A) HaCaT cells were treated with different concentrations of glucose for varying time periods, and then cell viability was measured with CCK-8. The data showed that glucose inhibited the cell viability of HaCaT cells in a dosage- and time-dependent way. (B) The annexin V-PI flow cytometry assay was utilized for the detection of the apoptosis rate of HaCaT cells, which were treated with 5.5, 25, or 30 mM glucose at different times. (C) The histogram results show that cell apoptosis increased following incubation with glucose. (D, E) Analysis of intracellular ROS levels using the flow cytometry assay. The histogram results show that the fluorescence intensity was increased following incubation with glucose. Control to 5.5 mM glucose, *P<0.05, **P<0.01, all results are representative of three separate experiments (means ± SD).*
## HG induced HaCaT cell injury by NLRP3 inflammasome-mediated pyroptosis
The ASC, NLRP3, and caspase-1 multiprotein complex NLRP3 inflammasome are widely recognized as the primary pathway of inflammation and pyroptosis following cell damage. To examine the impact of Exo on the activation of NLRP3 inflammasomes in vivo, we first needed to study the injury model produced by treating HaCaT cells with HG. We found that HG treatment enhanced the activation of caspase-1 and the production of IL-18 and IL-1β. The increased reactivity of HaCaT cells to the inflammasome was supported by NLRP3 inflammasomes ($P \leq 0.01$) (Figure 3A–3E). TUNEL staining results showed that HG-induced apoptosis increased in a time- and concentration-dependent way ($P \leq 0.01$) (Figure 3F, 3G). These findings strongly suggest that HG stimulates IL-1β production and caspase-1 activation regulated by the NLRP3 inflammasome in HaCaT cells.
**Figure 3:** *Effects of the activation of NLRP3 inflammasome and expression of pyroptosis-related proteins in HG-induced HaCaT cells. Western blotting (A), for caspase-1 (B), IL-1β (C), IL-18 (D), and NLRP3 (E) in HaCaT cells. (F, G), HaCaT apoptosis identified by the TUNEL assay (200×). *P<0.05 and **P<0.01versus the 5.5 mM cohort. Data are articulated as the mean±SD (n=6), and the experiment was redone independently three times.*
## HF-MSC-Exo remarkably attenuated HG-induced HaCaT cell death
Increasing research evidence indicates that paracrine signaling performs an instrumental function in stem cell-based therapy. To investigate the underlying mechanism through which paracrine signaling enhances cutaneous healing of wounds, we made HaCaT cells exposures to H2O2 and then to different concentrations of HFMSCs-Exo for various periods of time. CCK8 assay illustrated that the HaCaT cells cultivated in the presence of HFMSCs-Exo had considerably greater viability as opposed to that of the control cohort (0 ng/ml). Furthermore, the viability of HaCaT cells cultivated in HF-MSCs-Exo exhibited an increase in the dosage range from 500 ng/ml to 2000 ng/ml, and from 12-48 h (Figure 4A). Furthermore, we compared the effect of HF-MSCs-Exo on HG-induced increase in the levels of ROS and apoptotic rate of HaCaT cells with that of HFMSCs-dp-Exo (GW4869) ROS flow cytometry results showed that the mean fluorescence intensity of the cells was reduced more than HG cohort in the HF-MSC-Exo cohort (Figure 4B, 4D). Similar to ROS measurement, the percentage of apoptotic cells in the HF-MSCs-Exo cohort was considerably lesser than that in the HG cohort (Figure 5C, 5E). However, neither the percentage of apoptotic cells nor the intensity of ROS fluorescence showed any significant difference between the HG and HF-MSCs-dp-Exo cohorts ($P \leq 0.05$). We found that HF-MSC-Exo treatment reduced the activation of caspase-1 and the release of IL-18 and IL-1β. The reduction in the inflammasomes of HaCaT cells was supported by NLRP3 inflammasomes ($P \leq 0.01$) (Figure 4F–4J). These data clearly indicate that HF-MSC-Exo inhibited IL-1β, IL- 18 production, and activation of caspase-1 and reduced the NLRP3 inflammasome in HaCaT cells.
**Figure 4:** *HF-MSC-Exo inhibited HG-induced pyroptosis of HaCaT cells and promoted their proliferation. (A) HaCaT cells were treated with HG and then treated with HFMSC-dp-Exo or HF-MSC-Exo for 24-72 h. The CCK-8 assay findings illustrated that the cell viability was higher in the presence of HF-MSC-Exo than in the control and also higher than the HF-MSC-dp-Exo cohort. (B, C) The flow cytometric assay results showed that the HF-MSC-Exo can inhibit HG-induced apoptosis of HaCaT cells. (D, E) The flow cytometric assay results showed that the HF-MSC-Exo can reduce HG-induced fluorescence intensity in HaCaT cells. Western blotting (F) for caspase-1 (G), IL-1β (H), IL-18 (I), and NLRP3 (J) in HaCaT cells. Compared to control: *P<0.05 and **P<0.01 representing three separate experiments (means ± SD).*
## HF-MSC-Exo could transfer to HaCaT cells and promote HaCaT cell migration
To detect the internalization of HF-MSC-Exo by HaCaT cells, PHK26 fluorescent dye was used to label HF-MSC-Exo. Exosomes labeled with PKH26 were subjected to incubation with HaCaT cells. In a duration of 24 h, the fluorescence was observed in the cytoplasm and nucleus of the cells under the fluorescence microscope, suggesting that the HFMSC-Exo could be taken up by HaCaT cells and distributed intracellularly into the cytoplasm and nucleus (Figure 5A).
**Figure 5:** *HF-MSC-Exo can be uptaken by HaCaT cells and promote the migration of HG-treated HaCaT cells. (A) HF-MSC-Exo was labeled with PKH26, a lipid membrane-intercalating dye. The labeled exosomes were introduced to the culture medium of HaCaT cells. After 24 hours, HaCaT cells were fixed, counterstained with Hoechst 33342, and analyzed by fluorescent microscopy. The images showed that the labeled HF-MSC-Exo entered into the cytoplasm of HaCaT cells. Scale bar=200 μm. HaCaT cells were treated with HG and subsequently cultured in the presence of HF-MSC-Exo or HF-MSC-dp-Exo for 24 hours. The control cohort was the normal HaCaT cells. (B, C) HaCaT cells skin wound healing and cell migration experiment after treated with different culture medium for 24hours. (D, E) Quantified the proportion of wounded area closure and cell migration rate. (n=3); **P<0.01 Control vs HG; ##P<0.01, dp+Exo vs HG; &&P<0.01, Exo vs HG). Scale bar = 200 μm. All results are representative of three independent experiments (means ± SD).*
Keratinocyte migration has been known to be a crucial phase in the healing of cutaneous wounds. HaCaT cells were pretreated with HG, followed by incubation with HF-MSC-dp-Exo or HF-MSC-Exo for 24 hours to investigate the impact of HF-MSC-Exo on cellular migration. The cell scratch assay illustrated that the healing rate of wound areas was larger in the HF-MSC-Exo cohort contrasted with that in the HF-MSC-dp-Ex cohort ($P \leq 0.05$) (Figure 5B, 5C). Transwell assay indicated that incubation with HF-MSC-Exo dramatically increased the number of migrated cells relative to the HFMSC-dp-Exo cohort ($P \leq 0.05$) (Figure 5D, 5E).
## Construction of cell lines with stable lncRNA H19 expression
Existing research evidence has indicated that the lncRNA H19 may serve as a ceRNA to modulate miRNAs expression to regulate skin wound healing and glioma. Using qRT-PCR, we found that the levels of lncRNA H19 expression were low in the diabetic skin compared to the normal skin (Figure 6A). While it is highly expressed in HF-MSC-Ex (Figure 6B), the effects of lncRNA H19 on the biological characteristics of HG-treated HaCaT cells were examined by overexpressing or silencing lncRNA H19 expression. The efficiency of overexpressing or silencing lncRNA H19 in HFMSCs cells met the requirements for further experiments ($P \leq 0.05$) (Figure 6C). Exosomes derived from the transfected cells also showed overexpression or downregulation of lncRNA H19 expression. ( Figure 6D). RNA- FISH result illustrated that lncRNA H19 was present in the nucleus and cytoplasm of HFMSCs. ( Figure 6E).
**Figure 6:** *Production HF-MSCs and HF-MSC-Exo with lncRNA H19. (A) qRT-PCR was utilized to identify the lncRNA H19 expression in mouse diabetic skin. *P<0.05 WT vs Diabetes. (B) qRT-PCR was utilized to identify the lncRNA H19 expression in HF-MSCs and HF-MSC-Exo. *P<0.05 MSC vs MSC-Exo. (C, D) qRT-PCR was employed to ascertain the efficiency of overexpressing or silencing lncRNA H19 in HF-MSCs and HF-MSC-Exo.**P<0.01, OE-H19 vs OE-NC ##P<0.01, sh-H19 vs sh-NC. (E) Subcellular localization of lncRNA H19 in HFMSCs detected by FISH (400×). Data measured are articulated as mean ± SD.*
## HF-MSC-Exo carrying of lncRNA H19 regulates HaCaT proliferation, migration, and apoptosis via the NLRP3 pyroptosis signaling pathway
Considering that HF-MSCs-Exo can deliver lncRNA H19 to HaCaT cells through exosomes, the modulatory function of exosomes in proliferation, migration, and apoptosis of HaCaT cells was investigated by incubating HaCaT cells with normal HF-MSCs-Exo (NC) or HF-MSCs-Exo overexpressing H19 (OE-H19) or having downregulated levels H19 (sh-H19). The results showed that OE-H19 delivery promoted HaCaT cell proliferation and migration by suppressing apoptosis, while sh-H19 had the reverse effect ($P \leq 0.01$), implying that HF-MSC-Exo carrying of lncRNA H19 regulates fibroblast proliferation, migration, and apoptosis (Figure 7A–7G). NLRP3 is considered to be an essential inhibitor of the pyroptosis signaling pathway, which performs a critical function in regulating cell growth, differentiation, proliferation, and apoptosis. Treatment of HG-induced cells with different exosomes showed that OE-Exo inhibited the activation of caspase-1 and GSDMD cleavage. Therefore, lncRNA H19 is also essential for the modulation of the activation of NLRP3 inflammasome and cleavage of GSDMD in HaCaT cells (Figure 7H–7L). In summary, these findings illustrate that the anti-inflammatory benefit of lncRNA H19 in exosomes from HF-MSCs may be an important gene for skin wound healing.
**Figure 7:** *Exosomes overexpressing lncRNA H19 affected HaCaT cell proliferation, apoptosis, migration, and pyroptosis. (A) HaCaT cells were treated with HG and subsequently cultured in the presence of OE-H19-exosomes, sh-H19-exosomes, and NC-exosomes. The CCK-8 assay findings illustrated that cell viability was higher in the OE-H19-exosomes cohort than in the sh-H19-exosomes cohort. (B, C) HaCaT apoptosis identified by TUNEL assay (200×). *P<0.05 and **P<0.01versus the 5.5 mM cohort. Data are articulated as mean±SD (n=6), and the experiment was redone separately three times. (D, E) HaCaT cells were treated with OE-H19-exosomes, sh-H19-exosomes, and NC-Exo were exposed to a wound-healing assay and transwell migration assay for 12 hours. Scale bar =200μm. (F, G) Statistic the wound area closure and the number cell of per filed. Scale bar=200 μm. All results are representative of three separate experiments (means ± SD). Western blotting (H) for caspase-1 (I), IL-1β (J), IL-18 (K), and NLRP3 (L) expression in HaCaT cells. **P<0.01 OE-19 vs NC; ##P<0.01, sh-19 vs NC; &&P<0.01, OE-19 vs sh-19), representative of three independent experiments (means ± SD).*
## HF-MSC-Exo carrying of lncRNA H19 mediates pyroptosis to promote wound healing in diabetes mice
To examine the impact of HF-MSC-Exo in the process of wound repair in diabetic mice, we constructed a diabetic mouse skin wound model and injected the tissues surrounding the wound with HF-MSC-Exo overexpressing H19 (OE-H19) or with downregulated H19 (sh-H19), followed by immunohistochemical and H&E staining to observe wound healing. Exosomes overexpression of H19 (OE-H19) significantly accelerated the wound healing process in vivo, as confirmed by the presence of thicker granulation tissues and fewer inflammatory cells surrounding the wound (Figure 8A–8E). Western blot analysis of the skin wound tissue revealed that treatment with HF-MSC-Exo OE-H19 led to a lower level of caspase1, IL-1β, and TNF-α, illustrating that HF-MSC-Exo overexpressing H19 can mitigate the NLRP3 inflammation of diabetic skin wounds in mice. Therefore, HF-MSC-Exo carrying of overexpressing the lncRNA H19 quicken the process of wound healing in diabetes skin. OE-H19 exosomes protected against HG-induced injury in HaCaT cells by suppressing NLRP3 inflammasome-mediated pyroptosis by inhibiting the protein expression of IL-18, IL-1β, Caspase-1, and NLRP3 $P \leq 0.01$ (Figure 8F–8J).
**Figure 8:** *HF-MSC-Exo carrying lncRNA H19 promote mouse skin wound healing. (A) Representative images displaying mouse skin wound healing. (B) Wound histology after H&E staining. Tissue sections acquired from the wound site on day 14 after different injections were stained with antibodies against cytokeratin 14 and CD31. Scale bar = 200 μm. (C) Statistic the wound healing percent. (D) Quantitative analysis of the thickness of the new epidermis. (E) Quantitative analysis of the number of blood vessels. n = 3 per cohort. **P<0.01 OE-H19 contrasted with Control, ##P<0.01 OE-H19 compared with sh-H19. (F) Protein bond diagram of caspase-1 (I), IL-1β (J), IL-18 (K), and NLRP3 ascertained by western blot analysis. (G–J) Relative protein expression of caspase-1, IL-1β, IL-18, and NLRP3 normalized to GAPDH was evaluated by western blot analysis.*
## DISCUSSION
Wound healing is known to be an intricate and highly regulated process that is key to maintaining the function of the skin barrier [15]. Patients with diabetes mellitus suffer from slow or even nonunion, which can lead to diabetic foot and amputation [16]. It is the main complication causing a high disability rate among patients with diabetes and can threaten their lives in severe cases [17]. The healing of surface wounds requires the synergy of many factors to restore the barrier function of injured skin [18]. The diabetic wound is a complex condition that is characterized by oxidative stress, long-term chronic inflammation, neovascularization, peripheral neuropathy, extracellular matrix accumulation, and remodeling imbalance [19]. It is critical to research the fundamental processes of skin wound healing so as to create viable treatments. lncRNAs have been regarded as conserved, tissue-specific, and endogenous molecules that regulate a wide range of biological processes via the adsorption of miRNAs [20].
lncRNA H19 is a highly conserved gene that has gained extensive research attention in the fields of cancer and cardiovascular disease [21, 22]. Furthermore, it has recently been linked to the skin and smooth muscles, stem cell differentiation and embryonic development [23, 24]. lncRNA H19 has been found to be expressed in abundance in embryonic tissues, demonstrating its prospective role in shaping the growth of embryonic cells [25]. By conducting a series of in vivo and in vitro experimentations, Chang et al. have discovered the existence of MSC-derived exosomal [9]. Downmodulation of lncRNA H19 has also been strongly linked to the occurrence of dysplasia and the progression of hip dislocation [26]. Notably, long non-coding RNAs (lncRNAs) have been found in exosomes, indicating that lncRNAs could be transferred from one cell to another via exosomes in the process of cell-to-cell contact and additionally modulate gene expression in host cells [27, 28]. In this research, we investigated the mechanism that underlie HF-MSC-derived exosomal lncRNA H19 and determined by virtue of a blend of in vivo and in vitro experimentations that it functions in the NLRP3 pyroptosis signaling pathway. Depending on the available data, it is reliable to infer that HF-MSC-Exo carrying of high levels of lncRNA H19 may stimulate the process of wound healing in mice with DFU.
As an important accessory organ of the skin, hair follicles not only have a series of basic physiological functions but also play an important role in repairing skin damage during the period of skin damage. Hair follicle stem cells are abundant in sources and can only be obtained by hair extraction. The main types of stem cells obtained from hair follicles are melanocytes, mesenchymal stem cells, and keratin stem cells. These cells are located in the specific microenvironment of hair follicles, and their unique periodic activities are regulated by multiple signaling pathways [29]. In vivo, treatment of wounds with HF-MSC resulted in shorter wound length and faster fibroblasts differentiation to myofibroblasts in the wound, which contributed to a shortened proliferation stage [30]. Exosomes, which have good therapeutic characteristics, have been found in MSC-conditioned media. Exosomes have received a lot of interest due to their paracrine signaling characteristics and a new form of noncellular treatment [31, 32]. MSCs secreted exosomes (MSCs-Exo) have been demonstrated to have protective effects against liver fibrosis in vivo and also modify epithelial to mesenchymal transition (EMT) markers by elevating the count of E-cadherin-positive cells while decreasing the count of N-cadherin- and vimentin-positive following MSCs-Ex transplantation in mice [33, 34]. In this research, we effectively harvested exosomes from HF-MSCs, and we demonstrated that HF-MSC-Exo can enhance HG-induced HaCaT proliferation, migration and inhibit apoptosis. However, the underlying mechanism is not clear.
Numerous research reports have illustrated that a decline in the inflammatory response contributes to a delay in the repair rate of skin wounds in chronic wound healing models [35]. Pyroptosis occurs simultaneously with an inflammatory reaction, where inflammasomes, gasdermin D, pro-inflammatory cytokines (IL-18, and IL-1β), and caspase-1 could be seen [8]. We demonstrated that HaCaT cells can be induced to undergo pyroptosis with caspase-1, NLPR3, ILβ, and IL18. NALP3 signaling is necessary for the natural process of wound healing. A number of pro-inflammatory cytokines, such as TNF-α, IL-1β, and IL-6, play a key role in the healing of skin wounds. It has been reported that NALP3 signaling is crucial for the optimum healing of skin wounds [36, 37]. The inflammasomes activation via the topical injection of extracellular ATP considerably enhanced skin wound healing by upregulating the inflammatory response in the initial wound healing phase. Topical ATP delivery might contribute to the development of novel, successful therapies for speeding skin wound healing [36]. In our study, HF-MSC-Exo significantly enhanced HaCaT cell proliferation, migration and inhibited apoptosis. The results showed that Exo can inhibit NLRP3 related proteins including caspase-1, ILβ, and IL18. A prior study found that H19 had anti-inflammatory properties [38] and illustrated that H19 was inhibited in sepsis patients and that the H19 overexpression might have a reversal effect on the LPS-induced myocardial dysfunction and production of anti-inflammatory cytokines both in vitro and in vivo. H19 has been reported to perform a function in modulating mitochondrial apoptosis in the process of cardiac ischemia-reperfusion injury via the miR-877-3p/Bcl-2 axis [39]. Therefore, it is apparent that H19 has a protective function in skin wounds. However, whether H19 can mitigate skin wounds in diabetic mice by inhibiting pyroptosis in chronic skin wound healing requires further evidence. Next, we found that lncRNAH19 was more highly expressed in diabetic mouse skin than in normal mouse skin. We overexpressed lncRNA H19 in HF-MSCs and harvested the exosomes, OE-H19-Exo, which were found to enhance HaCaT proliferation, migration, and inhibit apoptosis following high expression of the inflammation factor NLRP3, caspase-1, ILβ, and IL18 in the model of the injury cell model.
In summation, these reasonable findings enabled us to postulate a molecular mechanism that underlie DFU therapy by which HF-MSC-released exosomal lncRNA H19 suppresses inflammation, enhances proliferation and migration, and suppresses apoptosis of fibroblasts, thus ameliorating the damage of DFU and speeding up the process of wound healing. To conclude, HF-MSC-Exo lncRNA H19 might serve as an auspicious treatment strategy for DFU. However, the present study is still in the preclinical stage, and more research is necessary to investigate the mechanism of action.
## CONCLUSIONS
Our study revealed a significant role of HF-MSC-Exo that overexpress lncRNA H19 in the development of strategies against diabetic skin wounds. HF-MSC-Exo overexpressing lncRNA H19 promoted HaCaT proliferation, migration, and pyroptosis suppression both in vitro and in vivo.
## Isolation, culture, and osteogenic, adipogenic, and chondrogenic differentiation of HF-MSCs
HF-MSC isolation was performed as previously reported [40]. Succinctly, a minimum of 20 hairs with intact follicles was manually plucked from the occipital area of volunteers’ scalp. The hair was rinsed three times in phosphate-buffered saline (PBS; Life Technologies, Carlsbad, CA, USA) comprising 1 percent penicillin/streptomycin solution (P/S; 100 IU/ml penicillin, 100 IU/ml streptomycin; HyClone, Victoria, Australia). Following the cutting of the hair shafts, the hair follicles were seeded at the base of a 24-well plate (Corning, Tewksbury, MA, USA), with each well containing one piece of hair. They were then subjected to culturing in 100 ml of Dulbecco’s modified Eagle medium: nutrient mixture F-12 (DMEM/F12; Life Technologies) comprising 10 percent fetal bovine serum (FBS; HyClone), 1 percent P/S, and 2 ng/ml basic fibroblast growth factor (bFGF; Life Technologies) and subsequently subjected to incubation at 37° C over the night in a humidified environment comprising $5\%$ CO2. Each well received 400 μl of culture media the following day, and the medium was replaced after every 3 days. When proliferating HF-MSCs achieved $80\%$ confluence, sub-culturing was performed HF-MSCs were employed for experimentations at passages 3–8.
The expression of embryonic stem cell (ESC) markers and MSC surface markers was evaluated utilizing immune fluorescence, as described earlier [41]. Concisely, fixing of HF-MSCs was done using 4 percent paraformaldehyde for 15 minutes at a temperature of 25° C, followed by blocking using 1 percent bovine serum albumin (BSA; Roche Diagnostics, France), and incubation using primary mouse anti-human antibodies against CD31, CD105, CD90, CD34, (eBioscience, San Diego, CA, USA), AIF (Cell Signaling Technology, Danvers, MA, USA) at a dilution ratio of 1:200 at a temperature of 4° C over the night. After washing thrice with PBS, HF-MSCs were subjected to incubation with Alexa Fluor $\frac{488}{555}$-conjugated anti-mouse secondary antibodies (dilution ratio of 1:400; Cell Signaling Technology) at 25° C for 60 minutes in darkness. HF-MSCs were then rinsed thrice in PBS, followed by counterstaining with Hoechst 33342 (dilution ratio of 1:10,000; Life Technologies) at a temperature of 25° C for five minutes in darkness. Subsequently, fluorescence microscopy (Olympus, Japan) was utilized to visualize the cells’ images. To perform flow cytometry analysis, HF-MSCs were obtained via centrifugation and subsequently treated using similar methods applied in immunofluorescence staining before being evaluated utilizing a FACS Calibur flow cytometer (BD Biosciences, San Jose, CA, USA). Cell Quest Software (BD Biosciences) was used to analyze the data.
Osteogenic and adipogenic differentiation of HF-MSCs was conducted as earlier reported [42]. Succinctly, culturing of the cells was done in either adipogenic differentiation medium comprising of high-glucose DMEM (H-DMEM, Life Technologies) comprising 10 percent FBS, 1 mM dexamethasone, 0.5 mM isobutyl-methylxanthine (Sigma-Aldrich, St. Louis, MO, USA), 10 mM insulin (Sigma-Aldrich), and 200 mM indomethacin (Sigma-Aldrich). Oil Red O staining (Sigma-Aldrich) was utilized two weeks following adipogenic stimulation to detect intracellular lipid droplets.
With regards to osteogenic differentiation assays, culturing of HF-MSCs was done in H-DMEM comprising 10 percent FBS, 0.1 mM, and 10 nM β-glycerophosphate (Sigma-Aldrich) for four weeks. Alizarin red S staining (Sigma-Aldrich) was carried out at the latest stage of the culturing to assess the development of calcium nodules.
With respect to chondrogenic differentiation, HF-MSCs spheres were produced by hanging drop culture using 20 μL of 8 × 106 cells/mL; spheres were then subjected to culturing in chondrogenic-induction medium comprising of (D)MEM, 10 percent FBS, 6.25 μg/mL insulin, 50nM of ascorbate-2-phosphate (Sigma-Aldrich), and 10 ng/mL transforming growth factor-beta 1 (PeproTech, London, UK). After every three days, the culture media were changed. Cartilages were identified three weeks following chondrogenic stimulation utilizing toluidine blue staining (Dingguo, Beijing, China) as per the guidelines stipulated by the manufacturer.
## Isolation and identification of MSC-Exo
The exosomes were isolated from MSCs as earlier reported [43]. Succinctly, HF-MSCs at passage 4 were cultured in DMEM comprising 10 ng ml−1 bFGF and 10 percent (w/v) FBS deprived of exosomes by centrifugation at 120,000×g over the night at a temperature of 4° C. The HF-MScs were cultivated in DMEM with 2 percent (w/v) exosome-FBS for 24 hours after achieving 80 percent confluence. To pellet the cells, the culture media were obtained and centrifugated at 300×g for 10 minutes at a temperature of 4° C. Next, the supernatant was extracted and centrifugated at 16,500×g (Optima™ L-100XP ultracentrifuge; Beckman Coulter, Palo Alto, CA, USA) at a temperature of 4° C for 20 minutes, and filtered using a 0.22-μm filter to get rid of cell debris. This medium was set as the conditioned medium (HF-MSCs-CM). The filtrate was centrifugated at 120,000×g at a temperature of 4° C for 90 minutes. The exosomes were extracted and assigned as HF-MSC-derived exosomes (HFMSCs-Exo). HFMSCs-Exo were resuspended in PBS and kept at a temperature of −80° C. A bovine calf albumin (BCA) kit (Beyotime, Shanghai, China) was used to determine the protein content in the HF-MSCs-Exo. Nanoparticle tracking analysis was adopted to assess the concentration and size distribution of exosomes utilizing ZetaView particle tracker from Particle Metrix (Germany), where each of the Nanoparticle Tracking Analysis (NTA) assessments for the various procedures for each participant was done in triplicate. Transmission electron microscopy (TEM; FEI Tecnai 12; Philips, Amsterdam, The Netherlands) was utilized to examine the morphology of HF-MSCs-Exo. Western blotting was employed in evaluating the expression levels of TSG101 (1:500, ProteinTech, Chicago, IL, USA), Alix (1:1,000, Abcam, Cambridge, MA, USA), and CD63 (1:500, Millipore, Temecula, CA, USA), in exosomes. The exosome-free medium was assigned as exosome-deprived HF-MSCs conditioned media with exosomes inhibitor GW4689 (HF-MSCs-dp-Ex).
## RNA-FISH
The subcellular localization of lncRNA H19 in HaCaT cells (Immortalized human epidermal cells) was detected using fluorescence in situ hybridization (FISH) techniques, following the guidelines described in the Ribo lncRNA FISH Probe Mix (red) kit (lnc10000; Ribobio, Guangzhou, Guangdong, China). Coverslips were put onto the wells of a 24-well plate, and cells were grown at a concentration of 6 × 104 cells in each well. Upon achieving approximately $80\%$ confluence, the cells were fixed using 1 ml 4 percent paraformaldehyde followed by treatment with 2 mg/ml acetylation reagent, glycine, and proteinase K. The cells were subsequently prehybridized for 1 hour at a temperature of 42° C using a 250-ml prehybridization solution followed by hybridization over the night at 42° C using a 250-ml hybridization solution comprising 300 ng/ml probe lncRNA H19. Staining of the nucleus was done for five minutes using DAPI diluted by PBS comprising Tween 20 at 1:800. After using an anti-fluorescence-quenching agent to mount the cells, they were examined utilizing a microscope and then photographed.
## Cell proliferation assay
The proliferation of HaCaT cells was identified utilizing Cell Counting Kit-8 (CCK-8; Dojindo Molecular Technologies, Tokyo, Japan) [44]. Succinctly, 2 × 103 cells were placed in triplicate in 96-well plates and subsequently cultured in L-DMEM medium supplemented with [25, 30] μM glucose (Sigma-Aldrich) and $2\%$ (w/v) FBS, or different concentrations of exosomes (0, 500, 1,000, 2,000 ng/ml). After 12, 24, 48, 72, or 96 h, each well received CCK-8 reagent and plates were further incubated for an extra 1 h. After the incubation approached completion, a microplate reader (Synergy H1; Biotek, USA) was employed to determine the absorbance of each well’s supernatant at 450 nm. The outcomes are articulated as the mean ± standard deviation from 3 separate experiments.
## Apoptosis assays and ROS detection
Annexin V-PI Apoptosis Detection Kit (Sangene, Tianjin, China) and a ROS assay kit (EMD Millipore, Billerica, MA, USA) were utilized to perform apoptosis analysis as per the protocols provided by the manufacturer [45, 46]. Concisely, HaCaT cells were plated into 6-well or 96-well tissue culture plates (triplicate) at a concentration of 5 × 104cells/cm2, followed by culturing in DMEM comprising 10 percent FBS for 24 hours. After cultivation, aspiration of the culture medium was carried out, and the cells were rinsed in PBS and subjected to culturing in DMEM comprising 2 percent FBS in the presence or absence of (control) 30 mM glucose for an extra 12-72 h. At the latest stage of cultivation, propidium iodine-Annexin V staining (San Jian, Tianjin, China) and reactive oxygen species (ROS) staining (Millipore) were conducted to determine apoptosis and ROS generation, respectively, at the specified time points following the manufacturer’s guidelines. Flow cytometry (BD Biosciences) was subsequently used to analyze the cells. For the detection of ROS, the probe dilution was done to a final concentration of 10 μM using a serum-free medium. The probe was introduced to the cells, placed in an incubator at 37° C with $5\%$ CO2 for 30 minutes, and then washed with a serum-free medium. Flow cytometry (BD Biosciences) was then employed to examine the cells.
## TUNEL assay for cell apoptosis
HaCaT cells induced by HG and then added into different mediums including exosomes overexpressing Lnc H19 or sh-H19 were added to the cell cultures. TUNEL staining was performed utilizing a TUNEL Apoptosis Assay Kit (MK500, TaKaRa Bio Inc, Shiga, Japan), as described previously [47]. TUNEL+ cells were manually quantified with an inverted fluorescence microscope (Carl Zeiss, Oberkochen, German).
## Protein extraction and Western blot
Extraction of total proteins from the wound edge tissue and HaCaT cells was done using Radio immunoprecipitation assay buffer (RIPA) buffer (Boster Biological Technology, China) as per the protocol stipulated by the manufacturer. The protein concentration was ascertained utilizing the BCA protein assay kit (Boster Biological Technology, China). Extracted proteins (100 μg) were isolated by a 10 percent SDS polyacrylamide gel (Bio-Rad, Hercules, CA, USA). Blocking of the membranes was done with 5 percent nonfat milk for 2 h at 25° C ensued by the incubation of the PVDF membrane with the antibodies over the night at 4° C. The rinse of the membranes was done thrice for 8 minutes each in moderate volume of 1 × TBST and then incubated with secondary antibody in a 37° C incubator for 2 hours. Protein visualization was done by an enhanced chemiluminescence system using a Fluor Chem FC system (Alpha Innotech, San Leandro, CA, USA). Image J densitometry analysis was conducted to analyze protein bands, and the fold expression is designated as the relative protein expression. Antibodies information is in Table 1.
**Table 1**
| Name of antibody | Antibody dilution | Company | Catalogue number |
| --- | --- | --- | --- |
| Caspase1 | 1:1000 | Cell signaling | 4542s |
| NLRP3 | 1:1000 | Cell signaling | 9532s |
| IL-18 | 2.5 μg/ml | IBL | 1C-205 |
| IL-1β | 1:1000 | Cell signaling | 2556s |
| GADPH | 1:1000 | Cell signaling | 4280s |
| CD31 | 1:200 | Santa Cruz | SC-133494 |
| CK4 | 1:500 | Santa Cruz | SC-1060 |
| CD90 | 1:200 | eBioscience | 2154875 |
| CD105 | 1:200 | eBioscience | 2205006 |
| CD73 | 1:200 | Invitrogen | 2105949 |
| CD44 | 1:200 | eBioscience | 2087672 |
| CD31 | 1:200 | eBioscience | 2152408 |
| CD45 | 1:200 | eBioscience | 31011A20 |
| CD34 | 1:200 | Cell signaling | 35695 |
## Cell transfection
H19 overexpression plasmid (pcDNA3.1-H19) was purchased from Shanghai Gene Pharma, and H19 smart silencer was purchased from RiboBio (Shanghai, China). HF-MSCs ($60\%$–$70\%$ confluency) were plated into six-well plates. Lipofectamine3000 transfection reagent (Thermo, USA) was employed to conduct cell transfection with the pcDNA3.1-H19 plasmid. The smart silencer was transfected utilizing Thermo’s Lipofectamine RNAiMAX transfection reagent (80 nM). The culture medium was changed with a new one 6 hours following incubation.
## Transwell assay and scratch assay
To examine cell migration, transwell and cell scratch tests were conducted [48]. Succinctly, HaCaT cells were plated in the upper chamber of a transwell (8-μm pore filters; Corning, Corning, NY, USA) as per the guidelines provided by the manufacturer. HaCaT cells were plated in 6-well plates at a seeding concentration of 5 × 104 cm2, followed by culturing over the night in DMEM comprising 10 percent (w/v) FBS. In the day that followed, the cells were rinsed thrice in PBS before being cultured in 30 mM glucose DMEM comprising 2 percent (w/v) exosome-free FBS for 24 hours. A 200-μl micropipette tip was utilized to scratch HaCaT cells grown in the 6-well plates. The HaCaT cells cultivated in the 6-well plates and Transwell were rinsed three times in PBS before being cultured for 24hours in DMEM comprising 2 percent (w/v) exosome-deprived FBS, HF-MSCs-dp-Exo, or HF-MSCs-Exo in 6-well plates or 100 μl DMEM comprising 2 percent (w/v) exosome-deprived FBS in the upper Transwell chamber. In the lower Transwell chamber, 600 μl DMEM comprising 2 percent exosome-free FBS, HFMSC-dp-Exo, or 1,000 ng/ml HFMSCs-Exo were added. After the Transwells were taken away from the tissue culture plate, the culture medium in the upper chamber of the Transwell was aspirated ensued by fixing of the HaCaT cells in the Transwells using methanol, drying with a laminar flow hood, and staining with Hoechst 33342 (1:10,000, 25° C, 2 min; Life Technologies, Carlsbad, CA, USA). The HaCaT cells that were left in the upper Transwell chamber were extracted with a cotton tab and washed thrice in PBS. A fluorescent microscope (Olympus) was used to visualize the HaCaT cells that were remained on the lower surface of the top transwell chamber. ImageJ software (NIH, Bethesda, MD, USA) was used to examine HaCaT cells in five view areas chosen at random. Photographs were taken for HaCaT cells cultivated by 6-well plates. Migration under the above-mentioned culture conditions was measured. Computation of the migration rates was done as indicated below: (distance of cells at scratching time point − distance of cells 24 h post-scratching)/(distance of cells at scratching time point).
## RNA isolation and quantitation
TRIzol reagent (Invitrogen, Carlsbad, CA, USA) was utilized to obtain total RNA from HFMSCs, exosomes, and diabetes mouse skin or normal mouse skin as per the protocols stipulated by the manufacturer. The Prime Script RT Reagent Kit (Takara, Beijing, China) was employed for reverse transcription of RNA into cDNA. The amplification reactions were conducted using a qRT-PCR system (ABI 7500, Thermo Fisher Scientific, MA, USA) utilizing 1 μl cDNA and 1 μl primer (Sangon Biotech, Shanghai, China). A 20 μl reaction volume was employed, comprising amplification SYBR Premix Ex Taq kit (Takara) and primers. The reactions were done premised on the conditions illustrated below: denaturation at 95° C for 3 minutes, followed by 40 cycles of denaturation at 95° C for 3 seconds and 60° C for 30 seconds. All reactions were repeated in triplicates for each sample. The relative miRNA expression or mRNAs expression was assessed by the 2-ΔΔCt method and normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH), in that order. The primers are H19 (F:5’-TACAACCACTGCACTACCTG-3’, R:5’-TGGAATGCTTGAAGGCTGCT-3’), GAPDH (F: 5’-AATCCCATCACCATCTTCCA-3’, R: 5’-TGGACTCCACGACGTACTCA-3’).
## Diabetic model induction
All of the animal experiments were carried out in strict adherence to the standards approved by the Institutional Animal Care Committee and the China Association of Laboratory Animal Care. The male SPF C57 mice weighing between 20–25 g were procured from the Laboratory Animal Department of China Medical University and kept at the Institutional Animal Center of Jilin University, Jilin, China. A week was spent for mice to adapt to their new surroundings. After adaptation, they were initially fed with a high-fat diet for 8 weeks, followed by intraperitoneal injection (i.p.) with STZ solution (140 mg/kg, dissolved in 0.1 M sodium citrate buffer) [49, 50]. Following the injection, the mice were given an HF diet for an additional four weeks. The level of blood glucose (BG) was measured from the mice’s tail tip. Once the levels of BG greater than 11.2 mmol/L were achieved, the development of the diabetic mice model was judged effective.
## Full-thickness skin defects and diabetes mouse cutaneous wound healing
After the successful preparation of the diabetes model, surgical scissors were employed to create a full-thickness excisional wound measuring 0.8 cm × 0.8 cm on the dorsal skin of each mouse [51]. Three cohorts of mice, each comprising six animals, were grouped at random. Subcutaneous injections of 100 g OE-H19 100 L, 100 g sh-H19 100 L, or 100 g PBS were administered at the edges of each wound margin (PBS cohort). As pointed out earlier, the wounds were dressed in one oil gauze layer topped with three cotton gauze layers.
## Gross inspection, and immunofluorescence, and H&E staining
Images of the wound sites were obtained on days 7 and 14 following wounding for a thorough examination of wound healing. A transparent film was used to display the contour along the wound edge, and the wound closure rate was determined as indicated below: [(original wound area−new wound area)/original wound area] ×100. After sacrificing the mice using an overdose of anesthetic, we obtained skin specimens that included wounds and surrounding tissues. The specimens were subsequently fixed using 10 percent (w/v) buffered formaldehyde/PBS, embedded in paraffin, dissected at 5-μm thick slices in the middle of the wound, followed by staining with hematoxylin and eosin (H&E). Observation and photographing of the tissue sections were done with the aid of a microscope. The newly formed epidermis was designated as epidermal tissue delimited by hair follicle-free dermis. The computation of the histological wound healing rate was performed as illustrated below: (length of the newly formed epidermis)/(length of newly formed epidermis + non-healed epidermis).
In order to identify epidermal formation, dermal angiogenesis and scar tissue in injured skin, immunofluorescence staining was conducted. Concisely, skin slices were deparaffinized followed by rehydration, and eventual blocking in 1 percent (w/v) BSA/PBS at a temperature of 25° C for 30 minutes. The tissue sections were subjected to incubation over the night at a temperature of 4° C using rat anti-mouse primary antibodies against cytokeratin 14 (1: 100; Abcam) and CD31 (1:100; Cell Signaling Technology) and subsequently rinsed thrice with PBS. In order to track the nuclei, the sections were subjected to incubation for 30 minutes at 25° C with Alexa Fluor-$\frac{488}{555}$-conjugated anti-mouse secondary antibody, followed by counterstaining with Hoechst 33342 (Invitrogen). A fluorescent microscope equipped with a digital camera (Leica DFC500, Wetzlar, Germany) was used to examine and take pictures of the slices. Five fields in each tissue section were chosen at random and examined at a magnification of ×400. From each of the three mice in each cohort, three sections were selected. The averaged optical densities of the expression of CK14 and CD31 were determined utilizing Image-Pro Plus (Media Cybernetics, Rockville, MD, USA). At each time point, five fields chosen at random were inspected and utilized to compute the average optical density per unit area for each cohort.
## Statistical analysis
SPSS (version 17.0; IBM Corp., Armonk, NY, USA) was utilized to execute statistical analyses. Data are articulated as the mean ± standard deviation (SD) for ≥3 separate experiments. One-way ANOVA was utilized to compare multiple cohorts. The student’s t-test was utilized to compare paired cohorts. Statistical significance was adjusted at $P \leq 0.05.$
## References
1. Pombeiro I, Moura J, Pereira MG, Carvalho E. **Stress-Reducing Psychological Interventions as Adjuvant Therapies for Diabetic Chronic Wounds.**. *Curr Diabetes Rev* (2022) **18** e060821195361. DOI: 10.2174/1573399817666210806112813
2. Yazdanpanah L, Nasiri M, Adarvishi S. **Literature review on the management of diabetic foot ulcer.**. *World J Diabetes* (2015) **6** 37-53. DOI: 10.4239/wjd.v6.i1.37
3. Singer AJ, Clark RA. **Cutaneous wound healing.**. *N Engl J Med* (1999) **341** 738-46. DOI: 10.1056/NEJM199909023411006
4. Squillaro T, Peluso G, Galderisi U. **Clinical Trials With Mesenchymal Stem Cells: An Update.**. *Cell Transplant* (2016) **25** 829-48. DOI: 10.3727/096368915X689622
5. Lee BH, Park JN, Lee EJ, Moon YW, Wang JH. **Therapeutic Efficacy of Spherical Aggregated Human Bone Marrow-Derived Mesenchymal Stem Cells Cultured for Osteochondral Defects of Rabbit Knee Joints.**. *Am J Sports Med* (2018) **46** 2242-52. DOI: 10.1177/0363546518780991
6. He Q, Song J, Cui C, Wang J, Hu H, Guo X, Yang M, Wang L, Yan F, Liang K, Liu Z, Liu F, Sun Z. **Mesenchymal stem cell-derived exosomal miR-146a reverses diabetic β-cell dedifferentiation.**. *Stem Cell Res Ther* (2021) **12** 449. DOI: 10.1186/s13287-021-02371-0
7. Muralikumar M, Manoj Jain S, Ganesan H, Duttaroy AK, Pathak S, Banerjee A. **Current understanding of the mesenchymal stem cell-derived exosomes in cancer and aging.**. *Biotechnol Rep (Amst)* (2021) **31** e00658. DOI: 10.1016/j.btre.2021.e00658
8. Zhao M, Dai Y, Li P, Wang J, Ma T, Xu S. **Correction: Inhibition of NLRP3 inflammasome activation and pyroptosis with the ethyl acetate fraction of**. *Food Funct* (2021) **12** 8813. DOI: 10.1039/d1fo90071d
9. Li B, Luan S, Chen J, Zhou Y, Wang T, Li Z, Fu Y, Zhai A, Bi C. **The MSC-Derived Exosomal lncRNA H19 Promotes Wound Healing in Diabetic Foot Ulcers by Upregulating PTEN via MicroRNA-152-3p.**. *Mol Ther Nucleic Acids* (2020) **19** 814-26. DOI: 10.1016/j.omtn.2019.11.034
10. Yan L, Liu G, Wu X. **The umbilical cord mesenchymal stem cell-derived exosomal lncRNA H19 improves osteochondral activity through miR-29b-3p/FoxO3 axis.**. *Clin Transl Med* (2021) **11** e255. DOI: 10.1002/ctm2.255
11. Liu B, Lin L, Yu S, Xia R, Zheng L. **Long non-coding RNA H19 acts as a microRNA-194 sponge to inhibit the apoptosis and promote the proliferation of hypertrophic scar fibroblasts.**. *Can J Physiol Pharmacol* (2021) **99** 1288-97. DOI: 10.1139/cjpp-2021-0351
12. Kuai L, Jiang JS, Li W, Li B, Yin SY. **Long non-coding RNAs in diabetic wound healing: Current research and clinical relevance.**. *Int Wound J* (2022) **19** 583-600. DOI: 10.1111/iwj.13655
13. Zhu Q, Zheng M, Balakrishnan A, Karki R, Kanneganti TD. **Gasdermin D Promotes AIM2 Inflammasome Activation and Is Required for Host Protection against**. *J Immunol* (2018) **201** 3662-8. DOI: 10.4049/jimmunol.1800788
14. Liu P, Zhang Z, Li Y. **Relevance of the Pyroptosis-Related Inflammasome Pathway in the Pathogenesis of Diabetic Kidney Disease.**. *Front Immunol* (2021) **12** 603416. DOI: 10.3389/fimmu.2021.603416
15. Jin R, Cui Y, Chen H, Zhang Z, Weng T, Xia S, Yu M, Zhang W, Shao J, Yang M, Han C, Wang X. **Three-dimensional bioprinting of a full-thickness functional skin model using acellular dermal matrix and gelatin methacrylamide bioink.**. *Acta Biomater* (2021) **131** 248-61. DOI: 10.1016/j.actbio.2021.07.012
16. Song FC, Yuan JQ, Zhu MD, Li Q, Liu SH, Zhang L, Zhao C. **High glucose represses the proliferation of tendon fibroblasts by inhibiting autophagy activation in tendon injury.**. *Biosci Rep* (2021). DOI: 10.1042/BSR20210640
17. Mansano Pletsch AH, de Souza Borges NC, Villar DM, Franzini Sutilo AL, de Oliveira Guirro EC, de Paula FJA, de Jesus Guirro RR. **Does sensorimotor training influence neuromuscular responses, balance, and quality of life in diabetics without a history of diabetic distal polyneuropathy?**. *J Bodyw Mov Ther* (2021) **27** 148-56. DOI: 10.1016/j.jbmt.2021.01.012
18. Zhang L, Wang S, Tan M, Zhou H, Tang Y, Zou Y. **Efficacy of Oxidized Regenerated Cellulose/Collagen Dressing for Management of Skin Wounds: A Systematic Review and Meta-Analysis.**. *Evid Based Complement Alternat Med* (2021) **2021** 1058671. DOI: 10.1155/2021/1058671
19. Wang L, Wang C, Chen H, Chen Y, Li J, Xiao L, Zhu D, Yang C. **The Combined Use of a Non-Invasive Skin-Stretching Device and the Negative-Pressure Wound Therapy Technique in the Treatment of Postoperative Diabetic Foot.**. *Diabetes Metab Syndr Obes* (2021) **14** 3523-32. DOI: 10.2147/DMSO.S322757
20. Yan B, Liu T, Yao C, Liu X, Du Q, Pan L. **LncRNA XIST shuttled by adipose tissue-derived mesenchymal stem cell-derived extracellular vesicles suppresses myocardial pyroptosis in atrial fibrillation by disrupting miR-214-3p-mediated Arl2 inhibition.**. *Lab Invest* (2021) **101** 1427-38. DOI: 10.1038/s41374-021-00635-0
21. Liu F, Yang XC, Chen ML, Zhuang ZW, Jiang Y, Wang J, Zhou YJ. **LncRNA H19/Runx2 axis promotes VSMCs transition via MAPK pathway.**. *Am J Transl Res* (2020) **12** 1338-47. PMID: 32355545
22. Zhang Y, Li Y, Hu Q, Xi Y, Xing Z, Zhang Z, Huang L, Wu J, Liang K, Nguyen TK, Egranov SD, Sun C, Zhao Z. **The lncRNA H19 alleviates muscular dystrophy by stabilizing dystrophin.**. *Nat Cell Biol* (2020) **22** 1332-45. DOI: 10.1038/s41556-020-00595-5
23. Han H, Tian T, Huang G, Li D, Yang S. **The lncRNA H19/miR-541-3p/Wnt/β-catenin axis plays a vital role in melatonin-mediated osteogenic differentiation of bone marrow mesenchymal stem cells.**. *Aging (Albany NY)* (2021) **13** 18257-73. DOI: 10.18632/aging.203267
24. Wang L, Qi L. **The role and mechanism of long non-coding RNA H19 in stem cell osteogenic differentiation.**. *Mol Med* (2021) **27** 86. DOI: 10.1186/s10020-021-00350-y
25. Liao J, Yu X, Hu X, Fan J, Wang J, Zhang Z, Zhao C, Zeng Z, Shu Y, Zhang R, Yan S, Li Y, Zhang W. **lncRNA H19 mediates BMP9-induced osteogenic differentiation of mesenchymal stem cells (MSCs) through Notch signaling.**. *Oncotarget* (2017) **8** 53581-601. DOI: 10.18632/oncotarget.18655
26. Wang CL, Zuo B, Li D, Zhu JF, Xiao F, Zhang XL, Chen XD. **The long noncoding RNA H19 attenuates force-driven cartilage degeneration via miR-483-5p/Dusp5.**. *Biochem Biophys Res Commun* (2020) **529** 210-7. DOI: 10.1016/j.bbrc.2020.05.180
27. Kim NH, Choi SH, Kim CH, Lee CH, Lee TR, Lee AY. **Reduced MiR-675 in exosome in H19 RNA-related melanogenesis via MITF as a direct target.**. *J Invest Dermatol* (2014) **134** 1075-82. DOI: 10.1038/jid.2013.478
28. Li X, Liu R, Huang Z, Gurley EC, Wang X, Wang J, He H, Yang H, Lai G, Zhang L, Bajaj JS, White M, Pandak WM. **Cholangiocyte-derived exosomal long noncoding RNA H19 promotes cholestatic liver injury in mouse and humans.**. *Hepatology* (2018) **68** 599-615. DOI: 10.1002/hep.29838
29. Wang J, Qu J, Li Y, Feng Y, Ma J, Zhang L, Chu C, Hu H, Wang Y, Ji D. **miR-149-5p Regulates Goat Hair Follicle Stem Cell Proliferation and Apoptosis by Targeting the CMTM3/AR Axis During Superior-Quality Brush Hair Formation.**. *Front Genet* (2020) **11** 529757. DOI: 10.3389/fgene.2020.529757
30. Khandpur S, Gupta S, Gunaabalaji DR. **Stem cell therapy in dermatology.**. *Indian J Dermatol Venereol Leprol* (2021). DOI: 10.25259/IJDVL_19_20
31. Kadota T, Fujita Y, Araya J, Watanabe N, Fujimoto S, Kawamoto H, Minagawa S, Hara H, Ohtsuka T, Yamamoto Y, Kuwano K, Ochiya T. **Human bronchial epithelial cell-derived extracellular vesicle therapy for pulmonary fibrosis via inhibition of TGF-β-WNT crosstalk.**. *J Extracell Vesicles* (2021) **10** e12124. DOI: 10.1002/jev2.12124
32. Fan Z, Jiang C, Wang Y, Wang K, Marsh J, Zhang D, Chen X, Nie L. **Engineered extracellular vesicles as intelligent nanosystems for next-generation nanomedicine.**. *Nanoscale Horiz* (2022) **7** 682-714. DOI: 10.1039/d2nh00070a
33. Huang Y, Yang L. **Mesenchymal stem cell-derived extracellular vesicles in therapy against fibrotic diseases.**. *Stem Cell Res Ther* (2021) **12** 435. DOI: 10.1186/s13287-021-02524-1
34. Zhang E, Geng X, Shan S, Li P, Li S, Li W, Yu M, Peng C, Wang S, Shao H, Du Z. **Exosomes derived from bone marrow mesenchymal stem cells reverse epithelial-mesenchymal transition potentially via attenuating Wnt/β-catenin signaling to alleviate silica-induced pulmonary fibrosis.**. *Toxicol Mech Methods* (2021) **31** 655-66. DOI: 10.1080/15376516.2021.1950250
35. Coombe RF, Hamed H. **An update on granulomatous mastitis: a rare and complex condition.**. *Br J Hosp Med (Lond)* (2021) **82** 1-7. DOI: 10.12968/hmed.2020.0718
36. Ito H, Kanbe A, Sakai H, Seishima M. **Activation of NLRP3 signalling accelerates skin wound healing.**. *Exp Dermatol* (2018) **27** 80-6. DOI: 10.1111/exd.13441
37. Dai J, Zhang X, Wang Y, Chen H, Chai Y. **ROS-activated NLRP3 inflammasome initiates inflammation in delayed wound healing in diabetic rats.**. *Int J Clin Exp Pathol* (2017) **10** 9902-9. PMID: 31966879
38. Song G, Zhou J, Song R, Liu D, Yu W, Xie W, Ma Z, Gong J, Meng H, Yang T, Song Z. **Long noncoding RNA H19 regulates the therapeutic efficacy of mesenchymal stem cells in rats with severe acute pancreatitis by sponging miR-138-5p and miR-141-3p.**. *Stem Cell Res Ther* (2020) **11** 420. DOI: 10.1186/s13287-020-01940-z
39. Li X, Luo S, Zhang J, Yuan Y, Jiang W, Zhu H, Ding X, Zhan L, Wu H, Xie Y, Song R, Pan Z, Lu Y. **lncRNA H19 Alleviated Myocardial I/RI via Suppressing miR-877-3p/Bcl-2-Mediated Mitochondrial Apoptosis.**. *Mol Ther Nucleic Acids* (2019) **17** 297-309. DOI: 10.1016/j.omtn.2019.05.031
40. Zhao G, Liu F, Liu Z, Zuo K, Wang B, Zhang Y, Han X, Lian A, Wang Y, Liu M, Zou F, Li P, Liu X. **MSC-derived exosomes attenuate cell death through suppressing AIF nucleus translocation and enhance cutaneous wound healing.**. *Stem Cell Res Ther* (2020) **11** 174. DOI: 10.1186/s13287-020-01616-8
41. Zhao G, Liu F, Lan S, Li P, Wang L, Kou J, Qi X, Fan R, Hao D, Wu C, Bai T, Li Y, Liu JY. **Large-scale expansion of Wharton’s jelly-derived mesenchymal stem cells on gelatin microbeads, with retention of self-renewal and multipotency characteristics and the capacity for enhancing skin wound healing.**. *Stem Cell Res Ther* (2015) **6** 38. DOI: 10.1186/s13287-015-0031-3
42. Jiang Y, Liu F, Zou F, Zhang Y, Wang B, Zhang Y, Lian A, Han X, Liu Z, Liu X, Jin M, Wang D, Li G, Liu J. **PBX homeobox 1 enhances hair follicle mesenchymal stem cell proliferation and reprogramming through activation of the AKT/glycogen synthase kinase signaling pathway and suppression of apoptosis.**. *Stem Cell Res Ther* (2019) **10** 268. DOI: 10.1186/s13287-019-1382-y
43. Xu H, Zhao G, Zhang Y, Jiang H, Wang W, Zhao D, Hong J, Yu H, Qi L. **Mesenchymal stem cell-derived exosomal microRNA-133b suppresses glioma progression via Wnt/β-catenin signaling pathway by targeting EZH2.**. *Stem Cell Res Ther* (2019) **10** 381. DOI: 10.1186/s13287-019-1446-z
44. Zhang C, Xiao C, Ren G, Cai D, Long L, Li J, Li K, Tang Y, Huang T, Deng W. **C-terminal-truncated hepatitis B virus X protein promotes hepatocarcinogenesis by activating the MAPK pathway.**. *Microb Pathog* (2021) **159** 105136. DOI: 10.1016/j.micpath.2021.105136
45. Coperchini F, Croce L, Pignatti P, Ricci G, Gangemi D, Magri F, Imbriani M, Rotondi M, Chiovato L. **The new generation PFAS C6O4 does not produce adverse effects on thyroid cells**. *J Endocrinol Invest* (2021) **44** 1625-35. DOI: 10.1007/s40618-020-01466-4
46. Huang Y, Mei X, Jiang W, Zhao H, Yan Z, Zhang H, Liu Y, Hu X, Zhang J, Peng W, Zhang J, Qi Q, Chen N. **Mesenchymal Stem Cell-Conditioned Medium Protects Hippocampal Neurons From Radiation Damage by Suppressing Oxidative Stress and Apoptosis.**. *Dose Response* (2021) **19** 1559325820984944. DOI: 10.1177/1559325820984944
47. Wu Q, Miao X, Zhang J, Xiang L, Li X, Bao X, Du S, Wang M, Miao S, Fan Y, Wang W, Xu X, Shen X. **Astrocytic YAP protects the optic nerve and retina in an experimental autoimmune encephalomyelitis model through TGF-β signaling.**. *Theranostics* (2021) **11** 8480-99. DOI: 10.7150/thno.60031
48. Schilb AL, Ayat NR, Vaidya AM, Hertz LM, Hall RC, Scheidt JH, Sun D, Sun Z, Gopalakrishnan R, Lu ZR. **Efficacy of Targeted ECO/miR-200c Nanoparticles for Modulating Tumor Microenvironment and Treating Triple Negative Breast Cancer as Non-invasively Monitored by MR Molecular Imaging.**. *Pharm Res* (2021) **38** 1405-18. DOI: 10.1007/s11095-021-03083-z
49. Damond N, Thorel F, Moyers JS, Charron MJ, Vuguin PM, Powers AC, Herrera PL. **Blockade of glucagon signaling prevents or reverses diabetes onset only if residual β-cells persist.**. *Elife* (2016) **5** e13828. DOI: 10.7554/eLife.13828
50. Marques LS, Zborowski VA, Heck SO, Fulco BCW, Nogueira CW. **4,4’-Dichloro-diphenyl diselenide modulated oxidative stress that differently affected peripheral tissues in streptozotocin-exposed mice.**. *Can J Physiol Pharmacol* (2021) **99** 943-51. DOI: 10.1139/cjpp-2020-0652
51. Guo Q, Liu Z, Zheng J, Zhao H, Li C. **Substances for regenerative wound healing during antler renewal stimulated scar-less restoration of rat cutaneous wounds.**. *Cell Tissue Res* (2021) **386** 99-116. DOI: 10.1007/s00441-021-03505-9
|
---
title: Epigenetic age and lung cancer risk in the CLUE II prospective cohort study
authors:
- Dominique S. Michaud
- Mei Chung
- Naisi Zhao
- Devin C. Koestler
- Jiayun Lu
- Elizabeth A. Platz
- Karl T. Kelsey
journal: Aging (Albany NY)
year: 2023
pmcid: PMC9970317
doi: 10.18632/aging.204501
license: CC BY 3.0
---
# Epigenetic age and lung cancer risk in the CLUE II prospective cohort study
## Abstract
Background: Epigenetic age, a robust marker of biological aging, has been associated with obesity, low-grade inflammation and metabolic diseases. However, few studies have examined associations between different epigenetic age measures and risk of lung cancer, despite great interest in finding biomarkers to assist in risk stratification for lung cancer screening.
Methods: A nested case-control study of lung cancer from the CLUE II cohort study was conducted using incidence density sampling with 1:1 matching of controls to lung cancer cases ($$n = 208$$ matched pairs). Prediagnostic blood samples were collected in 1989 (CLUE II study baseline) and stored at −70°C. DNA was extracted from buffy coat and DNA methylation levels were measured using Illumina MethylationEPIC BeadChip Arrays. Three epigenetic age acceleration (i.e., biological age is greater than chronological age) measurements (Horvath, Hannum and PhenoAge) were examined in relation to lung cancer risk using conditional logistic regression.
Results: We did not observe associations between the three epigenetic age acceleration measurements and risk of lung cancer overall; however, inverse associations for the two Hannum age acceleration measures (intrinsic and extrinsic) were observed in men and among younger participants, but not in women or older participants. We did not observe effect modification by time from blood draw to diagnosis.
Conclusion: Findings from this study do not support a positive association between three different biological age acceleration measures and risk of lung cancer. Additional studies are needed to address whether epigenetic age is associated with lung cancer in never smokers.
## INTRODUCTION
Lung cancer remains the leading cause of cancer deaths in the US and worldwide [1]. Substantial effort has been devoted to identifying heritable genomic markers that could aid in classification of high-risk individuals for screening purposes [2, 3]. While results from these studies are promising, predictive modeling using genomic markers (in addition to age and smoking) are currently not sufficiently discriminatory or calibrated to be useful in clinical settings for risk prediction. Identifying high risk groups could improve efficiency in lung cancer screening (with low-dose computed tomography) and reduce racial inequalities associated with the current recommendations for screening based on smoking history [4]. Thus, there is an urgent need to identify biomarkers that can reflect biological processes in lung cancer development and that could, eventually, be incorporated into models for risk stratification.
Variation in DNA methylation levels in peripheral blood leukocytes reflect genetic imprinting, environmental exposures, and the lineage differentiation that gives rise to immune cell subtypes [5]. Recent studies using epigenetic markers in blood have identified differentially methylated regions in smokers [6, 7]; DNA methylation levels in these regions remained strongly associated with lung cancer mortality after adjusting for smoking history [6]. Epigenetic aging measures or “clocks” have also been developed to reflect biological age in tissue and blood [8]. These epigenetic clocks are highly correlated with chronological age, but can deviate from chronological age, reflecting changes in immunity and cellular senescence, which are closely aligned with health and disease. Epigenetic age acceleration is the difference between the predicted biological age (based on the epigenetic measurements) and the given chronological age. Recent studies have linked epigenetic age acceleration to a range of disease outcomes, including all-cause mortality [9, 10], cardiovascular disease (CVD) incidence [11], coronary heart disease (CHD) mortality [12], cancer incidence [13] and cancer mortality [12]. Epigenetic age acceleration estimated using the Horvath and Hannum clocks, known as “first generation clocks”, is highly heritable (~0.4 [9]) and has been associated with CVD and cancer risk factors, including obesity, low-grade inflammation [14], and metabolic syndrome [15]. The newer generation of clocks, such as PhenoAge clock, have been developed based on associations with age, all-cause mortality, and several clinical biomarkers [16].
To date, studies evaluating epigenetic age acceleration and lung cancer risk have been inconsistent. The first nested case-control study conducted in the Women’s Health Initiative (WHI) observed a strong positive association [17], while a larger nested case-control study (Melbourne Collaborative Consortium Study; MCCS) reported no associations for the Horvath and Hannum clocks [18] but positive association with PhenoAge clock [13]. Additionally, stratified analyses by time since blood drawn were performed in the MCCS, and the results showed no significant differences in the positive associations between PhenoAge acceleration and lung cancer risk by time since blood drawn (≤5 years, 5–10 years, or >10 years) [13]. In the WHI study, the positive associations were stronger among women developing lung cancer at 70 or more years and among current smokers. In the MCCS [18], men and women were combined, and no stratified analyses were conducted by sex, to inform whether the association was restricted to women.
It is important to examine whether epigenetic age is associated with lung cancer risk across multiple prospective studies to determine its utility as a potential biomarker to be considered for risk stratification in the selection of high-risk individuals for lung cancer screening. Thus, for this analysis, we conducted a nested case-control analysis of 208 lung cancer cases and 208 matched controls with archived pre-diagnostic blood samples (from 1989). The case-control study is nested in the CLUE II cohort study, a predominantly White cohort of men and women, based in Maryland, USA.
## Study population
Individuals included in this analysis were selected from participants in the CLUE II study, a prospective cohort study initiated in Washington County, Maryland, in 1989 [19, 20]. The CLUE II study was an outgrowth of a previous study (CLUE I) that had been conducted in the same region in 1974. Some of the participants in CLUE II had been participants in CLUE I (about a third), but this was not a requirement for recruitment into the overall CLUE II cohort. At the baseline visit (1989 for CLUE II), brief medical histories, blood pressure readings, and blood samples were collected on 32,894 participants (25,076 of which were residents of Washington County). Mobile office trailers were used to recruit participants and to collect blood samples. Blood was drawn into 20 ml heparinized Vacutainers (Becton- Dickinson, Rutherford, NJ), kept at 4°C until the plasma was separated, usually within 2–6 h, and divided into aliquots of plasma, buffy coat, and red blood cells. All samples are stored at −70°C. Comparisons with published figures from the 1990 Census indicated that approximately 30 percent of adult residents had participated: $98.3\%$ were White, reflecting the population of this county, and $59\%$ were female, with the better-educated and the age group 45 to 70 years having higher participation rates. Self-reported attained education, weight and height, cigarette smoking status, number of cigarettes smoked per day, and cigar/pipe smoking status were recorded for each participant at baseline.
The Institutional Review Board at the Johns Hopkins Bloomberg School of Public Health and the Tufts University Health Sciences Campus Institutional Review Board approved this study.
## Lung cancer cases and matched controls
Incident lung cancer cases were ascertained from linkage to the Washington Co. Cancer Registry (1989-January 2018) and the Maryland Cancer Registry (1992-January 2018). The Maryland Cancer *Registry is* certified by the North American Association of Central Cancer Registries as being more than $95\%$ complete. Compared with the Maryland Cancer Registry, the Washington County Cancer Registry captured $98\%$ of the lung cancer cases diagnosed in Washington County residents in 1998. Cancer deaths were identified from state vital statistics, next of kin, and obituaries and confirmed on death certificates; underlying cause of death was obtained from the death certificates. Between 1989 and January 2018, a total of 241 eligible incident first primary lung cancer cases were ascertained from CLUE II participants with blood samples, and who had also previously participated in CLUE I (a requirement based on a shared study population). All 241 lung cancer cases (ICD 9 162 and ICD10 C34) were confirmed by pathology report.
Controls were selected from among CLUE II participants who had also participated in CLUE I. Matching was conducted using incidence density sampling such that a control had to be alive and free of cancer at the time the matched case was diagnosed with lung cancer. One control was matched to each case on the following factors: age (±3 year), sex, race, cigarette smoking status and intensity, cigar/pipe smoking status, and date of blood draw (±4 months). Controls who later became cases were also included as cases with their new matched controls.
## DNA methylation measurements
Extracted DNA was bisulfite-treated using the EZ DNA Methylation Kit (Zymo), and DNA methylation was measured with the 850K Illumina Infinium MethylationEPIC BeadChip Arrays (Illumina, Inc, CA, USA). All samples and all array measurements were performed blinded to case-control status. Details on DNA methylation measurements, data preprocessing processing and quality control assessment/screening have been published [21]. Due to lack of remaining DNA, 8 of the 241 incident cases were removed from the dataset before matching.
## Estimation of peripheral blood leukocyte composition
Peripheral blood leukocyte subtypes proportions were estimated using a newly expanded reference-based deconvolution library EPIC IDOL-Ext [22]. This library used the IDOL methodology [23] to optimize the currently available six-cell reference library [24] to deconvolute the proportions of 12 leukocyte subtypes in peripheral blood (neutrophils, eosinophils, basophils, monocytes, naïve and memory B cells, naïve and memory CD4+ and CD8+ cells, natural killer, and T regulatory cells).
## Data processing
All methylation data preprocessing and normalization steps were performed using the Bioconductor packages. The raw IDAT files from methylation array were processed using the minfi Bioconductor package [25, 26]. Within-array correction for background fluorescence and dye-biases were performed using the Noob methodology via the function “preprocessNoob” in the minfi Bioconductor package [27]. The QCinfo function in ENmix Bioconductor [28] package was then used to identify and remove poor quality samples and probes. Samples were excluded if: 1) more than $5\%$ of probes had quality issues as addressed using the detection p-value, 2) the bisulfite conversion intensity was lower than 3 standard deviations from the mean, or 3) the mean average intensity and/or the mean average beta values were more than 3 times IQR from the upper quartile or less than 3 time IQR from the lower quartile of the average intensity values or beta value across the samples. In addition, we excluded probes that had detection p-values exceeding 1 × 10−6 (compared to the negative background probes) in more than $5\%$ of the samples. After sample- and probe- level quality control, we corrected the type II probe bias to make the methylation distribution of type II feature comparable to the distribution of type I feature using the beta mixture quantile dilation intra-sample normalization method [29], implemented using “BMIQ” function in the wateRmelon Bioconductor package [30]. Principal components analysis (PCA) was performed on the BMIQ-adjusted values and the top K principal components (K determined using a previously described random matrix theory approach [31]) to detect whether the microarray dataset had the batch effect. Then ctrlsva function in ENmix Bioconductor packages [32] was used to estimate the surrogate variables of batch effects [33]. The estimated surrogate variables were used in downstream analyses to adjust for batch effects and other unwanted technical sources of variation.
## Smoking methylation score
A smoking methylation score was calculated to estimate individual pack-years of smoking based on known smoking-related DNA methylation alterations [34]. The smoking methylation score was first developed to predict smoking pack-years using smoking ‘signatures’ reported by large-scale epigenome-wide association meta-analyses [34]. The score correlates with gene expression changes affected by smoking and can be utilized in lieu of self-reported smoking data.
## Estimation of epigenetic age
Three DNAm clocks (Hannum [35], Horvath [36], and PhenoAge [16]) were used to estimate subjects’ DNAm age (using ENMix Biocondontor package: https://rdrr.io/bioc/ENmix/man/methyAge.html). For each of the three DNAm clocks, DNAm age acceleration (AA) was defined by regressing DNAm age on chronologic age and calculating the difference between the observed chronological age and the fitted DNAm age (i.e., the residual). Additionally, intrinsic epigenetic age acceleration (IEAA) metrics were calculated using the residuals from the linear regression fit to DNAm age on chronologic age, adjusted for estimated blood cell composition [37, 24] (for comparability with prior studies, we did not update the reference library for the IEAA measurements). Three subjects with an absolute value of the age acceleration estimate greater than 3 standard deviations (SDs) from the mean were excluded from the regression analyses; sensitivity analyses conducted retaining these 3 subjects did not materially modify the results.
## Statistical analyses
Given the 1:1 case-control matching present in our study, conditional logistic regression models were used to examine the association between epigenetic age acceleration and lung cancer risk. As age, sex, and smoking status (never, former, current), smoking intensity (cigarettes/day) and cigar/pipe smoking were matching factors, these were implicitly adjusted for when using conditional regression. In the conditional regression model, we additionally adjusted for BMI as a continuous variable, batch effect (for methylation arrays), and a previously described methylation-predicted variable to capture pack-years smoked [34]. We conducted stratified analyses by sex, median age, lung cancer histology (non-small cell lung cancer [NSCLC], small cell lung cancer [SCLC]), and length of time between blood draw and diagnosis (≤10, >10 years) to evaluate potential effect modification. We did not adjust for methylation-derived cell proportions given that the intrinsic epigenetic age (IEAA) measures already account for immune cells. Pearson’s correlation was used to examine correlation between epigenetic age acceleration and methylation predicted immune cell proportions. All statistical analyses were performed in R (version 3.5.1).
## Availability of data materials
The data cannot be deposited into a controlled access database due to a State of Maryland law that established the Maryland Cancer Registry (where the lung cancer data was obtained).
## RESULTS
The final analysis consisted of 208 cases and matched 208 controls. As a result of matching, lung cancer cases and controls had similar age (mean age: 55.9 years among controls, 58.3 years among cases), sex distribution ($54.3\%$ females in cases and controls), smoking status ($51\%$ current smokers in cases and controls; $39\%$ former smokers in cases and controls), smoking intensity (25 cigarettes per day in current smokers among cases; 24 cigarettes/day in current smokers among controls) and cigar or pipe smoking ($15\%$ ever in cases and controls). Only 3 cases, and no controls, were non-White individuals. Cases and controls were also similar with respect to BMI (mean, in kg/m2, BMI 26.0 cases, 26.2 controls). Most lung cancer cases were NSCLC ($74\%$). Cases were diagnosed a mean of 14 years post-blood donation (median 14 years; range >0–29 years; all cases were incident cases).
In this population, men, and cases with a shorter time between blood draw and cancer diagnosis, were more likely to have age acceleration (vs. deceleration) in all 3 epigenetic clock measures (Table 1). Other characteristics, including smoking and BMI were very similar for acceleration and deceleration of epigenetic age in all 3 measures.
**Table 1**
| Unnamed: 0 | Overall | AA_Hannum | AA_Hannum.1 | AA_Horvath | AA_Horvath.1 | AA_Pheno | AA_Pheno.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | Overall | Acceleration | Deceleration | Acceleration | Deceleration | Acceleration | Deceleration |
| N | 416 | 210 | 203 | 208 | 205 | 220 | 193 |
| Case | 208 (50.0%) | 104 (49.5%) | 101 (49.8%) | 99 (47.6%) | 107 (52.2%) | 114 (51.8%) | 91 (47.2%) |
| Control | 208 (50.0%) | 106 (50.5%) | 102 (50.2%) | 109 (52.4%) | 98 (47.8%) | 106 (48.2%) | 102 (52.8%) |
| Age (yrs), mean (SD) | 57.1 (10.0) | 57.3 (9.3) | 56.6 (10.5) | 57.1 (8.9) | 56.9 (11.0) | 56.8 (9.4) | 57.2 (10.7) |
| Female, n (%) | 226 (54.3%) | 90 (42.9%) | 133 (65.5%) | 97 (46.6%) | 127 (62.0%) | 108 (49.1%) | 115 (59.6%) |
| Time Difference (yrs), mean (SD) | 14.9 (7.9) | 13.9 (7.3) | 16.2 (8.2) | 14.5 (7.6) | 15.4 (8.1) | 14.7 (7.7) | 15.3 (8.2) |
| Smoking status | | | | | | | |
| Never | 44 (10.6%) | 23 (11.0%) | 21 (10.3%) | 28 (13.5%) | 16 (7.8%) | 22 (10.0%) | 21 (10.9%) |
| Former | 160 (38.5%) | 86 (41.0%) | 73 (36.0%) | 80 (38.5%) | 79 (38.5%) | 86 (38.6%) | 74 (38.3%) |
| Current | 212 (50.9%) | 101 (48.0%) | 109 (53.7%) | 100 (48.0%) | 110 (53.7%) | 112 (51.4%) | 98 (50.8%) |
| BMI (kg/m2), mean (SD) | 26.1 (4.4) | 26.2 (4.3) | 26.1 (4.5) | 26.7 (4.5) | 25.6 (4.2) | 26.0 (4.4) | 26.2 (4.4) |
| Normal and underweight | 179 (43.0%) | 87 (41.4%) | 90 (44.3%) | 76 (36.5%) | 100 (48.8%) | 94 (42.3%) | 82 (42.5%) |
| Overweight | 171 (41.1%) | 89 (42.4%) | 81 (39.9%) | 91 (43.8%) | 80 (39.0%) | 95 (43.2%) | 76 (39.4%) |
| Obese | 66 (15.9%) | 34 (16.2%) | 32 (15.8%) | 41 (19.7%) | 25 (12.2%) | 31 (14.4%) | 35 (18.1%) |
Overall, we did not observe any associations between the 3 epigenetic age acceleration measures and lung cancer risk using both continuous and categorical variables for age acceleration (Table 2). Associations were similar when stratified by time between blood drawn and cancer diagnosis (Table 3).
Given that in a prior study positive associations were modified by age and smoking status and were only reported for women [17], we conducted stratified analyses by age (<65 years, ≥65 years), smoking status (current vs. former smokers) and sex. There was an inverse trend for the Hannum measurements in men but not in women (Table 4). Associations for all three age acceleration measures were statistically significantly inversely associated with lung cancer in the younger (<65 years) but not older age group (Supplementary Table 1). Associations were similar among current and former smokers after adjusting for methylation predicted pack-years; associations were not estimated among never smokers due to small numbers ($$n = 22$$ matched pairs; Supplementary Table 1). Finally, to examine whether associations might vary by histology, we separated NSCLC and SCLC; no associations were observed in either subgroup (data not shown).
**Table 4**
| Unnamed: 0 | Female | Female.1 | Male | Male.1 |
| --- | --- | --- | --- | --- |
| | OR (95% CI) | p-value | OR (95% CI) | p-value |
| AA_Hannum | | | | |
| Q1 | 1 (ref.) | | 1 (ref.) | |
| Q2 | 0.74 (0.28, 2.00) | 0.56 | 0.78 (0.30, 2.00) | 0.60 |
| Q3 | 1.18 (0.49, 2.84) | 0.71 | 0.50 (0.21, 1.19) | 0.12 |
| Q4 | 0.79 (0.33, 1.88) | 0.60 | 0.29 (0.10, 0.80) | 0.017 |
| | | P trend = 0.78 | | P trend =0.015 |
| AA_Horvath | | | | |
| Q1 | 1 (ref.) | | 1 (ref.) | |
| Q2 | 0.92 (0.40, 2.13) | 0.84 | 0.72 (0.30, 1.73) | 0.47 |
| Q3 | 0.82 (0.34, 1.99) | 0.66 | 0.51 (0.21, 1.24) | 0.14 |
| Q4 | 0.92 (0.40, 2.15) | 0.85 | 0.50 (0.19, 1.35) | 0.17 |
| | | P trend = 0.80 | | P trend = 0.13 |
| AA_Pheno | | | | |
| Q1 | 1 (ref.) | | 1 (ref.) | |
| Q2 | 0.70 (0.28, 1.76) | 0.44 | 1.11 (0.43, 2.86) | 0.83 |
| Q3 | 0.89 (0.38, 2.05) | 0.78 | 0.90 (0.36, 2.26) | 0.83 |
| Q4 | 0.93 (0.40, 2.16) | 0.86 | 0.60 (0.21, 1.74) | 0.35 |
| | | P trend = 0.97 | | P trend = 0.30 |
| IEAA_Hannum | | | | |
| Q1 | 1 (ref.) | | 1 (ref.) | |
| Q2 | 1.27 (0.52, 3.08) | 0.60 | 0.31 (0.11, 0.82) | 0.018 |
| Q3 | 1.17 (0.47, 2.93) | 0.74 | 0.37 (0.14, 0.96) | 0.041 |
| Q4 | 0.92 (0.40, 0.10) | 0.84 | 0.35 (0.13, 0.92) | 0.034 |
| | | P trend = 0.68 | | P trend = 0.03 |
| IEAA_Horvath | | | | |
| Q1 | 1 (ref.) | | 1 (ref.) | |
| Q2 | 0.85 (0.36, 2.00) | 0.70 | 0.85 (0.34, 2.15) | 0.73 |
| Q3 | 0.73 (0.31, 1.75) | 0.48 | 0.83 (0.33, 2.11) | 0.70 |
| Q4 | 0.21 (0.54, 2.74) | 0.64 | 0.53 (0.17, 1.58) | 0.25 |
| | | P trend = 0.65 | | P trend = 0.28 |
| IEAA_Pheno | | | | |
| Q1 | 1 (ref.) | | 1 (ref.) | |
| Q2 | 1.01 (0.43, 2.36) | 0.98 | 1.34 (0.51,3.52) | 0.55 |
| Q3 | 1.10 (0.46, 2.63) | 0.83 | 1.18 (0.46, 3.05) | 0.73 |
| Q4 | 1.04 (0.46, 2.35) | 0.92 | 0.63 (0.21, 1.82) | 0.39 |
| | | P trend = 0.88 | | P trend = 0.38 |
Prior studies suggest that epigenetic age acceleration may be strongly linked to the immune response, and specifically CD8 and CD4 naïve cells [38]. In this study, all three epigenetic clock measures were strongly associated with CD8 and CD4 naïve immune subsets in control subjects (Pearson correlations ranging from −0.22 to −0.41; Supplementary Table 2). The only statistically significant correlation between the clock measures and NK cells was for PhenoAge (r = −0.16); however, the NK cells in the reference library do not differentiate naïve and memory NK cells, so it is possible the associations would be different for NK naïve cells. The CD8 memory cells were positively associated with Hannum and Horvath clocks but not with PhenoAge. The IEAA measures (for each clock) were not associated with CD8 memory cells but were still strongly associated with CD8 naïve cells, which can be explain by the lack of adjustment for naïve and memory T cells as these fractions were not available in earlier deconvolution libraries (and memory cell represent a larger proportion of total T cells in older adults).
## DISCUSSION
In this nested case-control study on incident lung cancer, we observed no positive associations between lung cancer risk and epigenetic age acceleration using three different measures (Horvath, Hannum and PhenoAge) with two adjustment approaches for each, i.e., intrinsic and extrinsic measures. We observed inverse associations for men and subjects below the median age, but not in women or older subjects.
Our null findings for epigenetic age acceleration associations with lung cancer risk using the Horvath and Hannum clocks are consistent with those reported in a nested case-control study in the Melbourne Collaborative Cohort Study (MCCS; 332 cases) [18]. Our null findings differ from those reported in the Women’s Health Initiative (WHI), where a $50\%$ increase in risk of lung cancer was observed for every unit increase in intrinsic epigenetic age acceleration using the Horvath epigenetic age measure ($$p \leq 3.4$$ × 10−3) [17]; it is worth noting that the number of lung cancer cases included in the WHI analysis was small ($$n = 43$$). Our results for age acceleration based on the PhenoAge measure were also null, whereas a positive association was observed for PhenoAge and lung cancer risk in the MCCS (OR = 1.25, $95\%$ CI = 1.05–1.49, for a 1 SD increase) [13] and in the WHI (HR = 1.05, $$p \leq 0.031$$) [16]. The PhenoAge measure was derived using immune and inflammatory phenotypes, in contrast to the other two epigenetic measures. Differences in the two populations may explain the different findings, such as smoking prevalence, although we could not confirm that as the MCCS analysis did not provide characteristics for the lung cancer case-control study (only for the pooled population). We observed that adjusting for methylation predicted pack-years attenuated the associations for PhenoAge in our analysis (among current smokers: before adjustment OR = 2.15, $95\%$ CI = 0.92–5.04 for Q4 vs. Q1; after adjustment OR = 1.40, $95\%$ CI = 0.52–3.76; overall: before adjustment OR 1.22, $95\%$ CI = 0.70–2.13; after adjustment OR = 0.86, $95\%$ CI = 0.46–1.62). Thus, it is possible that elevated risk associated with the PhenoAge age acceleration in some studies is a measure of the residual effect of smoking, which is captured with the methylation markers for pack-years smoked.
The inverse associations between age acceleration for several epigenetic clock measurements and lung cancer risk we observed in men and subjects less than 65 years of age were unexpected. However, our findings are consistent with results from a large study using Mendelian randomization (MR) methods conducted to examine the causal link between several epigenetic clocks and cancer, including lung cancer. In the MR analysis, the genetically predicted intrinsic Horvath Age acceleration was associated with a decrease in lung cancer risk (the association was statistically significant, $$p \leq 0.03$$, prior to multiple comparisons correction). Alternatively, these results may be due to a selection bias that occurred as a results of survivor bias in enrollment into our cohort (this could have occurrent if individuals with poor health and epigenetic age acceleration were less likely to participate).
Recent studies have begun to elucidate the biological processes that explain age acceleration associations detected using epigenetic clocks [38]. In a functional genomics study, changes in proportions of naive and activated immune blood cells were strongly associated with the Hannum and Horvath age acceleration measures [38]. We confirmed these relationships in our dataset using new immune cell reference libraries allowing the deconvolution of naive immune T cells [22]. The strong inverse correlations observed between naive T and B cells and the three age acceleration measurements suggest that they are strongly linked to changes in the immune response, which is not surprising, given that reduction of naïve T cells is a component of immunosenescence [39]. Of interest, the intrinsic epigenetic age acceleration (IEAA) measurements remained strongly associated with the CD8 naïve cells; future analyses using intrinsic measures of age acceleration should adjust for these cells.
The strengths of our study include the prospective nature of the analysis with a long follow-up period, thus removing the potential for spurious associations that may be driven by the cancer progression (i.e., reverse causation), a relatively large sample size for methylation analyses, and tight adjustment for smoking. In addition, the cancer ascertainment for the CLUE II cohort is very high, given the quality of cancer registry data. The limitations of our analysis include one-time point for epigenetic measurements and lack of data on non-Whites, thus limiting generalizability.
To our knowledge, this is the third prospective study examining the association between epigenetic aging, measured in peripheral blood, and risk of lung cancer. Findings from this study suggest that there are no strong positive associations between biological aging, measured an average of 15 years prior to cancer, and lung cancer risk. The majority of cases in this study were ever smokers ($90\%$), and smoking history was well controlled for, suggesting that biological aging, independent of smoking, is not associated with an increased risk of lung cancer, at least among smokers. Our data also suggest that prior associations with PhenoAge and lung cancer might have been due to residual effects of smoking. Future studies should include more racially diverse populations and examine associations among never smokers.
## References
1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. **Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries.**. *CA Cancer J Clin* (2021) **71** 209-49. DOI: 10.3322/caac.21660
2. Bossé Y, Amos CI. **A Decade of GWAS Results in Lung Cancer.**. *Cancer Epidemiol Biomarkers Prev* (2018) **27** 363-79. DOI: 10.1158/1055-9965.EPI-16-0794
3. Lebrett MB, Crosbie EJ, Smith MJ, Woodward ER, Evans DG, Crosbie PAJ. **Targeting lung cancer screening to individuals at greatest risk: the role of genetic factors.**. *J Med Genet* (2021) **58** 217-26. DOI: 10.1136/jmedgenet-2020-107399
4. Colson YL, Shepard JO, Lennes IT. **New USPSTF Guidelines for Lung Cancer Screening: Better but Not Enough.**. *JAMA Surg* (2021) **156** 513-4. DOI: 10.1001/jamasurg.2021.0242
5. Michaud DS, Kelsey KT. **DNA Methylation in Peripheral Blood: Providing Novel Biomarkers of Exposure and Immunity to Examine Cancer Risk.**. *Cancer Epidemiol Biomarkers Prev* (2021) **30** 2176-8. DOI: 10.1158/1055-9965.EPI-21-0866
6. Baglietto L, Ponzi E, Haycock P, Hodge A, Bianca Assumma M, Jung CH, Chung J, Fasanelli F, Guida F, Campanella G, Chadeau-Hyam M, Grankvist K, Johansson M. **DNA methylation changes measured in pre-diagnostic peripheral blood samples are associated with smoking and lung cancer risk.**. *Int J Cancer* (2017) **140** 50-61. DOI: 10.1002/ijc.30431
7. Heikkinen A, Bollepalli S, Ollikainen M. **The potential of DNA methylation as a biomarker for obesity and smoking.**. *J Intern Med* (2022) **292** 390-408. DOI: 10.1111/joim.13496
8. Horvath S, Zhang Y, Langfelder P, Kahn RS, Boks MP, van Eijk K, van den Berg LH, Ophoff RA. **Aging effects on DNA methylation modules in human brain and blood tissue.**. *Genome Biol* (2012) **13** R97. DOI: 10.1186/gb-2012-13-10-r97
9. Marioni RE, Shah S, McRae AF, Chen BH, Colicino E, Harris SE, Gibson J, Henders AK, Redmond P, Cox SR, Pattie A, Corley J, Murphy L. **DNA methylation age of blood predicts all-cause mortality in later life.**. *Genome Biol* (2015) **16** 25. DOI: 10.1186/s13059-015-0584-6
10. Chen BH, Marioni RE, Colicino E, Peters MJ, Ward-Caviness CK, Tsai PC, Roetker NS, Just AC, Demerath EW, Guan W, Bressler J, Fornage M, Studenski S. **DNA methylation-based measures of biological age: meta-analysis predicting time to death.**. *Aging (Albany NY)* (2016) **8** 1844-65. DOI: 10.18632/aging.101020
11. Roetker NS, Pankow JS, Bressler J, Morrison AC, Boerwinkle E. **Prospective Study of Epigenetic Age Acceleration and Incidence of Cardiovascular Disease Outcomes in the ARIC Study (Atherosclerosis Risk in Communities).**. *Circ Genom Precis Med* (2018) **11** e001937. DOI: 10.1161/CIRCGEN.117.001937
12. Perna L, Zhang Y, Mons U, Holleczek B, Saum KU, Brenner H. **Epigenetic age acceleration predicts cancer, cardiovascular, and all-cause mortality in a German case cohort.**. *Clin Epigenetics* (2016) **8** 64. DOI: 10.1186/s13148-016-0228-z
13. Dugué PA, Bassett JK, Wong EM, Joo JE, Li S, Yu C, Schmidt DF, Makalic E, Doo NW, Buchanan DD, Hodge AM, English DR, Hopper JL. **Biological Aging Measures Based on Blood DNA Methylation and Risk of Cancer: A Prospective Study.**. *JNCI Cancer Spectr* (2021) **5** pkaa109. DOI: 10.1093/jncics/pkaa109
14. Huang RC, Lillycrop KA, Beilin LJ, Godfrey KM, Anderson D, Mori TA, Rauschert S, Craig JM, Oddy WH, Ayonrinde OT, Pennell CE, Holbrook JD, Melton PE. **Epigenetic Age Acceleration in Adolescence Associates With BMI, Inflammation, and Risk Score for Middle Age Cardiovascular Disease.**. *J Clin Endocrinol Metab* (2019) **104** 3012-24. DOI: 10.1210/jc.2018-02076
15. Nannini DR, Joyce BT, Zheng Y, Gao T, Liu L, Yoon G, Huan T, Ma J, Jacobs DR, Wilkins JT, Ren J, Zhang K, Khan SS. **Epigenetic age acceleration and metabolic syndrome in the coronary artery risk development in young adults study.**. *Clin Epigenetics* (2019) **11** 160. DOI: 10.1186/s13148-019-0767-1
16. Levine ME, Lu AT, Quach A, Chen BH, Assimes TL, Bandinelli S, Hou L, Baccarelli AA, Stewart JD, Li Y, Whitsel EA, Wilson JG, Reiner AP. **An epigenetic biomarker of aging for lifespan and healthspan.**. *Aging (Albany NY)* (2018) **10** 573-91. DOI: 10.18632/aging.101414
17. Levine ME, Hosgood HD, Chen B, Absher D, Assimes T, Horvath S. **DNA methylation age of blood predicts future onset of lung cancer in the women's health initiative.**. *Aging (Albany NY)* (2015) **7** 690-700. DOI: 10.18632/aging.100809
18. Dugué PA, Bassett JK, Joo JE, Jung CH, Ming Wong E, Moreno-Betancur M, Schmidt D, Makalic E, Li S, Severi G, Hodge AM, Buchanan DD, English DR. **DNA methylation-based biological aging and cancer risk and survival: Pooled analysis of seven prospective studies.**. *Int J Cancer* (2018) **142** 1611-9. DOI: 10.1002/ijc.31189
19. Comstock GW, Helzlsouer KJ, Bush TL. **Prediagnostic serum levels of carotenoids and vitamin E as related to subsequent cancer in Washington County, Maryland.**. *Am J Clin Nutr* (1991) **53** 260S-4. DOI: 10.1093/ajcn/53.1.260S
20. Braun MM, Helzlsouer KJ, Hollis BW, Comstock GW. **Colon cancer and serum vitamin D metabolite levels 10-17 years prior to diagnosis.**. *Am J Epidemiol* (1995) **142** 608-11. DOI: 10.1093/oxfordjournals.aje.a117682
21. Zhao N, Ruan M, Koestler DC, Lu J, Salas LA, Kelsey KT, Platz EA, Michaud DS. **Methylation-derived inflammatory measures and lung cancer risk and survival.**. *Clin Epigenetics* (2021) **13** 222. DOI: 10.1186/s13148-021-01214-2
22. Salas LA, Zhang Z, Koestler DC, Butler RA, Hansen HM, Molinaro AM, Wiencke JK, Kelsey KT, Christensen BC. **Enhanced cell deconvolution of peripheral blood using DNA methylation for high-resolution immune profiling.**. *Nat Commun* (2022) **13** 761. DOI: 10.1038/s41467-021-27864-7
23. Koestler DC, Jones MJ, Usset J, Christensen BC, Butler RA, Kobor MS, Wiencke JK, Kelsey KT. **Improving cell mixture deconvolution by identifying optimal DNA methylation libraries (IDOL)**. *BMC Bioinformatics* (2016) **17** 120. DOI: 10.1186/s12859-016-0943-7
24. Salas LA, Koestler DC, Butler RA, Hansen HM, Wiencke JK, Kelsey KT, Christensen BC. **An optimized library for reference-based deconvolution of whole-blood biospecimens assayed using the Illumina HumanMethylationEPIC BeadArray.**. *Genome Biol* (2018) **19** 64. DOI: 10.1186/s13059-018-1448-7
25. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, Irizarry RA. **Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays.**. *Bioinformatics* (2014) **30** 1363-9. DOI: 10.1093/bioinformatics/btu049
26. Fortin JP, Triche TJ, Hansen KD. **Preprocessing, normalization and integration of the Illumina HumanMethylationEPIC array with minfi.**. *Bioinformatics* (2017) **33** 558-60. DOI: 10.1093/bioinformatics/btw691
27. Triche TJ, Weisenberger DJ, Van Den Berg D, Laird PW, Siegmund KD. **Low-level processing of Illumina Infinium DNA Methylation BeadArrays.**. *Nucleic Acids Res* (2013) **41** e90. DOI: 10.1093/nar/gkt090
28. Xu Z, Niu L, Li L, Taylor JA. **ENmix: a novel background correction method for Illumina HumanMethylation450 BeadChip.**. *Nucleic Acids Res* (2016) **44** e20. DOI: 10.1093/nar/gkv907
29. Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, Beck S. **A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data.**. *Bioinformatics* (2013) **29** 189-96. DOI: 10.1093/bioinformatics/bts680
30. Pidsley R, Y Wong CC, Volta M, Lunnon K, Mill J, Schalkwyk LC. **A data-driven approach to preprocessing Illumina 450K methylation array data.**. *BMC Genomics* (2013) **14** 293. DOI: 10.1186/1471-2164-14-293
31. Teschendorff AE, Zhuang J, Widschwendter M. **Independent surrogate variable analysis to deconvolve confounding factors in large-scale microarray profiling studies.**. *Bioinformatics* (2011) **27** 1496-505. DOI: 10.1093/bioinformatics/btr171
32. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. **The sva package for removing batch effects and other unwanted variation in high-throughput experiments.**. *Bioinformatics* (2012) **28** 882-3. DOI: 10.1093/bioinformatics/bts034
33. Leek JT, Storey JD. **Capturing heterogeneity in gene expression studies by surrogate variable analysis.**. *PLoS Genet* (2007) **3** 1724-35. DOI: 10.1371/journal.pgen.0030161
34. Sugden K, Hannon EJ, Arseneault L, Belsky DW, Broadbent JM, Corcoran DL, Hancox RJ, Houts RM, Moffitt TE, Poulton R, Prinz JA, Thomson WM, Williams BS. **Establishing a generalized polyepigenetic biomarker for tobacco smoking.**. *Transl Psychiatry* (2019) **9** 92. DOI: 10.1038/s41398-019-0430-9
35. Hannum G, Guinney J, Zhao L, Zhang L, Hughes G, Sadda S, Klotzle B, Bibikova M, Fan JB, Gao Y, Deconde R, Chen M, Rajapakse I. **Genome-wide methylation profiles reveal quantitative views of human aging rates.**. *Mol Cell* (2013) **49** 359-67. DOI: 10.1016/j.molcel.2012.10.016
36. Horvath S. **DNA methylation age of human tissues and cell types.**. *Genome Biol* (2013) **14** R115. DOI: 10.1186/gb-2013-14-10-r115
37. Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, Nelson HH, Wiencke JK, Kelsey KT. **DNA methylation arrays as surrogate measures of cell mixture distribution.**. *BMC Bioinformatics* (2012) **13** 86. DOI: 10.1186/1471-2105-13-86
38. Jonkman TH, Dekkers KF, Slieker RC, Grant CD, Ikram MA, van Greevenbroek MMJ, Franke L, Veldink JH, Boomsma DI, Slagboom PE, Heijmans BT. **Functional genomics analysis identifies T and NK cell activation as a driver of epigenetic clock progression.**. *Genome Biol* (2022) **23** 24. DOI: 10.1186/s13059-021-02585-8
39. Santoro A, Bientinesi E, Monti D. **Immunosenescence and inflammaging in the aging process: age-related diseases or longevity?**. *Ageing Res Rev* (2021) **71** 101422. DOI: 10.1016/j.arr.2021.101422
|
---
title: miRNA-874-3p inhibits the migration, invasion and proliferation of breast cancer
cells by targeting VDAC1
authors:
- Housheng Yang
- Zhiwen Wang
- Shuang Hu
- Lu Chen
- Wei Li
- Zhongyi Yang
journal: Aging (Albany NY)
year: 2023
pmcid: PMC9970320
doi: 10.18632/aging.204474
license: CC BY 3.0
---
# miRNA-874-3p inhibits the migration, invasion and proliferation of breast cancer cells by targeting VDAC1
## Abstract
Breast cancer is an important cause of crisis for women’s life and health. Voltage-dependent anion channel 1 (VDAC1) is mainly localized in the outer mitochondrial membrane of all eukaryotes, and it plays a crucial role in the cell as the main interface between mitochondria and cellular metabolism. Through bioinformatics, we found that VDAC1 is abnormally highly expressed in breast cancer, and the prognosis of breast cancer patients with high VDAC1 expression is poor. Through in vivo and in vitro experiments, we found that VDAC1 can promote the proliferation, migration and invasion of breast cancer cells. Further research we found that VDAC1 can activate the wnt signaling pathway. Through analysis, we found that miR-874-3p can regulate the expression of VDAC1, and the expression of miR-874-3p is decreased in breast cancer, resulting in the increase of VDAC1 expression. Our findings will provide new targets and ideas for the prevention and treatment of breast cancer.
## INTRODUCTION
As the most prevalent malignant solid tumor in the world, breast cancer has a devastating impact on the lives of women [1, 2]. With the advancement of medical technology and the rising popularity of early breast cancer screening, the 5-year survival rate for early breast cancer patients after surgery, radiotherapy, and chemotherapy can reach 90 percent. However, the late breast tumor by direct infiltration or with the blood and lymph flow, transfer to other tissues or organs, and growth result in a poor prognosis for advanced breast cancer patients, with a 5-year survival rate of only about $26\%$, which has a significant impact on patients’ survival time and quality [3–5]. Moreover, due to the lack of early tumor-specific indicators or negligible initial symptoms of breast cancer, patients are diagnosed at stage III or IV and even miss the optimal treatment period, resulting in varied degrees of physical and mental burden [6, 7]. The molecular mechanism behind the onset and progression of breast cancer is still unclear. Consequently, research into the pathogenesis is critical for the development of effective therapies for breast cancer.
miRNA is an endogenous single-stranded non-coding RNA molecule composed of more than 20 nucleotides [8]. MiRNA mainly plays the role of gene regulation at the post-transcriptional level, and plays the role of tumor suppressor gene or oncogene by targeting mRNA degradation or inhibiting its translation process [9]. MiRNAs can affect the molecular expression of tumor cells through a variety of ways, thus affecting the ability of cell adhesion, changing the cytoskeleton structure, and participating in the regulation of cell-extracellular matrix interaction [10, 11]. Some scholars have demonstrated that miRNA can affect tumor bioactivity by regulating the signaling pathway of target genes. Cancer-related miRNAs are usually divided into two types. The first type of carcinogenic miRNA is often highly expressed in tumors, where it can promote the incidence and development of tumors and play a crucial role in tumor phenotypic maintenance. By controlling cell proliferation, apoptosis, and other processes, the second class of tumor suppressor microRNA decreases the occurrence, development, and medication resistance of cancers. In malignant tumors, miRNAs can therefore play either synergistic or antagonistic roles [12, 13]. It has been reported that miR-100, miR-122, miR-145, and miR-205 are down-regulated in breast cancer [14–17], while miR-1228, miR-150, miR-155, and miR-330-3p are up-regulated in breast cancer [18–21]. Recent study has shown that miR-874-3p expression is aberrant in liver cancer, osteosarcoma, colon cancer, and ovarian cancer [22–25], but our data shows that the expression of miR-874-3p is lowered in breast cancer.
Voltage-dependent anion channel 1 (VDAC1) is primarily positioned in the outer mitochondrial membrane of all eukaryotes and is the fundamental link between mitochondria and cellular metabolism [26]. VDAC1 has been demonstrated to be related with a number of disorders in the current investigation, and it is significantly expressed in a variety of tumors [27]. VDAC1 supports their metabolism via the transfer of diverse metabolites and mitochondrial ATP binding, resulting in mitochondrial regulation of glycolytic flow through the TCA cycle and the action of ATP synthase to meet tumor demand for metabolites or metabolite precursors [28]. By interacting with the anti-apoptotic proteins Bcl-XL, Bcl-2, and HK, VDAC1 regulates cancer cell apoptosis and shields tumor cells from cell death [29]. In addition, hyperglycemia increases VDAC1 expression in pancreatic -cells and kidneys, and VDAC1 levels are elevated in mouse coronary endothelial cells isolated from diabetic mice, because glucose-stimulated insulin secretion is dependent on the production of ATP and other metabolites in mitochondria, and VDAC1 regulates energy and metabolism [30]. Consequently, VDAC1 is necessary for insulin secretion.
We detected miR-874-3p and VDAC1 expression in breast cancer cells and tissues in our investigation. The targeted regulatory link between miR-874-3p and VDAC1 was confirmed, along with the associations between miR-874-3p expression, patient clinicopathology, and prognosis.
## VDAC1 expression is elevated in breast cancer
To determine the expression of VDAC1 in human breast cancer, western blot was utilized to identify the expression of VDAC1 in human breast cancer and normal breast tissues. Western blotting revealed that VDAC1 protein expression was considerably higher in breast cancer tissues than in surrounding tissues (Figure 1A, 1B). Simultaneously, we examined the expression of VDAC1 in human breast cancer cells and human mammary epithelial cells, and the findings revealed that, compared with normal mammary epithelial cells HMEC, the expression of VDAC1 was higher in human breast cancer cells BT549 and MCF-7 (Figure 1C). By studying the TCGA database, we discovered that the expression of VDAC1 is elevated in breast cancer patients, and that these patients had a bad prognosis (Figure 1D, 1E). All of these results indicate that VDAC1 may be associated with the incidence of breast cancer.
**Figure 1:** *VDAC1 is highly expressed in breast cancer. (A, B) Analysis of VDAC1 expression in 18 breast cancer tissues and adjacent paired normal breast tissue samples. Representative western blotting images of VDAC1 levels in three breast cancer tissues and three normal breast tissues (A). VDAC1 and GAPDH protein levels were determined using ImageJ (B). (C) RT-qPCR was used to detect the mRNA expression of VDAC1 in normal breast cells and breast cancer cells. (D) Expression of VDAC1 in normal tissues and breast cancer tissues from the TCGA database. (E) Kaplan-Meier survival analysis of breast patients with positive or negative VDAC1 expression. The data are presented as the mean ± S.D. *P < 0.05, **P < 0.01.*
## VDAC1 promotes the proliferation, migration and invasion of breast cancer cells
Next, we evaluated the impact of VDAC1 on breast cancer in vitro by transfecting BT549 and McF-7 cells with VDAC1 overexpression plasmid plvx-VDAC1 and control plasmid plvx-Con, respectively. Overexpression of VDAC1 significantly boosted the proliferative potential of breast cancer cells compared to the control group, as determined by CCK8 (Figure 2A). In addition, EdU measurements revealed that in the VDAC1 overexpression group, the fraction of cells in the replication phase was considerably elevated (Figure 2B). The findings of the Transwell experiment indicated further that VDAC1 overexpression enhances the capacity of breast cancer cells to move and invade (Figure 2C, 2D). All of these data demonstrated that VDAC1 can enhance breast cancer malignancy in vitro.
**Figure 2:** *VDAC1 promotes the proliferation, migration and invasion of breast cancer. The VDAC1 overexpression plasmid plvx-VDAC1 or blank plasmid plvx-Con was transfected into BT549 and MCF-7 cells. (A, B) Cell proliferation was detected with CCK8 assay (A) and EDU assay (B). (C, D) Transwell assay was used to detect cell migration and invasion ability. The data are presented as the mean ± S.D. *P < 0.05, **P < 0.01.*
## VDAC1 promotes the occurrence and development of breast cancer in vivo
To investigate the influence of VDAC1 on breast cancer in vivo, MCF-7 cells stably overexpressed with VDAC1 and control cells were subcutaneously injected into immunodeficient mice to provide a subcutaneous tumor-bearing model of human breast cancer. The observation was started on the seventh day. Every four days, tumor growth and size were recorded. After 28 days, mice were killed and tumors were removed. The tumors generated in the VDAC1 overexpression group (plvx-VDAC1) were considerably larger than those in the control group (plvx-Con), as measured by tumor volume (Figure 3A–3C). In vitro and in vivo studies demonstrate that VDAC1 promotes breast cancer.
**Figure 3:** *VDAC1 promotes breast cancer growth in vivo. Subcutaneous xenografts of MCF7 cells infected with VDAC1 overexpressing lentivirus or control lentivirus. (A) Images of the tumors at autopsy from nude mice are presented. (B, C) Tumor volumes (B) and average weight (C) of xenografted tumors were measured. Data represent the means ± S.D. **P < 0.01.*
## miR-874-3p directly binds to the VDAC1 3’-UTR
To explore the reasons for the abnormally high expression of VDAC1 in breast cancer, we predicted the miRNAs that bind to the 3’UTR region of VDAC1 mRNA through a bioinformatics website (Figure 4A). In accordance with our predictions, RT-qPCR detection revealed that only miR-874-3p was decreased in breast cancer (Figure 4B). Additionally, we discovered that miR-874-3p was downregulated in breast cancer tissues and was negatively linked with the expression of VDAC1 (Figure 4C, 4D). To establish the targeting interaction between miR-874-3p and VDAC1, luciferase reporter-tagged wild-type and mutant plasmids were generated (Figure 4E). The findings demonstrated that miR-874-3p mimics had no effect on the luciferase activity of mutant plasmids in breast cancer cells, but reduced the luciferase activity of wild-type plasmids (Figure 4F).
**Figure 4:** *WTX is a direct target of miR-874-3p. (A) The four-way Venn diagram reveals the numbers of overlapping miRNAs obtained using four publicly available bioinformatics algorithms and the microarray-based VDAC1 signature. (B) RT-qPCR was used to detect the relative expression of miR-874-3p in normal breast cells and breast cancer cells. (C) RT-qPCR was used to detect the expression in 18 breast cancer tissues and adjacent paired normal breast tissue samples. (D) Correlation between miR-874-3p levels and VDAC1 levels in 18 breast cancer tissues. (E) Nucleotide predicted miR-874-3p-binding site in the VDAC1 mRNA 3′-UTR. (F) MCF-7 and BT549 cells were transfected with reporter plasmids containing WT-pmir-VDAC1 or MUT-pmir-VDAC1 and miR-874-3p mimic or miR-874-3p mimic NC, and luciferase activity was detected. Data represent the means ± S.D. **P < 0.01.*
## miR-874-3p inhibits the malignancy of breast cancer cells
During this period, breast cancer cells were transfected with miR-874-3p mimics or controls without mimics. Compared with the control group, the proliferation ability of MCF-7 and BT549 cells transfected with miR-874-3p mimic was significantly inhibited (Figure 5A, 5B). The Transwell experiment showed that transfection of miR-874-3p mimics made BT549 and MCF-7 cells less likely to move and invade compared to the blank control (Figure 5C, 5D).
**Figure 5:** *miR-874-3p inhibits proliferation, migration and invasion of breast cancer cells. MCF-7 and BT549 cells were transduced with miR-874-3p mimic NC or miR-874-3p mimics. (A, B) Cell proliferation was detected with CCK8 assay (A) and EDU assay (B). (C, D) Transwell assay was used to detect cell migration and invasion ability. The data are presented as the mean ± S.D. *P < 0.05, **P < 0.01.*
## miR-874-3p-VDAC1 axis regulates Wnt/β-catenin signaling in breast cells
The Wnt signaling pathway is a classic signaling pathway in tumors. Abnormal activation of wnt signaling often leads to tumorigenesis. Will VDAC1 activate wnt signaling? In MCF-7 and BT549 cells, VDAC1 can increase the activity of the β-catenin reporter gene (Figure 6A), although miR-874-3p has the opposite effect (Figure 6B). In addition, we discovered that the expression of β-catenin and downstream Cyclin D1 was reduced, while the expression of p-β-catenin was elevated, following transfection with miR-874-3p mimics (Figure 6C). These findings suggest that VDAC1 enhances the onset and progression of breast cancer via stimulating the wnt signaling pathway.
**Figure 6:** *miR-874-3p-VDAC1 axis regulates Wnt/β-catenin signaling in breast cancer cells. (A) β-catenin reporter assay in MCF-7 and BT549 cells with VDAC1 overexpression (A) or miR-874-3p overexpression (B). (C) Effects of miR-874-3p on protein levels of total β-catenin, phosphorylated β-catenin (Ser33/37/Thr41) and cyclin D1. (D) Schematic of the mechanism. The data are presented as the mean ± S.D. *P < 0.05, **P < 0.01.*
## DISCUSSION
Breast cancer is a common malignant tumor, accounting for $20\%$ of female malignant tumors, the latest research reports show that breast cancer has become the highest incidence of cancer in the world, seriously affecting the survival of patients [31]. Our previous study discovered a link between VDAC1 and invasion and metastasis of breast cancer cell proliferation, but VDAC1 in the molecular regulation mechanism of breast cancer is unclear, so the VDAC1 in the incidence of breast cancer, development of molecular mechanism, and study the related regulation can help to find new breast cancer treatment targets, and promote the development of precision medical treatment of breast cancer. *This* gene has been discovered to be overexpressed in numerous malignancies, including lung cancer, cervical cancer, gastric cancer, pancreatic cancer, and laryngeal cancer [32–36]. In 2012, Brahimi-Horn et al., in their research on the drug resistance of lung cancer, found that cell lines with high expression of VDAC1 gene had higher resistance to apoptosis induced by staurosporine and etoposide, and silenced VDAC1 gene. Afterwards, the cell line regained its sensitivity to the above-mentioned drugs. In addition, high expression levels of VDAC1 gene can be detected in advanced lung cancer and tissues with larger lung cancer. It is concluded that VDAC1 may become a novel marker for early diagnosis and prognosis evaluation of lung cancer [37]. In 2010, Lan et al. found in their study on the mechanism of apoptosis in gastric cancer that the up-regulation of complexes and the down-regulation of proteins play a key role in the process of mediated apoptosis [34]. These results suggest that it is necessary to explore the role of VDAC1 in cancer.
Non-coding RNAs (ncRNAs) are RNA molecules that are produced by transcription but do not code for proteins; these molecules mostly consist of LncRNAs, CircRNAs, miRNAs, etc. At the post-transcriptional level, they play a significant role in gene expression regulation and epigenetic regulation [38]. This research investigated the function of microRNAs in the onset and progression of breast cancer. miRNAs are directly associated to the advancement of breast cancer, and they play varied roles at different phases of breast cancer development due to their involvement in numerous cancer processes. We believe miRNAs are involved in the regulation of malignant proliferation, escape from growth inhibition, cellular senescence, and genomic instability, despite the fact that their functions in numerous cancer processes have not been demonstrated. Several microRNAs have been discovered to be inappropriately expressed in breast tumors during the earliest stages of malignancy. VDAC1 is a miR-874-3p target gene, and miR-874-3p inhibits its expression in breast cancer cells. miR-874-3p has been shown to affect the expression of a variety of target genes, including GDPD5, FOXM1, FAM84A, and ADAM19 [22, 24, 39, 40]. However, no articles have reported the targeting link between miR-874-3p and VDAC1 and the role of miR-874-3p in breast cancer.
According to this study, miR-874-3p suppresses the proliferation, migration, and invasion of breast cancer cells as well as the expression of VDAC1, regulates the cell cycle, and promotes the death of breast cancer cells, as shown in Figure 6D. Therefore, miR-874-3p may be used as a biomarker for the early diagnosis of breast cancer. VDAC1 was validated as a target gene of miR-874-3p in the breast cancer molecular regulation mechanism, and its molecular regulation mechanism was investigated. We will improve the molecular regulatory mechanism of VDAC1 in the future. The findings imply that miR-874-3p may also serve as a novel therapeutic target for breast cancer, shedding fresh light on the investigation of miRNA expression profiles in breast cancer.
## Tissue specimens
The breast cancer tissue specimens used in this study were obtained from patients who underwent breast cancer resection in Yueyang People’s Hospital.
## Cell lines and cell culture
The BT549 and MCF-7 breast cancer cell lines were obtained from the Shanghai Cell Bank of the Chinese Academy of Sciences (Shanghai, China). MCF-7 was grown in DMEM medium (BI, Israel) with $10\%$ fetal bovine serum, whereas BT549 was grown in RPMI 1640 medium (BI, Israel) with $10\%$ fetal bovine serum. All cells were grown at 37° C in an incubator containing $5\%$ CO2.
To construct cell lines stably overexpressing VDAC1, we infected MCF-7 cells with lentivirus containing VDAC1 and used 500 ng/mL puromycin for resistance selection. Cells were harvested a week later for verification.
## qRT-PCR
Firstly, Trizol was added to the collected cells or tissues for mixing, then a quarter volume of chloroform was added, and then the supernatant was centrifuged, and the same volume of isopropanol was added for RNA deposition, all at 4° C. 1ugRNA was added to gDNA Wiper Mix solution to remove genomic DNA, and then HiScript qRT superMix solution was added for reverse transcription to obtain cDNA. Then take 2 μL of cDNA and add qPCR SuperMix solution for qRT-PCR. Primer sequences are as follows: VDAC1 forward, 5′-GGTGCTCTGGTGCTAGGTTA-3′ and reverse, 5′-CAGCGGTCTCCAACTTCTTG-3′ and GAPDH forward, 5′-ACCCACTCCTCCACCTTTGAC-3′ and reverse, 5′-TGTTGCTGTAGCCAAATTCGTT-3′.
## Plasmids, miRNAs, and transfection
The plvx-VDAC1 plasmid and miR-874-3p mimics were obtained from Origene (Beijing, China). miR-874-3p mimics sequence: 5′-ACUGCGUUGAAACAUGGGUA-3′, and the sequence of the miR-mimic NC was 5′-UUGAGGCUUCAAUCGACGUTT-3′. All transfections were carried out utilizing Lipo2000 per the directions. First, cells were seeded in 6-well plates, then transfected 24 hours later, followed by additional functional testing 24 hours following transfection.
## Transwell invasion and migration assays
36 hours after transfection, 5 × 104 cells were planted to the transwell chamber, the cells were incubated in the incubator for 36 hours, washed with PBS solution, cleaned with sterile cotton swabs, fixed with methanol, stained with crystal violet, and dried in a ventilation cabinet. Under an inverted microscope, photographs were taken and numbered. In preparation of the cell invasion assay, matrigel was spread in the Transwell chamber and dried overnight.
## CCK8 assay
Transfected cells were inoculated into 96-well plates and absorbance was measured at 0, 1, 3, 5 and 7 days, respectively. Before each measurement, 10 μL CCK8 (Sigma) solution was added to each well and incubated for 2 h. At a wavelength of 450 nm, the absorbance of each well was measured.
## EdU assay
The transfected cells were counted and then seeded on cell slides, EdU working solution was added at 24 hours and incubated for 3 hours, fixed with paraformaldehyde, and then EdU reaction solution was added, and photographed and analyzed under a fluorescence microscope.
## Western blotting
The cells or tissues were treated with RIPA lysate, properly mixed, centrifuged, and the supernatant was added to electrophoresis loading buffer. The separated proteins in the gel were transferred to PVDF membranes, coated with primary antibody overnight, incubated with secondary antibody the next day, and then exposed and developed using a gel imaging system. Primary antibodies used were anti-VDAC1 (A19707; Abclonal), anti-β-Catenin (A0316; Abclonal), phosphorylated β-catenin (AP1076; Abclonal), anti-Cyclin D1 (A0310; Abclonal), anti-GAPDH (A19056; Abclonal).
## Animal studies
In this study, 10 Balb/ C female nude mice (4 weeks old) were purchased from Beijing Spaifu Company and kept under SPF (temperature 25° C, humidity $55\%$). The Yueyang People’s Hospital Animal Ethics Committee authorized the trial. Mice were randomly assigned into two groups corresponding to the MCF-7-control group and MCF-7-overexpression group, and then the cells in different treatment groups were mixed with matrix glue at 1:1 and adjusted to a cell suspension of 5×107/ mL cells. Mice were subcutaneously sown with cells using a standard 1 ml syringe (100μl, 5×106 cells per mouse).
## Statistical analysis
This study repeated each experiment at least three times. GraphPad Prism 8.0 was utilized to process and assess the experimental data. The data are given as the standard deviation of the mean. Student’s t-test, chi-square test, and Kaplan *Meier analysis* were utilized in the statistical study. $P \leq 0.05$ signifies that the outcomes of the experiment are statistically significant.
## References
1. Liang Y, Zhang H, Song X, Yang Q. **Metastatic heterogeneity of breast cancer: Molecular mechanism and potential therapeutic targets.**. *Semin Cancer Biol* (2020) **60** 14-27. DOI: 10.1016/j.semcancer.2019.08.012
2. Chodosh LA. **Breast cancer: current state and future promise.**. *Breast Cancer Res* (2011) **13** 113. DOI: 10.1186/bcr3045
3. Fahad Ullah M. **Breast Cancer: Current Perspectives on the Disease Status.**. *Adv Exp Med Biol* (2019) **1152** 51-64. DOI: 10.1007/978-3-030-20301-6_4
4. Rossi L, Mazzara C, Pagani O. **Diagnosis and Treatment of Breast Cancer in Young Women.**. *Curr Treat Options Oncol* (2019) **20** 86. DOI: 10.1007/s11864-019-0685-7
5. McDonald ES, Clark AS, Tchou J, Zhang P, Freedman GM. **Clinical Diagnosis and Management of Breast Cancer.**. *J Nucl Med* (2016) **57** 9S-16S. DOI: 10.2967/jnumed.115.157834
6. Veronesi U, Boyle P, Goldhirsch A, Orecchia R, Viale G. **Breast cancer.**. *Lancet* (2005) **365** 1727-41. DOI: 10.1016/S0140-6736(05)66546-4
7. Low SK, Zembutsu H, Nakamura Y. **Breast cancer: The translation of big genomic data to cancer precision medicine.**. *Cancer Sci* (2018) **109** 497-506. DOI: 10.1111/cas.13463
8. Chen L, Heikkinen L, Wang C, Yang Y, Sun H, Wong G. **Trends in the development of miRNA bioinformatics tools.**. *Brief Bioinform* (2019) **20** 1836-52. DOI: 10.1093/bib/bby054
9. Lu TX, Rothenberg ME. **MicroRNA.**. *J Allergy Clin Immunol* (2018) **141** 1202-7. DOI: 10.1016/j.jaci.2017.08.034
10. Saliminejad K, Khorram Khorshid HR, Soleymani Fard S, Ghaffari SH. **An overview of microRNAs: Biology, functions, therapeutics, and analysis methods.**. *J Cell Physiol* (2019) **234** 5451-65. DOI: 10.1002/jcp.27486
11. Liu B, Li J, Cairns MJ. **Identifying miRNAs, targets and functions.**. *Brief Bioinform* (2014) **15** 1-19. DOI: 10.1093/bib/bbs075
12. Hill M, Tran N. **miRNA interplay: mechanisms and consequences in cancer.**. *Dis Model Mech* (2021) **14** dmm047662. DOI: 10.1242/dmm.047662
13. He B, Zhao Z, Cai Q, Zhang Y, Zhang P, Shi S, Xie H, Peng X, Yin W, Tao Y, Wang X. **miRNA-based biomarkers, therapies, and resistance in Cancer.**. *Int J Biol Sci* (2020) **16** 2628-47. DOI: 10.7150/ijbs.47203
14. Xie H, Xiao R, He Y, He L, Xie C, Chen J, Hong Y. **MicroRNA-100 inhibits breast cancer cell proliferation, invasion and migration by targeting FOXA1.**. *Oncol Lett* (2021) **22** 816. DOI: 10.3892/ol.2021.13077
15. Wang B, Wang H, Yang Z. **MiR-122 inhibits cell proliferation and tumorigenesis of breast cancer by targeting IGF1R.**. *PLoS One* (2012) **7** e47053. DOI: 10.1371/journal.pone.0047053
16. Zou C, Xu Q, Mao F, Li D, Bian C, Liu LZ, Jiang Y, Chen X, Qi Y, Zhang X, Wang X, Sun Q, Kung HF. **MiR-145 inhibits tumor angiogenesis and growth by N-RAS and VEGF.**. *Cell Cycle* (2012) **11** 2137-45. DOI: 10.4161/cc.20598
17. Zhang H, Fan Q. **MicroRNA-205 inhibits the proliferation and invasion of breast cancer by regulating AMOT expression.**. *Oncol Rep* (2015) **34** 2163-70. DOI: 10.3892/or.2015.4148
18. Lin L, Liu D, Liang H, Xue L, Su C, Liu M. **MiR-1228 promotes breast cancer cell growth and metastasis through targeting SCAI protein.**. *Int J Clin Exp Pathol* (2015) **8** 6646-55. PMID: 26261546
19. Huang S, Chen Y, Wu W, Ouyang N, Chen J, Li H, Liu X, Su F, Lin L, Yao Y. **miR-150 promotes human breast cancer growth and malignant behavior by targeting the pro-apoptotic purinergic P2X7 receptor.**. *PLoS One* (2013) **8** e80707. DOI: 10.1371/journal.pone.0080707
20. Shaker O, Maher M, Nassar Y, Morcos G, Gad Z. **Role of microRNAs -29b-2, -155, -197 and -205 as diagnostic biomarkers in serum of breast cancer females.**. *Gene* (2015) **560** 77-82. DOI: 10.1016/j.gene.2015.01.062
21. Mesci A, Huang X, Taeb S, Jahangiri S, Kim Y, Fokas E, Bruce J, Leong HS, Liu SK. **Targeting of CCBE1 by miR-330-3p in human breast cancer promotes metastasis.**. *Br J Cancer* (2017) **116** 1350-7. DOI: 10.1038/bjc.2017.105
22. Feng X, Xue H, Guo S, Chen Y, Zhang X, Tang X. **MiR-874-3p suppresses cell proliferation and invasion by targeting ADAM19 in nasopharyngeal carcinoma.**. *Panminerva Med* (2021) **63** 238-9. DOI: 10.23736/S0031-0808.19.03682-6
23. Liu WG, Zhuo L, Lu Y, Wang L, Ji YX, Guo Q. **miR-874-3p inhibits cell migration through targeting RGS4 in osteosarcoma.**. *J Gene Med* (2020) **22** e3213. DOI: 10.1002/jgm.3213
24. Wang Y, Chen H, Wei X. **Circ_0007142 downregulates miR-874-3p-mediated GDPD5 on colorectal cancer cells.**. *Eur J Clin Invest* (2021) **51** e13541. DOI: 10.1111/eci.13541
25. Xia B, Lin M, Dong W, Chen H, Li B, Zhang X, Hou Y, Lou G. **Upregulation of miR-874-3p and miR-874-5p inhibits epithelial ovarian cancer malignancy via SIK2.**. *J Biochem Mol Toxicol* (2018) **32** e22168. DOI: 10.1002/jbt.22168
26. Shoshan-Barmatz V, Nahon-Crystal E, Shteinfer-Kuzmine A, Gupta R. **VDAC1, mitochondrial dysfunction, and Alzheimer’s disease.**. *Pharmacol Res* (2018) **131** 87-101. DOI: 10.1016/j.phrs.2018.03.010
27. Magrì A, Reina S, De Pinto V. **VDAC1 as Pharmacological Target in Cancer and Neurodegeneration: Focus on Its Role in Apoptosis.**. *Front Chem* (2018) **6** 108. DOI: 10.3389/fchem.2018.00108
28. Shoshan-Barmatz V, De S. **Mitochondrial VDAC, the Na+/Ca2+ Exchanger, and the Ca2+ Uniporter in Ca2+ Dynamics and Signaling.**. *Adv Exp Med Biol* (2017) **981** 323-47. DOI: 10.1007/978-3-319-55858-5_13
29. Shoshan-Barmatz V, Krelin Y, Chen Q. **VDAC1 as a Player in Mitochondria-Mediated Apoptosis and Target for Modulating Apoptosis.**. *Curr Med Chem* (2017) **24** 4435-46. DOI: 10.2174/0929867324666170616105200
30. Zhang E, Mohammed Al-Amily I, Mohammed S, Luan C, Asplund O, Ahmed M, Ye Y, Ben-Hail D, Soni A, Vishnu N, Bompada P, De Marinis Y, Groop L. **Preserving Insulin Secretion in Diabetes by Inhibiting VDAC1 Overexpression and Surface Translocation in β Cells.**. *Cell Metab* (2019) **29** 64-77.e6. DOI: 10.1016/j.cmet.2018.09.008
31. Tray N, Taff J, Adams S. **Therapeutic landscape of metaplastic breast cancer.**. *Cancer Treat Rev* (2019) **79** 101888. DOI: 10.1016/j.ctrv.2019.08.004
32. Zerbib E, Arif T, Shteinfer-Kuzmine A, Chalifa-Caspi V, Shoshan-Barmatz V. **VDAC1 Silencing in Cancer Cells Leads to Metabolic Reprogramming That Modulates Tumor Microenvironment.**. *Cancers (Basel)* (2021) **13** 2850. DOI: 10.3390/cancers13112850
33. Zhang C, Hua Y, Qiu H, Liu T, Long Q, Liao W, Qiu J, Wang N, Chen M, Shi D, Yan Y, Xie C, Deng W. **KMT2A regulates cervical cancer cell growth through targeting VDAC1.**. *Aging (Albany NY)* (2020) **12** 9604-20. DOI: 10.18632/aging.103229
34. Lan CH, Sheng JQ, Fang DC, Meng QZ, Fan LL, Huang ZR. **Involvement of VDAC1 and Bcl-2 family of proteins in VacA-induced cytochrome c release and apoptosis of gastric epithelial carcinoma cells.**. *J Dig Dis* (2010) **11** 43-9. DOI: 10.1111/j.1751-2980.2009.00412.x
35. Wang W, Zhang T, Zhao W, Xu L, Yang Y, Liao Q, Zhao Y. **A Single Talent Immunogenic Membrane Antigen and Novel Prognostic Predictor: voltage-dependent anion channel 1 (VDAC1) in Pancreatic Cancer.**. *Sci Rep* (2016) **6** 33648. DOI: 10.1038/srep33648
36. Feng X, Ching CB, Chen WN. **EBV up-regulates cytochrome c through VDAC1 regulations and decreases the release of cytoplasmic Ca2+ in the NPC cell line.**. *Cell Biol Int* (2012) **36** 733-8. DOI: 10.1042/CBI20110368
37. Brahimi-Horn MC, Ben-Hail D, Ilie M, Gounon P, Rouleau M, Hofman V, Doyen J, Mari B, Shoshan-Barmatz V, Hofman P, Pouysségur J, Mazure NM. **Expression of a truncated active form of VDAC1 in lung cancer associates with hypoxic cell survival and correlates with progression to chemotherapy resistance.**. *Cancer Res* (2012) **72** 2140-50. DOI: 10.1158/0008-5472.CAN-11-3940
38. Anastasiadou E, Jacob LS, Slack FJ. **Non-coding RNA networks in cancer.**. *Nat Rev Cancer* (2018) **18** 5-18. DOI: 10.1038/nrc.2017.99
39. Zhang Z, Yang W, Li N, Chen X, Ma F, Yang J, Zhang Y, Chai X, Zhang B, Hou X, Luo S, Hua Y. **LncRNA MCF2L-AS1 aggravates proliferation, invasion and glycolysis of colorectal cancer cells via the crosstalk with miR-874-3p/FOXM1 signaling axis.**. *Carcinogenesis* (2021) **42** 263-71. DOI: 10.1093/carcin/bgaa093
40. Ding Y, Wu L, Zhuang X, Cai J, Tong H, Si Y, Zhang H, Wang X, Shen M. **The direct miR-874-3p-target FAM84A promotes tumor development in papillary thyroid cancer.**. *Mol Oncol* (2021) **15** 1597-614. DOI: 10.1002/1878-0261.12941
|
---
title: 'Incidence of new-onset in-hospital and persistent diabetes in COVID-19 patients:
comparison with influenza'
authors:
- Justin Y. Lu
- Jack Wilson
- Wei Hou
- Roman Fleysher
- Betsy C. Herold
- Kevan C. Herold
- Tim Q. Duong
journal: eBioMedicine
year: 2023
pmcid: PMC9970376
doi: 10.1016/j.ebiom.2023.104487
license: CC BY 4.0
---
# Incidence of new-onset in-hospital and persistent diabetes in COVID-19 patients: comparison with influenza
## Body
Research in contextEvidence before this studyWe searched PubMed and medRxiv with the search terms “new-onset diabetes”, “post-COVID-19 sequelae”, “persistent diabetes”, “hyperglycemia”, “diabetes mellitus”, “SARS-CoV-2” and “influenza” for articles published between Dec 8, 2020 and Jul 7, 2022. Patients infected with SARS-CoV-2 are at higher risk of developing new persistent diabetes. Data on the new persistent diabetes associated with SARS-CoV-2 infection compared to a similar respiratory virus (influenza) and identification of risk factors for persistent diabetes in COVID-19 patients may draw clinical attention for the need for careful follow-up. Added value of this studyNew-onset type-2 diabetes at follow-up is seen in $16.7\%$ of patients who are hospitalized for COVID-19 and who did not have a prior history of pre-diabetes and diabetes. The rates of diabetes are higher in hospitalized compared to non-hospitalized patients with COVID-19. Older patients who are male and with underlying major comorbidities are more likely to have new-onset diabetes persist after hospitalization. After adjusting for demographic factors and severity of illness, the incidence of post-infectious diabetes is similar between COVID-19 and influenza patients. The severity of illness rather than the respiratory virus per se is likely responsible for persistent diabetes and the increased incidence of newly diagnosed persistent diabetes likely reflects the incidence of severe COVID-19 observed particularly during the first wave of the pandemic. Implications of all the available evidenceNew-onset diabetes at follow-up is seen in $16.7\%$ of patients who are hospitalized with COVID-19 and who did not have a prior history of pre-diabetes and diabetes. The incidence of new-onset diabetes during hospitalization was 3.96 times (adjusted odds ratio) higher in patients with COVID-19 compared to those with influenza but only 1.24 (adjusted odds ratio) times higher at follow-up. Our findings suggest that the portion of increased risk of diabetes associated with COVID-19 is mediated through disease severity, which plays a dominant role in the development of this post-acute infection sequela. Identification of risk factors for P-DM could enable the need for careful follow-up in COVID-19 patients.
## Summary
### Background
This study investigated the incidences and risk factors associated with new-onset persistent type-2 diabetes during COVID-19 hospitalization and at 3-months follow-up compared to influenza.
### Methods
This retrospective study consisted of 8216 hospitalized, 2998 non-hospitalized COVID-19 patients, and 2988 hospitalized influenza patients without history of pre-diabetes or diabetes in the Montefiore Health System in Bronx, New York. The primary outcomes were incidences of new-onset in-hospital type-2 diabetes mellitus (I-DM) and persistent diabetes mellitus (P-DM) at 3 months (average) follow-up. Predictive models used $80\%$/$20\%$ of data for training/testing with five-fold cross-validation.
### Findings
I-DM was diagnosed in $22.6\%$ of patients with COVID-19 compared to only $3.3\%$ of patients with influenza ($95\%$ CI of difference [0.18, 0.20]). COVID-19 patients with I-DM compared to those without I-DM were older, more likely male, more likely to be treated with steroids and had more comorbidities. P-DM was diagnosed in $16.7\%$ of hospitalized COVID-19 patients versus $12\%$ of hospitalized influenza patients ($95\%$ CI of difference [0.03,0.065]) but only $7.3\%$ of non-hospitalized COVID-19 patients ($95\%$ CI of difference [0.078,0.11]). The rates of P-DM significantly decreased from $23.9\%$ to $4.0\%$ over the studied period. Logistic regression identified similar risk factors predictive of P-DM for COVID-19 and influenza. The adjusted odds ratio (0.90 [$95\%$ CI 0.64,1.28]) for developing P-DM was not significantly different between the two viruses.
### Interpretation
The incidence of new-onset type-2 diabetes was higher in patients with COVID-19 than influenza. Increased risk of diabetes associated with COVID-19 is mediated through disease severity, which plays a dominant role in the development of this post-acute infection sequela.
### Funding
None.
## Evidence before this study
We searched PubMed and medRxiv with the search terms “new-onset diabetes”, “post-COVID-19 sequelae”, “persistent diabetes”, “hyperglycemia”, “diabetes mellitus”, “SARS-CoV-2” and “influenza” for articles published between Dec 8, 2020 and Jul 7, 2022. Patients infected with SARS-CoV-2 are at higher risk of developing new persistent diabetes. Data on the new persistent diabetes associated with SARS-CoV-2 infection compared to a similar respiratory virus (influenza) and identification of risk factors for persistent diabetes in COVID-19 patients may draw clinical attention for the need for careful follow-up.
## Added value of this study
New-onset type-2 diabetes at follow-up is seen in $16.7\%$ of patients who are hospitalized for COVID-19 and who did not have a prior history of pre-diabetes and diabetes. The rates of diabetes are higher in hospitalized compared to non-hospitalized patients with COVID-19. Older patients who are male and with underlying major comorbidities are more likely to have new-onset diabetes persist after hospitalization. After adjusting for demographic factors and severity of illness, the incidence of post-infectious diabetes is similar between COVID-19 and influenza patients. The severity of illness rather than the respiratory virus per se is likely responsible for persistent diabetes and the increased incidence of newly diagnosed persistent diabetes likely reflects the incidence of severe COVID-19 observed particularly during the first wave of the pandemic.
## Implications of all the available evidence
New-onset diabetes at follow-up is seen in $16.7\%$ of patients who are hospitalized with COVID-19 and who did not have a prior history of pre-diabetes and diabetes. The incidence of new-onset diabetes during hospitalization was 3.96 times (adjusted odds ratio) higher in patients with COVID-19 compared to those with influenza but only 1.24 (adjusted odds ratio) times higher at follow-up. Our findings suggest that the portion of increased risk of diabetes associated with COVID-19 is mediated through disease severity, which plays a dominant role in the development of this post-acute infection sequela. Identification of risk factors for P-DM could enable the need for careful follow-up in COVID-19 patients.
## Introduction
The clinical course of SARS-CoV-2 infection is well-documented to be more severe in patients with pre-existing diabetes.1, 2, 3, 4, 5, 6, 7 Metabolic decompensation occurs frequently in COVID-19 patients with diabetes and tightening metabolic control improves outcomes.8 Conversely, SARS-CoV-2 infection has also been proposed to trigger new-onset diabetes.3, 4, 5,9 Several reports have drawn attention to new-onset diabetes among patients hospitalized with COVID-19,10, 11, 12, 13 which some have speculated may be the direct result of viral infection of insulin producing β cells, although SARS-CoV-2 viral particles have not been identified in β cells.14 The more common observation of diabetes in older hospitalized patients with COVID-19 suggests that inflammatory responses to infection together with obesity leads to insulin resistance and metabolic decompensation. In addition, COVID-19 treatments (e.g., glucocorticoids), may unmask latent diabetes because of their metabolic effects.
It is unclear whether new-onset diabetes diagnosed during COVID-19 persists after resolution of the acute infection. The effects of inflammatory mediators on β-cell dysfunction or insulin resistance should resolve with clinical improvement. Alternatively, there may be metabolic memory and thus the effects may persist even after the acute infection resolution. It is also not known whether new-onset diabetes that persists or presents following recovery from the acute viral illness is a consequence of SARS-CoV-2 virus, patient clinical profile, and/or hospital course and whether the incidence of this post-infectious sequela differs from what occurs following other severe respiratory viral infections such as influenza. The course of new-onset diabetes in COVID-19 patients may suggest pathologic mechanisms that contribute to its appearance.
The goals of this study were to determine whether COVID-19 related new-onset type-2 diabetes occurred more frequently among patients with COVID-19 compared to influenza, whether it was transient or persistent among those diagnosed in-hospital, and whether it also presented as a post-COVID-19 sequela. Outcomes were adjusted with covariates (age, sex, and major comorbidities) using odds ratios as well as compared with propensity matched controls. We also analyzed the incidence of diabetes across the pandemic and during the peak of each COVID-19 wave, and between hospitalized and non-hospitalized COVID-19 patients. Predictive models were used to identify and compare risk factors associated with post-COVID-19 or post-influenza new-onset diabetes.
## Ethics
This study was approved by the Einstein-Montefiore Institutional Review Board (#2021-13658) with an exemption for informed consent.
## Data sources
Health data came from the Montefiore Health System with 15 hospitals and medical centers located in New York *Metropolitan area* in the Bronx and the lower Westchester County (∼10 miles diameter), which serves a large diverse patient population including many patients with lower social economic status. Electronic medical records were extracted automatically as described previously.15, 16, 17, 18, 19, 20 De-identified health data were obtained for research after standardization to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) version 6. OMOP CDM represents healthcare data from diverse sources, which are stored in standard vocabulary concepts,21 allowing for the systematic analysis of disparate observational databases, including data from the electronic medical record (EMR), administrative claims, and disease classifications systems (e.g., ICD-10, SNOWMED, LOINC, etc.). ATLAS, a web-based tool developed by the Observational Health Data Sciences and Informatics (OHDSI) community that enables navigation of patient-level, observational data in the CDM format, was used to search vocabulary concepts and facilitate cohort building. Data were subsequently exported and queried as SQLite database files using the DB Browser for SQLite (version 3.12.0). For the variables extracted, chart reviews of a subset (N > 100) of data were performed to verify data accuracy and completeness.
## Participants
From March 11, 2020 to Feb 20, 2022, there were 35,644 COVID-19 positive patients, identified by polymerase-chain-reaction (PCR) test. From Jan 2018 to Feb 20, 2022, there were 12,354 hospitalized patients who tested positive for influenza without a positive COVID-19 PCR test. Using pre-COVID-19 pandemic data, we excluded patients with type-2 diabetes or prediabetes ICD10 diagnosis codes, on diabetes medications regardless of diabetes or prediabetes diagnosis, with A1c of 5.7–$6.5\%$ (pre-DM) or ≥$6.5\%$ (DM) prior to admission, two fasting glucoses of 100–125 mg/dl (pre-DM), a random glucose of 140–199 mg/dl (pre-DM), two fasting glucose readings ≥126 (DM) or two random glucoses of ≥200 mg/dl prior to admission (DM).
## Variables
Demographic data included age, sex, race, and ethnicity were collected via EMR. Preexisting comorbidities included body mass index (BMI), congestive heart failure (CHF), chronic kidney disease (CKD), hypertension, chronic obstructive pulmonary disease (COPD) and asthma that were designated by ICD10 codes at admission or prior. Steroid treatment, hospitalization status, intensive-care-unit (ICU) admission, and mortality were also extracted. Admission vital signs and laboratory data collected from hospitalized patients included temperature, systolic blood pressure (SBP), oxygen saturation (SPO2), lactate dehydrogenase (LDH), brain natriuretic peptide (BNP), creatinine (Cr), C-reactive protein (CRP), ferritin (FERR), D-dimer (DDIM), troponin-T (TNT), alanine aminotransferase (ALT), white-blood-cell count (WBC), lymphocyte count (Lymph), and prothrombin time (PT).
Overall, $64\%$ of patients in this study returned to the health system ∼3 months after diagnosis (mean = 83 and 87 days for COVID-19 and influenza patients, respectively). Data was collected at admission and the follow-up visit.
## Analysis of COVID-19 waves/strains
Predominant SARS-CoV-2 waves were estimated based on New York State Department of *Health data* https://coronavirus.health.ny.gov/covid-19-variant-data (assessed Nov 22, 2022). Waves were defined by daily test positivity $5\%$ above baseline that lasted at least 10 days in Bronx, New York.22 *By this* definition, the first wave spanned from March 8, 2020, to May 25, 2020, the alpha wave from December 6, 2020 to April 5, 2021, the Delta wave from July 6, 2021 to December 14, 2021, and the Omicron wave from December 15, 2021 to January 24, 2022.
## Outcomes
The primary outcomes analyzed were the incidence of new-onset type-2 diabetes mellitus while in the hospital (I-DM), and, for patients who returned to the hospital system, new-onset persistent diabetes mellitus (P-DM) at ∼3-month follow-up using above-defined criteria, grouped by SARS-CoV-2 or influenza infection. Outcomes by months and by waves across the pandemic were also analyzed.
## Predictive model/sample size
Logistic regression was used to build the predictive model. Sample size was based on availability of subjects. Univariable analysis was performed using each variable separately (demographics, comorbidities, and lab values, except FERR, BNP and A1c which had data missing from >$15\%$ of patients). All data were used except those with missing data >$15\%$. Imputation was done for data missing <$15\%$. The top 8 laboratory variables out of 14 (all extracted laboratory variables in Table 1) were first identified based on P-values. These top 8 laboratory variables out of 14 were then combined with all demographics, comorbidities collected to the logistic model to predict P-DM. This approach was adopted to avoid overfitting. Model performance was evaluated using area under the receiver operating characteristic curve with five-fold cross validation.17,23Table 1COVID-19 patient characteristics of (A) patients during hospitalization (B) hospitalized patients at follow-up, (C) hospitalized patients with in-hospital new-onset diabetes at follow-up, and (D) non-hospitalized patients at follow-up. COVID-19(A) Hospitalized patients during hospitalization ($$n = 8216$$)(B) All hospitalized patients at follow-up ($$n = 4982$$)(C) Subset of hospitalized patients with in-hospital DM at follow-up ($$n = 1034$$)(D) Non-hospitalized patients at follow-up ($$n = 1942$$)New-onset I-DM ($$n = 1854$$, $22.6\%$)No I-DM ($$n = 6362$$, $77.4\%$)Post-COVID P-DM ($$n = 834$$, $16.7\%$)No P-DM ($$n = 4148$$, $83.3\%$)Post-COVID P-DM ($$n = 383$$, $37.0\%$)No P-DM ($$n = 651$$, $63.0\%$%)Post-COVID P-DM ($$n = 142$$, $7.3\%$)No P-DM ($$n = 1800$$ $92.7\%$)Age, yo, median (IQR)66 [55, 78]41 [30, 59]62 [50, 75]45 [32, 62]66 [55, 77]63 [53, 76]54.9 ± 18.340.9 ± 16.0Female, n (%)818 ($44.1\%$)3896 ($61.2\%$)405 ($48.6\%$)2519 ($60.7\%$)161 ($42.0\%$)311 ($47.8\%$)74 ($52.1\%$)1250 ($69.4\%$)White, not hispanic191 ($10.3\%$)610 ($9.6\%$)92 ($11.0\%$)373 ($9.0\%$)50 ($13.1\%$)70 ($10.8\%$)18 ($12.7\%$)245 ($13.6\%$)Black, not hispanic572 ($30.9\%$)1930 ($30.3\%$)280 ($33.6\%$)1266 ($30.5\%$)129 ($33.7\%$)227 ($34.9\%$)43 ($30.3\%$)475 ($26.4\%$)Hispanic791 ($42.7\%$)2957 ($46.5\%$)354 ($42.4\%$)1941 ($46.8\%$)171 ($44.6\%$)272 ($41.8\%$)48 ($33.8\%$)594 ($33.0\%$)Other300 ($16.2\%$)865 ($13.6\%$)108 ($12.9\%$)568 ($13.7\%$)33 ($8.6\%$)82 ($12.6\%$)33 ($23.2\%$)486 ($27.0\%$)BMI29.0 (25.1, 34.3)28.0 (23.9, 32.8)28.9 (24.8, 34.4)∗∗28.7 (24.6, 33.2)29.1 (25.1, 34.9)29.8 (26.1, 34.4)29.5 (26.6, 33.9)28.2 (24.6, 32.1)Comorbidities, n (%) CHF221 ($11.9\%$)∗∗∗384 ($6.0\%$)167 ($20.0\%$)∗∗∗279 ($6.7\%$)88 ($23.0\%$)∗∗∗72 ($11.1\%$)15 ($10.6\%$)∗∗∗49 ($2.7\%$) CKD106 ($5.7\%$)∗∗283 ($4.4\%$)81 ($9.7\%$)∗∗∗204 ($4.9\%$)34 ($8.9\%$)∗32 ($4.9\%$)10 ($7.0\%$)∗∗∗43 ($2.4\%$) Hypertension487 ($26.3\%$)∗∗∗1436 ($22.6\%$)352 ($42.2\%$)∗∗∗1068 ($25.7\%$)154 ($40.2\%$)∗∗196 ($30.1\%$)55 ($38.7\%$)∗∗∗295 ($16.4\%$) COPD/Asthma122 ($6.6\%$)∗∗∗1088 ($17.1\%$)109 ($13.1\%$)∗∗726 ($17.5\%$)31 ($8.1\%$)61 ($9.4\%$)19 ($13.4\%$)245 ($13.6\%$)Steroid usage462 ($24.9\%$)∗∗∗423 ($6.6\%$)125 ($15.0\%$)∗∗∗430 ($10.4\%$)87 ($22.7\%$)187 ($28.7\%$)ICU admission77 ($4.2\%$)270 ($4.2\%$)21 ($2.5\%$)∗∗∗222 ($5.4\%$)17 ($4.4\%$)43 ($6.6\%$)In-hospital death, n (%)313 ($16.9\%$)∗∗∗167 ($2.6\%$)nanananaLab at admission, median (IQR) CRP8.3 (3.6, 16.0)∗∗∗3.5 (0.9, 8.9)6.1 (2.2, 13.7)∗∗∗4.6 (1.2, 10.5)6.8 (2.6, 5.1)7.7 (3.08, 15.2) Ferritin647 [308, 1305]∗∗355 [131, 921]559 [235, 1302]∗441 [168, 1053]631 [292, 1292]615 [276, 1235] LDH370 [276, 530]∗∗∗283 [213, 393]330 [251, 477]∗∗304 [227, 429]351 [260, 508]360 [267, 516] BNP60 [15, 243]60 [14, 219]86 [27, 647]∗∗∗60 [13, 171]66 [22, 410]60 [14, 187] Cr1.06 (0.81, 1.50)∗∗∗0.83 (0.70, 1.06)1.00 (0.80, 1.36)∗∗∗0.84 (0.70, 1.11)1.06 (0.81, 1.52)1.02 (0.79, 1.37) D-dimer1.33 (0.74, 2.80)∗0.94 (0.50, 1.98)1.22 (0.66, 2.87)∗∗∗1.00 (0.55, 2.05)1.31 (0.72, 3.19)1.29 (0.75, 2.46) TNT0.01 (0.01, 0.03)0.01 (0.01, 0.01)0.01 (0.01, 0.02)0.01 (0.01, 0.01)0.01 (0.01, 0.03)0.01 (0.01, 0.02) ALT31 [20, 51]∗∗∗23 [15, 37]26 [17, 43]25 [16, 41]28 [18, 47]31 [20, 51] LYMPH1.1 (0.7, 1.4)∗∗∗1.2 (0.8, 1.8)1.1 (0.8, 1.5)1.2 (0.8, 1.7)1.1 (0.8, 1.5)1.1 (0.7, 1.4) WBC7.2 (5.4, 10.0)∗∗∗6.4 (4.8, 8.6)6.8 (5.0, 9.5)∗∗∗5.7 (5.4, 6.2)7.5 (5.6, 10.5)∗6.9 (5.2, 9.6) PT13.8 (13.1, 14.9)∗∗13.6 (13.0, 14.4)13.9 (13.3, 15.1)∗∗∗13.6 (13.1, 14.4)14.0 (13.3, 15.1)13.8 (13.1, 14.7)∗ SBP131 [116, 147]129 [117, 144]134 [120, 150]∗∗∗129 [116, 144]133 [119, 150]131 [116, 147]∗ SPO295 [91, 98]∗∗∗98 [96, 99]97 [94, 98]∗∗∗98 [96, 99]95 [92, 98]95 [92, 98] TEMP98.7 (98.1, 99.7)∗∗∗98.6 (98.2, 99.4)98.7 (98.1, 99.6)∗98.7 (98.2, 99.5)98.7 (98.1, 99.7)98.7 (98.2, 99.7)Laboratory test data were obtained at admission. SD: standard deviation. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$ between new diabetes and no diabetes within each group.
To further investigate potential causal effects of virus type on outcomes (P-DM), we performed inverse probability weighting-based mediation analysis using the “MEDIATION” package in R software. CRP, LDH and DDIM were evaluated as the mediators in three separate models. ORs with $95\%$ confidence intervals based on 1000 nonparametric Bootstrap simulations were calculated. Average direct effects of virus type and mediated proportions were calculated.
## Exploratory analysis
Interrupted time series analysis was performed to investigate whether COVID-19 vaccine rollout affected I-DM and P-DM incidence.24 *Sensitivity analysis* of model performance metrics was also performed for different levels of prevalence of P-DM.
## Propensity score matching
Given the age differences between some groups, the adjustment with age as a covariate might not be adequate. Thus, ORs were calculated using propensity score matching on age and sex. We used 5:1 nearest neighbor propensity score matching without replacement with a propensity score estimated using logistic regression of the virus type on age and sex (R package “matchit”). After matching, 306 flu patients were successfully matched to 1530 COVID patients. Standardized mean differences for age and sex were 0.11 and 0.02 respectively indicating adequate balance.
## Statistical analysis methods
Statistical analysis was performed using Python, R and SAS (Version 9.4, Cary, NC, USA). Group comparison for categorical variables used χ2 or Fisher's exact tests, and for continuous variables used Mann–Whitney U test. P values less than 0.05 were considered statistically significant. Odds ratio was calculated with age, sex, ICU status, and admission CRP, LDH and D-Dimer as surrogates for severity as covariates using logistic regression. Comparisons of variables across different waves used ANOVA. P values for laboratory values were not adjusted for multiple comparisons due to the exploratory nature of this study.
## Role of funder
Not applicable.
## Rates of I-DM and P-DM are higher following COVID-19 than influenza viral infection
Fig. 1 shows the patient selection flowchart. Among 19,472 patients hospitalized with COVID-19 who met the inclusion criteria, 8216 were identified as not having a history of pre-diabetes or diabetes using the predefined criteria. Among these, 1854 ($22.6\%$) were diagnosed with type-2 diabetes during their hospitalization (I-DM). 1034 of those with a diagnosis of I-DM returned for follow-up and 383 ($37.0\%$) still had diabetes (P-DM). Notably, among the hospitalized patients with COVID-19 who did not have diabetes while in the hospital, 3948 returned and an additional 451 ($11.4\%$) were diagnosed with P-DM. Combining patients with and without I-DM, P-DM at follow-up was documented in $16.7\%$ ($$n = 834$$). There were 2998 non-hospitalized COVID-19 patients without history of prediabetes/diabetes and 1942 returned; only 142 ($7.3\%$) had P-DM at follow-up, which was significantly less than the rate among hospitalized patients with COVID-19 ($7.3\%$ vs 16.7, $95\%$ CI of difference [0.078, 0.11], $p \leq 0.001$).Fig. 1Patient selection flowchart. PMH: past medical history, DM: diabetes. Hospitalized influenza patient demographics pre (2018 and 2019) and post (2020 to Feb 2021) pandemic were similar and were thus combined. Note that there were more hospitalized ($$n = 8216$$) than non-hospitalized COVID-19 patients ($$n = 2998$$) in our cohort because the exclusion/inclusion criteria used to confirm no history of preDM and DM and many non-hospitalized patients did not have these detailed data and thus excluded.
We identified 2988 hospitalized influenza patients who met the inclusion criteria. Only 100 ($3.3\%$) had I-DM. Thus, hospitalized patients with COVID-19 were much 6.8 times more likely to develop I-DM compared to the influenza counterparts ($22.6\%$ vs $3.3\%$, $p \leq 0.001$, with OR = 3.96, [$95\%$ CI:3.2, 4.96], $p \leq 0.001$, adjusted for age, sex, comorbidities). However, 52 of the 100 who had influenza returned for follow-up and diabetes persisted in 24 ($46.2\%$), which was higher than observed with COVID-19 ($37\%$ vs $46.2\%$, $$p \leq 0.0035$$). Among the hospitalized influenza patients who did not have I-DM, 2157 returned and 241 ($11.2\%$) had P-DM, which is comparable to the rate of P-DM among COVID-19 patients without I-DM ($11.4\%$). When those with or without I-DM are combined, hospitalized patients with COVID-19 were more likely to develop P-DM compared to the influenza counterparts (1.39 times from $16.7\%$ vs $12.0\%$, $p \leq 0.0001$, with adjusted OR = 1.24, [$95\%$ CI:1.07, 1.45], $p \leq 0.001$).
Overall, $64\%$ of patients returned to the health system ∼3 months after diagnosis (mean = 83 and 87 days for COVID-19 and influenza patients, respectively). There were no significant differences in demographics, race, ethnicity, and major comorbidities ($p \leq 0.05$) between patients who did or did not return for follow up visits.
## Features of COVID-19 patients with and without I-DM
Table 1A compares the COVID-19 patient characteristics with and without I-DM during hospitalization. Patients with I-DM were older, less likely to be female, had a higher prevalence of CHF, CKD, and hypertension, a lower prevalence of COPD/asthma (all $p \leq 0.001$), but the BMI was not significantly different ($$p \leq 0.063$$) vs those without I-DM. More I-DM patients were treated with steroids (OR = 3.97 [$95\%$ CI: 3.33–4.74, $p \leq 0.05$]) after adjustment for age, sex and ICU admission. Mortality rate was higher among I-DM patients ($p \leq 0.001$).
Patients with I-DM had higher CRP, ferritin, LDH, Cr, D-dimer ALT, WBC, and PT ($p \leq 0.05$) and lower SPO2 and lymphocyte counts ($p \leq 0.001$) on admission compared to those without I-DM, findings consistent with greater disease severity.
## Features of COVID-19 patients with P-DM
Table 1B shows data for the hospitalized COVID-19 patients who returned for follow-up. Patients with P-DM were older, less likely to be female, and had higher prevalence of CHF, CKD, hypertension, and a lower prevalence of COPD/asthma. The BMI was higher in those with P-DM compared to those without. More P-DM patients had been treated with steroids, but the (OR =1.19 [$95\%$ CI: 0.93–1.52], $p \leq 0.05$) was not significantly different after adjustment for age, sex and ICU admission. Patients who were subsequently diagnosed with P-DM also had significantly higher admission laboratory data reflecting inflammation and disease severity (CRP, ferritin, LDH, BNP, Cr, D-dimer, WBC, PT, and SBP) and lower SPO2 compared to those without P-DM (all $p \leq 0.05$).
Table 1C compares the COVID-19 patients who returned for follow-up limited to the subgroup who were diagnosed with I-DM to identify factors that might predict those whose diabetes was persistent. Notably, I-DM was transient in most patients ($63.0\%$). Those in whom diabetes persisted had higher prevalence of CHF, CKD, and hypertension. However, age, sex, steroid use, and most laboratory values (at admission) were not significantly different between those with transient versus persistent DM.
Similar to hospitalized patients, non-hospitalized COVID-19 patients with P-DM were older, less likely female, and had higher prevalence of CHF, CKD, and hypertension (all $p \leq 0.001$), but a similar prevalence of COPD/asthma and BMI compared to those without P-DM (Table 1D). Laboratory tests were generally not performed for non-hospitalized patients.
## Features of hospitalized influenza patients with and without I-DM and/or P-DM
Table 2A shows the influenza patient characteristics during hospitalization. Similar to the patients with COVID-19 and I-DM, those with influenza and I-DM were markedly older and had a higher prevalence of most major comorbidities compared to those without. More I-DM patients were treated with steroids ($p \leq 0.001$). ICU admission rates were low but not statistically different between groups. The I-DM cohort had a higher mortality rate. Laboratory data was not analyzed because the sample size of influenza patients with I-DM was small. Table 2Hospitalized influenza patient characteristics (A) during hospitalization and (B) at follow-up. Influenza(A) During hospitalization ($$n = 2988$$)(B) at follow-up ($$n = 2209$$)New-onset I-DM ($$n = 100$$, $3.3\%$)No I-DM ($$n = 2888$$, $96.7\%$)Post-influenza P-DM ($$n = 265$$, $12.0\%$)No P-DM ($$n = 1944$$, $88.0\%$)Age, yo, median (IQR)62.5 [51, 72]27 [12, 41]50 [30, 63]26 [11, 39]Female, n (%)48 ($48.0\%$)1178 ($40.8\%$)180 ($67.9\%$)1203 ($61.9\%$)White, not hispanic9 ($9.0\%$)149 ($5.2\%$)19 ($7.2\%$)100 ($5.1\%$)Black, not hispanic38 ($38.0\%$)770 ($26.7\%$)93 ($35.1\%$)518 ($26.6\%$)Hispanic37 ($37.0\%$)1488 ($51.5\%$)128 ($48.3\%$)966 ($49.7\%$)Other16 ($16.0\%$)481 ($16.7\%$)25 ($9.4\%$)360 ($18.5\%$)BMI29.4 (24.1, 33.6)∗∗∗25.6 (19.8, 30.9)29.6 (25.1, 34.7)∗∗∗25.0 (19.2, 30.4)Comorbidities, n (%) CHF17 ($17.0\%$)∗∗∗87 ($3.0\%$)34 ($12.8\%$)∗∗∗40 ($2.1\%$) CKD8 ($8.0\%$)∗∗82 ($2.8\%$)22 ($8.3\%$)∗∗∗62 ($3.2\%$) Hypertension31 ($31.0\%$)∗∗∗403 ($14.0\%$)118 ($44.5\%$)∗∗∗263 ($13.5\%$) COPD/Asthma34 ($34.0\%$)865 ($30.0\%$)114 ($43.0\%$)∗∗∗630 ($32.4\%$)Steroid32 ($32.0\%$)∗∗∗190 ($6.6\%$)42 ($15.8\%$)∗∗∗120 ($6.2\%$)ICU admission0 ($0\%$)3 ($0.1\%$)1 ($0.4\%$)2 ($0.1\%$)In-hospital death, n (%)6 ($6.0\%$)∗∗∗15 ($0.5\%$)nanaLab at admission, median (IQR) CRP0.65 (0.525, 7.37)∗∗∗1.55 (0.575, 5.7) Ferritin506 [113, 1043]∗∗∗140 [41, 630] LDH178 [174, 183]∗∗∗322 [239, 403] BNP196 (68.2, 525)95 [60, 368] Cr0.81 (0.7, 1.1)∗0.78 (0.6, 1) D-dimer0.535 (0.485, 0.575)∗∗∗0.78 (0.665, 0.885) TNT0.01 (0.01, 0.01)0.01 (0.01, 0.01) ALT17 [13, 28]∗∗20 [13, 28] LYMPH0.9 (0.6, 1.5)0.9 (0.6, 1.5) PT13.8 (13.2, 14.6)13.9 (13.4, 14.5) SBP132 [115, 146]∗∗∗117 [108, 129] SPO297.5 [96, 99]∗∗∗98 [97, 100] TEMP99 (98.3, 101)99.2 (98.4, 101) WBC6.8 (4.8, 9.0)6.5 (5.0, 8.5)*Positive influenza* test (but no positive COVID-19 PCR tests) from Jan 1, 2018 to Feb 20, 2022. ∗$p \leq 0.05$, ∗∗$p \leq 0.01$, ∗∗∗$p \leq 0.001$ between new diabetes and no diabetes within each group. Laboratory test data were obtained at admission.
Table 2B shows the influenza patient characteristics at follow-up. Patients with P-DM were older and had higher prevalence of most major comorbidities. More P-DM patients were treated with steroids. Inflammatory markers including CRP, ferritin, LDH, D-Dimer were significantly higher at admission in the patients with influenza diagnosed with P-DM compared to those without P-DM.
The similarities of demographic and laboratory features with P-DM following COVID-19 (Table 1B) and influenza (Table 2B) suggested that differences in the rates might not be a feature of the virus itself but other clinical factors. Consistent with this notion we found that the OR of P-DM was not significantly different in hospitalized COVID-19 compared to influenza patients after adjusting for age and sex (OR = 1.14 [$95\%$ CI: 0.74–1.77], $$p \leq 0.61$$) without or with laboratory surrogates of disease severity (LDH, CRP and D-dimer) (OR = 0.90 [$95\%$ CI = 0.64,1.28], $$p \leq 0.56$$) (unmatched analysis). When analysis was performed using propensity score matching, the corresponding results were similar, namely, OR = 1.29 ($95\%$ CI: 0.93–1.76, $$p \leq 0.11$$) and 1.32 ($95\%$ CI = 0.94, 1.77, $$p \leq 0.13$$), respectively. P-DM ORs greater than unity, but with only trending p values toward significance suggest that small sample size and heterogeneity could contribute to non-significance.
## Model development/specification
Logistic regression was used to build the predictive model. The top 8 laboratory variables were first identified based on P-values. These top 8 laboratory variables were then combined with demographics, comorbidities to the logistic model to predict P-DM.
## Model performance
Predictive models were used to identify clinical variables associated with P-DM for COVID-19 or influenza patients who were hospitalized (Table 3). Using admission data, the significant top predictors were I-DM, age, CHF, steroid, ICU, and D-dimer for hospitalized COVID-19 patients, and age, I-DM, lymphocyte, BMI, and ALT for hospitalized influenza patients. Prediction of P-DM for the COVID-19 cohort yielded $71.42\%$ AUC ($95\%$ CI [0.70,0.84]), 78.28 accuracy ($95\%$ CI [0.77,0.83]), 96.95 specificity ($95\%$ CI [0.96,0.99]), 14.05 sensitivity ($95\%$ CI [0.11,0.26]), and $51.66\%$ positive predictive value ($95\%$ CI [0.58,0.91]). Prediction of P-DM for the influenza cohort yielded 66.70 AUC ($95\%$ CI [0.57,0.89]), 74.44 accuracy ($95\%$ CI [0.61,0.81]), $92.26\%$ specificity ($95\%$ CI [0.83,0.97]), 22.14 sensitivity ($95\%$ CI [0.05,0.37]) and $48.0\%$ positive predictive value ($95\%$ CI [0.14,0.79]). AUC curves for the test datasets are shown in Supplemental Fig. S1. Sensitivity analysis of the predictive models for different levels of prevalence of P-DM from $15\%$ to $40\%$ (Supplemental Table S1) showed that performance metrics were similar across different prevalence of P-DM.Table 3Prediction model performance indices and top predictors of P-DM for (A) hospitalized COVID-19 patients and (B) hospitalized influenza patients.(A) Hosp. COVID-19 pts(B) Hosp. influenza ptsVariablesP ValuesOR [$95\%$ CI]VariablesP ValuesOR [$95\%$ CI] I-DM<0.000103.32 [2.62,4.21]Age0.000691.02 [1.01,1.04] Age<0.000101.02 [1.01,1.02]I-DM0.00289.15 [0.3,272.71] CHF<0.000102 [1.49,2.67]LYMPH0.00501.44 [1.12,1.87] Steroid0.0180.71 [0.53,0.94]BMI0.00571.07 [1.02,1.12] ICU0.0220.51 [0.28,0.88]ALT0.0170.98 [0.96,0.99] D-dimer0.0231.03 [1.00,1.06]CHF0.0542.13 [0.98,4.64] BMI0.0811.01 [1.00,1.02]COPD0.0761.7 [0.94,3.08] Hypertension0.0831.24 [0.97,1.58]SBP0.131.01 [1,1.02] PT0.0921.02 [1.00,1.04]SPO20.130.94 [0.87,1.02] Sex0.140.85 [0.68,1.06]ICU0.141.35 [0.72,2.5] CKD0.161.33 [0.88,1.99]WBC0.160.94 [0.85,1.02] TEMP0.311.04 [0.96,1.13]Steroid0.210.63 [0.29,1.29] LDH0.400.99 [0.99,1.01]CKD0.280.61 [0.24,1.48]Values in bold indicate $p \leq 0.05.$ Note that prediction was not done for hospitalized influenza with I-DM because of small sample size.
For the mediation analysis, average direct effects of virus type were not significant with $95\%$ confidence intervals, ranging from −0.03 to 0.06 for disease severity (CRP, LDH and DDIM as surrogates) as mediators (all $p \leq 0.45$), and all the mediated proportions were also not significant (all $p \leq 0.4$). Therefore, the mediation effects of CRP, LDH, and DDIM are not significant.
## Rates of diabetes rates across the COVID-19 pandemic
To further distinguish the importance of disease severity or the virus itself leading to P-DM, we analyzed data over time during the pandemic. The prevalent strains varied in their clinical severity: original strain > Alpha > Delta > Omicron (Fig. 2A). To investigate whether COVID-19 vaccine rollout affected I-DM and P-DM incidence, interrupted time series analysis was performed. The I-DM incidence was high and time invariant till roughly when vaccine became available (Fig. 2B) and incidence of I-DM was significantly different before and after COVID-19 vaccine rolled out ($p \leq 0.001.$ Incidence of P-DM decreased across the pandemic more steadily across waves (Fig. 2C), but incidence of P-DM was not significantly different before and after COVID-19 vaccine rolled out ($$p \leq 0.34$$).Fig. 2Incidences across the pandemic. ( A) Number of hospitalized COVID-19 patients, interrupted time series analysis for (B) I-DM and (C) P-DM incidence. Blue lines indicate the date of the first vaccine rollout (Dec 14, 2021). P values indicate statistical significance value between incidence rate of diabetes before and after COVID-19 vaccine rollout.
There were general decreases in the incidence of I-DM and P-DM across the 4 waves ($p \leq 0.001$, Table 4). Age, sex, and comorbidities were not significantly different across waves. There were some differences in race and steroid use across waves. In the I-DM cohort, ICU admission was highest in the Omicron wave ($p \leq 0.0001$). Importantly, mortality rate decreased across waves ($p \leq 0.0001$).Table 4Incidence of I-DM and P-DM and their patient profiles grouped by different waves. I-DMP-DMOriginalAlphaDeltaOmicronP valueOriginalAlphaDeltaOmicronP valueIncidence of new DM$\frac{538}{1863}$ ($28.9\%$)$\frac{657}{1903}$ ($34.5\%$)$\frac{161}{878}$ ($18.3\%$)$\frac{198}{2329}$ ($8.5\%$)<$\frac{0.0001356}{1474}$ ($24.2\%$)$\frac{260}{1413}$ ($18.4\%$)$\frac{49}{547}$ ($9.0\%$)$\frac{26}{689}$ ($3.8\%$)<0.0001Age, median (IQR) (yo)54 [39, 68]54 [36, 68]39 [28, 55]37 [27, 54]57 [41, 71]56 [41, 71]40 [29, 58]37 [26, 55]Female, n (%)213 ($39.6\%$)312 ($47.5\%$)73 ($45.3\%$)89 ($44.9\%$)167 ($46.9\%$)136 ($52.3\%$)27 ($55.1\%$)10 ($38.5\%$)White, not Hispanic39 ($7.2\%$)76 ($11.6\%$)19 ($11.8\%$)27 ($13.6\%$)35 ($9.8\%$)42 ($16.2\%$)11 ($22.4\%$)3 ($6.1\%$)Black, not Hispanic201 ($37.4\%$)166 ($25.3\%$)49 ($30.4\%$)70 ($35.4\%$)146 ($41.0\%$)70 ($26.9\%$)18 ($36.7\%$)13 ($26.5\%$)Hispanic209 ($38.8\%$)281 ($42.8\%$)71 ($44.1\%$)82 ($41.4\%$)145 ($40.7\%$)117 ($45.0\%$)16 ($32.7\%$)7 ($14.3\%$)Other89 ($16.5\%$)134 ($20.4\%$)22 ($13.7\%$)19 ($9.6\%$)30 ($8.4\%$)31 ($11.9\%$)4 ($8.2\%$)3 ($6.1\%$)BMI29.1 (25.2,33.7)28.7 (24.5,33.3)28.9 (24.2,33.3)28.1 (23.6,32.3)0.7428.9 (25.0, 33.6)28.5 (24.4, 33.1)28.8 (24.8, 33.5)27.5 (23.5, 32.3)0.22Comorbidities, n (%) CHF61 ($11.3\%$)84 ($12.8\%$)11 ($6.8\%$)25 ($12.6\%$)0.2072 ($20.2\%$)56 ($21.5\%$)4 ($12.2\%$)6 ($23.1\%$)0.18 CKD45 ($8.4\%$)30 ($4.6\%$)4 ($2.5\%$)13 ($6.6\%$)0.00935 ($9.8\%$)22 ($8.5\%$)4 ($8.2\%$)4 ($15.4\%$)0.68 Hypertension136 ($25.3\%$)186 ($28.3\%$)31 ($19.3\%$)45 ($22.7\%$)0.08155 ($43.5\%$)118 ($45.4\%$)16 ($12.2\%$)6 ($23.1\%$)0.07 COPD/Asthma24 ($4.5\%$)53 ($8.1\%$)10 ($6.2\%$)9 ($4.5\%$)0.0542 ($11.8\%$)48 ($18.5\%$)4 ($6.1\%$)3 ($11.5\%$)0.06Steroid116 ($21.6\%$)183 ($27.9\%$)32 ($19.9\%$)48 ($24.2\%$)0.0443 ($12.1\%$)50 ($19.2\%$)11 ($12.2\%$)6 ($23.1\%$)0.03ICU admission10 ($1.9\%$)24 ($3.7\%$)5 ($3.1\%$)20 ($10.1\%$)<0.00014 ($1.1\%$)5 ($1.9\%$)2 ($4.1\%$)2 ($7.7\%$)0.07In-hospital death, n (%)149 ($27.7\%$)90 ($13.7\%$)18 ($11.2\%$)24 ($12.1\%$)<0.0001nananananaP values were obtained by Mann–Whitney U test and ANOVA analysis.
## Discussion
This study investigated the incidence of new-onset diabetes associated with acute SARS-CoV-2 and afterwards, compared to influenza in the Montefiore Health System in the Bronx. Our study sample is diverse and includes many with lower social economic status. The major findings are: i) new-onset diabetes is common at follow-up among COVID-19 patients rendering it a major post-acute sequalae of COVID-19 (PASC); ii) COVID-19 patients are 3.96 times more likely to develop new-onset diabetes compared to influenza during hospitalization but only 1.24 times at follow-up; iii) P-DM is more common among older adults, males, hospitalized patients, patients with preexisting co-morbidities, and those with abnormal admission laboratory tests reflecting increased disease severity and an inflammatory response; iv) predictive models identify risk factors of P-DM in COVID-19 patients with 71.42 ± $2.98\%$ AUC, 78.28 ± $1.72\%$ accuracy; and v) I-DM significantly decreased after COVID-19 vaccine rollout. Although the incidence of new-onset DM across the pandemic could be affected by differences in vaccination rate, strains, COVID-19 testing rate, and population profile, among others, increased risk of diabetes associated with COVID-19 is mediated through disease severity, which plays a dominant role in the development of this post-acute infection sequela.
While COVID-19 causes more severe disease and is affecting older individuals, our findings suggest that the higher rates of P-DM following SARS-CoV-2 compared to influenza ($16.7\%$ vs $12.0\%$) are related to greater clinical severity of the acute infection rather than virus types. The rates of P-DM were similar or even greater after influenza among those with I-DM ($37\%$ vs $46.2\%$) and did not differ significantly from COVID-19 after adjusting for age, sex and markers of severity. Furthermore, the rates of P-DM decreased as severity decreased with each subsequent wave of COVID-19. The decline in severity with each wave may be multifactorial since not only did the virus change but new treatments and vaccines became available, and there may have been immunologic memory even in non-vaccinated patients from prior exposures. Other comorbidities that were more common in patients with COVID-19 such as obesity, pulmonary, cardiac, and renal disease, may have affected disease severity and contributed to the rates of P-DM.
The interrupted time series analysis suggests that the introduction of COVID-19 vaccines, which are associated with a decrease in COVID-19 disease severity, may have played a role in the reduced incidence of I-DM and P-DM although other factors including natural immunity from prior SARS-CoV-2 exposure and differences in viral variants could also contribute to the decreased incidence. Indeed, when grouping data by waves/variants, we observed decreases in the incidence of I-DM and P-DM across the four waves, but no differences in age, sex, and major comorbidities. This could reflect differences in strain virulence, although population profile, COVID-19 testing rates, improvement in available treatments, and as noted above natural immunity or introduction of vaccines in the community could have also contributed.
Incidence of P-DM differed from that of national averages under non-covid pandemic conditions because our cohort excluded patients with pre-DM or those with a prior diagnosis of DM. P-DM in our study may have been impacted by the association with COVID-19 hospitalization and other major comorbidities as well as the demographics of our cohort, which included a large proportion of Blacks and Hispanics, including those who were underserved, who might be at higher risk.
Glucocorticoids were used in patients with severe COVID-19 and those with diabetes and influenza most likely reflecting the high rates of underlying COPD in the patients with influenza. In both viral illnesses their use was strongly associated with I-DM, consistent with the effects of steroids on impairing insulin sensitivity and enhancing hepatic gluconeogenesis. In addition, inflammatory cytokines, which are frequently elevated in the serum of patients with COVID-19, can impair insulin sensitivity and insulin secretion.25 Persistent effects of the host response to COVID-19 on adipocytes including the production of adipokines or other inflammatory mediators may account for persistent of DM in some patients as well as the delayed development of P-DM following hospitalization. In addition, beta cell failure or direct damage to hepatocytes may have occurred because of the toxic effects of inflammatory cytokines.26 Measurements of these cytokines and adipokines were not available to directly test this hypothesis.
Our findings differ from a previous report that suggested a direct relationship between COVID-19 infection and persistent, post-acute diabetes.9 *This previous* study was limited to Veterans who were predominantly male, whereas we found sex to be a contributing factor in P-DM. Importantly, this previous study did not control for factors such as the severity of the viral illness that seems to be the most significant determinant of post-hospitalization diabetes. To date, most other studies reporting new-onset P-DM were of perspective, case or case serious studies, or cohort studies with no comparison with appropriate controls, making direct comparison challenging. A novelty of our study is that we quantitatively compared new-onset P-DM of SARS-CoV-2 infection with that of influenza, a similar respiratory virus, in the same catchment area. We used predictive model to identify risk factors for developing new-onset P-DM.
There have been several reports suggesting that the SARS-CoV-2 might directly cause diabetes, and some have suggested that the virus might destroy insulin producing β cells directly or indirectly by infecting adipose cells, which produce inflammatory adipokines and enhance insulin resistance.26 Clinical experience has been consistent with this notion as large doses of insulin are frequently needed to manage patients in the hospital with diabetes. Based on this concept, it would be expected that diabetes would remit when the acute respiratory illness has resolved, even among those in whom diabetes was discovered in the hospital. Indeed, although the rate of P-DM is higher after COVID-19 than influenza, so was the severity of disease, and most who developed I-DM did not have P-DM at follow-up. This suggests that β cells had not been destroyed but that transient mechanisms associated with inflammatory responses lead to I-DM can resolve in the majority after acute SARS-CoV-2 infection.
## Limitations
There are several limitations in our analysis. Patients who did not return to our health system could not be studied. While it is possible that returned patients were more likely to have more severe COVID-19, our patient data obtained via EMR included those who returned for any medical reasons, including but not limited to regular checkups. Some patients with undiagnosed diabetes or pre-diabetes which could result in some patients being misclassified.
The high percentage of hospitalization is because many COVID-19 patients who came to the emergency department with COVID-19 were hospitalized, especially in the early pandemic. The non-hospitalized patients (enrolled as they presented to our health system and clinics for any medical reason, including regular checkup) were lower than expected because patients without A1c and multiple glucose measures, which were required to confirm DM status, were excluded. This could result in exclusion of some healthier patients and our cohort might not be representative the general COVID-19 patients. We could not distinguish patients hospitalized for COVID-19 from patients who were hospitalized for other indications with incidental COVID-19 because reasons for hospital admission were entered as free text clinical notes and could not be extracted automatically, although incidental COVID-19 was likely low, especially early in the pandemic.
Influenza patients were used as controls as opposed to COVID-19 negative patients to avoid bias by patients who were likely admitted for other serious medical issues (such as trauma, stroke, among others). Influenza testing was not performed on all hospitalized COVID-19 patients, but patients who tested positive for both viruses were rare, consistent with the very low incidence of influenza during the COVID-19 pandemic in 2020 and 2021. This confound is unlikely to alter our overall conclusions.
Incidence of new-onset DM across the pandemic could be affected by vaccination rate, strain, COVID-19 testing rate, population profile, and disease severity, among others. Vaccine status was not reliably recorded if patients received vaccine outside our healthcare system. Vaccines were availability in multiple stages based on age, and multiple doses, types (some requiring one or two shots). Boosters were also administered in the population. Thus, vaccine status is difficult to analyze in detail with respect to outcomes. Although there are published rate of vaccine and waves/strains at the population level in New York City, they are not applicable to our Montefiore cohort which was further filtered by the inclusion and exclusion criteria. Moreover, published data on websites are often bundled together without granular details to allow quantitatively analysis with our data. In addition, COVID-19 testing rate and the profile of the patients across the pandemic could also affect I-DM and P-DM incidence. The effects of these confounds and bias on outcomes are complex, difficult to assess, and not readily discernable from one another.
While our results suggest that hyperinflammation may play a role in new diabetes, data on cytokines and other inflammatory mediators were sparse and not analyzed. We followed patients for ∼3 month after diagnosis on average but recognize that a longer follow-up study is needed. Although our sample sizes were large compared to current published literature, as with any retrospective study, there could be other unintended patient selection bias and unaccounted confounds.
## Conclusions
New-onset diabetes at follow-up (P-DM) was detected in $16.7\%$ of patients who were hospitalized with COVID-19 and who did not have a prior history of pre-diabetes and diabetes. The rates of diabetes were higher in hospitalized compared to non-hospitalized patients with COVID-19. Older patients who were male and with underlying major comorbidities were more likely to have diabetes persist after hospitalization. After adjusting for demographic factors and measures of the severity of illness, the rate of post-infectious diabetes is similar after COVID-19 and influenza. The severity of illness rather than the respiratory virus per se is likely responsible for persistent diabetes. The high incidence of newly diagnosed persistent diabetes during the first wave of the COVID-19 pandemic likely reflects the incidence of severe COVID-19 during this time. This is an exploratory study and prospective studies are needed to confirm these findings. The shear number of COVID-19 patients in the world suggests that new-onset diabetes could be a major public health issue for years to come. Identification of risk factors for P-DM may draw clinical attention for the need for careful follow-up.
## Contributors
J. Wilson – concept and design, collected/verified data, analyzed data, created tables and figures, drafted paper.
J. Lu – concept and design, collected/verified data, analyzed data, created tables and figures, drafted paper.
W. Hou, R. Fleysher - collected/verified data, analyzed data, validated data.
B. Herold, K. Herold – concept and design, edited paper.
T.Q. Duong – concept and design, supervised, edited paper.
## Data sharing statement
Data used is available from the corresponding author upon reasonable request.
## Declaration of interests
Authors declare no conflicts of interest.
## Supplementary data
Supplemental Table S1 Supplemental Fig. S1 IRB-covid-subprojects
## References
1. Drucker D.J.. **Diabetes, obesity, metabolism, and SARS-CoV-2 infection: the end of the beginning**. *Cell Metab* (2021) **33** 479-498. PMID: 33529600
2. Holman N., Knighton P., Kar P.. **Risk factors for COVID-19-related mortality in people with type 1 and type 2 diabetes in England: a population-based cohort study**. *Lancet Diabetes Endocrinol* (2020) **8** 823-833. PMID: 32798471
3. Shrestha D.B., Budhathoki P., Raut S.. **New-onset diabetes in COVID-19 and clinical outcomes: a systematic review and meta-analysis**. *World J Virol* (2021) **10** 275-287. PMID: 34631477
4. Sathish T., Anton M.C., Sivakumar T.. **New-onset diabetes in "long COVID"**. *J Diabetes* (2021) **13** 693-694. PMID: 33893710
5. Rubino F., Amiel S.A., Zimmet P.. **New-onset diabetes in Covid-19**. *N Engl J Med* (2020) **383** 789-790. PMID: 32530585
6. Zhao Z., Chen A., Hou W.. **Prediction model and risk scores of ICU admission and mortality in COVID-19**. *PLoS One* (2020) **15**
7. Li X., Ge P., Zhu J.. **Deep learning prediction of likelihood of ICU admission and mortality in COVID-19 patients using clinical variables**. *PeerJ* (2020) **8**
8. Barron E., Bakhai C., Kar P.. **Associations of type 1 and type 2 diabetes with COVID-19-related mortality in England: a whole-population study**. *Lancet Diabetes Endocrinol* (2020) **8** 813-822. PMID: 32798472
9. Xie Y., Al-Aly Z.. **Risks and burdens of incident diabetes in long COVID: a cohort study**. *Lancet Diabetes Endocrinol* (2022) **10** 311-321. PMID: 35325624
10. Birabaharan M., Kaelber D.C., Pettus J.H., Smith D.M.. **Risk of new-onset type 2 diabetes in 600 055 people after COVID-19: a cohort study**. *Diabetes Obes Metab* (2022) **24** 1176-1179. PMID: 35112782
11. Chandrashekhar Joshi S., Pozzilli P.. **COVID-19 induced Diabetes: a novel presentation**. *Diabetes Res Clin Pract* (2022) **191**
12. Guo Y., Bian J., Chen A.. **Incidence trends of new-onset diabetes in children and adolescents before and during the COVID-19 pandemic: findings from Florida**. *Diabetes* (2022) **71** 2702-2706. PMID: 36094294
13. Metwally A.A., Mehta P., Johnson B.S., Nagarjuna A., Snyder M.P.. **COVID-19-Induced new-onset diabetes: trends and technologies**. *Diabetes* (2021) **70** 2733-2744. PMID: 34686519
14. van der Heide V., Jangra S., Cohen P.. **Limited extent and consequences of pancreatic SARS-CoV-2 infection**. *Cell Rep* (2022) **38**
15. Hoogenboom W.S., Pham A., Anand H.. **Clinical characteristics of the first and second COVID-19 waves in the Bronx, New York: a retrospective cohort study**. *Lancet Reg Health Am* (2021) **3**
16. Hoogenboom W.S., Fleysher R., Soby S.. **Individuals with sickle cell disease and sickle cell trait demonstrate no increase in mortality or critical illness from COVID-19 - a fifteen hospital observational study in the Bronx, New York**. *Haematologica* (2021) **106** 3014-3016. PMID: 34348452
17. Lu J.Q., Lu J.Y., Wang W.. **Clinical predictors of acute cardiac injury and normalization of troponin after hospital discharge from COVID-19**. *eBioMedicine* (2022) **76**
18. Lu J.Y., Hou W., Duong T.Q.. **Longitudinal prediction of hospital-acquired acute kidney injury in COVID-19: a two-center study**. *Infection* (2022) **50** 109-119. PMID: 34176087
19. Iosifescu A.L., Hoogenboom W.S., Buczek A.J., Fleysher R., Duong T.Q.. **New-onset and persistent neurological and psychiatric sequelae of COVID-19 compared to influenza: a retrospective cohort study in a large New York City healthcare network**. *Int J Methods Psychiatr Res* (2022) **31**
20. Lu J.Y., Boparai M.S., Shi C.. **Long-term outcomes of COVID-19 survivors with hospital AKI: association with time to recovery from AKI**. *Nephrol Dial Transplant* (2023)
21. Hripcsak G., Duke J.D., Shah N.H.. **Observational Health Data Sciences and Informatics (OHDSI): opportunities for observational researchers**. *Stud Health Technol Inform* (2015) **216** 574-578. PMID: 26262116
22. Seaton C.L., Cohen A., Henninger E.M.. **Coronavirus disease 2019 (COVID-19) perinatal outcomes across the pandemic at an academic medical center in New York city**. *Obstet Gynecol* (2022) **141** 144-151. PMID: 36302251
23. Hou W., Zhao Z., Chen A., Li H., Duong T.Q.. **Machining learning predicts the need for escalated care and mortality in COVID-19 patients from clinical variables**. *Int J Med Sci* (2021) **18** 1739-1745. PMID: 33746590
24. Fretheim A., Zhang F., Ross-Degnan D.. **A reanalysis of cluster randomized trials showed interrupted time-series studies were valuable in health system evaluation**. *J Clin Epidemiol* (2015) **68** 324-333. PMID: 25499983
25. Emamalipour M., Seidi K., Jahanban-Esfahlan A., Jahanban-Esfahlan R.. **Implications of resistin in type 2 diabetes mellitus and coronary artery disease: impairing insulin function and inducing pro-inflammatory cytokines**. *J Cell Physiol* (2019) **234** 21758-21769. PMID: 31187484
26. Sathish T., Tapp R.J., Cooper M.E., Zimmet P.. **Potential metabolic and inflammatory pathways between COVID-19 and new-onset diabetes**. *Diabetes Metab* (2021) **47**
|
---
title: Early-life dietary exposures mediate persistent shifts in the gut microbiome
and visceral fat metabolism
authors:
- Tiffany M. Newman
- Kenysha Y. J. Clear
- Adam S. Wilson
- David R. Soto-Pantoja
- Heather M. Ochs-Balcom
- Katherine L. Cook
journal: American Journal of Physiology - Cell Physiology
year: 2022
pmcid: PMC9970661
doi: 10.1152/ajpcell.00380.2021
license: CC BY 4.0
---
# Early-life dietary exposures mediate persistent shifts in the gut microbiome and visceral fat metabolism
## Abstract
In utero dietary exposures are linked to the development of metabolic syndrome in adult offspring. These dietary exposures can potentially impact gut microbial composition and offspring metabolic health. Female BALB/c mice were administered a lard, lard + flaxseed oil, high sugar, or control diet 4 wk before mating, throughout mating, pregnancy, and lactation. Female offspring were offered low-fat control diet at weaning. Fecal 16S sequencing was performed. Untargeted metabolomics was performed on visceral adipose tissue (VAT) of adult female offspring. Immunohistochemistry was used to determine adipocyte size, VAT collagen deposition, and macrophage content. Hippurate was administered via weekly intraperitoneal injections to low-fat and high-fat diet-fed female mice and VAT fibrosis and collagen 1A (COL1A) were assessed by immunohistochemistry. Lard diet exposure was associated with elevated body and VAT weight and dysregulated glucose metabolism. Lard + flaxseed oil attenuated these effects. Lard diet exposures were associated with increased adipocyte diameter and VAT macrophage count. Lard + flaxseed oil reduced adipocyte diameter and fibrosis compared with the lard diet. Hippurate-associated bacteria were influenced by lard versus lard + flax exposures that persisted to adulthood. VAT hippurate was increased in lard + flaxseed oil compared with lard diet. Hippurate supplementation mitigated VAT fibrosis pathology. Maternal high-fat lard diet consumption resulted in long-term metabolic and gut microbiome programming in offspring, impacting VAT inflammation and fibrosis, and was associated with reduced VAT hippurate content. These traits were not observed in maternal high-fat lard + flaxseed oil diet-exposed offspring. Hippurate supplementation reduced VAT fibrosis. These data suggest that detrimental effects of early-life high-fat lard diet exposure can be attenuated by dietary omega-3 polyunsaturated fatty acid supplementation.
## INTRODUCTION
The Developmental Origins of Health and Disease Theory states that the maternal intrauterine environment influences fetal development into adulthood, affecting metabolic disease risk [1]. A key component of the maternal intrauterine environment is the nutrition that is provided to the growing fetus through the placenta. Both maternal over- and undernutrition as well as a macronutrient imbalance can have adverse effects on the offspring.
Maternal obesity is linked to fetal overgrowth, a condition that increases the offspring’s likelihood of developing obesity later in life [2]. Maternal obesity correlates with hyperphagia, increased adiposity, insulin resistance, and hypertension in adult offspring [3]. In the absence of obesity, maternal high-fat diet has been linked to the development of the metabolic syndrome in offspring later in life, correlating with increased liver mass and triglyceride content, insulin resistance, increased visceral fat mass, hepatic steatosis, and adipocyte hypertrophy [3].
The detrimental effects of maternal diet are not limited to fat content. A gestational high-sugar diet containing simple sugars (dextrose and maltodextrin) altered the epigenetic modification of lipogenic and adipogenic genes in the white adipose tissue of adult rats, leading to an increased prevalence of obesity in males [4]. A gestational diet containing $50\%$ fructose produced hyperglycemia in male and female adult offspring [5]. The standard “Western Diet” is high in refined sugars and saturated fats. When a high-fat, high-sucrose diet is used as a model for maternal diets, a higher prevalence of type II diabetes, obesity, and metabolic syndrome was observed in the adult offspring of mice, rats, and humans (1, 6–11).
The gut microbiome composition varies in the presence of obesity [12]. Furthermore, the composition of the gut microbiota regulates the ability to harvest energy from food and the amount of energy stored as fat within the host [12]. During the first year of life, breastfeeding constitutes a major portion of the infant gut microbiome [13]. A review of 44 studies with a total of 2,655 human mothers identified the main bacterial genera in human breast milk as Staphylococcus, Lactobacillus, Streptococcus, Pseudomonas, Corynebacterium, Bifidobacterium, Enterococcus, Rothia, Acinobacter, Cutibacterium, Veillonella, and Bacteroides [14]. A study of 107 U.S. mothers demonstrated that the dominant phylum in breast milk was Proteobacteria (Enterobacteriaceaea and Pseudomonaceae), whereas the areolar skin primarily consisted of Firmicutes (Staphylococcaceae and Streptococcaceae) [15]. This study concluded that the breast milk microbial population contributed a mean of $27.7\%$ to the infant gut microbiome, whereas the areolar skin contributed $10.3\%$ [15]. Evidence suggests that the bacterial composition of breast milk may vary according to the dietary intake and BMI of the mother but currently available studies are limited and contradictory findings were reported [14]. The shift to a microbiome resembling an adult occurs with the cessation of breastfeeding rather than the introduction of alternative foods [13]. Breastmilk serves not only as a source of microbial diversity but also as a means of modulation of the gut microbiome by providing the infant with lactoferrin, lysozyme, milk glycans, and soluble immunoglobulins [16]. Certain families of microbes such as the Lachnospiraceae family are inflammatory and adipogenic [17]. Species of Lachnospiraceae and a proportional elevated abundance of the *Firmicute phyla* are elevated in overweight or obese children born to overweight or obese mothers [17]. Collectively, this evidence supports the hypothesis that the overweight or obese phenotype can be transmitted to offspring via modulation of gut microbiome diversity in utero and postnatally via breastfeeding.
To determine whether in utero and early-life (up to weaning) maternal diets have long-term impacts on offspring gut microbiome, adiposity, and metabolism, we placed dams on a low-fat, low-sugar control diet (control), a high sucrose, low-fat diet (HS), a high-fat lard-based diet [Lard obesity-inducing diet (OID)], and a high-fat lard + flaxseed oil diet (Flaxseed OID) beginning 1 mo before pregnancy. Dams were on the experimental diets throughout pregnancy, birth, and up to offspring weaning at postnatal day 21. Offspring were then placed on a low-fat, low-sugar control diet until adulthood (13 wk of age, 10 wk of control diet administration).
Our current study demonstrates that early-life exposure to a high-fat lard diet results in persistent microbiota shifts present even after 10 wk of control diet feeding that were not observed in offspring of lard + flaxseed oil OID-fed dams. Moreover, untargeted metabolomics of visceral fat from adult mice that were exposed in utero to a lard + flaxseed oil OID highlighted increased bacterially processed metabolites such as hippurate. Treatment of low-fat and high-fat diet-consuming mice with hippurate reduced visceral adipose tissue (VAT) profibrotic histopathology, suggesting a critical interaction between dietary-programmed microbiota-processed metabolites and extracellular matrix protein signaling. Taken together, we show that maternal high-fat lard diet consumption results in long-term gut microbiome programming in offspring that impacts metabolic function and inflammation that can be prevented by omega-3 polyunsaturated fatty acid supplementation.
## Materials
Hippuric acid was purchased from Sigma Aldrich (Cat. No. 112003). Antibodies were obtained for CD68 (Abcam; ab31630), phospho-SMAD2 (Cell Signaling Technologies; Cat. No. 3108), SMAD2 (Cell Signaling Technologies; Cat. No. 3103), transforming growth factor-β (TGF-β) (Cell Signaling Technologies; Cat. No. 3711), and collagen 1A (COL1A) (Cell Signaling Technologies; Cat. No. 91144). Harris Hematoxylin was purchased from Newcomer Supply (Cat. No. 1201). Xylene was purchased from Fisher Scientific (Cat. No. 05082-4). Eosin-Phloxine Working Solution was obtained from Newcomer Supply (Cat. No. 1082). Glacial acetic acid was purchased from Newcomer Supply (Cat. No. 10010 A). Sirius Red F3B (Cat. No. 36-554-8), a saturated aqueous solution of picric acid (Cat. No. P6744), and solid picric acid (Cat. No. 239801) were purchased from Sigma. Histoclear was obtained from National Diagnostics (Cat. No. HS-200). The DAB Staining kit (Ref: K4065), Protein Block reagent (Ref: X0909), and Antibody Diluent (Ref: 80809) were purchased from Dako. Cytoseal XYL Mounting Media was purchased from Thermo Scientific (Ref: 8312-4).
## Animals
Female 8-wk-old BALB/c mice were purchased from Jackson Laboratories. All Teklad custom diets were purchased from Envigo. Female mice were placed on a control diet ($10\%$ kcal from fat, TD.08806), a high sugar (TD.160065), lard obesity-inducing diet (Lard OID, $60\%$ kcal from fat, TD.06414), or a Lard + Flaxseed oil diet (Flaxseed OID, $60\%$ kcal from fat, TD.160066) ad libitum. At 12 wk of age, mice were paired with males (consuming a low-fat control diet) for mating. Pregnancy was confirmed 1-wk post pairing and males were removed from cages. Dams were maintained on their respective diets throughout pregnancy and lactation until weaning. Pups were counted and weighed on postnatal day 4 (PND4). Female pups were weaned at postnatal day 21 (PND21) and were administered a control diet.
At weaning, pup body weights were recorded and glucose tolerance tests were administered. Glucose tolerance testing consisted of overnight fasting followed by measurement of fasting blood glucose in blood collected via tail snip and measured using a OneTouch Ultra2 (LifeScan, Inc. with GenUltimate! Test Strips; Cat. No. 100-10). An intraperitoneal injection of 2 g/kg sucrose solution was administered, and blood glucose measurements were repeated at 15-, 30-, 60-, and 120-min postinjection for comparison via blood glucose curves. Body weight was recorded weekly. At postnatal day 91 (PND91) or 13 wk of age, repeat body weight measurements and glucose tolerance tests were performed. Following the procedure, mice were humanely euthanized, and their livers, VAT, and mammary glands were harvested, weighed, and preserved for future analysis. See Fig. 1A for the study schematic. The protocol was approved by the Animal Care and Use Committee of the Wake Forest School of Medicine and all procedures were carried out in accordance with relevant guidelines and regulations.
**Figure 1.:** *Maternal diet nutrient composition impacts metabolism of both mother and female offspring. A: BALB/c mouse model of in utero and early-life dietary exposures. B: body weight of adult female BALB/c mice before mating. C: percentage of weight gain of BALB/c mice during pregnancy based on measurements taken every 7 days. D: average number of pups born to BALB/c mice. E: body weight of female BALB/c offspring measured at 4 days of age. F: body weight of female BALB/c pups measured at postnatal day 21 (PND21) days of age. G: blood glucose concentration of 21-day-old female offspring following overnight fasting. H: blood glucose concentration of female offspring over 2 h following intraperitoneal injection of 2 mg/kg sucrose solution. I: comparison of glucose tolerance test curves by calculation of the area under each curve. *P < 0.05. OID, obesity-inducing diet.*
## Fecal 16S Sequencing
Fecal samples were collected from pregnant BALB/c dams on day 10 of pregnancy and from female offspring at weaning (PND21; 3 wk of age) and adulthood (PND91; 13 wk of age) in a sterile hood. At collection, samples were stored in sterile cryovials and immediately placed on dry ice. Samples were stored at −80°C until they were submitted for 16S sequencing. Mouse fecal 16S sequencing was performed by Microbiome Insights Inc. (Vancouver, BC, Canada). In brief, DNA was isolated from the feces using the MoBio Powersoil extraction kit. 16S rRNA genes were PCR- amplified with dual-barcoded primers targeting the V4 region, as previously described (18–20). Amplicons were sequenced with an Illumina MiSeq using the 250-bp paired-end kit (v. 2). Bacterial sequences were denoised, taxonomically classified using Greengenes (v. 13_8), and clustered into similar operational taxonomic units (OTUs) with the mother software package (v. 1.39.5) following the recommended protocol (https://www.mothur.org/wiki/MiSeq_SOP). The resulting data set had 10564 OTUs (including those occurring once with a count of 1, or singletons). An average of 22,733 quality-filtered reads were generated per sample. The potential for contamination was addressed by cosequencing DNA amplified from specimens and from each of four template-free controls and extraction kit reagents processed the same way as the specimens. Two positive controls, consisting of cloned SUP05 DNA, were also included (number of copies = 2 × 106). Operational taxonomic units were considered putative contaminants (and were removed) if their mean abundance in controls reached or exceeded $25\%$ of their mean abundance in specimens.
## Untargeted Metabolomics
Snap-frozen VAT from adult daughters was subjected to untargeted metabolomics (Metabolon, Raleigh, NC). In brief, a Waters ACQUITY ultra-performance liquid chromatography (UPLC) system and a Thermo Scientific Q-Exactive mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer were used for analysis. Compounds were identified by comparison with library entries of purified standards and peaks were quantified by measuring the area under the curve. The informatics system consisted of the Laboratory Information Management System (LIMS), the data extraction and peak-identification software, data processing tools for QC and compound identification, and a collection of information interpretation. Log transformation and imputation of missing values were performed with the minimum observed value for each compound. One-way ANOVA with Tukey’s multiple-comparisons post hoc analysis was used to identify biochemicals that differed significantly between visceral adipose tissue samples from adult female offspring with different maternal diet exposures.
## Immunohistochemistry
VAT was harvested following euthanasia of the BALB/c offspring. Half of the tissue was fixed in $4\%$ paraformaldehyde for 24 h and then moved to $70\%$ ethanol. Fixed tissue was paraffin-embedded and 5 µm slices were placed on slides (2 tissue segments per slide). This process was repeated for five animals per group for a total of 40 tissue segments.
## Hematoxylin and Eosin Staining
VAT segments were stained with hematoxylin and eosin to visualize adipocyte size. Five images of each slide were taken at ×20 magnification. ImageJ software was used to measure the diameter at the widest point on 20 representative adipocytes per image. Adipocyte diameters were compared using two-way ANOVA (main column effect) with a Holm–Šídák’s multiple comparisons (GraphPad Prism Software). The average adipocyte diameter of each group (daughters of mothers receiving control, high sugar, lard, and lard + flaxseed oil diets) was compared with every other group.
## Picrosirius Red Collagen Staining
Picrosirius red collagen staining was performed on VAT sections, as previously described [21]. Five images of VAT were taken per slide at ×10 magnification (5 images per mouse, 5 mice per group). InForm analysis software, v. 2.4.6781.17769 was used to process images. A threshold of optical density = 0.196 was designated as the minimum positive value and the percent of positive pixels per image was calculated. Two-way ANOVA (main column effect) with a Holm–Šídák’s multiple comparisons was used to examine differences in percent positive staining according to diet exposure group. GraphPad Prism software was used to visualize results.
## CD68 DAB Macrophage Staining
DAB staining was performed following the manufacturer’s instructions. An additional step, the Protein Block solution, was added and the slides were incubated for 5 min following the Endogenous Enzyme Block. Anti-mouse CD68 antibody was diluted in Dako Antibody Diluent (1:100) and applied to slides (24-h incubation at 4°C). Slides were exposed to DAB+ chromagen substrate solution for 9 min. Five images were taken per slide at ×20 magnification (1 slide per animal, 5 animals per group). Images were processed using InForm analysis software to generate composites. CD68+ cells were counted manually in each image. Adipocytes were counted and CD68 positive cells were normalized to total adipocytes in each image (CD68+ cells/adipocytes). Two-way ANOVA (main column effect) with a Holm–Šídák’s multiple-comparisons post hoc test was performed to compare CD68 positivity between diet exposure groups. Graphs were generated using GraphPad Prism software.
## In Vivo Hippurate Model
Low-fat or high-fat diet-fed female BALB/c mice at 12 wk of age were randomized into saline or hippurate treatment groups ($$n = 3$$ or 4 mice/group). At 14 wk of age, mice received a series of three intraperitoneal injections of either 100 µL saline (control) or 100 µL 10 mM hippurate at 0, 24, and 48 h. At 72 h, mice were humanely euthanized and VAT was fixed in $4\%$ paraformaldehyde for 24 h and then moved to $70\%$ ethanol. Fixed tissue was paraffin-embedded and 5 µm slices were placed on slides. Tissue sections were stained for fibrosis (Picrosirius red; see protocol above) or for collagen 1A (Cell Signaling, Cat. No. 91144, dilution: 1:100) using a DAB-chromogen staining protocol following the manufacturer’s instructions. Snap-frozen VAT was collected for protein analysis to measure the expression of transforming growth factor-β, phosphorylated SMAD$\frac{2}{3}$, SMAD $\frac{2}{3}$, and collagen 1A using Western blot analysis.
## Ex Vivo Hippurate Model
Control diet-fed female 12-wk-old BALB/c mice ($$n = 3$$) were humanely euthanized and VAT was harvested for ex vivo experimentation. Tissue was sectioned into 10 mg segments and 4–5 segments were placed in each well of a 6-well tissue culture plate containing 3 mL RPMI media per well. VAT from each animal was left untreated (control), treated with 10 µM hippurate, or treated with 100 µM hippurate. The plate was incubated overnight at 37°C with $5\%$ CO2. Protein was harvested the following morning in RIPA buffer with protease inhibitors. Western blotting was performed for quantification of pSmad2, TGF-β, and COL1A.
## Western Blot Hybridization
VAT from the in vivo and ex vivo hippurate models was sonicated in RIPA buffer to homogenize tissue. Protein was size fractionated by gel electrophoresis and transferred to a nitrocellulose membrane. Membranes were placed in blotto for 30 min to block, and then transferred to primary antibody solution for overnight incubation at 4°C (phospho-SMAD2, SMAD2, TGF-β, and COL1A from Cell Signaling Technology, dilution: 1:1,000). Membranes were washed and incubated with polyclonal horseradish peroxidase-conjugated secondary antibodies. Immunoreactive products were visualized by chemiluminescence with SuperSignal Femto and were quantified using the Bio-Rad digital densitometry software. Western blots are shown in figures as cropped images.
## Maternal Diet Nutrient Composition Impacts the Metabolic Health of BALB/c Dams and Female Offspring
Female BALB/c mice were randomized into diet groups (control, HS, Lard OID, or Flaxseed OID) at 8 wk of age. After 4 wk of diet administration, prepregnancy BALB/c body weights were recorded (Fig. 1B). Dams receiving control diet had a mean body weight of ∼20 g; both Lard OID (25 g) and Flaxseed OID (30 g) were higher by comparison ($P \leq 0.05$). Following mating, dam body weight was monitored at weekly intervals, and % weight gain during pregnancy was reported (Fig. 1C). Dams consuming either a Lard OID ($70\%$) or HS diet ($55\%$) had a greater percentage of weight gain throughout pregnancy than control diet mice ($45\%$, $P \leq 0.05$). Mice receiving Flaxseed OID did not significantly vary from the control group in pregnancy weight gain.
To reduce postpartum stress, dams were given four undisturbed days with pups before initial observations. At postnatal day 4 (PND4), pups were counted (Fig. 1D). Dams in all groups birthed an average of 6–8 pups; no significant difference in litter size was observed. Pup body weight was recorded at PND4 (Fig. 1E). Pups born to dams consuming either Lard OID or Flaxseed OID weighed an average of 0.3 g more than pups birthed by control-fed dams ($P \leq 0.05$).
At PND21, female offspring born to lard OID-fed mothers were observed to have an average body weight 2 g higher than that of daughters of control diet-fed mice ($P \leq 0.05$, Fig. 1F). PND21 fasting blood glucose was significantly elevated in daughters of Lard OID and Flaxseed OID-fed dams (60 mg/dL and 30 mg/dL higher than control, respectively, Fig. 1G). A bolus of sucrose solution was administered via intraperitoneal injection and blood glucose was monitored at 15-, 30-, 60-, and 120-min post-injection (Fig. 1H). Female offspring born to dams consuming lard OID had an average area under the curve of 10,000 mg/dL × min higher than female offspring of control-fed animals ($P \leq 0.05$, Fig. 1I).
## In Utero and Early-Life Macronutrient Exposures Initiate Long-Term Impact on the Metabolic Health of Female Offspring
After 10 wk of low-fat control diet exposures (at PND91), female offspring born to Lard OID and Flaxseed OID-fed dams had an average of 3 and 2 g higher body weight than offspring of control diet-fed mice, respectively ($P \leq 0.05$, Fig. 2A). At this time, the glucose tolerance testing procedure was repeated and blood glucose curves did not vary between groups (Fig. 2B). Mice born to Lard OID-fed mothers had a 1.25-fold elevated fasting blood glucose compared with both daughters of control-fed and Flaxseed OID-fed mice ($P \leq 0.05$, Fig. 2C). No significant differences in liver weight were observed (Fig. 2D). Adult offspring of dams fed Lard OID had a 1.5-fold higher VAT weight compared with daughters of control diet-fed mice ($P \leq 0.05$, Fig. 2E). Mammary glands from offspring of HS and Lard OID-fed dams weighed 1.3-fold and 1.5-fold higher, respectively, than those isolated from control-fed dams ($P \leq 0.05$, Fig. 2F). The mammary gland weight of daughters of Lard OID-fed dams also significantly differed from those of Flaxseed OID-fed dams, where Lard OID-fed dams had an average 1.5-fold higher mammary gland weight ($P \leq 0.05$).
**Figure 2.:** *In utero and early-life macronutrient exposures have long-term impacts on the metabolic health of female offspring. A: body weights of female BALB/c offspring at 91 days of age. B: blood glucose concentration of postnatal day 91 (PND91)-day old female offspring over 2 h following intraperitoneal injection of 2 mg/kg sucrose solution. C: comparison of glucose tolerance test curves by calculation of the area under each curve seen in B. D: wet weight of murine liver tissue at necropsy. E: VAT weight of female BALB/c mice at necropsy. F: wet weight of BALB/c L4/5 mammary glands at necropsy. *P < 0.05. OID, obesity-inducing diet.*
## In Utero and Early-Life Macronutrient Exposures Impact Long-Term VAT Morphology, Fibrosis, and Macrophage Populations in Female Offspring
Hematoxylin and eosin staining was performed on VAT, and adipocyte diameter was measured (Fig. 3A). Maternal consumption of HS, Lard OID, and Flaxseed OID was associated with 1.2-, 1.3-, and 1.1-fold adipocyte hypertrophy compared with the consumption of daughters of control diet-fed mice ($P \leq 0.05$). However, mice exposed to Lard OID had a 1.1-fold higher adipocyte diameter compared with those exposed to Flaxseed OID ($P \leq 0.05$). Sections of the VAT were stained with the pan-collagen stain, Picrosirius red, to compare adipose tissue fibrosis (Fig. 3B). Exposure to in utero and early-life HS diet was associated with a fourfold elevated VAT collagen deposition compared with daughters of control diet-fed mice ($P \leq 0.05$). In addition, the offspring of Lard OID-fed mothers exhibited 2.5-fold more VAT collagen content compared with offspring born to Flaxseed OID-fed mothers ($P \leq 0.05$). IHC staining for the pan-macrophage marker CD68 was performed on VAT sections (Fig. 3C). Maternal consumption of HS or lard OID diet but not a Flaxseed OID diet mediated an approximate threefold increase in the VAT macrophage population normalized to adipocyte number.
**Figure 3.:** *In utero and early-life macronutrient exposures impact long-term VAT morphology and macrophage populations in female offspring. A: hematoxylin and eosin staining of VAT harvested from adult female offspring following euthanasia. Adipocyte diameter measurements consist of 20 representative adipocytes in each of five images taken per mouse and n = 5 mice. B: Picrosirius red (PicRed) staining of collagen in VAT of adult female offspring. Quantification of fibrosis indicates the percentage of pixels that are positive for PicRed staining in five images per animal of n = 5 animals. C: hematoxylin-DAB staining for CD68 (a pan-macrophage marker) immunoreactivity. Macrophages were counted manually per image for a total of five images each of n = 5 or 6 animals. *P < 0.05; **P < 0.01; ****P < 0.0001 averages of groups were compared using two-way ANOVA (main column effect) with a Holm–Šídák’s multiple-comparisons post hoc test. Error bars indicate SEM and the line indicates the mean. H&E, hematoxylin and eosin; VAT, visceral adipose tissue.*
## Maternal Macronutrient Consumption Modulates Long-Term Gut Microbial Proportional Abundance of Hippurate-Associated Bacteria in Female Offspring
16S sequencing of fecal DNA samples from pregnant BALB/c mice consuming either control, HS, Lard OID, or Flaxseed OID diets at pregnancy day 10 and from female offspring at PND21 and PND91 was performed; principal coordinate analysis (PCoA) demonstrate bacterial community shifts over time and with maternal diet exposures (Supplemental Fig. S1, A–C), there were no significant differences in Shannon α-diversity between subjects and timepoint (data not shown), and the proportional abundance of identified bacterial phyla were reported (Supplemental Fig. S2A). At the phylum level, decreased Bacteroidetes and elevated Firmicutes proportional abundance were observed at weaning (PND21) in the daughters from Lard OID-fed dams. However, by adulthood, after a 10-wk administration of the control diet, the phylum level changes were no longer observed (Supplemental Fig. S2, B and C).
Although broad-spectrum shifts to the phyla were not maintained, we did observe significant shifts to several gut microbiota populations on the family and genus level that persisted to adulthood even after control diet administration (Fig. 4A). Female offspring from mothers consuming a lard OID displayed twofold elevated Lachnospiraceae_unclassified at PND21 that increased to nine fold at PND91 compared with controls ($P \leq 0.05$, Fig. 4B). Female offspring from Flaxseed OID-fed dams displayed fourfold increased Rikenellaceae (family) compared with offspring from Lard OID-fed dams at adulthood (PND91) that was not observed at PND21 ($P \leq 0.05$, Fig. 4C). Female offspring of Flaxseed OID-consuming mothers also displayed sevenfold elevated Clostridium (Fig. 4D) and Oscillospira (Fig. 4E) at weaning (PND21) that persisted to 10- and 3.5-fold increases at adulthood (PND91), respectively, compared with female offspring of Lard OID-fed mothers ($P \leq 0.05$).
**Figure 4.:** *Maternal macronutrient consumption modulates long-term gut microbial levels of hippurate-associated bacteria in female offspring. A: proportional abundance of bacterial genera quantified via 16S sequencing in pregnant dams, female offspring at weaning, and adult female offspring (n = 3 dams, n = 8 daughters). B: proportional abundance of Lachnospiraceae (f)_unclassified bacteria that are negatively associated with hippurate concentrations are elevated in female offspring from Lard OID-exposed mothers identified in fecal samples by 16S sequencing. Proportional abundance of fecal bacteria that are positively associated with hippurate concentration [C: Rikenellaceae (family); D: Clostridium; E: Oscillospira] are elevated in female offspring from Flaxseed OID-fed dams. *P < 0.05; averages of groups were compared using one-way ANOVA with an uncorrected Fisher’s LSD post hoc test. Error bars indicate SEM. OID, obesity-inducing diet.*
Previous studies using germ-free and antibiotic-treated mice demonstrated a microbial-mediated metabolite signature where bacteria presence was necessary for the formation of certain metabolites, including hippurate (22–26). Untargeted metabolomics was performed on VAT samples collected from adult female offspring (Fig. 5). In utero and early-life exposure to Flaxseed OID was associated with enriched levels of 4-hydroxycinnamate sulfate (35-fold), catechol sulfate (10-fold), hippurate (10-fold), and indolelactate (1.7-fold) compared with female offspring exposed to Lard OID ($P \leq 0.05$). With the exception of indolelactate, all metabolites were also significantly enriched in offspring of Flaxseed OID-fed mice compared with control diet-fed mice.
**Figure 5.:** *Maternal macronutrient consumption impacts VAT microbial-modified metabolites adult BALB/c female offspring. A: dietary phenolic and aromatic amino acids can be converted to form bioactive compounds through a combination of microbiota and host metabolic processes (simplified schematic created using BioRender.com). Scaled data collected via untargeted metabolomics detected parental amino acids [tryptophan (B), phenylalanine (C), and tyrosine (D), and benzoate (E)]. Microbial-associated metabolic products such as 4-hydroxycinnamate sulfate (F), catechol sulfate (G), hippurate (H), and indolelactate (I) are all elevated in VAT of adult daughters from Flaxseed OID-fed dams. *P < 0.05. n = 7 or 8. Averages of groups were compared using one-way ANOVA with a Tukey’s multiple-comparisons test. Error bars indicate SEM. OID, obesity-inducing diet; VAT, visceral adipose tissue.*
## Hippurate Supplementation Modulates Fibrosis in the VAT of BALB/c Mice
Female 12-wk-old BALB/c mice were fed a low-fat control diet ($$n = 6$$) or a high-fat diet ($$n = 8$$) for 2 wk before being randomized into groups receiving either 10 mM hippurate or a saline placebo ($$n = 3$$ or 4). Supplementation with exogenous hippurate was associated with an approximate 0.3-fold change in VAT fibrosis as determined by Picrosirius red pixel positivity in both low-fat and high-fat diet-consuming mice (Fig. 6, A and B). More specifically, VAT sections were stained against collagen 1A (COL1A). Supplementation with exogenous hippurate significantly reduced VAT COL1A deposition in high-fat diet consuming-mice (∼0.5-fold change), whereas there was only a modest trend for reduced COL1A in VAT of low-fat diet-fed subjects (Fig. 6, C and D). In vivo administration of hippurate in low-fat diet-fed animals decreased the mature TGF-β1 homodimer (25 kDa) (0.5-fold change) and COL1A (0.3-fold change) in VAT as determined by protein densitometry ($P \leq 0.05$, Supplemental Fig. S3, A–C).
**Figure 6.:** *In vivo assessment of microbial-mediated metabolite hippurate on visceral adipose tissue (VAT) pathology. Low-fat or high-fat diet-fed female BALB/c mice were given 3 × 1-wk intraperitoneal injections of 100 µL saline or 10 mM hippurate. Visceral adipose tissue was fixed and embedded in paraffin for physiological examination. VAT was stained using Picrosirius red protocol to visualize fibrosis. A: representative images of Picrosirius stained VAT from saline and hippurate-treated low-fat and high-fat diet-fed animals. B: quantification of Picrosirius red-stained VAT from saline and hippurate-treated low-fat and high-fat diet-fed animals. Quantification of fibrosis indicates the percentage of pixels that are positive for PicRed staining in five images per animal of n = 3 or 4 animals. C: representative images of collagen 1A (COL1A)-stained VAT from saline and hippurate-treated low-fat and high-fat diet-fed animals. D: quantification of collagen 1A-stained VAT from saline and hippurate-treated low-fat and high-fat diet-fed animals. Quantification of COL1A indicates the percentage of pixels that are DAB chromogen positive in five images per animal of n = 3 or 4 animals. *P < 0.05; **P < 0.01; ***P < 0.001; averages of groups were compared using two-way ANOVA (main column effect) with a Holm–Šídák’s multiple-comparisons post hoc test. Error bars indicate SEM.*
Ex vivo samples of VAT excised from low-fat diet-consuming mice were either untreated, treated with 10 µM hippurate, or treated with 100 µM hippurate (Supplemental Fig. S3D). High-dose exposure to hippurate (100 µM) was associated with reduced pSmad2 (0.3-fold change), TGF-β (0.3-fold change), and COL1A (0.45-fold change) compared with untreated tissue ($P \leq 0.05$, Supplemental Fig. S3, E–G). In addition, low-dose hippurate exposure (10 µM) modulated pSmad2 (0.3-fold) and COL1A (0.45-fold) ($P \leq 0.05$, Supplemental Fig. S3, E and G).
## DISCUSSION
Early-life programming in response to maternal diet exposure has been implicated as a risk factor for several diseases including neuroendocrine conditions, hypertension, and insulin resistance [27, 28]. In previous studies, lard-based maternal high-fat diet (HFD) influenced male rat offspring metabolic health through programming of liver fatty acid metabolism, whereas prenatal exposure to low-fat diet (LFD) protectively programmed the hepatic epigenome to reduce the effects of offspring HFD consumption [29, 30]. Maternal HFD’s effects on male offspring liver metabolism have also been observed in murine models where maternal diet influenced postweaning liver metabolite levels [31]. In addition, maternal flaxseed oil consumption during lactation enhanced bone development in male rat pups [32]. Although most of the previous studies investigated male offspring as a model of maternal diet, some studies identified sexually dimorphic responses to prenatal exposure, indicating the need for further research investigating the response to maternal diet in female offspring. For instance, although both male and female murine offspring developed an elevated body fat percentage in response to maternal HFD, only female offspring had reduced oxidative phosphorylation capabilities [33]. In addition, an investigation using a Wistar rat model reported that maternal OID consumption is associated with accelerated offspring metabolic aging in all offspring but is differentially regulated in males and females [34]. The maternal consumption of flaxseed had sexually dimorphic effects in rat offspring, including differences in glucose metabolism, pancreatic cellular morphology, and aortic remodeling (35–38). In a study using SM/J mice exposed to maternal HFD or LFD and then weaned onto either HFD or LFD, female HFD-fed offspring of HFD-fed mothers had an exacerbated response to their diets and quickly gained more weight and adipose tissue than other groups [39]. In our study, we provide a novel contribution to the field by examining the impact of in utero and early-life macronutrient exposure on gut microbial populations and their associated metabolites as a potential mechanism for maternal diet-associated adverse metabolic conditions in female offspring.
In the Healthy Start study, a poor diet during pregnancy was correlated with increased neonatal adiposity in humans regardless of maternal BMI before pregnancy [40]. We report that exposure to Lard and Flaxseed OID was associated with an elevated birth weight of mouse pups, echoing these results. Studies using C57BL/6 mice demonstrated that the offspring of obese murine dams are often heavier than those of nonobese dams, even up to 6 mo of age [27]. Our study uses a BALB/c model, which is known to be an obesity-resistant model. However, we were able to replicate the long-term impacts of in utero and early-life exposure to a Lard OID and Flaxseed OID on body weight in this model. Maternal diet-induced obesity in C57BL/6 mice correlated with insulin resistance of offspring in a model using a lard-based OID with supplementation of sweetened condensed milk [27]. An examination of monosodium glutamate (MSG)-induced obesity, primarily in male mice, demonstrated that prenatal flaxseed exposure through maternal diet blocked glucose intolerance, insulin resistance, pancreatic islet dysfunction, and elevated fat content associated with obesity [41]. Similarly, we report impaired glucose metabolism in female offspring of BALB/c mice exposed to maternal Lard OID. Our results indicate that alteration of the n-6:n-3 polyunsaturated fatty acid (PUFA) ratio by addition of flaxseed oil to a Lard OID (Flaxseed OID) is protective against this effect. This result corroborates an existing study in which female offspring of HFD and streptozocin-induced diabetic rat dams fed a high flaxseed oil diet had reduced fasting blood glucose compared with daughters of diabetic mothers consuming a control diet or HFD [38].
A study using a C57BL/6 mouse model reported increased adiposity in adult offspring associated with maternal consumption of a lard-based OID [27]. In addition, results in murine models reported inflamed adipose tissue in male offspring in response to maternal macronutrient consumption that was reduced by intervention with an LFD during pregnancy [42]. Our study expands this knowledge at the tissue level by characterizing the adipose tissue and reporting altered hypertrophy, fibrosis, and macrophage content in response to in utero and early-life macronutrient exposures. Of these effects, both fibrosis and hypertrophy correlate with obesity and increased the risk of metabolic disease, specifically diabetes, in humans with obesity [43, 44].
A study of male and female Sprague–Dawley rat pups of obese dams treated with a prebiotic supplement indicated that control of the maternal gut microbiota can influence metabolic factors in both mothers and offspring [45]. Prenatal exposure to the oligofructose prebiotic ameliorated glucose tolerance, insulin sensitivity, and hepatic steatosis in offspring of dams consuming a high fat/sucrose diet [46]. At weaning, offspring were offered a high-fat/sucrose diet, and gut microbial composition was examined. The study identified initial maternal diet-associated mediation of gut bacterial taxa (Akkermansia muciniphila, Bifidobacterium, Enterobacteriaceae, and Lactobacillus) that were not observed in fecal DNA at 11 wk of age, whereas maternal diet had a long-term impact on gut colonization of *Clostridium leptum* and Roseburia. This study supports our observation that maternal diet impacts gut microbiome shifts that can persist long term that are associated with metabolic outcomes. Weight gain in the offspring of murine dams fed a lard-based HFD was previously associated with gut microbial taxa (Lachnospiraceae, Clostridiaceae, Aldercreutzia, Coprococcus, and Lactococcus) and male offspring specifically had significantly enriched gut Firmicutes that are positively correlated with obesity risk [47]. Studies of direct flaxseed oil consumption in rats have indicated favorable changes in gut microbial populations including a reduced Firmicutes/Bacteroidetes ratio [48, 49]. However, maternal dietary consumption of flaxseed oil has not previously been linked to gut microbial populations of adult offspring. We further examined the influence of maternal diet on offspring gut microbial populations and indicate a novel effect of in utero and early-life exposure to Lard and Flaxseed OID by reporting their effects on the gut microbial contents of female offspring. Hippurate is a glycine conjugate derived from the metabolism of polyphenols to benzoate by the gut microbiome (50–53). We identified specific regulation of hippurate-associated bacterial taxa and report that Lard OID positively correlates with fecal Lachnospiraceae, which were previously negatively correlated with hippurate concentration and positively associated with weight gain in offspring of HFD-fed dams [47, 54]. We show that elevated fecal Lachnospiraceae in daughters of Lard OID-fed dams had decreased hippurate concentrations in the VAT supporting this previous association. In addition, we indicate that supplementation of flaxseed oil into a Lard OID (Flaxseed OID) is in turn positively associated with bacterial taxa (Rikenellaceae, Oscillospira, and Clostridium) that were seen to be positively correlated with VAT hippurate concentration. Of these taxa, Rickenellaceae has been positively correlated with hippurate concentration in human samples [54]. High levels of hippurate were reported to be favorable; hippurate is associated with high fruit and grain intake and is positively correlated with gut microbial diversity [54]. Elevated circulating hippurate is associated with the expression of neuroglobin by the adipose tissue, which has been shown to protect cells from hypoxia and oxidative stress in other tissues [55, 56], Low hippurate levels have also been reported to be unfavorable in inflammatory and microbial-associated conditions, as both Crohn’s disease and obesity are reported to be negatively correlated with hippurate concentration (57–60). We hypothesized that this effect may indicate that hippurate is the metabolite responsible for the protective effects of Flaxseed OID versus Lard OID that we observed and examined this using in vivo and ex vivo models.
In a murine model of HFD-fed mice receiving chronic subcutaneous hippurate infusion, hippurate was associated with improved glycemic control, improved insulin secretion, reduced liver inflammation, and reduced liver fibrosis [61]. We report that hippurate supplementation reduces the expression of TGF-β and COL1A in VAT. TGF-β is associated with immune cell activity through its functions in wound-healing and fibrosis by helping to initiate the chemotaxis of macrophages and fibroblasts to a body site [62, 63]. Reduction of TGF-β in response to hippurate is a potential mechanism by which Flaxseed OID may reduce VAT fibrosis and macrophage infiltration in our model. Similarly, the reduced collagen observed in the adipose tissue of hippurate-treated BALB/c mice is correlated with the reduced fibrosis observed in the VAT of mice receiving a Flaxseed OID compared with Lard OID. VAT fibrosis is common in chronic obesity, where fibrotic tissue can accumulate in response to chronic inflammation; it is characterized by reduced VAT plasticity and poor angiogenesis, leading to further dysregulation of insulin signaling [43]. Ex vivo analysis confirmed the results observed in our hippurate treatment model and revealed a reduction in pSmad2 mediated by hippurate exposure. Smad family proteins participate in TGF-β signal transduction pathways and are similarly correlated with inflammation and fibrotic responses when activated by phosphorylation (64–66). Direct consumption of flaxseed oil by rats was previously shown to have protective effects against pulmonary fibrosis, whereas maternal diet exposure to a flaxseed oil diet correlated with reduced kidney fibrosis [67, 68]. However, our study is the first to report an impact of maternal flaxseed oil exposure on VAT fibrosis.
In conclusion, our study indicates that in utero and early-life macronutrient exposure yields long-term effects on the metabolic and gut microbial health of female offspring. We also report that decreasing the n-6:n-3 PUFA ratio by supplementation of a lard-based high-fat diet with flaxseed oil (Flaxseed OID) is protective against some of the observed unfavorable effects. We propose microbial effects as a mechanism for this, specifically an increase in bacteria positively correlated with hippurate. Finally, we report that hippurate supplementation can reduce VAT fibrosis, suggesting exogenous administration of hippurate may be efficacious at reducing adverse health effects associated with obesity.
Our future studies will examine the mechanisms related to the effects observed with early-life exposure to a high sugar diet using our model (adipose fibrosis, adipocyte diameter, mammary gland weight). We will also further examine the mammary glands collected from these animals. A previous study in female Sprague–Dawley rats reported that a maternal western-style diet enhanced the effects of chemically induced mammary tumors through changes to the transcriptome [69]. In a C57BL/6J mouse model of 7,12-Dimethylbenz[a]anthracene-induced mammary tumors in daughters prenatally exposed to flaxseed, safflower, or fish oil diets, the daughters of mothers consuming n-3 PUFA (flaxseed oil and fish oil) had reduced breast cancer risk compared with those consuming safflower oil [70]. Our results also indicate that maternal macronutrient consumption may predispose female offspring to develop breast cancer risk factors. Elevated body weight and adipose tissue inflammation are risk factors for breast cancer [71]. In addition, both TGF-β and Smad2 are indicators of poor prognosis in human breast cancers (72–75). For these reasons, we propose examining the impact of maternal macronutrient consumption on female offspring breast cancer risk as our forthcoming focus for this project.
A potential limitation of our study is the use of only female offspring. Sex differences are observed in the development of obesity, and what we demonstrate in our female cohorts may be differentially regulated in male cohorts. Another important point to discuss is the distinction between gestational and lactational exposures. In the current study, we have grouped subjects by maternal diet as early-life exposures since this encompasses both in utero and lactational exposures. Breast milk has been shown to contain microbiota and microbial metabolites that can influence gut colonization of offspring. Therefore, determining gestational versus lactational differences is key for future studies using a dam cross-fostering approach with different dietary exposures. Finally, the associations between gut microbiome species and VAT hippurate concentrations are largely correlative. Future experimentation involving antibiotic administration and fecal microbiota transplants is needed to definitively demonstrate causality between the gut microbiome and hippurate in our preclinical models.
## DATA AVAILABILITY
All data generated or analyzed during the described study are included in this published article and supplemental materials.
## GRANTS
This work was supported by the American Cancer Society Research Scholar Grants (RSG-16-204-01-NEC to K.L.C. and 133727-RSG-19-150-01-LIB to D.R.S.-P.), a grant from the Susan G. Komen Foundation (CCR18547795 to K.L.C.), National Institutes of Health National Cancer Institute (NIH NCI) (R21CA249349 to D.R.S.-P.), American Society for Radiation Oncology-Breast Cancer Research Foundation (ASTRO-BCRF) Career Development Award (637969 to D.R.S.-P.), and Breakthrough Award from the Department of Defense Breast Cancer Research Program (W81XWH-20-1-0014 to K.L.C.). K.Y.J.C. is a recipient of a T32CA247819 training award. Shared Resource services were provided by the Wake Forest Baptist Comprehensive Cancer Center’s NCI Cancer Center Support Grant P30CA012197.
## DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
## AUTHOR CONTRIBUTIONS
K.L.C. conceived and designed research; T.M.N., K.Y.J.C., and A.S.W. performed experiments; T.M.N., K.Y.J.C., A.S.W., and K.L.C. analyzed data; A.S.W., D.R.S.-P., H.M.O.-B., and K.L.C. interpreted results of experiments; T.M.N., K.Y.J.C., and K.L.C. prepared figures; T.M.N. and D.R.S.-P. drafted manuscript; D.R.S.-P., H.M.O.-B., and K.L.C. edited and revised manuscript; T.M.N., K.Y.J.C., A.S.W., D.R.S.-P., H.M.O.-B., and K.L.C. approved final version of manuscript.
## References
1. Kereliuk S, Brawerman G, Dolinsky V. **Maternal macronutrient consumption and the developmental origins of metabolic disease in the offspring**. *Int J Mol Sci* (2017) **18**. DOI: 10.3390/ijms18071451
2. Masuyama H, Mitsui T, Nobumoto E, Hiramatsu Y. **The effects of high-fat diet exposure in utero on the obesogenic and diabetogenic traits through epigenetic changes in adiponectin and leptin gene expression for multiple generations in female mice**. *Endocrinology* (2015) **156** 2482-2491. DOI: 10.1210/en.2014-2020
3. Williams L, Seki Y, Vuguin PM, Charron MJ. **Animal models of in utero exposure to a high fat diet: a review**. *Biochim Biophys Acta* (2014) **1842** 507-519. DOI: 10.1016/j.bbadis.2013.07.006
4. Borengasser SJ, Zhong Y, Kang P, Lindsey F, Ronis MJJ, Badger TM, Gomez-Acevedo H, Shankar K. **Maternal obesity enhances white adipose tissue differentiation and alters genome-scale DNA methylation in male rat offspring**. *Endocrinology* (2013) **154** 4113-4125. DOI: 10.1210/en.2012-2255
5. Jen KL, Rochon C, Zhong SB, Whitcomb L. **Fructose and sucrose feeding during pregnancy and lactation in rats changes maternal and pup fuel metabolism**. *J Nutr* (1991) **121** 1999-2005. DOI: 10.1093/jn/121.12.1999
6. Dupont C, Cordier AG, Junien C, Mandon-Pépin B, Levy R, Chavatte-Palmer P. **Maternal environment and the reproductive function of the offspring**. *Theriogenology* (2012) **78** 1405-1414. DOI: 10.1016/j.theriogenology.2012.06.016
7. Pereira TJ, Fonseca MA, Campbell KE, Moyce BL, Cole LK, Hatch GM, Doucette CA, Klein J, Aliani M, Dolinsky VW. **Maternal obesity characterized by gestational diabetes increases the susceptibility of rat offspring to hepatic steatosis via a disrupted liver metabolome**. *J Physiol* (2015) **593** 3181-3197. DOI: 10.1113/JP270429
8. Clausen TD, Mathiesen ER, Hansen T, Pedersen O, Jensen DM, Lauenborg J, Schmidt L, Damm P. **Overweight and the metabolic syndrome in adult offspring of women with diet-treated gestational diabetes mellitus or type 1 diabetes**. *J Clin Endocrinol Metab* (2009) **94** 2464-2470. DOI: 10.1210/jc.2009-0305
9. Navarro E, Funtikova AN, Fíto M, Schröder H. **Prenatal nutrition and the risk of adult obesity: long-term effects of nutrition on epigenetic mechanisms regulating gene expression**. *J Nutr Biochem* (2017) **39** 1-14. DOI: 10.1016/j.jnutbio.2016.03.012
10. Barker DJ, Gluckman PD, Godfrey KM, Harding JE, Owens JA, Robinson JS. **Fetal nutrition and cardiovascular disease in adult life**. *Lancet* (1993) **341** 938-941. DOI: 10.1016/0140-6736(93)91224-a
11. Khan I, Dekou V, Hanson M, Poston L, Taylor P. **Predictive adaptive responses to maternal high-fat diet prevent endothelial dysfunction but not hypertension in adult rat offspring**. *Circulation* (2004) **110** 1097-1102. DOI: 10.1161/01.CIR.0000139843.05436.A0
12. Ley RE, Bäckhed F, Turnbaugh P, Lozupone CA, Knight RD, Gordon JI. **Obesity alters gut microbial ecology**. *Proc Natl Acad Sci USA* (2005) **102** 11070-11075. DOI: 10.1073/pnas.0504978102
13. Bäckhed F, Roswall J, Peng Y, Feng Q, Jia H, Kovatcheva-Datchary P, Li Y, Xia Y, Xie H, Zhong H, Khan MT, Zhang J, Li J, Xiao L, Al-Aama J, Zhang D, Lee YS, Kotowska D, Colding C, Tremaroli V, Yin Y, Bergman S, Xu X, Madsen L, Kristiansen K, Dahlgren J, Wang J, Jun W. **Dynamics and stabilization of the human gut microbiome during the first year of life**. *Cell Host Microbe* (2015) **17** 690-703. DOI: 10.1016/j.chom.2015.04.004
14. Zimmermann P, Curtis N. **Breast milk microbiota: a review of the factors that influence composition**. *J Infect* (2020) **81** 17-47. DOI: 10.1016/j.jinf.2020.01.023
15. Pannaraj PS, Li F, Cerini C, Bender JM, Yang S, Rollie A, Adisetiyo H, Zabih S, Lincez PJ, Bittinger K, Bailey A, Bushman FD, Sleasman JW, Aldrovandi GM. **Association between breast milk bacterial communities and establishment and development of the infant gut microbiome**. *JAMA Pediatr* (2017) **171** 647-654. DOI: 10.1001/jamapediatrics.2017.0378
16. Pacheco AR, Barile D, Underwood MA, Mills DA. **The impact of the milk glycobiome on the neonate gut microbiota**. *Annu Rev Anim Biosci* (2015) **3** 419-445. DOI: 10.1146/annurev-animal-022114-111112
17. Tun HM, Bridgman SL, Chari R, Field CJ, Guttman DS, Becker AB, Mandhane PJ, Turvey SE, Subbarao P, Sears MR, Scott JA, Kozyrskyj AL. **Roles of birth mode and infant gut microbiota in intergenerational transmission of overweight and obesity from mother to offspring**. *JAMA Pediatr* (2018) **172** 368-377. DOI: 10.1001/jamapediatrics.2017.5535
18. Shively CA, Register TC, Appt SE, Clarkson TB, Uberseder B, Clear KYJ, Wilson AS, Chiba A, Tooze JA, Cook KL. **Consumption of Mediterranean versus Western diet leads to distinct mammary gland microbiome populations**. *Cell Rep* (2018) **25** 47-56.e3. DOI: 10.1016/j.celrep.2018.08.078
19. Soto-Pantoja DR, Gaber M, Arnone AA, Bronson SM, Cruz-Diaz N, Wilson AS, Clear KYJ, Ramirez MU, Kucera GL, Levine EA, Lelièvre SA, Chaboub L, Chiba A, Yadav H, Vidi PA, Cook KL. **Diet alters entero-mammary signaling to regulate the breast microbiome and tumorigenesis**. *Cancer Res* (2021) **81** 3890-3904. DOI: 10.1158/0008-5472.can-20-2983
20. Chiba A, Bawaneh A, Velazquez C, Clear KYJ, Wilson AS, Howard-McNatt M, Levine EA, Levi-Polyachenko N, Yates-Alston SA, Diggle SP, Soto-Pantoja DR, Cook KL. **Neoadjuvant chemotherapy shifts breast tumor microbiota populations to regulate drug responsiveness and the development of metastasis**. *Mol Cancer Res* (2020) **18** 130-139. DOI: 10.1158/1541-7786.MCR-19-0451
21. Cook KL, Metheny-Barlow LJ, Tallant EA, Gallagher PE. **Angiotensin-(1-7) reduces fibrosis in orthotopic breast tumors**. *Cancer Res* (2010) **70** 8319-8328. DOI: 10.1158/0008-5472.CAN-10-1136
22. Claus SP, Tsang TM, Wang Y, Cloarec O, Skordi E, Martin FP, Rezzi S, Ross A, Kochhar S, Holmes E, Nicholson JK. **Systemic multicompartmental effects of the gut microbiome on mouse metabolic phenotypes**. *Mol Syst Biol* (2008) **4**. DOI: 10.1038/msb.2008.56
23. Yap IKS, Li JV, Saric J, Martin F-P, Davies H, Wang Y, Wilson ID, Nicholson JK, Utzinger J, Marchesi JR, Holmes E. **Metabonomic and microbiological analysis of the dynamic effect of vancomycin-induced gut microbiota modification in the mouse**. *J Proteome Res* (2008) **7** 3718-3728. DOI: 10.1021/pr700864x
24. Swann JR, Tuohy KM, Lindfors P, Brown DT, Gibson GR, Wilson ID, Sidaway J, Nicholson JK, Holmes E. **Variation in antibiotic-induced microbial recolonization impacts on the host metabolic phenotypes of rats**. *J Proteome Res* (2011) **10** 3590-3603. DOI: 10.1021/pr200243t
25. Zheng X, Xie G, Zhao A, Zhao L, Yao C, Chiu NHL, Zhou Z, Bao Y, Jia W, Nicholson JK, Jia W. **The footprints of gut microbial-mammalian co-metabolism**. *J Proteome Res* (2011) **10** 5512-5522. DOI: 10.1021/pr2007945
26. Wikoff WR, Anfora AT, Liu J, Schultz PG, Lesley SA, Peters EC, Siuzdak G. **Metabolomics analysis reveals large effects of gut microflora on mammalian blood metabolites**. *Proc Natl Acad Sci USA* (2009) **106** 3698-3703. DOI: 10.1073/pnas.0812874106
27. Samuelsson A-M, Matthews PA, Argenton M, Christie MR, McConnell JM, Jansen EHJM, Piersma AH, Ozanne SE, Twinn DF, Remacle C, Rowlerson A, Poston L, Taylor PD. **Diet-induced obesity in female mice leads to offspring hyperphagia, adiposity, hypertension, and insulin resistance: a novel murine model of developmental programming**. *Hypertension* (2008) **51** 383-392. DOI: 10.1161/HYPERTENSIONAHA.107.101477
28. Sullivan EL, Riper KM, Lockard R, Valleau JC. **Maternal high-fat diet programming of the neuroendocrine system and behavior**. *Horm Behav* (2015) **76** 153-161. DOI: 10.1016/j.yhbeh.2015.04.008
29. Seet EL, Yee JK, Jellyman JK, Han G, Ross MG, Desai M. **Maternal high-fat-diet programs rat offspring liver fatty acid metabolism**. *Lipids* (2015) **50** 565-573. DOI: 10.1007/s11745-015-4018-8
30. Moody L, Shao J, Chen H, Pan YX. **Maternal low-fat diet programs the hepatic epigenome despite exposure to an obesogenic postnatal diet**. *Nutrients* (2019) **11**. DOI: 10.3390/nu11092075
31. Safi-Stibler S, Thévenot EA, Jouneau L, Jouin M, Seyer A, Jammes H, Rousseau-Ralliard D, Baly C, Gabory A. **Differential effects of post-weaning diet and maternal obesity on mouse liver and brain metabolomes**. *Nutrients* (2020) **12**. DOI: 10.3390/nu12061572
32. Pereira AD, Ribeiro DC, de Santana FC, de Sousa dos Santos A, Mancini-Filho J, do Nascimento-Saba CCA, Velarde LGC, da Costa CAS, Boaventura GT. **Maternal flaxseed oil during lactation enhances bone development in male rat pups**. *Lipids* (2016) **51** 923-929. DOI: 10.1007/s11745-016-4165-6
33. Khamoui AV, Desai M, Ross MG, Rossiter HB. **Sex-specific effects of maternal and postweaning high-fat diet on skeletal muscle mitochondrial respiration**. *J Dev Orig Health Dis* (2018) **9** 670-677. DOI: 10.1017/S2040174418000594
34. Rodríguez-González GL, Reyes-Castro LA, Bautista CJ, Beltrán AA, Ibáñez CA, Vega CC, Lomas-Soria C, Castro-Rodríguez DC, Elías-López AL, Nathanielsz PW, Zambrano E. **Maternal obesity accelerates rat offspring metabolic ageing in a sex-dependent manner**. *J Physiol* (2019) **597** 5549-5563. DOI: 10.1113/JP278232
35. Guarda DS, de Moura EG, Carvalho JC, Reis AMD, Soares PN, Lisboa PC, Figueiredo MS. **Maternal flaxseed oil intake during lactation changes body fat, inflammatory markers and glucose homeostasis in the adult progeny: role of gender dimorphism**. *J Nutr Biochem* (2016) **35** 74-80. DOI: 10.1016/j.jnutbio.2016.05.011
36. Correia-Santos AM, Vicente GC, Suzuki A, Pereira AD, dos Anjos JS, Lenzi-Almeida KC, Boaventura GT. **Maternal use of flaxseed oil during pregnancy and lactation prevents morphological alterations in pancreas of female offspring from rat dams with experimental diabetes**. *Int J Exp Pathol* (2015) **96** 94-102. DOI: 10.1111/iep.12126
37. Vicente GC, Correia-Santos AM, Chagas MA, Boaventura GT. **Sexual dimorphism on aortic remodelling in rats offspring from diabetic mothers and the role of flaxseed oil in this effect**. *J Dev Orig Health Dis* (2019) **10** 338-344. DOI: 10.1017/S2040174418000740
38. Suzuki A, Correia-Santos AM, Vicente GC, Velarde LGC, Boaventura GT. **Effects of maternal flaxseed supplementation on female offspring of diabetic rats in serum concentration of glucose, insulin, and thyroid hormones**. *Int J Vitam Nutr Res* (2019) **89** 45-54. DOI: 10.1024/0300-9831/a000259
39. Keleher MR, Zaidi R, Shah S, Oakley ME, Pavlatos C, El Idrissi S, Xing X, Li D, Wang T, Cheverud JM. **Maternal high-fat diet associated with altered gene expression, DNA methylation, and obesity risk in mouse offspring**. *PLoS One* (2018) **13**. DOI: 10.1371/journal.pone.0192606
40. Shapiro ALB, Kaar JL, Crume TL, Starling AP, Siega-Riz AM, Ringham BM, Glueck DH, Norris JM, Barbour LA, Friedman JE, Dabelea D. **Maternal diet quality in pregnancy and neonatal adiposity: the Healthy Start Study**. *Int J Obes (Lond)* (2016) **40** 1056-1062. DOI: 10.1038/ijo.2016.79
41. Martin JM, Miranda RA, Barella LF, Palma-Rigo K, Alves VS, Fabricio GS, Pavanello A, Franco C. C D S, Ribeiro TA, Visentainer JV, Banafé EG, Martin CA, de Freitas Mathias PC, de Oliveira JC. **Maternal diet supplementation with n-6/n-3 essential fatty acids in a 1.2: 1.0 ratio attenuates metabolic dysfunction in MSG-induced obese mice**. *Int J Endocrinol* (2016) **2016** 1-10. DOI: 10.1155/2016/9242319
42. Summerfield M, Zhou Y, Zhou T, Wu C, Alpini G, Zhang KK, Xie L. **A long-term maternal diet transition from high-fat diet to normal fat diet during pre-pregnancy avoids adipose tissue inflammation in next generation**. *PLoS One* (2018) **13**. DOI: 10.1371/journal.pone.0209053
43. Marcelin G, Silveira ALM, Martins LB, Ferreira AVM, Clément K. **Deciphering the cellular interplays underlying obesity-induced adipose tissue fibrosis**. *J Clin Invest* (2019) **129** 4032-4040. DOI: 10.1172/JCI129192
44. Muir LA, Neeley CK, Meyer KA, Baker NA, Brosius AM, Washabaugh AR, Varban OA, Finks JF, Zamarron BF, Flesher CG, Chang JS, DelProposto JB, Geletka L, Martinez-Santibanez G, Kaciroti N, Lumeng CN, O’Rourke RW. **Adipose tissue fibrosis, hypertrophy, and hyperplasia: Correlations with diabetes in human obesity**. *Obesity (Silver Spring)* (2016) **24** 597-605. DOI: 10.1002/oby.21377
45. Paul HA, Bomhof MR, Vogel HJ, Reimer RA. **Diet-induced changes in maternal gut microbiota and metabolomic profiles influence programming of offspring obesity risk in rats**. *Sci Rep* (2016) **6**. DOI: 10.1038/srep20683
46. Paul HA, Collins KH, Nicolucci AC, Urbanski SJ, Hart DA, Vogel HJ, Reimer RA. **Maternal prebiotic supplementation reduces fatty liver development in offspring through altered microbial and metabolomic profiles in rats**. *FASEB J* (2019) **33** 5153-5167. DOI: 10.1096/fj.201801551R
47. Wankhade UD, Zhong Y, Kang P, Alfaro M, Chintapalli SV, Piccolo BD, Mercer KE, Andres A, Thakali KM, Shankar K. **Maternal high-fat diet programs offspring liver steatosis in a sexually dimorphic manner in association with changes in gut microbial ecology in mice**. *Sci Rep* (2018) **8**. DOI: 10.1038/s41598-018-34453-0
48. Zhu L, Sha L, Li K, Wang Z, Wang T, Li Y, Liu P, Dong X, Dong Y, Zhang X, Wang H. **Dietary flaxseed oil rich in omega-3 suppresses severity of type 2 diabetes mellitus via anti-inflammation and modulating gut microbiota in rats**. *Lipids Health Dis* (2020) **19**. DOI: 10.1186/s12944-019-1167-4
49. Wang T, Sha L, Li Y, Zhu L, Wang Z, Li K, Lu H, Bao T, Guo L, Zhang X, Wang H. **Dietary α-linolenic acid-rich flaxseed oil exerts beneficial effects on polycystic ovary syndrome through sex steroid hormones—microbiota—inflammation axis in rats**. *Front Endocrinol (Lausanne)* (2020) **11**. DOI: 10.3389/fendo.2020.00284
50. Perreault M, Białek A, Trottier J, Verreault M, Caron P, Milkiewicz P, Barbier O. **Role of glucuronidation for hepatic detoxification and urinary elimination of toxic bile acids during biliary obstruction**. *PLoS One* (2013) **8**. DOI: 10.1371/annotation/ef13ed51-6848-419d-94d8-1bb62e7bcf52
51. Gonthier MP, Verny MA, Besson C, Rémésy C, Scalbert A. **Chlorogenic acid bioavailability largely depends on its metabolism by the gut microflora in rats**. *J Nutr* (2003) **133** 1853-1859. DOI: 10.1093/jn/133.6.1853
52. Walsh MC, Brennan L, Pujos-Guillot E, Sébédio J-L, Scalbert A, Fagan A, Higgins DG, Gibney MJ. **Influence of acute phytochemical intake on human urinary metabolomic profiles**. *Am J Clin Nutr* (2007) **86** 1687-1693. DOI: 10.1093/ajcn/86.5.1687
53. Temellini A, Mogavero S, Giulianotti PC, Pietrabissa A, Mosca F, Pacifici GM. **Conjugation of benzoic acid with glycine in human liver and kidney: a study on the interindividual variability**. *Xenobiotica* (1993) **23** 1427-1433. DOI: 10.3109/00498259309059451
54. Pallister T, Jackson MA, Martin TC, Zierer J, Jennings A, Mohney RP, MacGregor A, Steves CJ, Cassidy A, Spector TD, Menni C. **Hippurate as a metabolomic marker of gut microbiome diversity: modulation by diet and relationship to metabolic syndrome**. *Sci Rep* (2017) **7**. DOI: 10.1038/s41598-017-13722-4
55. Pallister T, Jackson MA, Martin TC, Glastonbury CA, Jennings A, Beaumont M, Mohney RP, Small KS, MacGregor A, Steves CJ, Cassidy A, Spector TD, Menni C, Valdes AM. **Untangling the relationship between diet and visceral fat mass through blood metabolomics and gut microbiome profiling**. *Int J Obes (Lond)* (2017) **41** 1106-1113. DOI: 10.1038/ijo.2017.70
56. Burmester T, Gerlach F, Hankeln T. **Regulation and role of neuroglobin and cytoglobin under hypoxia**. *Adv Exp Med Biol* (2007) **618** 169-180. DOI: 10.1007/978-0-387-75434-5_13
57. Williams HRT, Cox IJ, Walker DG, Cobbold JFL, Taylor-Robinson SD, Marshall SE, Orchard TR. **Differences in gut microbial metabolism are responsible for reduced hippurate synthesis in Crohn’s disease**. *BMC Gastroenterol* (2010) **10** 108-108. DOI: 10.1186/1471-230X-10-108
58. Shearer J, Duggan G, Weljie A, Hittel DS, Wasserman DH, Vogel HJ. **Metabolomic profiling of dietary-induced insulin resistance in the high fat-fed C57BL/6J mouse**. *Diabetes Obes Metab* (2008) **10** 950-958. DOI: 10.1111/j.1463-1326.2007.00837.x
59. Waldram A, Holmes E, Wang Y, Rantalainen M, Wilson ID, Tuohy KM, McCartney AL, Gibson GR, Nicholson JK. **Top-down systems biology modeling of host metabotype-microbiome associations in obese rodents**. *J Proteome Res* (2009) **8** 2361-2375. DOI: 10.1021/pr8009885
60. Calvani R, Miccheli A, Capuani G, Tomassini Miccheli A, Puccetti C, Delfini M, Iaconelli A, Nanni G, Mingrone G. **Gut microbiome-derived metabolites characterize a peculiar obese urinary metabotype**. *Int J Obes (Lond)* (2010) **34** 1095-1098. DOI: 10.1038/ijo.2010.44
61. Brial F, Chilloux J, Nielsen T, Vieira-Silva S, Falony G, Hoyles L, Neves AL, Rodriguez-Martinez A, Mouawad GI, Pons N, Forslund S, Le Chatelier E, Le Lay AM, Nicholson JK, Hansen T, Clément K, Bork P, Ehrlich D, Raes J, Pedersen O, Gauguier D, Dumas ME. **Human and preclinical studies of the host-gut microbiome co-metabolite hippurate as a marker and mediator of metabolic health**. (2021) **70** 2105-2114. DOI: 10.1136/gutjnl-2020-323314
62. Lichtman MK, Otero-Vinas M, Falanga V. **Transforming growth factor beta (TGF-β) isoforms in wound healing and fibrosis**. *Wound Repair Regen* (2016) **24** 215-222. DOI: 10.1111/wrr.12398
63. Yang L, Pang Y, Moses HL. **TGF-β and immune cells: an important regulatory axis in the tumor microenvironment and progression**. *Trends Immunol* (2010) **31** 220-227. DOI: 10.1016/j.it.2010.04.002
64. Heldin CH, Miyazono K, Ten Dijke P. **TGF-β signalling from cell membrane to nucleus through SMAD proteins**. *Nature* (1997) **390** 465-471. DOI: 10.1038/37284
65. Attisano L, Wrana JL. **Mads and Smads in TGFβ signalling**. *Curr Opin Cell Biol* (1998) **10** 188-194. DOI: 10.1016/S0955-0674(98)80141-5
66. Derynck R, Zhang Y, Feng XH. **Smads: transcriptional activators of TGF-β responses**. *Cell* (1998) **95** 737-740. DOI: 10.1016/S0092-8674(00)81696-7
67. Abidi A, Kourda N, Feki M, Ben Khamsa S. **Protective effect of tunisian flaxseed oil against bleomycin-induced pulmonary fibrosis in rats**. *Nutr Cancer* (2020) **72** 226-238. DOI: 10.1080/01635581.2019.1622741
68. Sankaran D, Bankovic-Calic N, Peng CYC, Ogborn MR, Aukema HM. **Dietary flax oil during pregnancy and lactation retards disease progression in rat offspring with inherited kidney disease**. *Pediatr Res* (2006) **60** 729-733. DOI: 10.1203/01.pdr.0000245912.78078.d5
69. Grassi TF, Bidinotto LT, Lopes GAD, Zapaterini JR, Rodrigues MAM, Barbisan LF. **Maternal western-style diet enhances the effects of chemically-induced mammary tumors in female rat offspring through transcriptome changes**. *Nutr Res* (2019) **61** 41-52. DOI: 10.1016/j.nutres.2018.09.009
70. Li J, Li K, Gao J, Guo X, Lu M, Li Z, Li D. **Maternal exposure to an n-3 polyunsaturated fatty acid diet decreases mammary cancer risk of female offspring in adulthood**. *Food Funct* (2018) **9** 5768-5777. DOI: 10.1039/c8fo01006d
71. Newman TM, Vitolins MZ, Cook KL. **From the table to the tumor: the role of Mediterranean and Western dietary patterns in shifting microbial-mediated signaling to impact breast cancer risk**. *Nutrients* (2019) **11**. DOI: 10.3390/nu11112565
72. Siegel PM, Massagué J. **Cytostatic and apoptotic actions of TGF-β in homeostasis and cancer**. *Nat Rev Cancer* (2003) **3** 807-820. DOI: 10.1038/nrc1208
73. Syed V. **TGF-β signaling in cancer**. *J Cell Biochem* (2016) **117** 1279-1287. DOI: 10.1002/jcb.25496
74. Koumoundourou D. **Prognostic significance of TGF‚ -1 and pSmad2/3 in breast cancer patients with T 1-2, N 0 tumours**. *Anticancer Res* (2007) **27** 2613-2620. PMID: 17695423
75. Qiu Q, Su Y, Zheng Y, Cai H, Wu S, Lu W, Zheng W, Shu XO, Cai Q. **Increased pSmad2 expression and cytoplasmic predominant presence of TGF-βRII in breast cancer tissue are associated with poor prognosis: results from the Shanghai Breast Cancer Study**. *Breast Cancer Res Treat* (2015) **149** 467-477. DOI: 10.1007/s10549-014-3251-9
|
---
title: 'Mental health professionals and telehealth in a rural setting: a cross sectional
survey'
authors:
- David Nelson
- Maxime Inghels
- Amanda Kenny
- Steve Skinner
- Tracy McCranor
- Stephen Wyatt
- Jaspreet Phull
- Agnes Nanyonjo
- Ojali Yusuff
- Mark Gussy
journal: BMC Health Services Research
year: 2023
pmcid: PMC9970689
doi: 10.1186/s12913-023-09083-6
license: CC BY 4.0
---
# Mental health professionals and telehealth in a rural setting: a cross sectional survey
## Abstract
### Background
Telehealth usage has been promoted in all settings but has been identified as a panacea to issues of access and equity in the rural context. However, uptake and widespread integration of telehealth across all parts of the health system has been slow, with a myriad of barriers documented, including in rural settings. The crisis of the COVID-19 pandemic, saw barriers rapidly overturned with the unprecedented and exponential rise in telehealth usage. The uniqueness of the crisis forced telehealth adoption, but as the urgency stabilises, pandemic learnings must be captured, utilised, and built upon in a post-pandemic world. The aim of this study was to document staff experiences and perceptions of delivering rural psychological therapies via telehealth during the pandemic and to capture learnings for future rural telehealth delivery.
### Methods
An online cross-sectional survey that explored mental health professional’s experiences, use, and perceptions of telehealth before and after pandemic-enforced changes to service delivery.
### Results
Sixty-two respondents completed the questionnaire (response rate $68\%$). Both the delivery of telehealth via telephone and online video conferencing significantly increased during the pandemic ($66\%$ vs $98\%$, $p \leq .001$ for telephone and $10\%$ vs $89\%$, $p \leq 0.001$ for online video). Respondents indicated that client’s access to services and attendance had improved with telehealth use but their attention and focus during sessions and non-verbal communication had been negatively affected. The challenges for older adults, people with learning and sensory disabilities, and residents in remote areas with poorer mobile/internet connectivity were identified. Despite these challenges, none of the respondents indicated a preference to return to fully face-to-face service delivery with most ($86\%$) preferring to deliver psychological therapies fully or mostly via telehealth.
### Conclusions
This study addresses three major gaps in knowledge: the experience of delivering local telehealth solutions to address rural mental health needs, the provision of strong rural-specific telehealth recommendations, and the dearth of rural research emanating from the United Kingdom. As the world settles into a living with COVID-19 era, the uniqueness of the rural telehealth context may be forgotten as urban myopia continues to dominate telehealth policy and uptake. It is critical that rural resourcing and digital connectivity are addressed.
## Background
This study documents staff experiences and perceptions of delivering rural psychological therapies via telehealth and captures critical learnings for future rural telehealth delivery. Three major gaps in knowledge are addressed: the experience of delivering telehealth solutions to meet rural mental health needs, the provision of strong rural specific telehealth recommendations, and the dearth of rural research emanating from the United Kingdom (UK).
For decades, telehealth has been central in the digital strategies of major countries to support long-distance healthcare, education, public health, and administration [1–4]. In the rural context, telehealth has been extensively promoted to ensure safe, effective, and equitable service delivery to people impacted by distance and isolation [5–16].
In the pre-pandemic period, urgent calls were made for radical approaches to technological use in healthcare [1–3, 17].Authors of systematic reviews on telehealth had demonstrated client satisfaction, including improved outcomes (varied definitions), cost savings, better communication and ease of use, and reductions in travel time [18]. There were pockets of excellence in telehealth delivery, however, uptake and widespread integration was slow [1–3, 17]. Barriers to uptake included a lack of infrastructure, fractured and complex health systems, and unwieldy funding models/jurisdictional boundaries. Challenging technology and commercial arrangements, risk adversity, limited resources and increased costs, issues with public trust and data security, lack of digital skills amongst workers at all levels and resistance to education, and lack of political and organisational leadership to drive change have all been documented [1–3, 17]. Specific rural barriers included lack of digital connectivity, coordination and turnover of senior management, lower levels of health literacy, lack of resourcing at the client end, challenges in disciplines where touch is a requirement, concerns about security, inadequate staffing, lack of equipment and other resources, inflexible billing arrangements, complex referral patterns, and lack of local control [5–16, 19]. In 2018, in the UK, harsh criticism was directed at an increasing gap between technological potential and usage [1].
## The impact of the COVID-19 pandemic
In 2020, the COVID-19 pandemic saw an unprecedented and rapid overturning of many barriers and major acceleration of telehealth usage [3, 4, 17, 20–22]. In May 2020, Dr. Sacha Bhatia stated, ‘… the COVID pandemic was the match that lit the fire around this revolution in virtual care’. [ 4] The US Congress rapidly overturned telehealth billing and reimbursement restrictions, enabled telehealth from people’s homes (including physicians), and expanded approved platforms [3]. Similar changes were made in Australia [21, 22], Canada [4, 17, 19], and the UK [23].
Rates of pandemic related anxiety, depression, post-traumatic stress disorder, psychological distress, and stress across the general population [24] [25, 26], saw increased mental health service demand. This demand and reductions in service delivery due to COVID-19 controls, forced service providers to pivot rapidly to telehealth to maintain care delivery.
Key articles have emerged on telehealth for rural mental health service delivery during the pandemic [27, 28]. In rural Pennsylvania, Svistova and colleagues [27] explored mental health provider’s experiences of using telehealth with youth and older populations. Positives included service continuation, greater parental involvement, decreased no-show rates, and easing of transport difficulties. They did, however, note challenges in digital access and recommended hybrid models of service delivery [27]. In a study investigating telehealth usage among people with mental illness in rural Louisiana, Sizer et al. [ 28] argued that the digital divide required elimination, especially for older people, those with lower levels of education, and those with serious and enduring mental illness. In Virginia, US, telehealth usage during the pandemic was explored for people with adjustment disorders, anxiety, and depression [29]. It was hypothesised that lower uptake of rural telehealth may have been the result of fewer resources to manage demand surge. In Australia, Chatterton et al. [ 30] demonstrated marked increases in mental health telehealth usage because of the pandemic. Limitations associated with small sample sizes impact their findings, but lack of digital infrastructure, lack of user-friendly platforms, and privacy concerns were reported [30].
Caffery et al. [ 31] (p.544) argue that the COVID-19 pandemic uncovered ‘a myopia we term urban paternalism in understanding and delivering rural health’. This was described as policy and practice driven from an urban stance. The uniqueness of the pandemic and associated lockdown measures resulted in urban policymakers and urban dwellers experiencing isolation and the inability to access health services. These issues have been documented in the rural context for decades.
While much is documented on rural health in countries with large geographic mass and dispersed populations there is a void in rural research in the UK. It is often forgotten that $85\%$ of the landmass in the UK is rural and is home to more than 10 million people. While parts of rural UK represent the bucolic idyll that many imagine, deprivation, poor outcomes, major inequalities, and unmet needs are often hidden due to a lack of fined-grained data at the county level. Within relatively small geographic areas, life expectancy can vary by 10 years [32]. There are similar issues shared with other rural areas including aging populations, limited access to health and social care, inequalities associated with travel and transport, and generational poverty exacerbated by inadequate housing and escalating fuel costs [32]. Social isolation and loneliness are major features [33]. There are major issues with rural digital exclusion which exacerbates access and social isolation [34].
In 2019, the UK Government released key policy on rural research priorities. They reinforced the diversity in rural UK and need for locally developed strategies, digital connectivity, and technology [34]. The study described in this paper is a report of a major local strategy to address both mental health and rural isolation so is important in the genesis of embryonic UK rural research.
## Study aim
The aim of this study was to document staff experiences and perceptions of delivering rural psychological therapies via telehealth during the pandemic and to capture learnings for future rural telehealth delivery. The research question was: what were mental health professionals’ experiences, use and perceptions of telehealth before and after pandemic enforced changes to service delivery?
## Setting
The setting for this study was the UK county of Lincolnshire. The predominantly rural county has a mixture of affluence and deprivation [35]. Within the county, high rates of smoking, alcohol and drug use, and poor physical and mental health are evident in coastal and deprived communities. Rural, seasonal and coastal populations provide significant challenges to traditional modes of service delivery often based on urban modelling [36].
The Improving Access to Psychological Therapies (IAPT) programme has been implemented nationally in the UK as an evidence-based approach to delivering psychological therapies for depression and anxiety disorders in line with the stepped-care clinical guidelines issued by the National Institute for Health and Care Excellence (NICE) [37, 38]. For people accessing IAPT with anxiety disorders, cognitive behavioural therapy (CBT) is recommended [38]. For those with depression, a wider range of treatments are recommended (CBT, counselling, couples therapy, interpersonal therapy, and psychodynamic therapy) [38]. In England, there are now over 200 IAPT services which render it the largest publicly funded and systematic implementation of evidence-based psychological care in the world [39]. It now serves as a model for similar systems in countries such as Australia, Canada, Norway, and Japan [40–43].
Steps2change (Lincolnshire Partnership NHS Foundation Trust) is the Improving Access to Psychological Therapies (IAPT) service for the county of Lincolnshire in the East Midlands of England. The service is delivered using a nationally developed stepped-care approach offering the least intrusive intervention first and monitoring people’s progress in outcome-focused supervision. Before the COVID-19 pandemic, the service operated mainly from nine sites offering a mixture of one-to-one and group face-to-face appointments, telephone-guided self-help, and to a lesser extent internet-enabled therapy such as computerised cognitive behavioural therapy (cCBT).
## Design
An online study-specific cross-sectional questionnaire was administered to staff within the IAPT service at Lincolnshire Partnership NHS Foundation Trust. The survey was designed using Qualtrics software [44].
## Participants
All ninety-one IAPT service staff were invited to participate by email from the IAPT Clinical Lead on behalf of the research team. The IAPT service staff consists of an interdisciplinary team of cognitive behavioural therapists, counsellors, employment advisors, interpersonal therapists, and psychological well-being practitioners. Staff provide a range of evidence based talking therapies and psychological treatments for people with depression, anxiety, panic attacks, post-trauma reaction, phobias and obsessive-compulsive disorders. Those with management/administration-only roles were excluded as they did not have direct experience in delivering telehealth services to people. Trainees and staff who had been recently employed were also excluded as they did not have sufficient experience in delivering IAPT pre and post COVID-19. To maximise responses two reminder emails were sent from the Clinical Lead for the IAPT service to potential participants.
## Questionnaire
The questionnaire was developed from the published literature [45–50] by the core research team in collaboration with IAPT service provider staff. The questionnaire was piloted with four members of the IAPT team to ensure readability and face validity before distributing to all eligible staff. Those who took part in the pilot suggested that free text questions should be added to specific survey items to allow the participants to elaborate on some of their choices. This amendment was made before distribution.
The final questionnaire included demographic questions and items that asked about respondent’s use of telehealth before and after COVID-19 related service changes, their own experiences of telehealth delivery, their perceptions of the level of their client’s satisfaction and perceived impact on the quality and quantity of service provided. There were twenty-four questions in total. Where participants were asked to compare telehealth services with service delivery prior to the Covid-19 pandemic a Likert scale with the following categories was utilised [1] much worse [2] somewhat worse [3] about the same [4] somewhat better and [5] much better. The force response option in Qualtrics was used to ensure that participants did not overlook or miss any of the questions that were vital to the analysis. There was no missing data.
Data were collected between April-June 2021. During this time there were significant restrictions on non-essential social contact imposed by the UK government.
## Analysis
Descriptive statistics (frequencies and percentages) were used to describe and summarise the data. Fisher’s Exact Tests [51] were used to test for significant differences between telehealth use prior to and since the pandemic, as well as changes in perceived skill level before and after the pandemic onset. Free text responses were reported to allow further context to participant’s responses. Data were analysed using R software version 4.1.1 [52].
## Sample demographics
A total of 62 respondents completed the questionnaire resulting in a response rate of $68\%$. Whilst there were responses across the different occupations that make up IAPT, cognitive behavioural therapists and psychological and wellbeing practitioners were the highest responders. The responses were a good representation of the different occupations that make up the IAPT service at the time of data collection. Full respondent characteristics are displayed in Table 1.Table 1Characteristics of respondents and non-respondentsTotal $$n = 62$$n (%)Variable Age 18-243 (4.8) 25-3425 (40.3) 35-449 (13.4) 45-5410 (14.5) 55-6414 (22.6) 65-741 (1.6) Gender Female50 (80.6) Male12 (19.4)Total $$n = 91$$No. eligible to take part n (%) Occupation Cognitive Behavioural Therapist22 (35.5)35 (38.4) Counsellor8 (12.9)14 (15.4) Employment Advisor9 (14.5)11 (12.1) Interpersonal Therapist3 (4.8)9 (9.9) Psychological Wellbeing Practitioner20 (32.3)22 (24.2)Note: *Total data* were not available for age and gender only occupation. This number excludes trainee staff
## Pre and post COVID-19 use of telehealth and reported skill level
Table 2 presents findings about telehealth provision and self-rated skill levels prior to and since COVID-19. Before the pandemic, two-thirds ($66\%$) of staff had delivered services via the telephone but only $10\%$ reported the use of online video conferencing for service delivery before the pandemic. The reported delivery of telehealth via both telephone and online video conferencing had significantly increased since the pandemic with $98\%$ reporting the use of telephone and $89\%$ video conferencing. Additionally, participant’s self-rated skill levels delivering telehealth via the telephone and using video conferencing had significantly improved compared to before the pandemic. Despite this, $58\%$ reported no formal training for using telehealth at the time of data collection (Table 3).Table 2Telehealth provision, and self-rated skill level within IAPT prior to and since COVID-19 (March 2020)Prior to COVID-19Since COVID-19Total $$n = 62$$n (%)p value*Telephone Provision of care using telephone consultation<.001 Yes41 (66.1)61 (98.4) No21 (33.9)1 (1.6) Skill level using telephone consultation<.001 Very good16 (25.8)32 (51.6) Good18 (29.0)25 (40.3) OK, but not brilliant24 (38.7)5 (8.1) Not at all good4 (6.5)0 (0.0)Online video conferencing Provision of care using online video conferencing<.001 Yes6 (9.7)55 (88.7) No56 (90.3)7 (11.3) Skill level using online video conferencing<.001 Very good4 (6.5)18 (29.0) Good11 (17.7)33 (53.2) OK, but not brilliant21 (33.9)8 (12.9) Not at all good26 (41.9)3 (4.8)*Fisher’s Exact Test was used to assess for significance between telehealth use and skill level prior to and since COVID-19 (from March 2020)Table 3IAPT Practitioner Experiences of Telehealth since COVID-19Total $$n = 62$$n (%)Example free text responsesPerceived patient satisfaction with telehealth compared to face-to-face Higher when compared to F2F14 (22.6)“More flexibility in times and being able to offer appointments that suit. No travel.” Similar levels of satisfaction compared to F2F40 (64.5) Lower levels of satisfaction compared to F2F8 (12.9)“Clients still hold a belief that face-to-face treatment will be better. It can be hard to overcome this thinking. ”Practitioner satisfaction with telehealth since Covid-19 Very satisfied35 (56.5) Satisfied22 (35.5) Neither satisfied nor unsatisfied4 (6.5) Unsatisfied1 (1.6)Preferred form of service delivery Fully Telehealth20 (32.3) Mostly Telehealth with some F2F33 (53.2) Mostly face-to-face with some Telehealth9 (14.5) Fully F2F0 (0.0)Thinks diary is easier to manage with telehealth Yes, it is easier when compared to F2F33 (53.2) *It is* about the same effort compared to F2F27 (43.5) No, it is not as easy compared to F2F2 (3.2)“*Admin is* taking longer. ”Thinks telehealth changes the way confidentiality is maintained Yes, it makes it more difficult13 (21.0)“When the individual doesn’t have privacy at home it can be difficult.” No, it makes little difference45 (72.6) Yes, it makes it easier4 (6.5)“Patients do not have to sit in the waiting room where they may know others. ”Aware of groups who have difficulty engaging with Telehealth Yes40 (64.5)“People who live in remote areas who don’t have good internet or phone signal.” No22 (32.5)Ever had telehealth related training since COVID-19 Yes26 (41.9)“Webinars and written guidance.” No36 (58.1)“We have some CPD on this, but no specific training by the Trust I don’t think. ”Ever had a technology/ICT issue with Telehealth since COVID-19 Yes48 (77.4)“Poor connection, either on telephone or on-line.” No14 (22.6)Thinks that telehealth requires different skills compared to F2F Yes48 (77.4)“I sometimes find it harder to close down inappropriate communications and interject when necessary while being unable to use visual techniques.” No9 (14.5) Not sure5 (8.1)Note: Free text responses were not available for all questions
## Experiences and perceptions of telehealth service delivery
Respondents were asked to compare telehealth services with services provided before COVID-19 and the move to telehealth in ten key areas (Fig. 1). Positive changes were reported in client’s access to services as well as attendance rates for telehealth consultations. More negative perceived impacts were the client’s ability to use and observe non-verbal communication as well as to maintain attention and focus. Mixed findings are seen in the perception of time required for each approach to service. Fig. 1Comparing Telehealth services with services you provided prior to COVID-19 and the move to Telehealth (March 2020), how would you rate the following Respondents largely believed client’s had similar or higher levels of satisfaction with telehealth services when compared to face-to-face (Table 3). Most practitioners were themselves satisfied or very satisfied with the delivery of IAPT services via telehealth and none wanted IAPT to return to a fully face-to-face service. There was a strong preference for future services to be mostly or fully online (53 and $32\%$ respectively). Many respondents ($58\%$) indicated that they had not had any formal training in telehealth approaches and/or technologies. This was despite a high level of agreement ($77\%$) that telehealth service delivery required different skills to traditional in-person consultations. When asked about concerns maintaining patient confidentiality when using telehealth approaches most ($73\%$) indicated little difference when compared to pre-COVID practices, although one-fifth reported it as more difficult.
Respondents indicated that there might be groups of people who could have difficulty engaging with telehealth and digital services. These were largely older adults, people with learning and sensory disabilities, as well as those in remote areas prone to poor phone signal and internet connectivity.
## Discussion
The aim of the study was to document staff experiences and perceptions of delivering rural psychological therapies via telehealth during the pandemic and to capture learnings for future rural telehealth delivery. Three major gaps in knowledge are addressed.
The dearth of robust research on rural service delivery, telehealth, and other technological solutions has been documented by the UK Government who have called for research to inform rural policy development and implementation [34]. The Government highlights the need to ‘draw on specific local development strategies and their effectiveness in promoting inclusive growth and welfare.’ [ 34] Steps2change (Lincolnshire Partnership NHS Trust), the Improving Access to Psychological Therapies (IAPT) service for the county of Lincolnshire in the East Midlands of England, provides an ideal case to illustrate local strategies.
In the pre-pandemic period, the promotion of telehealth, almost as a panacea to address rural geographic distance and health service inequities, did not gain enough traction for widespread, large-scale adoption [1–3, 17]. Much of the existing literature over the last decades has been about chronicling challenges of why telehealth was difficult to implement [1–3, 17]. The COVID-19 pandemic resulted in a seismic paradigm shift as urban dwellers suddenly experienced what rural people had reported for decades: isolation and inability to easily access healthcare [31]. Without diminishing the impact of the pandemic globally, some argue that the rapid overturning of barriers and unprecedented acceleration of telehealth usage [3, 4, 17, 20–22] was because of urban paternalism [31] and urban policymakers who had their myopic blinders about isolation and inequity rapidly blown off. As the world settles into living with COVID-19, there is a major risk that urban myopia will continue to dominate telehealth expansion and the uniqueness of the rural context will be forgotten, resulting in continued inequities across the rural/urban divide.
Our results offer valuable insight into the experiences and perceptions of mental health practitioners using telehealth to deliver psychological therapies during the COVID-19 pandemic. Practitioners believed telehealth approaches had greatly improved patients’ access to IAPT services. Telehealth addressed flexibility around people’s schedules, reduced travel and opportunity costs, and reduced risks of the stigma associated with physically attending healthcare premises for mental health treatment. None of the participants in the study wanted to return to a completely face-to-face model of delivery.
The positives of telehealth identified by practitioners in this study have been identified in studies in the UK, US, Canada and Australia [53–55]. In rural US, Svistova and colleagues [27] recommended hybrid models of service delivery rather than a return to all face-to-face. Additionally, these positives align with systematic reviews on the perspectives of clients, who see improved access and reductions in both the cost and time commitment major benefits of telehealth [18].
The lack of digital infrastructure identified in this study has consistently been reported internationally across a multitude of health systems. In 2018 in the United Kingdom [1], building the best technology into health systems and ensuring that digital systems and people’s needs aligned were identified as critical. It took a global pandemic to expedite action despite decades of calls to improve digital connectivity and ensure the building blocks of the right digital architecture are in place. Urban policymakers and urban dwellers would not tolerate the rural digital access and connectivity issues that have been well-documented for decades [31]. The UK *Government is* not unique in stating that inferior digital infrastructure is beyond acceptable [2, 3, 19], but these comments are about all areas of the NHS, not just rural. As the crisis of the pandemic diminishes there is a major risk that technology developments will be centred on large urban centres and rural areas will continue to experience digital connectivity issues that are poorer than some low-income countries. Major, targeted investment must be made in rural areas. If personal connectivity cannot be guaranteed, then investment should be made by government to work with technology developers to identify and implement practical solutions to address issues such as sub-standard bandwidth. As noted in this study, it is unacceptable that major sections of the community cannot even achieve a consistent internet connection.
This study demonstrated major gaps in education to ensure that health professionals had the knowledge and skill to deliver outstanding telehealth services. Major investment in the education of all health professionals in telehealth service delivery, no matter the context, has been identified as critical. New roles that span the interface between clinical and technical staff in supporting staff and clients and providing detailed analytics will be needed [56] However, at a fundamental level, rural health professionals face major barriers in accessing education targeted to their needs and locality. There is a dearth of literature on telehealth education for rural health professionals. In a recent scoping review on health professional education in isolated settings, Reeve et al. [ 57] reported on 40 studies. Digital and technological education was not identified in any study [57]. Massive Open Online Courses (MOOCs) have been proposed as one solution to digital education [56], however, the delivery of urban-centric training programs by online means to outpost rural settings is not without its challenges. Local high quality-education programs must be implemented to ensure high-quality outcomes for both staff and clients. There should be a focus on co-designed educational programs by health professionals and clients that align with their need, and shared learning of staff and clients should be a priority.
In this study, respondents identified groups who might have difficulty engaging with telehealth and digital services. This needs to be flipped. It is not the groups that have difficulty engaging, it is that telehealth and digital services are hard to engage with. Studies that show the most promising telehealth outcomes are largely with well-educated populations [28]. Challenges that older adults, people with mental illness, and people with learning and sensory disabilities face in accessing telehealth documented in this study have been reported in rural US [27, 28]. A simple solution would be to suggest that face-to-face service delivery is better for groups that already face deprivation due to their location [28]. However, studies show high acceptability of telehealth with populations that are well supported [18] so why should people who already face major inequities be further disadvantaged and denied solutions that might improve their quality of life? The need for new service models to be co-created with all stakeholders who use telehealth, including diverse members of the public, has been reinforced in major UK reports [56]. Simple solutions such as having a professional with a client at the user end of telehealth and co-consultation situations would go a long way to improving the experience for both the client and the professional. This, of course, would take funding commitments but would be a small cost to ensure vulnerable rural people facing deprivation have the same rights of access as urban dwellers.
In the UK, demands for urgent action to address health inequalities and poor health and well-being outcomes in rural and coastal areas are escalating [36], however, calls without concerted action are meaningless. Given the significant rural geography and millions of UK residents that reside outside urban cities, rural UK research and dissemination of learnings that demonstrate tangible and measurable health and wellbeing outcomes must be an urgent priority. This study makes a small contribution to embryonic rural UK research.
## Limitations
While the questionnaire was developed from key published literature and met the aims of this study, pilot testing only occurred with four members of the IAPT team. Further testing of the validity of the questionnaire is recommended for further research. There are inherent limitations in self-reported questionnaires, particularly related to social desirability bias. There is a risk that participants in this study felt compelled to answer positively, however, the results indicate that participants perceived benefits and barriers.
Data in this study were collected over a three-month period (April-June 2021) which provides an understanding of how practitioners felt after one year of working with telehealth approaches rather than more immediate responses. Staff may have adjusted to delivering IAPT services via telehealth which could explain the high levels of satisfaction reported in this survey.
Future studies should collect data from users of IAPT via telehealth as well as those who have had difficulty or challenges engaging with telehealth. The cross-sectional descriptive design of this study limits the extent to which the findings are generalisable to IAPT services elsewhere and needs to be tested with subsequent data collection in both rural and urban settings.
## Conclusions
While much of the geography in the UK is designated rural, the health and well-being of millions of rural UK residents has attracted limited interest. The idyllic representation of rural UK hides substantial pockets of deprivation, major issues regarding healthcare accessibility, and poor health and well-being outcomes. Telehealth has been promoted as a panacea for healthcare accessibility in rural communities, but a myriad of barriers has impacted timely uptake. This study makes a significant contribution to both rural and telehealth literature through descriptions of clinician’s experiences in delivering the world’s largest publicly funded, evidence-based psychological program in a rural country in the United Kingdom. The COVID-19 pandemic created a massive shift in the global uptake of telehealth, with this study demonstrating improved access for rural people. While telehealth has captured the attention of policymakers and practitioners internationally, there is a risk that going forward the focus of technological development will centre on the needs of urban centres. The need for fit-for-purpose infrastructure and resourcing, education and training, and support for those with complex needs living in rural communities must not be compromised through urban paternalism and policy and practice driven from an urban stance.
## References
1. 1.United Kingdom Government. The future of healthcare: our vision for digital, data and technology in health and care, https://www.gov.uk/government/publications/the-future-of-healthcare-our-vision-for-digital-data-and-technology-in-health-and-care/the-future-of-healthcare-our-vision-for-digital-data-and-technology-in-health-and-care (2018, accessed November 7 2022).
2. 2.The University of QueenslandDeveloping a policy strategy for telehealth in Australia2019BrisbaneThe University of Queensland. *Developing a policy strategy for telehealth in Australia* (2019.0)
3. Shaver J. **The state of Telehealth before and after the COVID-19 pandemic**. *Primary Care* (2022.0) **49** 517-530. DOI: 10.1016/j.pop.2022.04.002
4. 4.Virtual Care TaskforceReport of the Virtual Care Taskforce2022. *Report of the Virtual Care Taskforce* (2022.0)
5. Avey JP, Hobbs RL. **Dial in: fostering the use of telebehavioral health services in frontier Alaska**. *Psychol Serv* (2013.0) **10** 289-297. DOI: 10.1037/a0028231
6. Sundstrom B, DeMaria AL, Ferrara M. **"The closer, the better:" the role of Telehealth in increasing contraceptive access among women in rural South Carolina**. *Matern Child Health J* (2019.0) **23** 1196-1205. DOI: 10.1007/s10995-019-02750-3
7. St Clair M, Murtagh D. **Barriers to Telehealth uptake in rural, regional, remote Australia: what can be done to expand Telehealth access in remote areas?**. *Stud Health Technol Inform* (2019.0) **266** 174-182. DOI: 10.3233/SHTI190791
8. Scriven H, Doherty DP, Ward EC. **Evaluation of a multisite telehealth group model for persistent pain management for rural/remote participants**. *Rural Remote Health* (2019.0) **19** 4710. DOI: 10.22605/RRH4710
9. Pradhan T, Six-Workman EA, Law KB. **An Innovative Approach to Care: Integrating Mental Health Services Through Telemedicine in Rural School-Based Health Centers**. *Psychiatr Serv* (2019.0) **70** 239-242. DOI: 10.1176/appi.ps.201800252
10. Myers CR. **Using Telehealth to remediate rural mental health and healthcare disparities**. *Issues Ment Health Nurs* (2019.0) **40** 233-239. DOI: 10.1080/01612840.2018.1499157
11. Lee S, Black D, Held ML. **Factors associated with Telehealth service utilization among rural populations**. *J Health Care Poor Underserved* (2019.0) **30** 1259-1272. DOI: 10.1353/hpu.2019.0104
12. Hosseini H, Yilmaz A. **Using Telehealth to address pediatric obesity in rural Pennsylvania**. *Hosp Top* (2019.0) **97** 107-118. DOI: 10.1080/00185868.2019.1629365
13. Hatton JL, Rowlandson J, Beers A. **Telehealth-enabled auditory brainstem response testing for infants living in rural communities: the British Columbia Early Hearing Program experience**. *Int J Audiol* (2019.0) **58** 381-392. DOI: 10.1080/14992027.2019.1584681
14. Campbell J, Theodoros D, Russell T. **Client, provider and community referrer perceptions of telehealth for the delivery of rural paediatric allied health services**. *Aust J Rural Health* (2019.0) **27** 419-426. DOI: 10.1111/ajr.12519
15. Sabesan S, Senko C, Schmidt A. **Enhancing Chemotherapy Capabilities in Rural Hospitals: Implementation of a Telechemotherapy Model (QReCS) in North Queensland, Australia**. *J Oncol Pract* (2018.0) **14** e429-e437. DOI: 10.1200/JOP.18.00110
16. Levin JL, Philips BU. **Improving health in hard-to-reach communities**. *Tex Med* (2018.0) **114** 28-32. PMID: 29393963
17. Johnson C, Dupuis J, Goguen P. **Changes to telehealth practices in primary care in New Brunswick (Canada): a comparative study pre and during the COVID-19 pandemic**. *PLoS One* (2021.0) **16** e0258839. DOI: 10.1371/journal.pone.0258839
18. Kruse CS, Krowski N, Bea R. **Telehealth and patient satisfaction: a systematic review and narrative analysis**. *BMJ Open* (2017.0) **7** e016242. DOI: 10.1136/bmjopen-2017-016242
19. Falk W. *The state of virtual care in Canada as of wave three of the COVID-19 pandemic* (2021.0)
20. 20.Australian GovernmentProviding health care remotely during the COVID-19 pandemic2022. *Providing health care remotely during the COVID-19 pandemic* (2022.0)
21. 21.Taylor A, Caffery LJ, Gesesew HA, et al. How Australian Health Care Services Adapted to Telehealth During the COVID-19 Pandemic: A Survey of Telehealth Professionals. Frontiers in Public Health. 2021;9. Original Research. 10.3389/fpubh.2021.648009.
22. 22.Ang A. Australia working towards 'permanent' telehealth, https://www.healthcareitnews.com/news/anz/australia-working-towards-permanent-telehealth (2021, accessed November 7 2022).
23. Hutchings R. *The impact of Covid-19 on the use of digital technology in the NHS* (2020.0)
24. Xiong J, Lipsitz O, Nasri F. **Impact of COVID-19 pandemic on mental health in the general population: A systematic review**. *J Affect Disord* (2020.0) **277** 55-64. DOI: 10.1016/j.jad.2020.08.001
25. Pierce M, Hope H, Ford T. **Mental health before and during the COVID-19 pandemic: a longitudinal probability sample survey of the UK population**. *Lancet Psychiatry* (2020.0) **7** 883-892. DOI: 10.1016/S2215-0366(20)30308-4
26. 26.Nelson D, McGonagle I, Jackson C, Gussy M, Kane R. A rural-urban comparison of self management in people living with cancer following primary treatment: a mixed methods study. Psychooncology. 2022;31( 10):1660–70. 10.1002/pon.6011.
27. Svistova J, Harris C, Fogarty B. **Use of Telehealth amid the COVID-19 pandemic: experiences of mental health providers serving rural youth and elderly in Pennsylvania**. *Admin Pol Ment Health* (2022.0) **49** 530-538. DOI: 10.1007/s10488-021-01181-z
28. Sizer MA, Bhatta D, Acharya B. **Determinants of Telehealth Service Use among Mental Health Patients: A Case of Rural Louisiana**. *Int J Environ Res Public Health* (2022.0) **19** 06. DOI: 10.3390/ijerph19116930
29. Sweet H, Campion F, Bixler M. **Rural Tele-mental health use during the pandemic for adjustment disorders, anxiety, and depression**. *J Rural Ment Health* (2022.0) **46** 151-161. DOI: 10.1037/rmh0000209
30. Chatterton ML, Marangu E, Clancy EM. **Telehealth service delivery in an Australian regional mental health service during COVID-19: a mixed methods analysis**. *Int J Ment Heal Syst* (2022.0) **16** 43. DOI: 10.1186/s13033-022-00553-8
31. Caffery LJ, Muurlink OT, Taylor-Robinson AW. **Survival of rural telehealth services post-pandemic in Australia: a call to retain the gains in the ‘new normal’**. *Aust J Rural Health* (2022.0) **30** 544-549. DOI: 10.1111/ajr.12877
32. 32.Local Government Association and Public Health EnglandHealth and wellbeing in rural areas2017LondonLocal Government Association. *Health and wellbeing in rural areas* (2017.0)
33. 33.Hart J. Older people in rural areas: vulnerability due to loneliness and isolation paper. Shropshire: Rural England; 2016. https://ruralengland.org/wp-content/uploads/2016/04/Final-report-Loneliness-and-Isolation.pdf.
34. 34.United Kingdom GovernmentStatement of rural research priorities2019LondonUnited Kingdom Government. *Statement of rural research priorities* (2019.0)
35. Moore H, Siriwardena A, Gussy M. **Mental health emergencies and COVID-19: the impact of ‘lockdown’ in the east midlands of the UK**. *BJPsych Open* (2021.0) **7** e139. DOI: 10.1192/bjo.2021.973
36. Bird W. **Improving health in coastal communities**. *Br Med J* (2021.0) **374** n2214. DOI: 10.1136/bmj.n2214
37. Clark DM. **Implementing NICE guidelines for the psychological treatment of depression and anxiety disorders: the IAPT experience**. *Int Rev Psych* (2011.0) **23** 318-327. DOI: 10.3109/09540261.2011.606803
38. Clark DM, Canvin L, Green J. **Transparency about the outcomes of mental health services (IAPT approach): an analysis of public data**. *Lancet* (2018.0) **391** 679-686. DOI: 10.1016/S0140-6736(17)32133-5
39. Wakefield S, Kellett S, Simmonds-Buckley M. **Improving access to psychological therapies (IAPT) in the United Kingdom: a systematic review and meta-analysis of 10-years of practice-based evidence**. *Br J Clin Psychol* (2021.0) **60** e12259. DOI: 10.1111/bjc.12259
40. Cromarty P, Drummond A, Francis T. **NewAccess for depression and anxiety: adapting the UK improving access to psychological therapies program across Australia**. *Australasian Psychiatry* (2016.0) **24** 489-492. DOI: 10.1177/1039856216641310
41. Naeem F, Pikard J, Rao S. **Is it possible to provide low-intensity cognitive behavioral treatment (CBT lite) in Canada without additional costs to the health system? First-year evaluation of a pilot CBT lite program**. *Int J Ment Health* (2017.0) **46** 253-268. DOI: 10.1080/00207411.2017.1345039
42. Kobori O, Nakazato M, Yoshinaga N. **Transporting cognitive behavioral therapy (CBT) and the improving access to psychological therapies (IAPT) project to Japan: preliminary observations and service evaluation in Chiba**. *J Ment Health Train Educ Pract* (2014.0) **9** 155-166. DOI: 10.1108/JMHTEP-10-2013-0033
43. Knapstad M, Nordgreen T, Smith ORF. **Prompt mental health care, the Norwegian version of IAPT: clinical outcomes and predictors of change in a multicenter cohort study**. *BMC Psychiatry* (2018.0) **18** 260. DOI: 10.1186/s12888-018-1838-0
44. 44.Qualtrics. Provo, Utah: https://www.qualtrics.com, 2022.
45. Marshall D, Quinn C, Child S. **What IAPT services can learn from those who do not attend**. *J Ment Health* (2016.0) **25** 410-415. DOI: 10.3109/09638237.2015.1101057
46. Chan SWY, Adams M. **Service Use, Drop-Out Rate and Clinical Outcomes: A Comparison Between High and Low Intensity Treatments in an IAPT Service**. *Behav Cogn Psychother* (2014.0) **42** 747-759. DOI: 10.1017/S1352465813000544
47. Meisel SF, Drury H, Perera-Delcourt RP. **Therapists’ attitudes to offering eCBT in an inner-city IAPT service: a survey study**. *Cogn Behav Therap* (2018.0) **11** e11. DOI: 10.1017/S1754470X18000107
48. Cole CL, Waterman S, Stott J. **Adapting IAPT services to support frontline NHS staff during the Covid-19 pandemic: the Homerton Covid Psychological Support (HCPS) pathway**. *Cogn Behav Therap* (2020.0) **13** e12. DOI: 10.1017/S1754470X20000148
49. Uscher-Pines L, Sousa J, Raja P. **Suddenly becoming a “virtual doctor”: experiences of psychiatrists transitioning to telemedicine during the COVID-19 pandemic**. *Psychiatr Serv* (2020.0) **71** 1143-1150. DOI: 10.1176/appi.ps.202000250
50. Uscher-Pines L, Raja PV, Qureshi N. **Use of Tele-mental health in conjunction with in-person care: a qualitative exploration of implementation models**. *Psychiatr Serv* (2020.0) **71** 419-426. DOI: 10.1176/appi.ps.201900386
51. Kim H-Y. **Statistical notes for clinical researchers: chi-squared test and Fisher’s exact test**. *Restor Dent Endod* (2017.0) **42** 152-155. DOI: 10.5395/rde.2017.42.2.152
52. 52.R Core Team and R Foundation for Statistical Computing. R: a language and environment for statistical computing Vienna, Austria: https://www.R-project.org/, 2021.
53. Brewster L, Mountain G, Wessels B. **Factors affecting front line staff acceptance of telehealth technologies: a mixed-method systematic review**. *J Adv Nurs* (2014.0) **70** 21-33. DOI: 10.1111/jan.12196
54. Gajarawala SN, Pelkowski JN. **Telehealth benefits and barriers**. *J Nurse Pract* (2021.0) **17** 218-221. DOI: 10.1016/j.nurpra.2020.09.013
55. Moffatt JJ, Eley DS. **The reported benefits of telehealth for rural Australians**. *Aust Health Rev* (2010.0) **34** 276-281. DOI: 10.1071/AH09794
56. Foley T, Woollard J. *The digital future of mental healthcare and its workforce: a report on a mental health stakeholder engagement to inform the Topol review* (2019.0)
57. Reeve C, Johnston K, Young L. **Health profession education in remote or geographically isolated settings: a scoping review**. *J Med Educ Curric Dev* (2020.0) **7** 1-12. DOI: 10.1177/2382120520943595
|
---
title: 'The Efficacy of Focused Extracorporeal Shock Wave Therapy and Ultrasound Therapy
in the Treatment of Calcar Calcanei: A Randomized Study'
authors:
- Ivana Topalović
- Dejan Nešić
- Sindi Mitrović
- Vera Miler Jerković
- Ljubica Konstantinović
journal: BioMed Research International
year: 2023
pmcid: PMC9970705
doi: 10.1155/2023/8855687
license: CC BY 4.0
---
# The Efficacy of Focused Extracorporeal Shock Wave Therapy and Ultrasound Therapy in the Treatment of Calcar Calcanei: A Randomized Study
## Abstract
The prospective, simple randomized study assesses the effect of focused extracorporeal shock wave therapy (f-ESWT) on pain intensity and calcification size compared to the application of ultrasound physical therapy in treating patients with calcar calcanei. A total of 124 patients diagnosed with calcar calcanei were consecutively included in the study. The patients were divided into two groups: the experimental group ($$n = 62$$), which included the patients treated with f-ECWT, and the control group ($$n = 62$$), consisting of patients treated with the standard ultrasound therapy method. The experimental group's patients received ten therapy applications spaced seven days apart. The patients in the control group had ten ultrasound treatments on ten consecutive days over two weeks. All patients in both groups were tested using the Visual Analog Scale (VAS) to measure pain intensity before the beginning and at the end of treatment. The size of the calcification was assessed in all patients. The study hypothesizes that f-ESWT reduces the pain and the size of the calcification. Pain intensity reduction was registered in all patients. The calcification size in patients in the experimental group was reduced from the initial range of 2 mm–15 mm, to a content of 0.0 mm–6.2 mm. The calcification size in the control group ranged from 1.2 to 7.5 mm, without any change. None of the patients experienced any adverse reactions to the therapy. Patients treated with standard ultrasound therapy did not have a statistically significant reduction in the calcification size. In contrast, the patients in the experimental group treated with f-ESWT showed a substantial decrease in the calcification size.
## 1. Introduction
Calcar calcanei is one of the most frequent causes of foot pain in adults (15–$20\%$) [1]. The main symptom of calcar calcanei is heel pain during weight-bearing activities [2]. According to literature data, this disease occurs more frequently in women than men. Calcification occurs most commonly if the footwear is uncomfortable, if the patient is obese, or due to a preexisting foot deformity [3, 4].
Calcar calcanei is a disease manifested by the presence of calcification on the calcaneus. Histological analysis shows that the calcar calcanei consists of a core of mature lamellar bone and demonstrates evidence of degeneration and fibro-cartilaginous proliferation, along with one or more intramembranous, chondroid, and endochondral ossifications occurring at the surface [5]. Calcification is most often found at the point of attachment of the plantar fascia (PF) to the calcaneus, and it is closely related to plantar fasciitis (PFis). Excessive stresses lead to a chronic degenerative condition of PFis that is histologically characterized by fibroblastic hypertrophy, absence of inflammatory cells, disorganized collagen, and chaotic vascular hyperplasia with zones of avascularity and even ossification at the fascial insertion point on the calcaneus [6, 7]. The lack of elasticity of the affected PF due to degenerative processes could be demonstrated using shear wave elastography (SWE) [8].
Its treatment is multimodal and may include the following: the application of anti-inflammatory drugs, physical therapy, surgical treatment, kinesiotherapy, and the use of an orthosis. Generally speaking, there is no proof of the efficiency of any particular method, apart from glucocorticoid infiltration, which is associated with possible atrophy of the tissue in the foot heel or rupture of the plantar fascia [9].
Nowadays, in the literature, there is numerous evidence of the use of f-ESWT in the treatment of plantar fasciitis without or with (calcar calcanei) calcification. [ 10–13]. The effects of treatment with f-ESWT, which are exploited in this context, are the indirect effects of increased inflammatory indices, neovascularization, and tissue regeneration. The final effect of this treatment is the return of the biological and functional properties of the tissue.
The effect of the application of ultrasound therapy is twofold. The thermal effect manifests as a local increase in the temperature of the tissue located at the site of therapy application. The mechanical effect leads to micro-massage of the tissue being treated, resulting in increased permeability of cell membranes. This therapy has the effect of reducing swelling and tension in soft tissues, reducing pain and inflammation, and improving sensation.
Research on the efficiency of safe physical therapy methods is an important current issue. Before introducing shock wave therapy into clinical practice, different forms of therapy were applied in treating calcifications. The application of different forms of therapy used thus far, including the local administration of corticosteroids, results in reduced swelling of the soft tissue of the heel and sole of the foot and reduced pain. Surgical removal of the calcification may lead to side effects in the soft tissue of the heel [14]. f-ESWT therapy results in the reduction or complete disappearance of the calcification, while the level of the patient's discomfort experienced during application is the lowest compared to other treatment options [15].
Standard ultrasound therapy in calcar calcanei was the treatment of choice before introducing ESWT [16]. When ultrasound and shock wave therapy are compared, the latter has the advantage, as it improves the quality of life and brings about the breaking up of the calcification, in addition to reducing pain [17].
The study aims to determine the effect of f-ESWT on pain intensity and the calcification size in calcar calcanei compared to the application of standard ultrasound therapy.
## 2. Methods
The protocol applied in this study is based on the principles of the Helsinki Declaration. The study was approved by the Faculty of Medicine Ethics Committee of Belgrade University (No. 213). All of the patients involved in the study submitted written consent forms for inclusion in the study. The research was performed at the Clinic for Rehabilitation dr Miroslav Zotovic. The study was carried out between September 1, 2021, and September 31, 2022.
The study included 124 consecutive patients. With the method of equal randomization, the patients were divided into two groups. The first group was designated as the experimental group and had 62 patients treated with f-ESWT. The second group was selected as the control group; it included 62 patients treated with standard ultrasound therapy.
## 3. Criteria
A total of 190 patients were considered for the study, of whom 124 fulfilled the criteria for inclusion. The primary criterion for including patients in the study was diagnostically confirmed painful calcar calcanei diagnosed with X-rays up to six months before. Patients with the following contraindications were excluded from the study: coagulation disorder, anticoagulant use, carcinoma, pregnancy, sensory polyneuropathy, osteoporosis, a pacemaker, acute conditions, lesions of the sole skin, and cortisone treatment up to six weeks before the initial treatment session. Statistical analysis was performed on an “intention to treat” basis, and dropout was not registered.
## 4. The Procedure
The shock wave therapy machine used was the Masterpuls MP 200, serial No: BS 2058, manufactured in 2011 by STORZ MEDICAL, Switzerland. Shock wave therapy was applied locally, on the heel of the foot, at the site of the calcar calcanei. The application was performed with a focused probe with the following parameters: 16 Hz frequency, 1,600 shocks, and a pressure of 1.6 bars. The patients lay prone, and a cylinder was placed under their lower leg so the heel would be completely relaxed and more accessible to the doctor performing the therapy. The f-ESWT was applied once a week for ten weeks.
The patients in the control group underwent a series of ten physical ultrasound therapy treatments continuously every day. The machine applied was the SONOPULS 492 Enraf-Nonius. The application of standard ultrasound therapy was performed using the stable method, directly on the heel of the foot and the plantar fascia, for five minutes at an intensity ranging between 0.5 and 0.8 W/cm2. The patients were positioned in the same way as for shock wave therapy application. A licensed physician administered shock wave therapy. A licensed physical therapist administered standard ultrasound treatment to the patients in the control group according to the protocol defined by a physiatrist. None of the patients treated in either of the groups took any medication belonging to the group of analgesics or non-steroid antirheumatics, nor were they treated with corticosteroid drugs during the study. Additionally, none used an orthosis for their heel or applied any gels locally.
## 5. Outcome Parameters
The pain intensity was measured using the Visual Analog Scale (VAS) [18]. All patients in both groups were tested with the VAS scale to measure pain at the beginning and end of treatment. The Visual Analog Scale (VAS) is based on the analysis of pain when the patient feels no pain and is experiencing significant pain. As a diagnostic method, this scale has excellent potential for pain assessment and is the most commonly used one-dimensional pain rating scale in clinical practice. The enumeration of the scale enables the numerical quantification of pain intensity, ranging from 0 to 10. The VAS scale can be used repeatedly in the same patient to follow up on the pain evaluation during the treatment. In this study, the VAS scale was used for each patient, with a range of 0 to 10, whereby 10 represented the highest pain intensity and 0 represented no pain. The patients were tested with the VAS scale twice, at the beginning of the course of treatment at the end.
The calcification size was measured in millimeters with X-ray diagnostics. Imaging the calcar calcanei was performed by applying the standard X-ray technique for the ankle, with the patient standing upright and the X-ray beam spread vertically. Thus, the images show all the elements of the ankle (distal tibia, talus, and calcaneus). The size and angle of each abnormality of the calcaneus, which was registered as a calcar calcanei, were measured. The measurement of the size of the calcar calcanei was performed via the calcaneal inclination angle (CIA), the lateral talocalcaneal angle (LTCA), the Böhler angle (BA), and the Gissane angle [19, 20].
All the results were statistically processed with the IBM SPSS Statistics 22 (SPSS Inc., Chicago, IL, USA) software package.
## 5.1. Statistical Analysis
Descriptive statistics were used to describe data. The continuous variables are presented by mean ± standard deviation, while the categorical variables are presented by count (percentage).
The comparison between independent variables, which distributions were approximately the same as a normal distribution, we were done by the t-test for independent data in the case of equal variance or by the Welch t-test in the case of unequal variances. The Mann–Whitney test compares independent variables, which distributions are not the same as a normal distribution. The comparison between dependent variables, which distributions were approximately the same as a normal distribution, we were done by the t-test for dependent data. The comparison between dependent variables, which distributions were not the same as a normal distribution, we were done by the Wilcoxon test. The testing of the normality of the distribution of each variable was done by using descriptive statistics, Kolmogorov–Smirnov, and Shapiro–Wilks and graphical methods. The Cohen effect size [21] was calculated for any test. According to Cohen [16], the interpretation for a small, moderate, and large Cohen effect size is $d = 0.20$, 0.50, and 0.80. The Cliff's d effect size [22] was used for the Mann–Whitney test. According to Cliff [23], this effect size is appropriate for comparing two groups. The interpretation of Cliff's d for small, medium, and large is 0.15, 0.33, and 0.48. The rank biserial correlation coefficient rs [22] was used as the effect size for the Wilcoxon test. The interpretation for the effect size for small, medium, and large is 0.10, 0.30, and 0.50. The categorical variables were compared by using the Chi-Squared test. The Cramér's V [24] effect size was calculated for the Chi-Squared test. The interpretation for Cramer V effect size for small, medium, and large is 0.10, 0.30, and 0.50.
The relationship between variables for each group was analyzed separately by Pearson's correlations. The relationship between the two variables was analyzed by controlling the third variable using partial correlation. The partial correlation coefficient measures the strength of the linear relationship between two variables after entirely controlling for the effects of other variables. The logistic regression was performed for changes in a specific outcome.
The level of statistical significance was set at a two-tailed alpha level of 0.05. RStudio (version 1.4.1106) was used for statistical analysis.
## 5.2. Results
One hundred and ninety consecutive patients with diagnostically confirmed painful calcar calcanei were screened for eligibility criteria. One hundred and twenty-four patients fulfilled all eligibility criteria. Eighteen patients had sensory polyneuropathies, a lesion of the sole skin had ten participants, cortisone treatment up to six weeks before the initial treatment session had eight, eleven patients were medically unstable, and nineteen declined to participate in the clinical trial. The patients were divided by the method of equal randomization into two groups. The first group was designated as the experimental group and had 62 patients treated with shock wave therapy. The second group was selected as the control group; it included 62 patients treated with standard ultrasound therapy. Figure 1 shows a flow diagram of patient recruitment throughout the study.
Descriptive statistics of all outcomes for both groups are presented in Table 1.
The results of the analysis of comparisons between experimental and control groups are presented in Table 2. The statistically significant differences in baseline outcomes between the experimental group and control group were found at therapy duration ($$p \leq 0.005$$) with large effect size and calcification posttherapy ($p \leq 0.0001$) and medium effect size.
The results of the ‘pre-post' therapy analysis are presented in Table 3. The statistically significant differences were obtained in both groups at parameters calcification and VAS with large effect sizes.
Calcification and VAS change were calculated as subtraction, pretherapy, posttherapy, and VAS pretherapy and VAS posttherapy. The results of comparing changes in calcification and VAS, separately between the experimental and control groups are presented in Table 4. A statistically significant difference was obtained at calcification between subjects from the experimental and control group ($p \leq 0.0001$) with a large effect size.
Additionally, we observed the change in the VAS as greatly improved (> = 50), much improved (50 < ×< = 30), somewhat improved (30 < ×< = 10), about the same (10 < ×< = 1), and worse (>1). The change in VAS in both groups is presented in Table 5. The result (χ2 = 2.37, $$p \leq 0.864$$) was shown that there is no statistically significant difference between VAS improvement between experimental and control groups.
Additionally, we observed the change in the calcification as improved (>0), not improved (<=0)—the change of calcification in both groups presented in Table 6. The result (χ2 = 15.34, $$p \leq 0.0005$$) showed a statistically significant difference in calcification improvement between experimental and control groups.
The logistic regression was done for calcification improvement, and the results are presented in Table 7. The odds for subjects in the experimental group to have calcification improvement are 21.2 higher than for subjects in the control group.
The relationship between variables for each group was analyzed separately by Pearson's correlations. The results are shown in Table 8. The significant correlations ($p \leq 0.05$) were marked in bold italic. In the experimental group, the positive correlation between calcification change and age is moderate ($r = 0.36$, $$p \leq 0.004$$) and shows that improvement in calcification after therapy is more significant for older subjects; the positive correlation between calcification change and BMI is moderate ($r = 0.31$, $$p \leq 0.01$$) and shows that improvement in calcification after therapy is more significant for subjects with bigger BMI. In the control group: the positive correlation between calcification change and BMI is a moderate degree ($r = 0.38$, $$p \leq 0.002$$) and shows that improvement in calcification after therapy is more significant for subjects with bigger BMI; the negative correlation between VAS change and BMI is a low degree (r = −0.26, $$p \leq 0.04$$) and shows that improvement in VAS after therapy is more significant for subjects with lower BMI.
The analysis of the relationship between two variables, calcification changes and BMI, while controlling by the third variable, age, was done using the partial correlation. The partial correlation was statistically significant and positive with a small degree ($r = 0.26$, $$p \leq 0.01$$). The controlling variable, age, had little effect on the relationship between calcification changes and BMI.
The analysis of the relationship between two variables, calcification changes and age, while controlled by the third variable, BMI, was done using the partial correlation. The partial correlation was statistically significant and positive with a medium degree ($r = 0.32$, $$p \leq 0.01$$). The controlling variable, BMI, had little effect on the relationship between calcification changes and age.
The analysis of the relationship between two variables, calcification change and VAS change, but controlling a third variable (age, BMI, and therapy duration, separately) was done using the partial correlation. The partial correlation coefficient measures the strength of the linear relationship between two variables after entirely controlling for the effects of other variables. The results are presented in Table 9.
## 6. Discussion
The most prominent symptom of calcar calcanei is pain. In our study, none of the patients from either of the groups took any analgesics or nonsteroid antirheumatics. Analyzing the results obtained in treating the patients in the experimental and control groups, we received statistically similar data regarding the decrease in pain levels in patients. The application of f-ECWT in the experimental group of patients showed a demonstrable reduction in the calcification size and a decrease in pain. Patients from the control group treated with standard ultrasound therapy experienced a positive effect regarding pain intensity reduction. By analyzing the calcification size in millimeters, in patients of both tested groups, a result was obtained demonstrating the effect of f-ESWT on reducing the calcification. Correlation analyses did not show a connection between the size of calcification and pain, but some factors, such as BMI and the age of patients, moderate the therapeutic effects. The results show that the improvement in calcification size is more significant for the elderly and subjects with a higher BMI. There is no data in the literature on the moderating effect of other factors related to patient characteristics. Side effects of the application of shock wave therapy were not registered. Similarly, there were no registered side effects among the patients in the control group.
Many studies confirm that the f-ESWT method is noninvasive, safe, and without severe contraindications for breaking up calcifications [25–27]. The experience of various authors is that they used f-ESWT at an intensity ranging between 10 and 21 bars, a frequency of 20 Hz, and some shocks ranging from two to four thousand [28]. The suggestions regarding the frequency of f-ESWT application, i.e., the length of time that should elapse between two applications of shock wave therapy, vary in the literature.
The complete biological mechanism of shock wave therapy has yet to be precisely understood. It is believed that ESWT promotes healing through the process of mechano-transduction, acting as a mechanical stimulus [29, 30]. The studies have suggested that mechano-transduction is the central mechanism whereby ESWT triggers angiogenic and bone remodeling responses at cellular and molecular levels, generating beneficial therapeutic effects such as pain relief and tissue regeneration by stimulating vascularization and reducing calcium deposits in tissues [31]. Additionally, it may also alleviate pain through hyperstimulation analgesia [32]. Applying ESWT stimulates the proliferation, migration, and differentiation of stem cells, fibroblasts, tenocytes, bone cells, and their precursors [33–35]. To understand the effects shown in this study, studies that show the direct influence of periosteal stimulation on orthotopic bone regeneration and the reduction of osteoclast activity through the inhibition of pro-osteoclastogenic factors could be considered. [ 36]. The research results on tenocyte activity indicate increased production of collagen and decreased expression of metalloproteinases and inflammatory cytokines, essential in remodeling plantar fasciitis [37, 38]. Some of the biological responses on a molecular level include increased early expression of vascular endothelial growth factor (VEGF), endothelial nitric oxide synthase (eNOS), and proliferating cell nuclear antigen (PCNA), which leads to neovascularization and better blood supply. [ 39, 40]. At the cellular level, it has been shown that ESW can induce transient deformation of some cytoskeletal proteins (actin and tubulin, but not vimentin), but reorganization into their original cytoskeletal network appeared within 3 hours, with a pattern similar to the control [41]. However, the primary biological effect of ESWT is stimulative. In recent years, the interest of researchers in studying the regenerative effect of shock wave therapy has increased [31].
It is assumed that the application of f-ESWT induced changes at the cellular and molecular level, resulting in a reduction in the calcification size. An objective reduction in calcification size was also statistically proven in this study.
## 7. Study Limitations
The result should be considered against a few limitations. The patients are not homogeneous in the duration of symptoms and have not been monitored for a long time. A limitation is the lack of determination of the sample size since the patients were determined consecutively. Also, many factors other than BMI that could influence the effects of therapy were not monitored.
## 8. Conclusion
In conclusion, both types of therapy applied in the study contributed to reducing pain intensity. The application of f-ESWT significantly reduced the calcification size compared with standard ultrasound therapy.
## Data Availability
Data supporting the study results can be found in the medical data system of the Clinic for Rehabilitation dr Miroslav Zotovic.
## Conflicts of Interest
The authors have no conflict of interest to declare.
## References
1. Toomey E. P.. **Plantar heel pain**. (2009) **14** 229-245. DOI: 10.1016/j.fcl.2009.02.001
2. Rosenbaum A. J., DiPreta J. A., Misener D.. **Plantar heel pain**. (2014) **98** 339-352. DOI: 10.1016/j.mcna.2013.10.009
3. Irving D. B., Cook J. L., Young M. A., Menz H. B.. **Obesity and pronated foot type may increase the risk of chronic plantar heel pain: a matched case-control study**. (2007) **8**. DOI: 10.1186/1471-2474-8-41
4. Sullivan J., Pappas E., Burns J.. **Role of mechanical factors in the clinical presentation of plantar heel pain: implications for management**. (2020) **42, article 101636**. DOI: 10.1016/j.foot.2019.08.007
5. Kirkpatrick J., Yassaie O., Mirjalili S. A.. **The plantar calcaneal spur: a review of anatomy, histology, etiology and key associations**. (2017) **230** 743-751. DOI: 10.1111/joa.12607
6. Snider M. P., Clancy W. G., McBeath A. A.. **Plantar fascia release for chronic plantar fasciitis in runners**. (1983) **11** 215-219. DOI: 10.1177/036354658301100406
7. Wearing S. C., Smeathers J. E., Urry S. R., Hennig E. M., Hills A. P.. **The pathomechanics of plantar fasciitis**. (2006) **36** 585-611. DOI: 10.2165/00007256-200636070-00004
8. Schillizzi G., Alviti F., D'Ercole C.. **Evaluation of plantar fasciopathy shear wave elastography: a comparison between patients and healthy subjects**. (2021) **24** 417-422. DOI: 10.1007/s40477-020-00474-7
9. Li S., Wang K., Sun H.. **Clinical effects of extracorporeal shock-wave therapy and ultrasound-guided local corticosteroid injections for plantar fasciitis in adults: a meta-analysis of randomized controlled trials**. (2018) **97**. DOI: 10.1097/MD.0000000000013687
10. Dizon J. N., Gonzalez-Suarez C., Zamora M. T., Gambito E. D.. **RE**. (2014) **93** p. 735. DOI: 10.1097/PHM.0000000000000121
11. Hammer D. S., Rupp S., Kreutz A., Pape D., Kohn D., Seil R.. **Extracorporeal shockwave therapy (ESWT) in patients with chronic proximal plantar fasciitis**. (2002) **23** 309-313. DOI: 10.1177/107110070202300403
12. Krukowska J., Wrona J., Sienkiewicz M., Czernicki J.. **A comparative analysis of analgesic efficacy of ultrasound and shock wave therapy in the treatment of patients with inflammation of the attachment of the plantar fascia in the course of calcaneal spurs**. (2016) **136** 1289-1296. DOI: 10.1007/s00402-016-2503-z
13. Fansa A., Talsania A. J., Kennedy J. G., O'Malley M. J.. **Efficacy of unfocused medium-intensity extracorporeal shock wave therapy (MI- ESWT) for plantar fasciitis**. (2021) **60** 471-476. DOI: 10.1053/j.jfas.2020.08.027
14. Lou J., Wang S., Liu S., Xing G.. **Effectiveness of extracorporeal shock wave therapy without local anesthesia in patients with recalcitrant plantar Fasciitis**. (2017) **96** 529-534. DOI: 10.1097/PHM.0000000000000666
15. Feichtinger X., Heimel P., Tangl S.. **Improved biomechanics in experimental chronic rotator cuff repair after shockwaves is not reflected by bone microarchitecture**. (2022) **17**. DOI: 10.1371/journal.pone.0262294
16. Castro B. K. C., Corrêa F. G., Maia L. B., Oliveira V. C.. **Effectiveness of conservative therapy in tendinopathy-related shoulder pain: a systematic review of randomized controlled trials**. (2021) **49** 15-20. DOI: 10.1016/j.ptsp.2021.01.010
17. Yin M. C., Ye J., Yao M.. **Is extracorporeal shock wave therapy clinical efficacy for relief of chronic, recalcitrant plantar fasciitis? A systematic review and meta-analysis of randomized placebo or active-treatment controlled trials**. (2014) **95** 1585-1593. DOI: 10.1016/j.apmr.2014.01.033
18. Collins S. L., Moore R. A., McQuay H. J.. **The visual analogue pain intensity scale: what is moderate pain in millimetres?**. (1997) **72** 95-97. DOI: 10.1016/S0304-3959(97)00005-5
19. Thomas J. L., Kunkel M. W., Lopez R., Sparks D.. **Radiographic values of the adult foot in a standardized population**. (2006) **45** 3-12. DOI: 10.1053/j.jfas.2005.10.014
20. Başdelioğlu K.. **Radiologic and demographic characteristics of patients with plantar calcaneal spur**. (2021) **60** 51-54. DOI: 10.1053/j.jfas.2020.06.016
21. Cohen J.. (1988)
22. King B. M., Rosopa P., Minimum E. W.. (2000)
23. Cliff N.. **Dominance statistics: ordinal analyses to answer ordinal questions**. (1993) **114** 494-509. DOI: 10.1037/0033-2909.114.3.494
24. Wagner W. E., Gillespie B. J.. **Using and interpreting statistics in the social, behavioral, and health sciences**. (2019). DOI: 10.4135/9781071814284
25. Schroeder A. N., Tenforde A. S., Jelsing E. J.. **Extracorporeal shockwave therapy in the management of sports medicine injuries**. (2021) **20** 298-305. DOI: 10.1249/JSR.0000000000000851
26. Wheeler P. C., Dudson C., Calver R.. **Three sessions of radial extracorporeal shockwave therapy gives no additional benefit over "minimal-dose" radial extracorporeal shockwave therapy for patients with chronic greater trochanteric pain syndrome: a double-blinded, randomized, controlled trial**. (2022) **32** e7-e18. DOI: 10.1097/JSM.0000000000000880
27. Abo Al-Khair M. A., El Khouly R. M., Khodair S. A., Al Sattar Elsergany M. A., Hussein M. I., Eldin Mowafy M. E.. **Focused, radial and combined shock wave therapy in treatment of calcific shoulder tendinopathy**. (2021) **49** 480-487. DOI: 10.1080/00913847.2020.1856633
28. Dedes V., Stergioulas A., Kipreos G., Dede A. M., Mitseas A., Panoutsopoulos G. I.. **Effectiveness and safety of shockwave therapy in tendinopathies**. (2018) **30** 141-146. DOI: 10.5455/msm.2018.30.141-146
29. Visco V., Vulpiani M. C., Torrisi M. R., Ferretti A., Pavan A., Vetrano M.. **Experimental studies on the biological effects of extracorporeal shock wave therapy on tendon models. A review of the literature**. (2014) **4** 357-361. DOI: 10.32098/mltj.03.2014.15
30. Ingber D. E.. **Cellular mechanotransduction: putting all the pieces together again**. (2006) **20** 811-827. DOI: 10.1096/fj.05-5424rev
31. Agostino D., Frairia R., Romeo P.. **Extracorporeal shockwaves as regenerative therapy in orthopedic traumatology: a narrative review from basic research to clinical practice**. (2016) **30** 323-332. PMID: 27358117
32. Hausdorf J., Lemmens M. A., Heck K. D.. **Selective loss of unmyelinated nerve fibers after extracorporeal shockwave application to the musculoskeletal system**. (2008) **155** 138-144. DOI: 10.1016/j.neuroscience.2008.03.062
33. Wuerfel T., Schmitz C., Jokinen L. L. J.. **The effects of the exposure of musculoskeletal tissue to extracorporeal shock waves**. (2022) **10** p. 1084. DOI: 10.3390/biomedicines10051084
34. Hausdorf J., Sievers B., Schmitt-Sody M., Jansson V., Maier M., Mayer-Wagner S.. **Stimulation of bone growth factor synthesis in human osteoblasts and fibroblasts after extracorporeal shock wave application**. (2011) **131** 303-309. DOI: 10.1007/s00402-010-1166-4
35. Leone L., Raffa S., Vetrano M.. **Extracorporeal shock wave treatment (ESWT) enhances the in vitro-induced differentiation of human tendon-derived stem/progenitor cells (hTSPCs)**. (2016) **7** 6410-6423. DOI: 10.18632/oncotarget.7064
36. Kearney C. J., Hsu H. P., Spector M.. **The use of extracorporeal shock wave-stimulated periosteal cells for orthotopic bone generation**. (2012) **18** 1500-1508. DOI: 10.1089/ten.tea.2011.0573
37. Vetrano M., d'Alessandro F., Torrisi M. R., Ferretti A., Vulpiani M. C., Visco V.. **Extracorporeal shock wave therapy promotes cell proliferation and collagen synthesis of primary cultured human tenocytes**. (2011) **19** 2159-2168. DOI: 10.1007/s00167-011-1534-9
38. Chao Y. H., Tsuang Y. H., Sun J. S.. **Effects of shock waves on tenocyte proliferation and extracellular matrix metabolism**. (2008) **34** 841-852. DOI: 10.1016/j.ultrasmedbio.2007.11.002
39. Yu T., Junger W. G., Yuan C.. **Shockwaves increase T-cell proliferation and IL-2 expression through ATP release, P2X7 receptors, and FAK activation**. (2010) **298** C457-C464. DOI: 10.1152/ajpcell.00342.2009
40. Wang F. S., Wang C. J., Chen Y. J.. **Ras induction of superoxide activates ERK-dependent angiogenic transcription factor HIF-1**. (2004) **279** 10331-10337. DOI: 10.1074/jbc.M308013200
41. Moosavi-Nejad S. F., Hosseini S. H., Satoh M., Takayama K.. **Shock wave induced cytoskeletal and morphological deformations in a human renal carcinoma cell line**. (2006) **97** 296-304. DOI: 10.1111/j.1349-7006.2006.00172.x
|
---
title: Synergistic Antibacterial Effect of Ethyl Acetate Fraction of Vernonia amygdalina
Delile Leaves with Tetracycline against Clinical Isolate Methicillin-Resistant Staphylococcus
aureus (MRSA) and Pseudomonas aeruginosa
authors:
- Denny Satria
- Urip Harahap
- Aminah Dalimunthe
- Abdi Wira Septama
- Triani Hertiani
- Nasri Nasri
journal: Advances in Pharmacological and Pharmaceutical Sciences
year: 2023
pmcid: PMC9970709
doi: 10.1155/2023/2259534
license: CC BY 4.0
---
# Synergistic Antibacterial Effect of Ethyl Acetate Fraction of Vernonia amygdalina Delile Leaves with Tetracycline against Clinical Isolate Methicillin-Resistant Staphylococcus aureus (MRSA) and Pseudomonas aeruginosa
## Abstract
Multidrug-resistant bacteria have raised global concern about the inability to fight deadly infectious diseases. Methicillin-resistant *Staphylococcus aureus* (MRSA) and *Pseudomonas aeruginosa* are the most common resistant bacteria that are causing hospital infections. The present study was undertaken to investigate the synergistic antibacterial effect of the ethyl acetate fraction of *Vernonia amygdalina* Delile leaves (EAFVA) with tetracycline against the clinical isolates MRSA and P. aeruginosa. Microdilution was used to establish the minimum inhibitory concentration (MIC). A checkerboard assay was conducted for the interaction effect. Bacteriolysis, staphyloxanthin, and a swarming motility assay were also investigated. EAFVA exhibited antibacterial activity against MRSA and P. aeruginosa with a MIC value of 125 μg/mL. Tetracycline showed antibacterial activity against MRSA and P. aeruginosa with MIC values of 15.62 and 31.25 μg/mL, respectively. The interaction between EAFVA and tetracycline showed a synergistic effect against MRSA and P. aeruginosa with a Fractional Inhibitory Concentration Index (FICI) of 0.375 and 0.31, respectively. The combination of EAFVA and tetracycline induced the alteration of MRSA and P. aeruginosa, leading to cell death. Moreover, EAFVA also inhibited the quorum sensing system in MRSA and P. aeruginosa. The results revealed that EAFVA enhanced the antibacterial activity of tetracycline against MRSA and P. aeruginosa. This extract also regulated the quorum sensing system in the tested bacteria.
## 1. Introduction
The problem of antibiotic resistance is prominent worldwide. Multidrug-resistant bacteria make fatal infectious diseases untreatable, causing global concern [1, 2]. Methicillin-resistant *Staphylococcus aureus* (MRSA) is currently the most common resistant bacterium, especially in hospitals [3]. MRSA resistance to various antibiotics, including β-lactams, is caused by the acquired mecA gene, which overexpresses efflux pump and produces a β-lactamase enzyme [4]. In the case of Gram-negative bacteria, *Pseudomonas aeruginosa* has become a great concern as fatal bacteria resistance, due to changes in its target enzymes [5].
Several steps have been put in place to combat bacterial resistance. Thus, new antibacterial agents with unique targets and mechanisms of action are urgently needed. It is expensive and time-consuming to test new molecules in humans to make sure they are safe and effective in treating disease without fostering resistance [6]. Combining typical antibiotics with agents that enhance antimicrobial properties has been suggested as a possible solution to these problems [7]. On the other hand, it may be difficult to discover new antibacterial agents, and when the chemical is employed in clinical settings, new resistance mechanisms will develop. Combining two or more antimicrobials to increase their efficiency against resistant infections is a unique technique for combating resistance.
Numerous plant-based chemical compounds have been identified as a significant source of novel antibacterial. Several studies have identified chemical components, including flavonoids, fatty acids, sesquiterpene lactones, and steroidal saponins [8]. In addition, plant-based compounds also possess various biological activities for pharmaceutical action, such as anti-inflammation, antimalaria, antitumor, antiobesity, and antioxidant [9–12].
The *Vernonia amygdalina* Delile species is a member of the Asteraceae family [13] and is native to West Africa. Nonetheless, V. amygdalina leaf extract exhibited several pharmacological effects, including antibacterial activity [14]. However, this plant extract's effect on tetracycline's antibacterial activity against MRSA and P. aeruginosa is unknown. Thus, the present study aimed to examine the synergistic effect of the ethyl acetate fraction of V. amygdalina (EAFVA) extract and tetracycline on selected clinical isolates. In addition, the combination's effect on membrane cells and the virulence factor was studied [15].
## 2.1. Chemicals and Media
Tetracycline was obtained from Sigma-Aldrich, United Kingdom, phosphate buffered saline (PBS), anhydrous sodium sulfate, and crystal violet were obtained from Sigma-Aldrich, United Kingdom, brain heart infusion was obtained from BHI, and agar was obtained from Becton, Dickinson & Company, Franklin Lakes, New Jersey, USA.
## 2.2. Bacterial Strains
The clinical isolates obtained from the MERO Foundation (Marine Education and Research Organization Foundation in Bali, Indonesia) were MRSA and P. aeruginosa.
## 2.3. Preparation of Ethyl Acetate Fraction of Vernonia Amygdalina Delile Leaves
500 grams of V. amygdalina Delile were taken. The reflux technique extracted air-dried and powdered leaves with 1:10 absolute ethanol for five hours. We collected the filtrate, evaporated it at a lower pressure to produce a viscous extract, and finally dried it in a water bath [16, 17]. Using the process of liquid-liquid extraction, ethanol extract was fractionated with ethyl acetate [18, 19].
## 2.4. Determination of Minimum Inhibitory Concentration
In order to determine the MIC, a microdilution test was performed. The bacterial suspension was present and subsequently added to a 96-well microplate containing a two-fold dilution of V. amygdalina Delile. The bacterium suspension was prepared and adjusted to 0.5 Mc. Farland was equivalent to turbidity 1 × 108 CFU/mL and then diluted using saline solution to generate 1 × 106 CFU/mL [20]. Ethyl acetate fraction (EAFVA) was added to the leaves and cultured for 24 hours at 37°C [21]. It was thought that the MIC was the lowest concentration at which growth could be stopped [22].
## 2.5. Checkerboard Assay
On the x-axis of the 96-well plate, twice as much brain heart infusion (BHI) was added to the EAFVA. On the y-axis, an antibiotic dilution was created that was twice as potent as the previous one. Afterward, each well was given a suspension of bacteria at a concentration of approximately 1 × 106 CFU/mL, and the mixture was incubated at 37°C for a full day. Calculating the Fractional Inhibitory Concentration Index (FICI) allowed us to investigate the interaction between EAFVA and antibiotics. The formula for calculating the FICI is as follows:[1]FICI=MIC of EAFVA or antibiotics in combinationMIC of EAFVA or antibiotics alone.
This study was conducted to find out how EAFVA and antibiotics affected the tested bacteria [23]. The FICI calculation yields synergy when the FICI is less than 0.5. On the other hand, it was additive if the FICI value was in the range of 0.5 to 1, indifferent if the FICI value was 1 to 4, or antagonistic if the FICI value was greater than 4 [24].
## 2.6. Loss of 260 nm Absorbing Material
With a modest modification, the release of the UV-absorbing materials was accomplished, as previously reported. The overnight culture of the studied bacteria was washed and resuspended in the saline solution. To reach a final count of approximately 5 × 107 CFU/mL, various concentrations of the substance were administered to the cell at 125 μg/mL of EAFVA, 15.62 μg/mL of Tetracycline, and the combination (15.62 μg/mL EAFVA + 3.9 μg/mL tetracycline).
Control cells were untreated. Each sample was incubated at 37°C for 24 hours, diluted with saline (1: 100), and filtered through a 0.2 μm pore-size membrane. Spectrophotometer UV-VIS measured 260 nm optical density. This exam was taken three times [25].
## 2.7. Bacteriolysis Activity
The cultured bacterial suspensions were one night old on BHI media. Detection of bacteriolytic activity by this approach had been reported. 125 μg/mL EAFVA, 31.25 μg/mL tetracycline, and the combination (7.8 μg/mL EAFVA + 7.8 μg/mL tetracycline) were added to the cells. The final cell concentration was 5 × 107 CFU/mL, whereas the control cells were not treated with the test concentration as a negative control. Then, they incubated and measured absorbance at OD 620 nm, which showed a decrease in absorbance. The calculated yield value was the percentage of absorbance versus the OD of 620 nm at 24 hours. The test was carried out three times with no errors [25].
## 2.8. Staphyloxanthin Assay
The ability of EAFVA to stop the production of the golden-yellow pigment and staphyloxanthin was being studied. Bacterial cultures were rejuvenated overnight in a lactose broth (LB) medium. Then the bacterial suspension was diluted in a ratio of 1: 100 in new LB media, which already contained EAFVA and MRSA. It was incubated at 37°C for 24 hours. EAFVA and negative control without treatment were centrifuged for 15 minutes at a speed of 10.000 rpm [26].
## 2.9. Swarming Motility Assay
Agar plates containing 24 mM CaCl2 were prepared for swarming agar plates (M8) containing 0.1 percent casamino acid, 0.5 percent glucose, and EAFVA. 20 mL of ready-made M8 media was put into each Petri dish. Let stand at room temperature until the consistency of the media solidified before using. 1 mL of the bacterial suspension cultured overnight was taken and then spun down at 6000 rpm for 3 minutes. Again, the cell pellet was spun down while suspended in 1 mL PBS, discarding the supernatant. This washing technique was carried out twice. Furthermore, bacteria were placed in the center of M8 medium and incubated for 10 minutes at room temperature [27].
## 2.10. Phytochemical Constituent Analysis of EAFVA with LC-HRMS
EAFVA phytochemical analysis using TSQ Exactive (Thermo) (LSIH, Universitas Brawijaya) in a gradient fashion at a flow rate of 40 L/min using Hypersil GOLD aQ 50 in a column 1 mm by 1.9 m; analysis time was 70 minutes. The flow rate for the mobile phase can be seen in Table 1. The compound finding from the analysis results was analyzed using the mzCloud software [15, 28].
## 2.11. Data Analysis
The data were presented as the mean value with a standard deviation (SD) and analyzed using SPSS v.22 software. All tests were repeated in triplicate.
## 3.1. Minimum Inhibitory Concentration (MIC)
As shown in Table 2, EAFVA inhibited MRSA and P. aeruginosa with MIC values of 125 μg/mL. Positive control tetracycline inhibited MRSA and P. aeruginosa with MIC values of 15.62 and 31.25 μg/mL, respectively. This result indicated that EAFVA had a moderate antibacterial effect against selected clinical isolates (see Table 3).
## 3.2. Checkerboard Assay
The checkerboard assay was used to determine how EAFVA and tetracycline affected the tested bacteria. As presented in Table 4, the combination of EAFVA (15.62 μg/mL) and tetracycline (3.9 μg/mL) produced a 0.375 Fractional Inhibitory Concentration Index (FICI) for synergy against MRSA. On the other hand, the EAFVA enhanced the antibacterial activity of tetracycline against selected clinical isolates with a synergistic effect. It was found that EAFVA reduced the concentration of tetracycline, which suppressed the growth of MRSA. In contrast, the synergistic effect of the combination of EAFVA (7.8 μg/mL) with tetracycline (7.8 μg/mL) on P. aeruginosa was calculated to be 0.31. This showed that EAFVA could reduce the concentration of P. aeruginosa by 16 times (see Figure 1 for isobologram result).
## 3.3. Loss of 260 nm Absorbing Material
There was leakage of cells (nucleic acid components) in MRSA (Figure 2) and P. aeruginosa (Figure 3) from the EAFVA-treated supernatant. Leakage in cells treated with EAFVA and antibiotics alone showed absorbance at 260 nm in contrast to the untreated negative control without treatment. However, in the combination of EAFVA and tetracycline antibiotics, the leakage between nucleic acids or the release at 260 nm was increased in MRSA and P. aeruginosa.
## 3.4. Bacteriolysis Activity
Based on the results of the bacteriolysis test, the percentage value of crystal violet absorption in treatment with tetracycline antibiotics with a concentration of 15.62 μg/mL with a negative control showed an almost comparable value (Figures 4 and 5). However, compared to the negative control and single treatment with EAFVA plus tetracycline, the test with EAFVA treatment had a higher percentage of crystal violet absorption. Better results were shown in the combination of EAFVA with tetracycline antibiotics in both MRSA and P. aeruginosa, showing the same better results in the combination against these two test microorganisms.
## 3.5. Staphyloxanthin Assay
The staphyloxanthin test showed that its production could be seen and observed, which was marked by the formation of a golden-yellow color. With varying concentrations of samples treated with EAFVA, there is a decrease in the production of staphyloxanthin (Figure 6). In Figure 6(e), which is a microtube with a negative control cell, a golden-yellow color can be seen. A golden-yellow color was formed on the microtube at the lowest test concentration, 15.625 μg/mL (Figure 6(d)). In comparison, the EAFVA test concentration is 125 μg/mL (Figure 6(a)) and does not show the formation of a golden-yellow color on the microtube, likewise with the treatment at other concentrations. The staphyloxanthin test results revealed that its production might be seen inspected due to its golden-yellow color. In varying doses, staphyloxanthin production is diminished in the cell pellets isolated from EAFVA-treated samples (Figure 6). On the microtube of the negative control (Figure 6(e)), a golden-yellow hue can be noticed. At the lowest test concentration of 15.625 μg/mL (Figure 6(d)), the formation of a golden-yellow color was observed in the microtubes. In contrast, the treatment with EAFVA at a concentration of 125 μg/mL (Figure 6(a)) neither did not show the formation of a golden-yellow color in the microtubes nor in the treatment with other concentrations (see Figure 6).
## 3.6. Swarming Motility Assay
Based on the results of swarming motility, there were bacterial cells growing in the center of the Petri dish. In the medium containing EAFVA, there was little growth in the middle of the medium containing bacterial cells.
## 3.7. Phytochemical Constituent Analysis of EAFVA with LC-HRMS
The phytochemical analysis of EAFVA with LC-HRMS shows 15 constituents (Table 5).
## 4. Discussion
The minimum effective concentration is the amount of an antimicrobial compound that can stop bacterial growth under certain conditions and in a certain amount of time [29]. The results showed that at a concentration of 125 μg/mL EAFVA, there was antibacterial activity that was significantly able to stop the growth of the two test bacteria (MRSA and P. aeruginosa). Based on its biological activity, EAFVA has phenolic molecules, especially in the flavonoid compound group. Due to this group of compounds, it is possible to have antibacterial activity. Flavonoids are a low-molecular-weight class of polyphenolic chemicals [30]. In another investigation, plant extracts containing rich flavonoids and pure flavonoids were tested to suppress the growth of pathogenic bacteria. Several mechanisms have been reported, such as those resulting from cell complexes with additional adhesion and developments in microbial inhibition [31]. EAFVA is highly effective against MRSA and P. aeruginosa [25].
Combining two types of antimicrobials has been described as one of the potential techniques to avoid the problem of antimicrobial resistance (AMR). The checkerboard assay was tested to evaluate the resulting interaction, namely, the synergistic interaction between the extract and commercial antibiotics [32]. In this test, the FICI was determined to prove the interaction between the EAFVA test sample and the tetracycline antibiotic. Thus, in combination, the MIC of EAFVA and Tetracycline decreased.
Apart from that, it shows a synergistic effect with no antagonism between EAFVA and tetracycline. Tetracycline works by attaching to the bacterial ribosome and engaging with the 30S subunit target of the 16S ribosomal binding. Tetracyclines impede translation by sterically interfering with the RNA transfer during elongation [33]. When used in conjunction with EAFVA, which was shown to contain flavonoid components such as luteolin, quercetin, apigenin, and other substances, it had a more significant impact than tetracycline alone. Tetracycline antibiotics and EAFVA work together synergistically, as evidenced by the increased effectiveness of combination therapy. In this mechanism, it is possible that EAFVA with Tetracycline has a different scenario where the site of action at the target is different. Secondary metabolites in EAFVA extract, such as flavonoids, possess a mechanism that can compromise the permeability of cell membranes. Whenever there is a disruption in the permeability of the cell membrane, it allows tetracycline to enter the cell and eventually occupy its working place. Previous studies reported a synergistic effect on nisin when mixed with an aldehyde from cinnamon. In contrast, nisin alone had modest antibacterial activity. When administered in combination, it provided better antibacterial activity, and the combination was considered successful [34].
In order to determine the mechanism of synergistic action between EAFVA and tetracycline, this study also focused on protein leakage and alteration of membrane cells. It was known that at a wavelength of 260 nm, the bacterial supernatant was measured, which was characterized by the absorption of absorbance and an increase in absorption; it indicated the occurrence of bacterial cell leakage or loss of nucleic acid material [35]. The combination between EAFVA and tetracycline was also studied. Compared with a single treatment, EAFVA and antibiotics with a combination of both showed an increase in absorbance at a wavelength of 260 nm. The result was confirmed using bacteriolysis activity. EAFVA and tetracycline altered MRSA and P. aeruginosa membrane cells by increasing crystal violet uptake. According to previous studies, it has been proven in previous reports that the treatment of microbiological infections with natural products combined with synthetic antibiotic mixtures can improve treatment and prevent microbiological resistance [36]. Combining *Eucalyptus camaldulensis* essential oil with polymyxin B antibiotics showed a synergistic effect on treatment-resistant Acinetobacter burmanni isolates [37]. Another study showed antiacne efficacy by combining two essential oils with tretinoin [38].
In addition, α-mangosteen and lawsone methyl ether also enhanced the antibacterial effect of ampicillin against several pathogens including MRSA by disrupting membrane cell permeability [39]. In another study, essential oils extracted from C. maculatum were used to inhibit the pathogens *Escherichia coli* DH5a and P. aeruginosa PAOI. The results of this study were similar to those obtained using the antibiotic colistin [40].
Another strategy to overcome antimicrobial resistance is with resistance restoration agents to detect antivirulence compounds. It allows the use of existing drugs and does not cause intense selection pressures that accelerate resistant colony growth [41]. As a virulence factor, S. aureus, specifically MRSA, can produce staphyloxanthin pigment [42]. The staphyloxanthin pigment distinguishes colonies of S. aureus from other staphylococci and Gram-positive bacteria. C30 gold keratinoid is a series of metabolic reactions embedded in a unique membrane in this pathogen known as staphyloxanthin [43]. Another study showed decreased staphyloxanthin pigment synthesis by L-ascorbyl 2,6-dipalmitate treatment with a more severe survival rate on whole blood analysis sensitivity testing on MRSA cells [44, 45].
Many types of bacteria in the laboratory show swarming, which is the movement of multiple cells with flagella on solid surfaces [46]. Biofilm production, colonization of plant interior and exterior surfaces, and pathogenicity or protective action on plant-associated bacteria can be significantly affected by the ability of organisms to swarm [47].
The antimicrobial effects of flavonoids are increasingly being recognized. In traditional medicine, crude plant extracts have been tested in vitro for antibacterial activity [48]. For example, flavonoids with antibacterial action have been isolated and their structures studied previously, for example, apigenin [49, 50], quercetin [51, 52], and others combined. The components of previous studies, such as apigenin, quercetin, luteolin, apigetrin, curom anine, and others, were in accordance with the results of the phytochemical constituent analysis of EAFVA with LC-HRMS in this study. At 78 μg/mL, the luteolin component was associated with an antibacterial action against Trueperella pyogenes [53].
## 5. Conclusion
The antibacterial effect of EFVA with tetracycline on clinical isolates (MRSA and P. aeruginosa) revealed the antibacterial mechanism and activity with a synergistic effect.
## Data Availability
Data sets used in this study are available from the corresponding author upon reasonable request.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## Authors' Contributions
Conceptualization was done by U.H. and D.S.; methodology was provided by A.W.S.; software was provided by NN.; validation was done by T.H and A.W.S; formal analysis was performed by NN.; investigation was done by U.H.; resources were provided by D.S.; data curation was performed by NN.; original draft preparation was done by D.S., NN, and A.W.S.; T.H and U.H. reviewed and edited the manuscript; visualization was done by A.W.S.; supervision was done by U.H.; project administration was done by D.S.; and funding acquisition was done by U.H. All authors have read and agreed to the published version of the manuscript.
## References
1. Horna G., Amaro C., Palacios A., Guerra H., Ruiz J.. **High frequency of the exoU+/exoS+ genotype associated with multidrug-resistant “high-risk clones” of**. (2019) **9** 1-13. PMID: 30626917
2. Haro G., Iksen I., Nasri N.. **Identification, characterization and antibacterial potential of probiotic lactic acid bacteria isolated from naniura (A traditional Batak fermented food from carp) against**. (2020) **13** 464-468
3. Lozano C., Fernández-Fernández R., Ruiz-Ripa L., Gómez P., Zarazaga M., Torres C.. **Human mecC-carrying MRSA: clinical implications and risk factors**. (2020) **8** p. 1615. DOI: 10.3390/microorganisms8101615
4. Brahma U., Sharma P., Murthy S.. **Decreased expression of femXAB genes and fnbp mediated biofilm pathways in OS-MRSA clinical isolates**. (2019) **9** 16028-8. DOI: 10.1038/s41598-019-52557-z
5. Fraile-Ribot P. A., Zamorano L., Orellana R.. **Activity of imipenem-relebactam against a large collection of**. (2020) **64** 021655-e2219. DOI: 10.1128/AAC.02165-19
6. Sinsinwar S., Vadivel V.. **Catechin isolated from cashew nut shell exhibits antibacterial activity against clinical isolates of MRSA through ROS-mediated oxidative stress**. (2020) **104** 8279-8297. DOI: 10.1007/s00253-020-10853-z
7. Arya S. S., Sharma M. M., Das R. K., Rookes J., Cahill D., Lenka S. K.. **Vanillin mediated green synthesis and application of gold nanoparticles for reversal of antimicrobial resistance in**. (2019) **5**. DOI: 10.1016/j.heliyon.2019.e02021
8. Jisaka M., Ohigashi H., Takagaki T.. **Bitter steroid glucosides, vernoniosides A1, A2, and A3, and related B1 from a possible medicinal plant, Vernonia amygdalina, used by wild chimpanzees**. (1992) **48** 625-632. DOI: 10.1016/s0040-4020(01)88123-0
9. Eyong U. E., Item J. A., Esien D. O., Margaret A. A., Patrick E. E.. **Haematological and immunological effect of coadministration of extracts of Vernonia amygdalina and Azadirachta indica on normal and diabetic rats**. (2011) **10** 10258-10262. DOI: 10.5897/ajb11.786
10. Igile G. O., Oleszek W., Jurzysta M., Burda S., Fafunso M., Fasanmade A. A.. **Flavonoids from Vernonia amygdalina and their antioxidant activities**. (1994) **42** 2445-2448. DOI: 10.1021/jf00047a015
11. Sinisi A., Millan E., Abay S. M.. **Poly-electrophilic sesquiterpene lactones from Vernonia amygdalina: new members and differences in their mechanism of thiol trapping and in bioactivity**. (2015) **78** 1618-1623. DOI: 10.1021/acs.jnatprod.5b00179
12. Harahap U., Dalimunthe A., Hertiani T., Muhammad M., Nasri D.. **Antioxidant and antibacterial activities of ethanol extract of Vernonia amygdalina Delile. Leaves**. (2021) **2342**
13. Egedigwe C. A., Ejike C. E., Ijeh I. I., Herbert U., Onwuka G. I., Asumugha V. U.. **Anti-obesity potentials of aqueous and methanol extracts of Vernonia amygdalina Del. Leaves in high-fat diet fed rats**. (2016) **13** 86-93. DOI: 10.21010/ajtcam.v13i2.11
14. Kadiri O., Olawoye B.. **Vernonia amygdalina: an underutilized vegetable with nutraceutical potentials – a review**. (2016) **4** p. 763. DOI: 10.24925/turjaf.v4i9.763-768.570
15. Hasibuan P. A. Z., Harahap U., Sitorus P., Satria D.. **The anticancer activities of Vernonia amygdalina Delile. Leaves on 4T1 breast cancer cells through phosphoinositide 3-kinase (PI3K) pathway**. (2020) **6**. DOI: 10.1016/j.heliyon.2020.e04449
16. Harahap U., Hasibuan P. A. Z., Sitorus P., Arfian N., Satria D.. **Antimigration activity of an ethylacetate fraction of Zanthoxylum acanthopodium DC. fruits in 4T1 breast cancer cells**. (2018) **19** 565-569. DOI: 10.22034/APJCP.2018.19.2.565
17. Lifiani R., Harahap U., Hasibuan P. A. Z., Satria D.. **Anticancer effect of african leaves (vernonia amygdalina del). To T47D Cell Resistant**. (2018) **11** 4-7. DOI: 10.22159/ajpcr.2018.v11s1.26551
18. Satria D., Silalahi J., Haro G., Ilyas S., Hasibuan P. A. Z.. **Cell cycle inhibition of ethylacetate fraction of zanthoxylum acanthopodium DC. Fruit against T47D cells**. (2019) **7** 726-729. DOI: 10.3889/oamjms.2019.178
19. Dalimunthe A., Pertiwi D., Muhmmad M., Kaban V. E., Nasri N., Satria D.. **The effect of extraction methods towards antioxidant activity of ethanol extract of Picria fel-terrae Lour. Herbs**. (2022) **1115**. DOI: 10.1088/1755-1315/1115/1/012040
20. Rani Z., Nasution H. M., Kaban V. E., Nasri N., Karo N. B.. **Antibacterial activity of freshwater lobster (Cherax quadricarinatus) shell chitosan gel preparation against**. (2022) **13**
21. Septama A. W., Tasfiyati A. N., Kristiana R., Jaisi A.. **Chemical profiles of essential oil from Javanese turmeric (Curcuma xanthorrhiza Roxb.), evaluation of its antibacterial and antibiofilm activities against selected clinical isolates**. (2022) **146** 728-734. DOI: 10.1016/j.sajb.2021.12.017
22. Wayne P. A.. **Performance standards for antimicrobial susceptibility testing, ninth informational suppl. NCCLS doc. M100-S9**. (2008) **3** 120-126
23. Septama A. W., Panichayupakaranant P.. **Synergistic effect of artocarpin on antibacterial activity of some antibiotics against methicillin-resistant**. (2016) **54** 686-691. DOI: 10.3109/13880209.2015.1072566
24. Milne K. E. N., Gould I. M.. **Combination antimicrobial susceptibility testing of multidrug-resistant Stenotrophomonas maltophilia from cystic fibrosis patients**. (2012) **56** 4071-4077. DOI: 10.1128/aac.00072-12
25. Septama A. W., Panichayupakaranant P.. **Antibacterial activity of artocarpanone isolated from Artocarpus heterophyllus heartwoods against diarrheal pathogens and its mechanism of action on membrane permeability**. (2017) **7** 064-068
26. Abbas H. A., Elsherbini A. M., Shaldam M. A.. **Glyceryl trinitrate blocks staphyloxanthin and biofilm formation in**. (2019) **19** 1376-1384. DOI: 10.4314/ahs.v19i1.10
27. Yang A., Tang W. S., Si T., Tang J. X.. **Influence of physical effects on the swarming motility of**. (2017) **112** 1462-1471. DOI: 10.1016/j.bpj.2017.02.019
28. Dalimunthe A., Anjelisa Zaitun Hasibuan P., Silalahi J., Fatimah Sinaga S., Satria D.. **Antioxidant activity of alkaloid compounds from Litsea cubeba Lour**. (2018) **34** 1149-1152. DOI: 10.13005/ojc/340270
29. Van de Vel E., Sampers I., Raes K.. **A review on influencing factors on the minimum inhibitory concentration of essential oils**. (2019) **59** 357-378. DOI: 10.1080/10408398.2017.1371112
30. Farhadi F., Khameneh B., Iranshahi M., Iranshahy M.. **Antibacterial activity of flavonoids and their structure–activity relationship: an update review**. (2019) **33** 13-40. DOI: 10.1002/ptr.6208
31. Jantrapanukorn B., Pongpraritt S., Powthong P., Pheungphu T.. **The study of antibacterial activity in enteric pathogens of Roselle (Hibiscus sabdariffa Linn.) by broth micro-dilution method**. (2017) **7** 119-122
32. He J., Starr C. G., Wimley W. C.. **A lack of synergy between membrane-permeabilizing cationic antimicrobial peptides and conventional antibiotics**. (2015) **1848** 8-15. DOI: 10.1016/j.bbamem.2014.09.010
33. Grossman T. H.. **Tetracycline antibiotics and resistance**. (2016) **6**. DOI: 10.1101/cshperspect.a025387
34. Shi C., Zhang X., Zhao X.. **Synergistic interactions of nisin in combination with cinnamaldehyde against**. (2017) **71** 10-16. DOI: 10.1016/j.foodcont.2016.06.020
35. Nasri N., Harahap U., Silalahi J., Satria D.. **Antibacterial activity of lactic acid bacteria isolated from Dengke Naniura of Carp (**. (2021) **22**. DOI: 10.13057/biodiv/d220802
36. Ferreira G. R. S., Brito J., Procopio T. F.. **Antimicrobial potential of Alpinia purpurata lectin (ApuL): growth inhibitory action, synergistic effects in combination with antibiotics, and antibiofilm activity**. (2018) **124** 152-162. DOI: 10.1016/j.micpath.2018.08.027
37. Knezevic P., Aleksic V., Simin N., Svircev E., Petrovic A., Mimica-Dukic N.. **Antimicrobial activity of**. (2016) **178** 125-136. DOI: 10.1016/j.jep.2015.12.008
38. Mazzarello V., Gavini E., Rassu G.. **Clinical assessment of new topical cream containing two essential oils combined with Tretinoin in the treatment of acne**. (2020) **13** 233-239. DOI: 10.2147/CCID.S236956
39. Meah M. S., Lertcanawanichakul M., Pedpradab P.. **Synergistic effect on anti-methicillin-resistant**. (2020) **71** 510-519. DOI: 10.1111/lam.13369
40. Di Napoli M., Varcamonti M., Basile A., Bruno M., Maggi F., Zanfardino A.. **Anti-Pseudomonas aeruginosa activity of hemlock (Conium maculatum, Apiaceae) essential oil**. (2019) **33** 3436-3440. DOI: 10.1080/14786419.2018.1477151
41. Spengler G., Gajdacs M., Donadu M. G.. **Evaluation of the antimicrobial and antivirulent potential of essential oils isolated from juniperus oxycedrus L. ssp. macrocarpa aerial parts**. (2022) **10** p. 758. DOI: 10.3390/microorganisms10040758
42. Sun J., Zhang Y., Su J.. **Naftifine enhances photodynamic therapy against**. (2020) **179**. DOI: 10.1016/j.dyepig.2020.108392
43. Hall J. W., Yang J., Guo H., Ji Y.. **The**. (2017) **85** 008388-e916. DOI: 10.1128/iai.00838-16
44. Kannappan A., Balasubramaniam B., Ranjitha R.. **In vitro and in vivo biofilm inhibitory efficacy of geraniol-cefotaxime combination against Staphylococcus spp**. (2019) **125** 322-332. DOI: 10.1016/j.fct.2019.01.008
45. Sethupathy S., Vigneshwari L., Valliammai A., Balamurugan K., Pandian S. K.. **L-Ascorbyl 2, 6-dipalmitate inhibits biofilm formation and virulence in methicillin-resistant**. (2017) **7** 23392-23406. DOI: 10.1039/c7ra02934a
46. Venieraki A., Tsalgatidou P. C., Georgakopoulos D. G., Dimou M., Katinakis P.. **Swarming motility in plant-associated bacteria**. (2016) **9** 16-27. DOI: 10.1515/hppj-2016-0002
47. Xu J., Platt T. G., Fuqua C.. **Regulatory linkages between flagella and surfactant during swarming behavior: lubricating the flagellar propeller?**. (2012) **194** 1283-1286. DOI: 10.1128/jb.00019-12
48. Cushnie T. T., Lamb A. J.. **Antimicrobial activity of flavonoids**. (2005) **26** 343-356. DOI: 10.1016/j.ijantimicag.2005.09.002
49. Khanna P.. **Antimicrobial principles from tissue culture of some plant species**. (1980) **42** 113-117
50. Torrenegra R. D., Ricardo A. A., Pedrozo J. P., Fuentes O. C.. **Flavonoids from Gnaphalium gracile HBK**. (1989) **27** 22-24. DOI: 10.3109/13880208909053932
51. Basile A., Sorbo S., Giordano S.. **Antibacterial and allelopathic activity of extract from Castanea sativa leaves**. (2000) **71** S110-S116. DOI: 10.1016/s0367-326x(00)00185-4
52. Rauha J.-P., Remes S., Heinonen M.. **Antimicrobial effects of Finnish plant extracts containing flavonoids and other phenolic compounds**. (2000) **56** 3-12. DOI: 10.1016/s0168-1605(00)00218-x
53. Guo Y., Liu Y., Zhang Z.. **The antibacterial activity and mechanism of action of luteolin against Trueperella pyogenes**. (2020) **13** 1697-1711. DOI: 10.2147/idr.s253363
|
---
title: Extracellular Vesicles Derived from Three-Dimensional-Cultured Human Umbilical
Cord Blood Mesenchymal Stem Cells Prevent Inflammation and Dedifferentiation in
Pancreatic Islets
authors:
- Eunwon Lee
- Seungyeon Ha
- Gyuri Kim
- Jae Hyeon Kim
- Sang-Man Jin
journal: Stem Cells International
year: 2023
pmcid: PMC9970714
doi: 10.1155/2023/5475212
license: CC BY 4.0
---
# Extracellular Vesicles Derived from Three-Dimensional-Cultured Human Umbilical Cord Blood Mesenchymal Stem Cells Prevent Inflammation and Dedifferentiation in Pancreatic Islets
## Abstract
It is unclear whether extracellular vesicles (EVs) from mesenchymal stem cells (MSCs) have a direct protective effect on pancreatic islets. In addition, whether culturing MSCs in three dimensions (3D) instead of a monolayer (2D) can induce changes in the cargo of EVs that facilitate the polarization of macrophages into an M2 phenotype has not been investigated. We sought to determine whether EVs from MSCs cultured in 3D can prevent inflammation and dedifferentiation in pancreatic islets and, if so, whether the protective effect is superior to that of EVs from 2D MSCs. Human umbilical cord blood- (hUCB-) MSCs cultured in 3D were optimized according to cell density, exposure to hypoxia, and cytokine treatment based on the ability of the hUCB-MSC-derived EVs to induce the M2 polarization of macrophages. Islets isolated from human islet amyloid polypeptide (hIAPP) heterozygote transgenic mice were cultured in serum-deprived conditions with hUCB-MSC-derived EVs. EVs derived from 3D hUCB-MSCs had more abundant microRNAs involved in M2 polarization of macrophages and had an enhanced M2 polarization ability on macrophages, which was optimized when the 3D culture condition was 2.5 × 104 cells per spheroid without preconditioning with hypoxia and cytokine exposure. When islets isolated from hIAPP heterozygote transgenic mice were cultured in serum-deprived conditions with hUCB-MSC-derived EVs, the EVs derived from 3D hUCB-MSCs suppressed the expression of proinflammatory cytokines and caspase-1 in pancreatic islets and increased the proportion of M2-polarized islet-resident macrophages. They improved glucose-stimulated insulin secretion, reduced the expression of Oct4 and NGN3, and induced the expression of Pdx1 and FoxO1. The greater suppression of IL-1β, NLRP3 inflammasome, caspase-1, and Oct4 and induction of Pdx1 and FoxO1 were found in islets cultured with the EVs derived from 3D hUCB-MSCs. In conclusion, EVs derived from 3D hUCB-MSCs optimized for M2 polarization attenuated nonspecific inflammation and preserved β-cell identity of pancreatic islets.
## 1. Introduction
Intravenous injection of mesenchymal stem cells (MSCs) improves the function and survival of pancreatic islets and preserves β-cell identity in animal models of type 2 diabetes [1–3] and pancreatic islet transplantation [4–6]. An important mechanism of these benefits is the ability of MSCs to induce polarization of macrophages into the M2 subtype. In type 2 diabetes and primary islet graft failure after pancreatic islet transplantation, the progression of inflammatory reactions such as the increase in islet macrophage infiltration, the polarization of macrophages into the M1 subtype, and the inflammasome activation play an essential role in the progression of β-cell failure [7]. Clinical trials on the infusion of MSCs into patients with type 2 diabetes [8] and recipients of pancreatic islet autotransplantation have been conducted without severe adverse events [9].
The use of MSC-derived extracellular vesicles (EVs) offers a promising alternative to MSCs themselves by reproducing their biological function in delivering nucleic acids, proteins, and lipids to the local microenvironment of damaged cells or tissues [10, 11]. Moreover, EVs could decrease safety concerns because they are nonimmunogenic and not likely to cause maldifferentiation [12, 13], which is a clinical concern of therapies using MSCs. Interestingly, it has been suggested that EVs derived from MSCs, without additional effects from MSCs themselves or other components of their secretome, can improve pancreatic β-cell survival and insulin sensitivity in rodents with low-dose streptozotocin and high-fat diet-induced diabetes [14]. This indicates that the use of MSC-derived EVs could be a novel approach to delaying the progression of β-cell failure in type 2 diabetes and primary islet graft failure after pancreatic islet transplantation.
MSCs cultured using the three-dimensional method (3D MSCs), which better reflects an in vivo environment, consistently exhibit enhanced anti-inflammatory, angiogenic, and tissue reparative/regenerative effects with improved cell survival after transplantation [15]. Such benefits are mediated, at least in part, by EVs from MSCs [16]. In addition to the changes in gene expression profiles such as the expression of genes with cell stemness and migration ability in 3D MSCs, EV production increases with the three-dimensional culture method, and the resulting EVs can have a cargo profile that is different from 2D MSCs [17]. In a recent study, total RNA sequencing using next-generation sequencing platforms on human amnion-derived 2D and 3D MSCs revealed profound transcriptome changes, including enhanced secretion of C-C motif chemokine ligand 2 (CCL2), C-X-C motif chemokine ligand 12 (CXCL12), and bone morphogenetic protein 2 (BMP2), which could contribute to a microenvironment favouring polarization of macrophages into the M2 phenotype [18]. However, a previous study conducted in an animal model of bleomycin-induced lung fibrosis suggested that EVs produced from 3D MSCs did not demonstrate enhanced immunomodulatory properties compared with 2D MSC-derived EVs [19], indicating that optimization of 3D MSC-derived EVs in each disease model is required.
EVs from cytokine-preconditioned MSCs contain several microRNAs (miRs) that can induce macrophage polarization into the M2 subtype, in contrast to EVs from resting MSCs [20, 21]. However, whether the cargo changes in EVs derived from 3D MSCs include an enhanced capacity for the polarization of macrophages into the M2 phenotype has not been investigated, and there has been no specific study that compared the benefits of EVs from either 2D or 3D MSCs on pancreatic islets. Although it has been suggested that intraperitoneal injection of 3D MSCs instead of 2D MSCs in a multiple low-dose streptozotocin-induced diabetes model better attenuates inflammatory processes in pancreatic islets and improves glycemic control [22], the role of 3D MSC-derived EVs in this benefit, if any, has not been determined.
Therefore, we investigated whether EVs from MSCs cultured in 3D have direct protective effects against inflammation and dedifferentiation on pancreatic islets and, if so, whether the protective effects are superior to those of EVs from monolayer-cultured MSCs (2D MSCs).
## 2.1. Animals and Cell Culture
Mouse islets were isolated from heterozygous human islet amyloid polypeptide (hIAPP) transgenic (hIAPP+/-) 8-12-week-old FVB/N mice (Jackson Laboratory, Bar Harbor, ME, USA). The Institutional Animal Care and Use Committee of Samsung Biomedical Research Institute approved all animal experimental protocols in this study. Human umbilical cord blood- (hUCB-) MSCs were isolated according to a reported method [23]. Umbilical cord blood (UCB) units obtained from full-term deliveries were collected from the unborn placenta with the informed consent of the mothers. The human UCB-MSC isolation procedure was approved by the Institutional Review Board of Samsung Medical Center (IRB No. SMC 2019-11-026), and all participants provided informed consent for the use of the umbilical cord in this experimental study. The human monocyte cell line THP-1 was purchased from the Korean Cell Bank (Seoul, Korea) and maintained in complete RPMI 1640 media.
## 2.2. Culture of 3D hUCB-MSC Spheroids
The hUCB-MSCs were grown in minimum essential medium-alpha (MEM-α; Gibco, Waltham, MA, USA) supplemented with $10\%$ fetal bovine serum (FBS; Gibco) and $1\%$ penicillin/streptomycin (Gibco) at 37°C and $5\%$ CO2 until passage 5. To obtain 3D spheroids, hUCB-MSCs at three to five passages were seeded into round-bottom 96-well plates coated with poly(2-hydroxyethyl methacrylate) (pHEMA; Sigma-Aldrich, St. Louis, MO, USA) reagent prepared in $95\%$ ethanol. In each well containing 100 μL of MEM-α (Gibco) supplemented with $10\%$ FBS (Gibco) and $1\%$ penicillin/streptomycin (Gibco), 2.5 × 103, 6.25 × 103, and 25 × 103 MSCs/well were seeded and cultured at 37°C and $5\%$ CO2 for three days (2.5 × 103, 6.25 × 103, and 25 × 103 cells/spheroid referred to as 2.5 K, 6.25 K, and 25 K 3D hUCB-MSCs, respectively). The cell numbers were based on the number of seeded cells before spheroid formation. Then, the 3D spheroids were transferred into a Petri dish to obtain conditioned media (CM). Only one spheroid was formed per well of the 96-well plate, and all 96 spheroids were transferred into 15 mL of MEM-α (Gibco) supplemented with $10\%$ Exo-free FBS (Gibco) and $1\%$ penicillin/streptomycin (Gibco) in a 100 mm diameter Petri dish. The 3D spheroids were cultured at 37°C and $5\%$ CO2 for six days. Every 48 h, 13 mL of conditioned medium was collected and replaced with an equal volume of fresh medium. This resulted in a total of 39 mL of conditioned medium per Petri dish with a total of 96 hUCB-MSC spheroids.
## 2.3. Isolation and Characterization of EVs
Monolayer (2D)-cultured and 3D-cultured hUCB-MSC spheroids were incubated in MEM-α (Gibco) supplemented with $10\%$ exosome-depleted FBS (Gibco) for 72 h at 37°C in a fully humidified $5\%$ CO2 atmosphere. The CM was harvested and centrifuged at 2,000 × g for 10 min to remove cells and debris. Then, the supernatant was concentrated using an Amicon Ultra Centrifugal Filter (100 kDa cut-off) Unit (Merck, Burlington, VT, USA). EVs were isolated from concentrated CM using a Total Exosome Isolation Kit (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). The protein in EVs was quantified using a BCA Protein Assay (Thermo Fisher Scientific). The size distribution and concentration of EVs were measured by nanoparticle tracking analysis (NTA) using a NanoSight NS300 (Malvern, Worcestershire, UK). The sample was diluted to a number concentration of between 108 and 109 particles per milliliter. The morphology of the EVs was characterized using transmission electron microscopy (TEM; Hitachi HT7700, Hitachi Ltd., Tokyo, Japan). The EVs were fixed with $4\%$ paraformaldehyde and spotted onto a carbon-coated 300-mesh grid. Then, the samples were stained with $1\%$ uranyl acetate. Subsequently, the samples were examined at 100.0 keV.
## 2.4. Mouse Islet Isolation and Culture
Mouse islets were isolated from 8-12-week-old hIAPP+/- mice as described previously [24]. Briefly, 0.8 mg/mL collagenase P (Roche, Basel, Switzerland) in Hanks' balanced saline solution (HBSS; Sigma-Aldrich) was infused into the common bile duct for mouse pancreas digestion. Islets were purified from the digested pancreas using a Human PanColl (PAN-Biotech GmbH, Am Gewerbepark, Aidenbach, Germany) gradient and washed several times with HBSS (Sigma-Aldrich). Purified hIAPP+/- mouse islets were cultured in RPMI 1640 (Gibco) containing 10,000 U/mL penicillin-streptomycin (Gibco) and cultured at 37°C in a fully humidified $5\%$ CO2 atmosphere.
## 2.5. Coculture of Mouse Islets with hUCB-MSCs or hUCB-MSC-Derived EVs
Islets were hand-picked with a microscope at an average size of 150 μm and cultured in serum deprivation conditions (RPMI 1640 supplemented with bovine serum albumin, BSA, $0.625\%$) for 48 h. During ex vivo culture of islets, six experimental groups were designated for experiments using hIAPP+/- FVB/N mouse islets: medium supplemented with $10\%$ FBS (FBS group), medium supplemented with $0.625\%$ BSA (Qbiogene, Carlsbad, CA, USA) (BSA group), medium supplemented with 20 μg/mL 2D hUCB-MSC-derived EVs plus $0.625\%$ BSA (BSA+2D EV group), medium supplemented with 20 μg/mL 25 K 3D hUCB-MSC-derived EVs plus $0.625\%$ BSA (BSA+3D EV group), coculture with 2D hUCB-MSCs in medium supplemented with $0.625\%$ BSA (BSA+2D MSC group), and coculture with 25 K 3D hUCB-MSC spheroids in medium supplemented with $0.625\%$ BSA (BSA+3D MSC group). The numbers of 2D and 3D MSCs per media volume in the BSA+2D MSC and BSA+3D MSC groups were determined as the expected numbers required to yield equivalent amounts of EV per volume in their counterpart groups (BSA+2D EV and BSA+3D EV groups). For coculture of islets with 2D MSCs, 5 × 105 MSCs were seeded in six-well plates overnight before the islets were added. For coculture of islets with 25 K 3D spheroids, 28 spheroids were transferred to a low-attachment six-well plate. From each well, 50 islets were hand-picked onto a cell culture insert (Falcon, Corning, NY, USA) and incubated for 48 h at 37°C in a fully humidified $5\%$ CO2 atmosphere.
## 2.6. Coculture of Mouse Islets with Pancreatic Macrophages with or without MSC-Derived EVs
For coculture of islets with pancreatic macrophages, 1 × 106 isolated pancreatic macrophages were seeded in six-well plates and incubated for 2 h with complete RPMI 1640 media. Nonadherent cells were removed, and 50 islets were seeded above the cell culture insert with fresh complete RPMI 1640 media. Islets were cocultured only with macrophages (control group), 10 μg/mL 2D MSC EV treatment (2D hUCB-MSC-derived EV group), or 10 μg/mL 3D MSC EV treatment (3D hUCB-MSC-derived EV group). All samples were cultured in complete RPMI1640 media for 48 h.
## 2.7. Pancreatic Macrophage Isolation
To isolate pancreatic macrophages, STZ-induced type 1 diabetes was established by s.c. administration of STZ at the dose of 120 mg/kg body weight. At 4-day post-STZ injection, pancreas tissue was digested in Hanks balanced saline solution (Sigma-Aldrich) containing 2 mg/mL of collagenase P (Roche) for 15 min at 37°C. Digestion was quenched by FBS, and the digestate was filtered through 70 μm Nylon mesh and centrifuged at 350 g for10 min. Cell suspensions were resuspended, and anti-F$\frac{4}{80}$ microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) were used for negative selection of macrophages according to the manufacturer's instructions.
## 2.8. Real-Time Reverse Transcription-PCR
Total RNA was isolated from mouse islets using an RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. RNA (1 μg) was reverse-transcribed to obtain cDNA using PrimeScript RT Master Mix (Takara Bio Inc., Kusatsu, Japan). Real-time quantitative reverse transcription- (RT-) PCR was performed in triplicate using gene-specific primers. PCR was performed with a TB Green Premix Kit (Takara) and 7500 Fast Real-Time PCR System (Applied Biosystems, Waltham, MA, USA). The relative gene expression of TNF-α, interleukin- (IL-) 1 beta (1β), IL-18, NOD-like receptor pyrin domain-containing protein 3 (NLRP3), high mobility group box 1 (HMGB1), caspase-1, Oct4, neurogenin 3 (NGN3), forkhead box protein O1 (FoxO1), and pancreatic and duodenal homeobox 1 (Pdx1) was determined and normalized to that of mouse β-actin using gene-specific primer pairs (see Supplementary Materials, Table 1).
For the microRNA real-time PCR assay, total RNA was extracted from 30 μg EVs using a Total Exosome RNA and Protein Isolation Kit (Thermo Fisher Scientific). The relative expression (ΔΔCt) of microRNA suggested to induce polarization of macrophages into M1 (miR-127-3p and miR-155-5p) and M2 (miR-34a-5p and miR-146a-5p) phenotypes which were determined using a Mir-X™ miRNA qRT-PCR TB Green® Kit (Takara), according to manufacturer's instructions. These microRNAs were chosen based on a previous study that used exosomes from proinflammatory cytokine-stimulated adipose MSCs [20]. The entire sequence of the mature microRNA can be used as the microRNA-specific 5′ primer.
## 2.9. Western Blot Analysis
For protein extraction, hUCB-MSC-derived EVs were lysed with RIPA buffer. Equal amounts of proteins from each group were separated by $10\%$ sodium dodecyl sulfate (SDS) PAGE and transferred to a PVDF membrane. Anti-CD63 (System Biosciences, LLC, Palo Alto, CA, USA), anti-TSG101 (System Biosciences), anti-CANX (ABclonal, Woburn, MA, USA), and anti-GM130 (ABclonal) were used as primary antibodies. Horseradish peroxidase- (HRP-) conjugated IgG antibody (ABclonal) was used as the secondary antibody. Antibody-bound proteins were visualized using an enhanced chemiluminescent reagent.
## 2.10. Glucose-Induced Insulin Secretion (GSIS) Test
A GSIS test was performed to measure islet functionality as described previously [24]. Islets cultured with or without hUCB-MSC-derived EVs were cultured in a KRBB solution containing low- (60 mg/dL) or high-concentration glucose (300 mg/dL). Twenty islets were used for each condition. After incubating for 60 min at 37°C, the supernatant was collected for measuring insulin concentration by ultrasensitive insulin enzyme-linked immunosorbent assay (ELISA) (ALPCO, Salem, NH, USA).
## 2.11. Enzyme-Linked Immunosorbent Assay and Cytokine Array
To determine the expression level of cytokines, mouse islets (80 islets per well) were seeded onto a 6-well ultralow attachment plate (Corning, NY, USA). The culture medium was collected at 48 h without cellular debris. The concentrations of TNF-α, IL-1β, IL-6, and HMGB1 were quantified with an ELISA kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's protocol. A cytokine array was performed using a Proteome Profiler Human XL Cytokine Array Kit (R&D Systems) following the manufacturer's instructions.
## 2.12. Cytokine Profiling Array
A cytokine array was performed using a Proteome Profiler Human XL Cytokine Array Kit (R&D Systems) following the manufacturer's instructions. For cytokine arrays, the membrane was blocked in buffer and incubated with 100 μg of EVs overnight at 4°C on a rocking platform shaker. Then, the membrane was incubated with a detection antibody cocktail for 1 h on a shaker. After washing, the membrane was incubated with streptavidin-HRP. Finally, Chemi Reagent mix was spread on the membrane, and an autoradiograph was obtained. Pixel densities were analyzed using NIH ImageJ medical imaging software.
## 2.13. Flow Cytometry Analysis
Cells were washed and resuspended in staining buffer. Fluorochrome-conjugated monoclonal antibodies were added to cells and incubated for 30 min in the dark. An hUCB-MSC surface marker analysis with flow cytometry was performed using the following monoclonal antibodies: CD105, CD73, CD90, CD45, CD34, CD14, CD11b, CD19, HLA-DR, positive isotype control cocktail (mIgG1 FITC, mIgG1 PerCP-Cy5.5, and mIgG1 APC), and negative isotype control cocktail (mIgG1 PE and mIgG2a PE) purchased from BD Biosciences (San Jose, CA, USA). Phenotyping of macrophages with flow cytometry was performed using the antibodies of M2 macrophage differentiation markers (CD163, BD Biosciences; CD206, Abcam, Cambridge, UK) and M1 macrophage differentiation markers (CD80; Abcam). After staining of primary antibody, cells were washed and resuspended in 150 μL of the appropriate secondary antibody (Alexa Fluor® 488-conjugated donkey anti-rat, Alexa Fluor® 647-conjugated donkey anti-rabbit; Abcam). Data were acquired using BD LSRFortessa flow cytometers and analyzed with BD FACSDiva software (BD Biosciences, Franklin Lakes, NJ, USA).
## 2.14. Fluorescence Immunohistochemistry
Islets were harvested, fixed in $4\%$ paraformaldehyde, and immersed in $30\%$ sucrose solution to be embedded in a frozen block. Inflammatory markers and dedifferentiation markers were detected in cell sections by immunofluorescence. Anti-FoxO1 (Santa Cruz, Dallas, Texas, USA), anti-cleaved caspase 3 (Cell Signaling, Danvers, MA, USA), anti-Oct4 (Abcam, Cambridge, UK), anti-NGN3 (Abcam), anti-A11 oligomer (Invitrogen), anti-Pdx1 (Abcam), anti-IL-1β (Abcam), and anti-insulin (Abcam) were used as primary antibodies. M1/M2 macrophage markers were identified using the antibodies of M2 macrophage differentiation markers (anti-arginase-1; Cell Signaling) and M1 macrophage differentiation markers (CD80; Abcam). Sections were incubated with primary antibody diluted in blocking buffer overnight at 4°C. Secondary antibodies (Alexa Fluor® 488-conjugated goat anti-rabbit, Alexa Fluor® 488-conjugated goat anti-mouse, and Alexa Fluor® 594-conjugated goat anti-rabbit; Abcam) were added and incubated for 1 h at room temperature. Counterstaining was performed using DAPI (1: 10,000). Images were obtained from each section using a fluorescent microscope (Olympus, Shinjuku, Tokyo, Japan).
## 2.15. Statistical Methods
Results are reported as mean ± standard deviation. Statistical analysis was performed using GraphPad Software (GraphPad, San Diego, CA, USA). Continuous variables were compared using one-way analysis of variance (ANOVA) or the Mann–Whitney U test, as appropriate. P values < 0.05 were considered significant.
## 3.1. Characterization of EVs Derived from Umbilical Cord Blood MSCs
To confirm the identity of the isolated hUCB-MSCs, surface expression of CD90, CD44, CD73, and CD105 and the nonexpression of the negative marker cocktail (CD45, CD34, CD14, CD11b, CD19, and HLA-DR) were confirmed using flow cytometry (Figure 1(a)). After conditioned medium was obtained during 2D and 3D cultures of hUCB-MSCs, hUCB-MSC-derived EVs were isolated from each group using a polymer-based precipitation method. The shape of each EV was observed using TEM. The cup-shaped morphology of the isolated EVs was confirmed in both 2D and 3D hUCB-MSC-derived EVs (Figure 1(b)). There was no significant change in the size distribution of EVs according to the MSC culture method (Figure 1(c)). When the expression of EV markers was examined using western blotting, CD63 and TSG101 were expressed, and the ER and Golgi markers CANX and GM130 were not expressed (Figure 1(d) and Supplementary Figure S1).
## 3.2. hUCB-MSC-Derived EVs Promote M2 Polarization in Pancreatic Macrophages and THP-1 Monocytes
Next, we evaluated whether hUCB-MSC-derived EVs increase M2 polarization in isolated pancreatic macrophages from FVB mice. The expression of an M2 polarization marker (CD206) in pancreatic macrophage cells increased after treatment with 2D and 3D hUCB-MSC-derived EVs for 48 h. The results showed that hUCB-MSC-derived EVs could promote M2 polarization of macrophages in a concentration-dependent manner, showing 10 μg/mL as the likely optimal concentration (Figure 2(a)). To determine the M2 polarization activity of EVs, 2D and different sizes of 3D hUCB-MSC-derived EVs (10 μg/mL) were added to cultures of isolated macrophages, and the expression of M2 polarization markers (CD206) was identified using flow cytometry (Figure 2(b) and Supplementary Figure S2a). Compared to the macrophages treated with 2D hUCB-MSC-derived EVs with and without cytokines (IFN-γ and TNF-α, each 40 ng/mL), the M2 polarization marker CD206 was upregulated in the macrophages treated with 3D hUCB-MSC-derived EVs. When the macrophages were treated with 3D hUCB-MSC-derived EVs, the expression of CD206 was higher when the seeding density of hUCB-MSC was 25 K cells per spheroid than for 2.5 K cells per spheroid. When macrophages were treated with 3D hUCB-MSC-derived EVs, the expression of CD206 was significantly higher in macrophages cultured with 25 K 3D hUCB-MSC-derived EVs than with 2.5 K 3D hUCB-MSC-derived EVs (Figure 2(b)). Consistent findings were observed when M2 polarization markers (CD206 and CD163) of THP-1 monocytes were compared among groups (Supplementary Figure S3).
To evaluate whether the additional preconditioning of hUCB-MSCs, such as exposure to hypoxia and cytokine, further potentiates the M2 polarization ability of EVs, we obtained EVs from 3D cultures of hUCB-MSCs grown under hypoxia or cytokines (TNF-α and IFN-γ, each 40 ng/mL) at a seeding density of 25 × 103 cells per spheroid (Figure 2(c) and Supplementary Figure S2b). Preconditioning with cytokines did not further increase the M2 polarization ability of 3D hUCB-MSC-derived EVs. When cytokine secretion of the macrophages cultured in the presence of 2D and 3D hUCB-MSC-derived EVs was measured by ELISA, the IL-10 and TGF-β levels were significantly increased in macrophages cultured in the presence of 25 K 3D hUCB-MSC-derived EVs (Figure 2(d)).
A recent study suggested that EVs from cytokine-preconditioned MSCs contain several miRs that can induce macrophage polarization into the M2 subtype, in contrast to the EVs from resting MSCs [20, 21]. Thus, miRs in EVs isolated after preconditioning with IFN-γ and TNF-α in 2D- and 3D-cultured hUCB-MSC-derived EVs were quantified and compared with miRs in EVs isolated from 2D cultures without cytokine stimulation. When the cargoes of 2D and 3D (25 K) hUCB-MSC-derived EVs were analyzed, the level of miR-127-3p, a miR involved in M1 polarization [20], was decreased in 25 K 3D hUCB-MSC-derived EVs compared to 2.5 K and 2D hUCB-MSC EVs, resulting in a similar level to that in cytokine-treated 2D hUCB-MSC EVs (Figure 3(a)). There was no difference in the level of miR-155-5p between 25 K 3D hUCB-MSC-derived EVs and 2D hUCB-MSC-derived EVs. The level of miR-155-5p was increased in the 2.5 K 3D hUCB-MSC EVs (Figure 3(a)). In 25 K 3D hUCB-MSC-derived EVs, the levels of miR-34a-5p and miR-146a-5p, miRs involved in M2 polarization [20], were significantly greater than those in unstimulated 2D, cytokine-stimulated 2D, and other 3D hUCB-MSC-derived EVs. A significantly higher level of miR-34a-5p was observed in 2.5 K 3D hUCB-MSC-derived EVs than in unstimulated and cytokine-stimulated 2D hUCB-MSC-derived EVs (Figure 3(a)). Based on these results, 25 K 3D hUCB-MSC-derived EVs without exposure to hypoxia and cytokines were used for further experiments.
A comparison of the protein expression profiles of 2D MSC-derived EVs and 3D MSC-derived EVs revealed that cytokines related to angiogenesis and inflammation, such as IL-6, MCP-1, MIC-1, IL-11, G-CSF, CCL20, and IL-27, were expressed in higher quantities in 3D MSC-derived EVs than in 2D MSC-derived EVs (Figure 3(b)).
## 3.3. hUCB-MSC-Derived EVs Attenuate Transcription of Proinflammatory Cytokines in hIAPP+/- Mouse Pancreatic Islets during Serum-Deprived Culture
To examine whether hUCB-MSC-derived EVs could inhibit nonspecific inflammation and β-cell dysfunction provoked by serum deprivation, islets isolated from hIAPP+/- mice were cultured in media supplemented with BSA without FBS (BSA group), media supplemented with FBS (FBS group), and media supplemented with BSA without FBS but with EVs derived from 2D hUCB-MSCs (BSA+2D EV group) or 3D hUCB-MSCs (BSA+3D EV group) or cocultured with 2D hUCB-MSCs (BSA+2D MSC group) or 3D hUCB-MSCs (BSA+3D MSC group) in media supplemented with BSA without FBS. The mRNA transcription levels of IL-1β, IL-18, NLRP3, caspase-1, TNF-α, IL-6, and HMGB1 in the six groups were measured using quantitative real-time PCR (Figure 4(a)). The transcription levels of IL-1β, IL-18, NLRP3, caspase-1, TNF-α, IL-6, and HMGB1 were all significantly lower in the BSA+3D EV group than in the BSA group. The transcription levels of TNF-α, IL-18, IL-6, and HMGB1 but not IL-1β, NLRP3, and caspase-1 were also significantly lower in the BSA+2D EV group than in the BSA group. The transcription levels of TNF-α, IL-1β, IL-18, NLRP3, IL-6, and caspase-1 were lower in the BSA+3D EV group than in the BSA+2D EV group (Figure 4(a)). We then evaluated the effect of hUCB-MSC-derived EVs on GSIS in hIAPP+/- islets, as shown in Figure 4(b). High-glucose (300 mg/dL)-stimulated insulin secretion islet cells from 3D hUCB-MSC-derived EVs over a one-hour static incubation period were more numerous than in the BSA group.
At least in part, the decrease in the transcription levels of IL-18, NLRP3, caspase-1, TNF-α, IL-6, and HMGB1 in the BSA+3D MSC group was reproducible in the BSA+3D EV group. The decrease in medium concentrations of TNF-α, IL-1β, and HMGB1 in the BSA+3D MSC group and the BSA+3D EV group was comparable, and the decrease in medium concentration of IL-6 in the BSA+3D MSC group was partly reproducible in the BSA+3D EV group (Figure 4(c)). The decrease in transcription levels of TNF-α and IL-6 in the BSA+2D MSC group was partly reproducible in the BSA+2D EV group. The decrease in medium concentration of TNF-α and in part of the IL-6 in the BSA+2D MSC group was reproducible in the BSA+2D EV group. The decrease in medium concentration of IL-1β in the BSA+2D MSC group was not reproducible in the BSA+2D EV group (Figure 4(c)).
To determine whether these findings are consistent in the setting of interaction among pancreatic islets, macrophages, and EVs from MSCs, we cocultured pancreatic islets from hIAPP+/- mice and freshly isolated pancreatic macrophages using a transwell system, with or without medium supplementation of 2D or 3D hUCB-MSC-derived EVs. After 48 hours of coculture, the total RNA of the pancreatic islets was extracted and real-time RT-PCR was performed. The transcription levels of IL-1β, IL-18, caspase-1, TNF-α, IL-6, and HMGB1 were downregulated in the islets cocultured with pancreatic macrophages in the presence of 3D hUCB-MSC-derived EVs (Figure 5(a)). High-glucose (300 mg/dL)-stimulated insulin secretion over a 1 h static incubation period in islet cells from 3D hUCB-MSC-derived EVs was greater than in the control group or the 2D hUCB-MSC-derived EV group (Figure 5(b)). To evaluate the hUCB-MSC-derived EV-mediated M2 polarization, cocultured macrophages were stained with CD80 and CD206. The proportion of CD80-CD206+ cells was significantly greater in the 3D hUCB-MSC-derived EV groups than in the control and 2D hUCB-MSC-derived EV groups, suggesting that 3D hUCB-MSC-derived EVs promoted M2 polarization of macrophages in the setting of interaction among pancreatic islets, macrophages, and EVs from MSCs (Figure 5(c)). When the concentrations of TGF-β and IL-10 in the supernatant were determined by ELISAs in this setting, both significantly increased in the presence of 3D hUCB-MSC-derived EVs compared to the control group. The concentration of IL-10 in the 3D hUCB-MSC derived EV group was significantly higher than that of the 2D hUCB-MSC derived EV group (Figure 5(d)).
To confirm whether the inhibition of nonspecific inflammation caused by EVs derived from hUCB-MSCs was associated with the M1/M2 polarization of islet-resident macrophages, the proportion of ARG1+ and CD80+ cells in the islets of each group was compared. Although the proportion of ARG1+ cells in the islets of the BSA+2D EV group was not different from that of the BSA and FBS groups, the proportion of ARG1+ cells in the islets of the BSA+3D EV group was higher than in the other groups. Although the proportion of CD80+ cells in the islets of the BSA+2D EV group was not different from that of the FBS group, the proportion of ARG1+ cells in the islets of the BSA+3D EV group was lower than in the BSA, FBS, and BSA+2D EV groups (Figures 6(a) and 6(b)). The proportion of IL-1β+ cells in the islets of both the BSA+3D EV and BSA+2D EV groups was lower than that of the BSA group. A more potent reduction in the proportion of IL-1β+ cells in islets was observed in the BSA+3D EV group, which was similar to that of the FBS group (Figures 6(a) and 6(b)).
## 3.4. hUCB-MSC-Derived EVs Attenuate Dedifferentiation of hIAPP+/- Mouse Pancreatic Islets during Serum-Deprived Culture
Metabolic stress-induced proinflammatory cytokines can induce the dedifferentiation of islet cells [25, 26]. Therefore, the mRNA transcription levels of β-cell identity markers were compared among the four groups. The mRNA transcription levels of Oct4 and NGN3 in the BSA+2D EV and BSA+3D EV groups were significantly lower than those of the BSA group. In the BSA+3D EV group but not in the BSA+2D EV group, the mRNA transcription levels of Pdx1 and FoxO1 were significantly higher than those in the BSA group, with a significantly higher level in the BSA+3D EV group than in the BSA+2D EV group (Figure 7(a)). The protein expression of β-cell identity markers was evaluated using immunocytochemical staining. Similar to real-time RT-PCR assays, immunostaining showed that the expression levels of Pdx1 and FoxO1 were significantly higher in the BSA+3D EV group than in the BSA group. The proportion of FoxO1-positive cells was ~3-fold higher in both the EVs (20 μg/mL) derived from the 3D hUCB-MSC and 2D hUCB-MSC groups than in the BSA group. The proportion of islet cells expressing Oct4 and NGN3 was downregulated in FBS, 2D hUCB-MSC-derived EVs, and 3D hUCB-MSC-derived EVs. The proportion of islets expressing Oct4 was lower in the 3D hUCB-MSC-derived EVs than in the 2D hUCB-MSC-derived EVs.
In addition to alleviating the dedifferentiation of islets, the proportion of islets expressing cleaved caspase-3 was significantly lower in both the 3D hUCB-MSC-derived EV and 2D hUCB-MSC-derived EV groups than in the BSA group (Figures 7(d) and 7(e)), whereas there were no significant differences in the numbers of hIAPP oligomer-positive cells between the BSA and 2D or 3D hUCB-MSC-derived EV groups.
## 4. Discussion and Conclusions
In this study, miRs involved in the M2 polarization of macrophages were more abundant in EVs derived from 3D hUCB-MSCs than in 2D hUCB-MSC- and 3D hUCB-MSC-derived EVs that possessed an enhanced M2 polarization ability on THP-1 monocytes. The protective effects of the 3D hUCB-MSC-derived EVs on hIAPP heterozygote transgenic mouse islets were more potent than 2D hUCB-MSC-derived EVs in terms of reducing nonspecific inflammation and preserving β-cell identity, and this was associated with a higher proportion of islet-resident macrophages with M2 polarization markers. To the best of our knowledge, these results are the first evidence that 3D hUCB-MSC-derived EVs have a direct protective effect on hIAPP-expressing pancreatic islets, and the M2-polarizing ability of the EVs might be an important contributor to such protective effects.
The greater M2-polarizing ability of EVs derived from 3D hUCB-MSCs compared to those derived from 2D hUCB-MSCs is a novel finding of this study. Although a recent study revealed that human adipose mesenchymal stem cell- (AMSC-) derived EVs can directly induce the M2 polarization of macrophages through the direct effect of miRs related to M2 polarization, such as that of miR34a-5p and miR 146a-5p [20], it has not been determined if the levels of such miRs can be amplified in 3D cultures of hUCB-MSCs. In the optimized 3D culture conditions in this study, miR34a-5p and miR 146a-5p were more abundant in EVs derived from 3D hUCB-MSCs than in those derived from 2D hUCB-MSCs, even when the 2D MSCs were preconditioned by IFN-γ and TNF-α. In addition, miR127-3p was less abundant in 3D than in 2D hUCB-MSC-derived EVs. miR-34 inhibits the transcription of proinflammatory cytokines by targeting Notch1, and miR-146 targets NF-κB signaling mediators such as IRAK1 and TRAF6 to promote the expression of M2-associated genes [20, 27, 28]. These results would be relevant to the enhanced protective effects of 3D MSC-derived EVs on β-cells because macrophages are associated with inflammation in islet cells and β-cell dysfunction [7]. The injection of hUCB-MSCs into a type 2 diabetes mouse model polarized M1 macrophages to M2 macrophages in islets [3], reduced apoptosis of β-cells, and increased PDX-1 and MafA expressions. The results of our study indicate that similar effects can be reproduced by 3D hUCB-MSC-derived EVs alone, without the help of cellular components.
In addition to comparing miRs related to M2 polarization, we compared the protein expression profiles of 2D and 3D hUCB-MSC-derived EVs. Proteomic analysis of 3D hUCB-MSC-derived EVs (Figure 3(b)) demonstrated markedly upregulated expression of CHI3L1, IL-6, monocyte chemoattractant protein-1 (MCP-1), and IL-27. CHI3L1 is an inducer of the PI3K/AKT signaling pathways [29] and M2 mature differentiation of macrophages for Th2 inflammation [30], IL-6 and MCP-1 are inducers of M2 mature differentiation of macrophages [31], and IL-27 is a cytokine involved in anti-inflammatory and immune-regulatory functions and inhibits Th2, innate lymphoid cell-2 (ILC2), and Th17 responses [32]. Moreover, compared to 2D hUCB-MSC-derived EVs, there was decreased expression of dickkopf 1 (Dkk-1), an inhibitor of Wnt signaling [33–35]. Wnt/β-catenin plays a role in the development and function of insulin-producing β-cells [36]. These differences in protein expression profiles might have contributed to the greater protective effects of 3D hUCB-MSC-derived EVs on islet inflammation and dedifferentiation in this study.
Although the results of our study are in line with previous research suggesting the superior immunomodulatory effects of 3D MSCs over 2D MSCs [15, 16, 18, 22], some previous studies reported contradictory results [19, 37]. In one study, T cell-suppressive abilities of MSCs were observed only in 2D MSCs and not in 3D MSCs; in that study, the T cell-suppressive abilities of 3D MSCs were partly restored by addition of a corticosteroid [37]. Another previous study compared the immunomodulatory potency of 2D MSC-derived EVs and 3D MSC-derived EVs in vitro and then compared their anti-inflammatory and antifibrotic potentials in vivo using a bleomycin-induced lung fibrosis model [19]. In that study, the in vitro immunomodulatory potency of 2D MSC-derived EVs and 3D MSC-derived EVs was compared after IFN-γ stimulation. The T cell suppression ability in terms of indoleamine 2,3-dioxygenase activity after IFN-γ stimulation and macrophage phenotype in terms of phagocytosis activity were lower in 3D MSC experiments, indicating polarization into the M1 subtype, although they did not directly measure M1 or M2 markers on macrophages [19]. In our study, the M2-polarizing ability of 3D MSCs was optimized when 3D MSCs were generated with sufficient cell numbers, without hypoxia or cytokine stimulation. In 25 K 3D MSCs, cytokine stimulation reduced the M2-polarizing ability. Therefore, it is possible that the immunomodulatory properties of 3D MSC-derived EVs could be lost if 3D culture conditions are not optimized. We also used a different source of MSCs (human UCB-MSCs in the current study vs. human lung tissue-derived MSCs or bone marrow-derived MSCs in the previous study [19]) and different disease models (serum-deprived culture of isolated pancreatic islets and cocultivation of hIAPP-producing pancreatic islets, MSCs, and macrophages). In this context, it is reassuring that the uniform-sized 3D MSCs produced by nanopatterned culture plasticware prevented β-cell death in a multiple low-dose streptozotocin-induced diabetes model [22], in which immune cell infiltration and progressive loss of β-cells typically occurs. Whether 3D MSC-derived EVs alone without the cellular component would have similar benefit should be explored in future research.
An important translational potential in our study is the use of EVs derived from 3D hUCB-MSCs in the pretransplant cultures of isolated human islets before clinical islet transplantation. Our previous study showed that serum deprivation, as in pretransplant cultures in clinical intraportal islet transplantation to avoid the use of animal-derived materials, induces an inflammatory response in cultured hIAPP+/- islets even without the prolonged culture with hyperglycemia [24]. This study hypothesized that the addition of EVs derived from 3D hUCB-MSCs during the pretransplant culture of islets could attenuate the inflammatory response and loss of β-cell identity caused by serum deprivation. In this study, 3D hUCB-MSC-derived EVs promoted the M2 polarization of islet-resident macrophages and reduced the inflammasome activation induced by serum deprivation in hIAPP heterozygote mouse islets. This resulted in improved islet viability and insulin response to glucose, attenuation of proinflammatory cytokine expression, and preservation of β-cell identity. In addition to the enhanced M2 polarization ability of 3D hUCB-MSC-derived EVs, this therapeutic effect could result from differences in the cargo and the proteome profiles of EVs derived from 2D and 3D hUCB-MSCs, which include several angiogenesis-related cytokines.
Several limitations of this study should be discussed. First, the 3D culture condition has not been optimized in experiments with primates or human islets using 3D hUCB-MSC-derived EVs. The concentration of 3D hUCB-MSC-derived EVs should be optimized using human islets before clinical application. Second, whether the results of this study can be reproduced by the systemic use of 3D hUCB-MSC-derived EVs in type 2 diabetes or islet transplant recipients remains unclear. Third, we assessed only a limited number of miRs that promoted M2 polarization of macrophages in a previous study using cytokine-pretreated 2D MSCs, rather than a systemic comparison of all miRs related to macrophage M1/M2 polarization. Fourth, we did not investigate whether 3D MSC-derived EVs could modulate immune responses of T cells and other components of immune system as well. Although we speculate that 3D MSC-derived EVs might be helpful for prevention of islet graft rejection by alloimmune responses as well, the data presented in this study is insufficient to support the hypothesis.
In conclusion, we found that EVs derived from 3D hUCB-MSCs can protect hIAPP-expressing pancreatic islets against nonspecific inflammation and dedifferentiation, and these benefits were significantly greater than those of 2D hUCB-MSC-derived EVs. These results might be, at least in part, due to the enhanced M2 polarization ability of the EVs derived from 3D hUCB-MSCs compared to those derived from 2D hUCB-MSCs on islet-resident macrophages.
## Data Availability
The datasets during the current study are available from the corresponding authors on reasonable request.
## Ethical Approval
All experimental protocols in this study were approved by the Institutional Animal Care and Use Committee of Samsung Biomedical Research Institute.
## Consent
With the approval of the Institutional Review Board of Samsung Medical Center, all participants provided informed consent for the use of the umbilical cord in this experimental study.
## Conflicts of Interest
The authors have no conflicts of interest to declare.
## Authors' Contributions
Eunwon Lee and Seungyeon Ha have contributed equally to this work.
## References
1. Si Y., Zhao Y., Hao H.. **Infusion of mesenchymal stem cells ameliorates hyperglycemia in type 2 diabetic rats: identification of a novel role in improving insulin sensitivity**. (2012) **61** 1616-1625. DOI: 10.2337/db11-1141
2. Sun X., Hao H., Han Q.. **Human umbilical cord-derived mesenchymal stem cells ameliorate insulin resistance by suppressing NLRP3 inflammasome-mediated inflammation in type 2 diabetes rats**. (2017) **8** p. 241. DOI: 10.1186/s13287-017-0668-1
3. Yin Y., Hao H., Cheng Y.. **Human umbilical cord-derived mesenchymal stem cells direct macrophage polarization to alleviate pancreatic islets dysfunction in type 2 diabetic mice**. (2018) **9** p. 760. DOI: 10.1038/s41419-018-0801-9
4. Ito T., Itakura S., Todorov I.. **Mesenchymal stem cell and islet co-transplantation promotes graft revascularization and function**. (2010) **89** 1438-1445. DOI: 10.1097/TP.0b013e3181db09c4
5. Park K. S., Kim Y. S., Kim J. H.. **Trophic molecules derived from human mesenchymal stem cells enhance survival, function, and angiogenesis of isolated islets after transplantation**. (2010) **89** 509-517. DOI: 10.1097/TP.0b013e3181c7dc99
6. Borg D. J., Weigelt M., Wilhelm C.. **Mesenchymal stromal cells improve transplanted islet survival and islet function in a syngeneic mouse model**. (2014) **57** 522-531. DOI: 10.1007/s00125-013-3109-4
7. Eguchi K., Nagai R.. **Islet inflammation in type 2 diabetes and physiology**. (2017) **127** 14-23. DOI: 10.1172/JCI88877
8. Kong D., Zhuang X., Wang D.. **Umbilical cord mesenchymal stem cell transfusion ameliorated hyperglycemia in patients with type 2 diabetes mellitus**. (2014) **60** 1969-1976. DOI: 10.7754/Clin.Lab.2014.140305
9. Wang H., Strange C., Nietert P. J.. **Autologous mesenchymal stem cell and islet cotransplantation: safety and efficacy**. (2018) **7** 11-19. DOI: 10.1002/sctm.17-0139
10. Bagno L., Hatzistergos K. E., Balkan W., Hare J. M.. **Mesenchymal stem cell-based therapy for cardiovascular disease: progress and challenges**. (2018) **26** 1610-1623. DOI: 10.1016/j.ymthe.2018.05.009
11. Lou G., Chen Z., Zheng M., Liu Y.. **Mesenchymal stem cell-derived exosomes as a new therapeutic strategy for liver diseases**. (2017) **49**. DOI: 10.1038/emm.2017.63
12. Liew L. C., Katsuda T., Gailhouste L., Nakagama H., Ochiya T.. **Mesenchymal stem cell-derived extracellular vesicles: a glimmer of hope in treating Alzheimer's disease**. (2017) **29** 11-19. DOI: 10.1093/intimm/dxx002
13. Börger V., Bremer M., Ferrer-Tur R.. **Mesenchymal stem/stromal cell-derived extracellular vesicles and their potential as novel immunomodulatory therapeutic agents**. (2017) **18** p. 1450. DOI: 10.3390/ijms18071450
14. Sun Y., Shi H., Yin S.. **Human mesenchymal stem cell derived exosomes alleviate type 2 diabetes mellitus by reversing peripheral insulin resistance and relieving**. (2018) **12** 7613-7628. DOI: 10.1021/acsnano.7b07643
15. Cesarz Z., Tamama K.. **Spheroid culture of mesenchymal stem cells**. (2016) **2016** 11. DOI: 10.1155/2016/9176357
16. Miceli V., Pampalone M., Vella S., Carreca A. P., Amico G., Conaldi P. G.. **Comparison of immunosuppressive and angiogenic properties of human amnion-derived mesenchymal stem cells between 2D and 3D culture systems**. (2019) **2019** 16. DOI: 10.1155/2019/7486279
17. Fierabracci A., Del Fattore A., Luciano R., Muraca M., Teti A., Muraca M.. **Recent advances in mesenchymal stem cell immunomodulation: the role of microvesicles**. (2015) **24** 133-149. DOI: 10.3727/096368913X675728
18. Gallo A., Cuscino N., Contino F.. **Changes in the transcriptome profiles of human amnion-derived mesenchymal stromal/stem cells induced by three-dimensional culture: a potential priming strategy to improve their properties**. (2022) **23** p. 863. DOI: 10.3390/ijms23020863
19. Kusuma G. D., Li A., Zhu D.. **Effect of 2D and 3D culture microenvironments on mesenchymal stem cell-derived extracellular vesicles potencies**. (2022) **10, article 819726**. DOI: 10.3389/fcell.2022.819726
20. Domenis R., Cifu A., Quaglia S.. **Pro inflammatory stimuli enhance the immunosuppressive functions of adipose mesenchymal stem cells-derived exosomes**. (2018) **8** p. 13325. DOI: 10.1038/s41598-018-31707-9
21. Bulati M., Miceli V., Gallo A.. **The immunomodulatory properties of the human amnion-derived mesenchymal stromal/stem cells are induced by INF-**. (2020) **11** p. 54. DOI: 10.3389/fimmu.2020.00054
22. Lee S., Kim H. S., Min B. H.. **Enhancement of anti-inflammatory and immunomodulatory effects of adipose-derived human mesenchymal stem cells by making uniform spheroid on the new nano-patterned plates**. (2021) **552** 164-169. DOI: 10.1016/j.bbrc.2021.03.026
23. Bieback K., Kern S., Kluter H., Eichler H.. **Critical parameters for the isolation of mesenchymal stem cells from umbilical cord blood**. (2004) **22** 625-634. DOI: 10.1634/stemcells.22-4-625
24. Jin S. M., Shim W., Oh B. J.. **Anakinra protects against serum deprivation-induced inflammation and functional derangement in islets isolated from nonhuman primates**. (2017) **17** 365-376. DOI: 10.1111/ajt.13953
25. Rui J., Deng S., Arazi A., Perdigoto A. L., Liu Z., Herold K. C.. (2017) **25** 727-738. DOI: 10.1016/j.cmet.2017.01.005
26. Talchai C., Xuan S., Lin H. V., Sussel L., Accili D.. **Pancreatic**. (2012) **150** 1223-1234. DOI: 10.1016/j.cell.2012.07.029
27. Jiang P., Liu R., Zheng Y.. **MiR-34a inhibits lipopolysaccharide-induced inflammatory response through targeting Notch1 in murine macrophages**. (2012) **318** 1175-1184. DOI: 10.1016/j.yexcr.2012.03.018
28. Vergadi E., Vaporidi K., Theodorakis E. E.. **Akt2 deficiency protects from acute lung injury via alternative macrophage activation and miR-146a induction in mice**. (2014) **192** 394-406. DOI: 10.4049/jimmunol.1300959
29. Wang S., Hu M., Qian Y.. **CHI3L1 in the pathophysiology and diagnosis of liver diseases**. (2020) **131, article 110680**. DOI: 10.1016/j.biopha.2020.110680
30. Xu N., Bo Q., Shao R.. **Chitinase-3-like-1 promotes M2 macrophage differentiation and induces choroidal neovascularization in neovascular age-related macular degeneration**. (2019) **60** 4596-4605. DOI: 10.1167/iovs.19-27493
31. Chung H. Y., Kim J. H., Han I. H., Ryu J. S.. **Polarization of M2 macrophages by interaction between prostate cancer cells treated with Trichomonas vaginalis and adipocytes**. (2020) **58** 217-227. DOI: 10.3347/kjp.2020.58.3.217
32. Fabbi M., Carbotti G., Ferrini S.. **Dual roles of IL-27 in cancer biology and immunotherapy**. (2017) **2017** 14. DOI: 10.1155/2017/3958069
33. Semënov M. V., Tamai K., Brott B. K., Kühl M., Sokol S., He X.. **Head inducer Dickkopf-1 is a ligand for Wnt coreceptor LRP6**. (2001) **11** 951-961. DOI: 10.1016/S0960-9822(01)00290-1
34. Bafico A., Liu G., Yaniv A., Gazit A., Aaronson S. A.. **Novel mechanism of Wnt signalling inhibition mediated by Dickkopf-1 interaction with LRP6/arrow**. (2001) **3** 683-686. DOI: 10.1038/35083081
35. Mao B., Wu W., Li Y.. **LDL-receptor-related protein 6 is a receptor for Dickkopf proteins**. (2001) **411** 321-325. DOI: 10.1038/35077108
36. Murtaugh L. C.. **The what, where, when and how of Wnt/**. (2008) **4** 81-86. DOI: 10.4161/org.4.2.5853
37. Burand A. J., Di L., Boland L. K.. **Aggregation of human mesenchymal stromal cells eliminates their ability to suppress human T cells**. (2020) **11** p. 143. DOI: 10.3389/fimmu.2020.00143
|
---
title: Diminished AdipoR1/APPL1 Interaction Mediates Reduced Cardioprotective Actions
of Adiponectin against Myocardial Ischemia/Reperfusion Injury in Type-2 Diabetic
Mice
authors:
- Chao Ren
- Wei Yi
- Bo Jiang
- Erhe Gao
- Jiali Liang
- Bo Zhang
- Zhe Yang
- Dezhi Zheng
- Yong Zhang
journal: Stem Cells International
year: 2023
pmcid: PMC9970717
doi: 10.1155/2023/7441367
license: CC BY 4.0
---
# Diminished AdipoR1/APPL1 Interaction Mediates Reduced Cardioprotective Actions of Adiponectin against Myocardial Ischemia/Reperfusion Injury in Type-2 Diabetic Mice
## Abstract
### Background
Obesity-related diseases have important implications for the occurrence, severity, and outcome of ischemic heart disease. Patients with obesity, hyperlipidemia, and diabetes mellitus (metabolic syndrome) are at increased risk of heart attack with decreased plasma lipocalin levels, and lipocalin is negatively correlated with heart attack incidence. APPL1 is a signaling protein with multiple functional structural domains and plays an important role in the APN signaling pathway. There are two known subtypes of lipocalin membrane receptors, AdipoR1 and AdipoR2. AdioR1 is mainly distributed in skeletal muscle while AdipoR2 is mainly distributed in the liver.
### Objective
To clarify whether the AdipoR1-APPL1 signaling pathway mediates the effect of lipocalin in reducing myocardial ischemia/reperfusion injury and its mechanism will provide us with a new approach to treat myocardial ischemia/reperfusion injury using lipocalin as an intervention and therapeutic target.
### Methods
[1] Induction of hypoxia/reoxygenation in SD mammary rat cardiomyocytes to simulate myocardial ischemia/reperfusion; [2] downregulation of APPL1 expression in cardiomyocytes to observe the effect of lipocalin on myocardial ischemia/reperfusion and its mechanism of action.
### Results
[1] Primary mammary rat cardiomyocytes were isolated and cultured and induced to simulate MI/R by hypoxia/reoxygenation; [2] lipocalin inhibited H/R-induced apoptosis in cardiomyocytes; and [3] APN attenuated MI/R injury through AdipoR1-APPL1 and the possible mechanism.
### Conclusion
This study demonstrates for the first time that lipocalin can attenuate myocardial ischemia/reperfusion injury through the AdipoR1-APPL1 signaling pathway and that the reduction of AdipoR1/APPL1 interaction plays an important role in cardiac APN resistance to MI/R injury in diabetic mice.
## 1. Introduction
In recent years, the widespread use of arterial bypass surgery, thrombolytic therapy, percutaneous transluminal coronary angioplasty, cardiac surgery with cardiopulmonary bypass, and cardiopulmonary resuscitation has made it possible to restore blood flow to the ischemic heart in a short time, leading to recovery of damaged structures, cardiac function, and improved health [1–4].
APN is an adipocytokine synthesized and secreted by white and brown adipose tissue [5]. APN has three functions. First, it is an important regulatory hormone that prevents insulin sensitivity and tissue inflammation [6]. APN acts directly on the liver, skeletal muscle, and blood vessels, increasing insulin sensitivity by modulating AMPK activity and also inhibiting vascular inflammatory responses [7, 8]. APN levels are further decreased, and the antagonistic effect of APN on TNF-α in turn inhibits APN synthesis [9, 10]. Furthermore, thiazolidinediones (which antagonize TNF-α) can increase APN secretion from adipocytes [11, 12]. APN also has atherosclerotic effects, and recent reports indicate that APN plays an important regulatory role in the development and progression of several cardiac diseases [13, 14].
APPL1 is a signaling protein with multiple functional structural domains [15]. Two subtypes of lipoalkali receptors are known, AdipoR1 and AdipoR2 [16, 17]. Numerous studies have shown that APN in skeletal muscle has a protective effect, mainly because it activates AMPK via APPL1 [18, 19]. Extensive recent experimental results have shown that the APPL1-AMPK signaling axis mediates the beneficial metabolic effects of lipocalin in the heart. In the myocardium, lipocalin can regulate myocardial glycolipid metabolism by activating downstream signaling molecules through the AdipoR1-APPL1 signaling pathway. The protective role of APPL1 in the early stages of myocardial ischemia/reperfusion has not been reported, and the possible involvement of APPL1-dependent signaling pathways is also unknown. Interactions with androgen receptor, epidermal growth factor (EGF) receptor, follicle-stimulating hormone receptor, and APN receptor have been reported. Several studies have shown that the interaction between APPL1 and AdipoR mediates the metabolic modulatory and endothelial protective effects of APN. However, whether the interaction between APPL1 and AdipoR1 (the major APN receptor in cardiomyocytes) mediates the cardioprotective effects of APN against MI/R disorders and, therefore, is involved in the mechanism of cardiac APN resistance in type-2 DM remains to be examined.
Lipocalin is a new target for the treatment of insulin suppression, type-2 diabetes, and metabolism-related diseases. The study of APPL1 may enhance our current understanding of the lipocalin and insulin signaling pathways. Furthermore, it is indispensable for our study of the lipocalin signaling pathway and the insulin-sensitizing effects of lipocalin. Although the study of APPL1 has been progressed in some fields, its role in lipocalin-mediated myocardial ischemia/reperfusion injury is limited. Does APPL1 participate in the role of lipocalin in reducing myocardial ischemia/reperfusion injury? What are the molecular mechanisms through which APPL1 mediates the protection of lipocalin against ischemia/reperfusion myocardium? Therefore, we used this experiment to demonstrate the role of APPL1 in the protection of myocardial ischemia/reperfusion by lipocalin. This study will provide new targets for more effective clinical treatment of coronary artery disease and ischemic heart disease.
## 2.1. Experimental Protocols
Forty-eight hours after RNAi injection, mice were anesthetized with $2\%$ isoflurane and reperfused through an incision on the left side of the chest to induce a myocardial infarction (MI). Twenty minutes after MI, mice were randomly injected with APN (2 μg/g) as vehicle or globule. At 30 min after infarction, the soluble compound was released, and the myocardium was resuscitated for 3 to 24 h (to determine cardiac function and infarct size). Control mice were operated sham, but the suture under the left coronary artery was not tightened.
## 2.2. Western Blot Analysis
The tissue was obtained from the MI mice 3 hours after reperfusion. The expressions of various proteins (APPL1, P-AMPK, AMPK, P-ACC, ACC, iNOS, and Caspase-3) were detected by Western blot as we previously described.
## 2.3. Neonatal Rat Primary Cardiomyocyte Culture
Preparation and dilution: prepare 100 ml of double antibody (add 5 ml saline to 5 ml streptomycin, shake well, then add 5 ml saline to the penicillin vial to dissolve, transfer all 5 ml to 95 ml saline, and divide into 10 vials). Prepare 100 ml DMEM culture solution containing $15\%$ fetal bovine serum (85 ml DMEM, 15 ml fetal bovine serum, and 1 ml double antibody), and prepare 10 ml $0.1\%$ collagenase I (10 mg collagenase I dissolved in 10 ml PBS). The solution is usually prepared in advance and filtered through a 0.22 pore filterIrradiation with UV light for 20-30 min (2 culture flasks, 2 15 ml centrifuge tubes, and 4 lunch boxes were placed on the UV table)SD mice immersed in $75\%$ alcohol (5~10 min)Add two drops of medium (DMEM + serum + double antibodies) to one vial of medium, and leave the other vial of medium undisturbed and screwed to one side of the ultrathin table (note: this step can be performed with a 50 ml pipette)Take another dropper with a straight tip, rinse it, insert PBS, then remove the round container set, and add the appropriate amount of PBSTake the heart. Six milk mice can be sewn sequentially onto a plastic foam plate using a broad-headed needle, and then open the chest sequentially using scissors to reveal the heart. Take another knife and forceps, lift the heart, cut it out, place it in a round dish with prechilled PBS, rinse the heart 2 times with PBS, and cut it out (1 mm × 1 mm × 1 mm)Remove the drop and aspirate the PBS; then take the penicillin vial and transfer the excised myocardial tissue into the penicillin vial into the penicillin vial. Add 2 drops of collagenase I (approximately 3 ml) to the penicillin vial with dropper. After 10 minutes, add collagenase I (approximately 3 ml) to the dropper vial, close tightly, and leave for 10 minutes in a water bath at 37°C. After 10 minutes, add the digestion solution to the centrifuge tube with DMEM to complete the digestionCentrifugation (room temperature, 1000 rpm, 10 min/time)*Through a* 200-mesh sieve, after centrifugation, discard the supernatant, and use the medium in a third tube to resuspend the cells. After centrifugation, discard the supernatant, resuspend the cell pellet with the medium in the third centrifuge tube, take another set of round dishes, pass the cell suspension through a 200-mesh sieve, and then collect the sieved cells
## 2.4. Induced Cardiomyocyte Hypoxia/Reoxygenation (H/R) Simulating Myocardial Ischemia/Reperfusion (MI/R)
The culture medium was replaced with ischemic blood (sodium hypodisulfate 0.75, KCl 12, lactate 20 mmol/L, pH 6.5). Three different anoxia/reoxygenation times were selected for experimental observation, which were sequentially set to incubate in anoxic incubator at 37°C: 6 h, 12 h, and 24 h. Then, the normal incubation medium was replaced and incubated in $5\%$ CO2 incubator at 37°C for 3 h, 6 h, and 12 h, respectively, for testing.
## 3.1. The AdipoR1-APPL1 Interaction Was Decreased in HD Mice
As an important regulator of APN metabolic signaling, APPL1 binds to AdipoR1 and mediates AMPK activation in the axon. We and others have shown that the cardioprotective effects of APN are attenuated in a diabetic state characterized by APN resistance. And we have found that the AMPK pathway and the antinitrogen/oxidant pathway are at least partially inhibited. However, we did not investigate whether the AdipoR1-APPL1 interaction is reduced in mice with type-2 diabetes. To determine the mechanism of APN resistance, we first detected the AdipoR1-APPL1 interaction by coimmunoprecipitation in HD-induced type-2 diabetic mouse hearts. Compared with HD hearts, neither AdipoR1 nor APPL1 expression was significantly altered in the protein level in HD mouse hearts. However, the coimmunoprecipitation of AdipoR1 and APPL1 in HD hearts was dramatically reduced ($51.6\%$), suggesting that the interaction between APPL1 and AdipoR1 was greatly reduced in HD mice (Figure 1).
## 3.2. Knockdown APPL1 Protein Expression by Infecting Stealth RNAi In Vivo
To investigate the role of the AdiopR1-APPL1 interaction in the in vivo regulation of APN cardioprotection, we deleted the APPL1 adaptor protein by intramyocardial injection of APPL1 stealth siRNA, as previously described. As shown in Figure 2 compared with scrambled siRNA injection, APPL1 protein levels in APPL1 knockout hearts were reduced by approximately $73.6\%$, as determined by Western blotting.
## 3.3. Cardioprotective Effect of APN Was Abolished in APPL1 Knockdown Mice
Having successful knockdown APPL1 expression by infecting APPL1 specific stealth RNAi, we determined to confirm whether the cardioprotective effect of APN would be affected. Figure 3 shows that the cardioprotective effect of APN to MI/R injury was partially abolished on the APPL1 knockdown mice. Specifically, without the administration of APN, knockdown APPL1 before MI/R has no statistical significant difference from the group with scramble siRNA. The conclusion above can also be seen from the result of echocardiography (Figure 3(b)). These results confirmed that APPL1 is an important molecule in APN-mediated cardioprotection. Since AdipoR1-APPL1 interaction was reduced in HD mice, the cardioprotective effect of APN was diminished in APPL1 knockdown mice, and we propose that reduced AdipoR1-APPL1 interaction is the mechanism of APN resistance in type-2 DM.
## 3.4. Inhibition of APPL1 Expression Blocked APN's Phosphorylation of AMPK-ACC Axis
Previous studies have shown that APN stimulates AMPK/ACC phosphorylation [11]. To determine the mechanisms responsible for the APN-induced loss of cardiopulmonary protection in APPL1 knockout mice, we first assessed the phosphorylation stimulated by AMPK and ACC. Compared with controls, phosphorylation of AMPK or ACC was increased in all MI/R-treated groups. Phosphorylation of AMPK and ACC was significantly increased after MI/R, but there was no statistically significant difference between MI/R groups for APPL1 or scramble siRNA (Figures 4(a) and 4(b)). Compared with controls, administration of gAPN (with siRNA bracketed siRNA) further increased the phosphorylation of AMPK and ACC. In addition, gAPN treatment partially reduced phosphorylation of AMPK and ACC in APPL1 knockout hearts (Figures 4(a) and 4(b)), indicating that APPL1 is important for the APN-mediated AMPK signaling pathway.
## 3.5. Cardiac Specific Injection of APPL1 Stealth RNAi Blunted Antinitrative and Antiapoptotic Effects of APN
As an overproduction of peroxynitrite, inducible nitric oxide synthase (iNOS) indicates a lack of source. As shown in Figure 5(a), APPL1 deficiency treatment of APPL1 knockdown hearts with gAPN increased MI/R-induced iNOS expression to levels comparable to those observed in normal hearts with APPL1 with gAPN, suggesting that APPL1 knockdown expression may impair the antinitrative effect of APN.
Our aforementioned experimental results showed that APPL1 is required for both AMPK-dependent and AMPK-independent APN signaling. We further evaluated the antiapoptotic effect of APN. Myocardial apoptosis was determined by the caspase-3 expression. As shown in the figure, no significant difference in caspase-3 expression was observed between normal hearts and APPL1 beating hearts. Treatment of wild-type mice as well as APPL1 knockdown mice with gAPN significantly suppressed caspase-3 expression. However, the absolute level of caspase-3 expression remained higher in gAPN-treated APPL1 knockdown hearts than in normal APPL1-treated hearts (Figure 5(b)). Together with the data presented in Figure 5(b), the results demonstrated that the MI/R-induced decrease in caspase-3 expression under gAPN was partially abolished when APPL1 was knocked down, indicating that APPL1 acts on the antiapoptotic effects of APN.
Our previous studies have demonstrated that HD impaired APN-induced AMPK-ACC activation and antinitrative protection [2]. Moreover, blunted antiapoptotic effects of APN in APPL1 knockdown mice indirectly indicating the effect of APPL1 in HD induced type-2 diabetic mice. Having determined that the AdipoR1-APPL1 interaction was reduced in HD mice, demonstrating that reduced AdipoR1-APPL1 interaction is the molecular mechanism of the cardiac APN resistance in HD induced type-2 DM.
## 3.6. APPL1 Stealth RNAi Blunted Anti-Cell Death Action of gAPN in Neonatal Cardiomyocyte
Yet in vivo experimental results, we have demonstrated that the interaction of AdipoR1 and APPL1 was decreased in HD-induced DM and the effect of APPL1 in APN preventing MI/R injury in cardiac. To obtain more evidence to support our plan, an additional study was performed by cultured neonatal rat primary cardiomyocytes. Cardiomyocytes were treated with APPL1 siRNA or scrambled siRNA and hypoxia/reoxygenation to simulate ischemia/reperfusion (SI/R) in the presence or absence of APN. First, the effect of APPL1 siRNA in cardiomyocytes was assessed by APPL1 expression. The APPL1 protein level was decreased by approximately $50\%$ in the cardiomyocyte with APPL1 siRNA, which is a little higher than the results in vivo (Figure 6(a)). Since the APPL1 expression was successfully knockdown, the effect of gAPN on SI/R-induced cell death (Figure 6(b)) demonstrates that knockdown APPL1 is capable of blocking APN-induced cardiomyocyte anti-cell death.
## 3.7. Inhibition of APPL1 Expression Partially Mimicked APN Resistance in HGHL-Treated Neonatal Cardiomyocyte
Our aforementioned in vivo experiment results have demonstrated that AdipoR1-APPL1 interaction is the molecular mechanism of the cardiac APN resistance in HD-induced type-2 DM. To further confirm it in vitro, HGHL cardiomyocytes were performed to simulate induction of type-2 DM as described previously to see APN resistance. As shown in Figure 7, cardiomyocyte apoptosis was determined by TUNEL. However, the effect of APN is partially abolished on cardiomyocyte cultured with either HGHL or APPL1 siRNA, which indicates that inhibition of APPL1 expression partially mimicked APN resistance in HGHL-treated neonatal cardiomyocyte. From Figure 7, cardiomyocyte culture in HGHL showed significantly increased TUNEL-positive cells, as provided from our previous study: type-2 diabetics manifested greater MI/R injury suffering.
## 4. Discussion
Myocardial ischaemia/reperfusion is an important risk factor for heart failure, and many studies have been proposed to identify the main mechanisms leading to cardiac remodelling. In particular, apoptosis-induced cardiomyocyte depletion is an important factor contributing to the progressive worsening of left ventricular hypertrophy, which may eventually lead to end-stage cardiomyopathy. With the development of research, adipocytokines are considered important factors regulating pathophysiological changes in heart failure and are strongly associated with the occurrence, progression, and prognosis of cardiovascular disease. The role of lipocalin in reducing myocardial apoptosis has been clearly demonstrated in numerous studies, effectively reducing acute myocardial ischaemia/reperfusion injury as well as apoptosis induced by chronic coronary syndrome. Recently, lipocalin has been shown to reduce cardiomyocyte apoptosis via AMPK-dependent signaling pathways and to reduce tumour necrosis factor (TNF-α) production via COX-2/PGE2-dependent molecular signaling mechanisms; lipocalin also exerts these protective effects by reducing oxidative/nitrosative stress and Akt activation. However, the exact mechanism remains to be elucidated.
Lipocalin exerts its physiological effects by binding to lipocalin receptors on the cell membrane and regulating several metabolic processes. Studies [8] using a yeast two-hybrid approach identified the first protein, APPL1, which binds directly to lipocalin receptors expressed in insulin-sensitive tissues such as muscle, liver, and adipose tissue and contributes to lipocalin signaling. APPL1 has several functional structural domains, of which the PTB structural domain binds to lipocalin receptor I (AdipoR1) and plays an important role in the lipocalin signaling pathway. Studies [1] have shown that in skeletal muscle, lipocalin activates AMPK via an APPL1/LKB1-dependent signaling pathway. In turn, lipocalin stimulates APPL1 interaction with AdipoR1. Increasing mutant APPL1 levels without the ability to bind AdipoR1 or inhibiting APPL1 with siRNA significantly attenuated lipocalin-induced AMPK, p38MAPK and ACC phosphorylation, and fatty acid oxidation, suggesting that APPL1 plays an important role in lipocalin-regulated lipid metabolism in skeletal muscle cells; APN promotes LKB1, mainly through APPL1/LKB1 APN that promotes the intracytoplasmic localisation of LKB1, which ultimately leads to AMPK activation, mainly through APPL1/LKB1-dependent signaling pathways.
Studies [5] on the effects of lipocalin on myocardial metabolism using primary cardiomyocyte culture and isolated cardiac perfusate showed that lipocalin increases the interaction between AdipoR1 and APPL1, and finally, APPL1 binds to AMPK and is phosphorylated to inhibit ACC and increase fatty acid oxidation; using siRNA to decrease APPL1 expression in rat mammary cardiomyocytes, lipocalin function was inhibited. These experiments confirmed that APPL1 plays an important role in lipocalin-mediated myocardial metabolism. It is unclear whether lipocalin-stimulated AdipoR1 binding to APPL1 is involved in the role of lipocalin in reducing myocardial ischaemia/reperfusion.
In this study, we examined the important role of APPL1 in reducing myocardial injury during ischemia/reperfusion by isolating and culturing primary SD mouse cardiomyocytes and creating a hypoxia/reoxygenation model to simulate myocardial injury during ischemia/reperfusion. Primary isolation and culture of milk rat SD cardiomyocytes were used to simulate ischemia/reperfusion injury by hypoxia/reoxygenation of cells, and it was found that hypoxia/reoxygenation can significantly increase cardiomyocyte apoptosis and that after 12 h of hypoxia, a large number of cardiomyocytes undergo apoptosis, but that apoptosis is significantly reduced after lipocalin treatment, suggesting that lipocalin may play a cardioprotective role in I/RI. In addition, APPL1 expression in cardiomyocytes was significantly higher in the lipocalin-treated group compared with the control and H/R groups. To further test the role of APPL1 in the protective mechanism of apoptotic damage to cardiomyocytes by lipocalin, we found that the apoptosis index of cardiomyocytes was significantly increased after the inhibition of APPL1 expression in cardiomyocytes by RNA interference, which blocked the antiapoptotic effect of lipocalin on cardiomyocytes during I/RI, confirming that lipocalin inhibition of I/RI-induced cardiomyocyte apoptosis was associated with APPL1. In addition, we examined cell survival by MTT assay, which was significantly higher after lipocalin treatment compared with the H/R group and significantly lower in the interferon group compared with the H/R+g Ad group after inhibition of APPL1 expression. The expression levels of p-AMPK, p-e NOS and iNOS were determined by Western blot, and it was found that the expression levels of p-AMPK, p-e NOS, and iNOS were significantly reduced and increased after APPL1 inhibition. It is suggested that lipocalin may reduce myocardial injury during ischemia/reperfusion via the AdipoR1-APPL1-AMPK signaling pathway and that the AdipoR1-APPL1 signaling axis is also involved in the antioxidant/nitrosative stress effect of lipocalin.
## Data Availability
The experimental data used to support the findings of this study are available from the corresponding author upon request.
## Conflicts of Interest
The authors declared that they have no conflicts of interest regarding this work.
## References
1. Yi W., Sun Y., Yuan Y.. **C1q/tumor necrosis factor-related protein-3, a newly identified adipokine, is a novel antiapoptotic, proangiogenic, and cardioprotective molecule in the ischemic mouse heart**. (2012) **125** 3159-3169. DOI: 10.1161/CIRCULATIONAHA.112.099937
2. Holland W. L., Miller R. A., Wang Z. V.. **Receptor-mediated activation of ceramidase activity initiates the pleiotropic actions of adiponectin**. (2011) **17** 55-63. DOI: 10.1038/nm.2277
3. Ma Y., Liu Y., Liu S.. **Dynamic alteration of adiponectin/adiponectin receptor expression and its impact on myocardial ischemia/reperfusion in type 1 diabetic mice**. (2011) **301** E447-E455. DOI: 10.1152/ajpendo.00687.2010
4. Lee B., Shao J.. **Adiponectin and energy homeostasis**. (2014) **15** 149-156. DOI: 10.1007/s11154-013-9283-3
5. Wang Z. V., Scherer P. E.. **Adiponectin, the past two decades**. (2016) **8** 93-100. DOI: 10.1093/jmcb/mjw011
6. Xing Y., Sui Z., Liu Y.. **Blunting TRPML1 channels protects myocardial ischemia/reperfusion injury by restoring impaired cardiomyocyte autophagy**. (2022) **117** 1-26. DOI: 10.1007/s00395-022-00930-x
7. Li T., Tan Y., Ouyang S., He J., Liu L.. **Resveratrol protects against myocardial ischemia-reperfusion injury via attenuating ferroptosis**. (2022) **808, article 145968**. DOI: 10.1016/j.gene.2021.145968
8. Cho S. A., Joo H. J., Cho J. Y.. **Visceral fat area and serum adiponectin level predict the development of metabolic syndrome in a community-based asymptomatic population**. (2017) **12**. DOI: 10.1371/journal.pone.0169289
9. Van Berendoncks A. M., Conraads V. M.. **Functional adiponectin resistance and exercise intolerance in heart failure**. (2011) **8** 113-122. DOI: 10.1007/s11897-011-0056-6
10. Fang X., Palanivel R., Cresser J.. **An APPL1-AMPK signaling axis mediates beneficial metabolic effects of adiponectin in the heart**. (2010) **299** E721-E729. DOI: 10.1152/ajpendo.00086.2010
11. Yi W., Sun Y., Gao E.. **Reduced cardioprotective action of adiponectin in high-fat diet-induced type II diabetic mice and its underlying mechanisms**. (2011) **15** 1779-1788. DOI: 10.1089/ars.2010.3722
12. Luo Y., Liu M.. **Adiponectin: a versatile player of innate immunity**. (2016) **8** 120-128. DOI: 10.1093/jmcb/mjw012
13. Qi G. M., Jia L. X., Li Y. L., Li H. H., Du J.. **Adiponectin suppresses angiotensin II-induced inflammation and cardiac fibrosis through activation of macrophage autophagy**. (2014) **155** 2254-2265. DOI: 10.1210/en.2013-2011
14. Scherer P. E.. **The multifaceted roles of adipose tissue-therapeutic targets for diabetes and beyond: the 2015 Banting Lecture**. (2016) **65** 1452-1461. DOI: 10.2337/db16-0339
15. Wang Y., Lau W. B., Gao E.. **Cardiomyocyte-derived adiponectin is biologically active in protecting against myocardial ischemia-reperfusion injury**. (2010) **298** E663-E670. DOI: 10.1152/ajpendo.00663.2009
16. Shanmugam K., Prem P. N., Boovarahan S. R., Sivakumar B., Kurian G. A.. **FIsetin preserves interfibrillar mitochondria to protect against myocardial ischemia-reperfusion injury**. (2022) **80** 123-137. DOI: 10.1007/s12013-021-01026-4
17. Roger V. L., Go A. S., Lloyd-Jones D. M.. **Heart disease and stroke statistics--2011 update: a report from the American Heart Association**. (2012) **123** e18-e209
18. Wang Y., Wang X., Jasmin J. F.. **Essential role of caveolin-3 in adiponectin signalsome formation and adiponectin cardioprotection**. (2012) **32** 934-942. DOI: 10.1161/ATVBAHA.111.242164
19. Lau W. B., Tao L., Wang Y., Li R., Ma X. L.. **Systemic adiponectin malfunction as a risk factor for cardiovascular disease**. (2011) **15** 1863-1873. DOI: 10.1089/ars.2010.3743
|
---
title: Mechanism of Jiawei Zhengqi Powder in the Treatment of Ulcerative Colitis Based
on Network Pharmacology and Molecular Docking
authors:
- Chao Zhao
- ChenYang Zhi
- JianHua Zhou
journal: BioMed Research International
year: 2023
pmcid: PMC9970719
doi: 10.1155/2023/8397111
license: CC BY 4.0
---
# Mechanism of Jiawei Zhengqi Powder in the Treatment of Ulcerative Colitis Based on Network Pharmacology and Molecular Docking
## Abstract
### Objective
Ulcerative colitis is an intestinal condition that severely affects the life quality of a patient. Jiawei Zhengqi powder (JWZQS) has some therapeutic benefits for ulcerative colitis. The current study investigated the therapeutic mechanism of JWZQS for ulcerative colitis using a network pharmacology analytical approach.
### Methods
In this study, network pharmacology was used to investigate the potential mechanism of JWZQS in treating ulcerative colitis. The common targets between the two were identified, and a network map was created with the Cytoscape software. The Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses of JWZQS was performed using the Metascape database. Protein-protein interaction networks (PPI) was created to screen core targets and main components, and molecular docking was conducted between the main components and core targets. The expression levels of IL-1β, IL-6, and TNF-α were detected in animal experiments. Their effect on the NF-κB signaling pathway and the protective mechanism of JWZQS on the colon by tight junction protein were investigated.
### Results
There were 2127 potential ulcerative colitis targets and 35 components identified, including 201 non-reproducible targets and 123 targets shared by drugs and diseases. Following the analysis, we discovered 13 significant active components and 10 core targets. The first 5 active ingredients and their corresponding targets were molecularly docked, and the results showed a high level of affinity. GO analysis showed that JWZQS participate in multiple biological processes to treat UC. KEGG analysis showed that JWZQS may be involved in regulating multiple pathways, and the NF-κB signaling pathway was selected for analysis and verification. JWZQS has been shown in animal studies to effectively inhibit the NF-κB pathway; reduce the expression of IL-1β, TNF-α, and IL-6 in colon tissue; and increase the expression of ZO-1, Occludin, and Claudin-1.
### Conclusion
The network pharmacological study provides preliminary evidence that JWZQS can treat UC through multiple components and targets. JWZQS has been shown in animal studies to effectively reduce the expression levels of IL-1β, TNF-α, and IL-6, inhibit the phosphorylation of the NF-κB pathway, and alleviate colon injury. JWZQS can be used in clinical, but the precise mechanism of UC treatment requires further investigation.
## 1. Introduction
Ulcerative colitis (UC) is an inflammatory condition that primarily affects the mucosa and submucosa of the large intestine (rectum and colon). It is distinguished by large intestinal mucosal ulceration and persistent inflammation [1]. Mucous, diarrhea, hemafecia, pus, stomachache, and tenesmus were the most common clinical symptoms [2]. Medication, surgery, and antibody therapy are the three main types of current UC treatments. The most common existing therapies include aminosalicylates, corticosteroids, and immunosuppression, all of which have negative side effects and necessitate long-term medication [3]. Therefore, safer and more effective therapeutic approaches for UC are required.
In traditional Chinese medicine (TCM), there are several ways to treat UC, which is also known as diarrhea or dysentery. TCM provides a comprehensive understanding of a patient's physical condition based on their pulsation, tongue, signs and symptoms, and modern laboratory tests. Although therapeutic regimens vary depending on constitution, the general approach is to clear heat; expel dampness; strengthen the spleen, kidney, and qi; promote blood circulation; and dispel blood stagnation. JWZQS decoction contains Houpo, Chenpi, Banxia, Danggui, Chuanxiong, Huoxiang, Cangzhu, Xianhecao, and Shengma. The original prescription comes from a Ming Dynasty work by Xue Ji called the “Waike Jingyao.” It has Qi-widening, spleen-energizing, and dampening effects. The current study makes reference to Andrew L. Hopkins' theory of “network pharmacology” and Professor Li Shao's theory of “correlation between TCM and biomolecular network” [4, 5]. Network pharmacology, which combines pharmacology and systems biology, is an analytical system suitable for the study of complex components of traditional Chinese medicine.
The current study used the network pharmacology approach to construct the “drug-components-targets” network. For the core target and pathway verification, the KEGG and GO enrichment analyses were performed. Molecular docking technique was used to study the affinity between active components and key targets in JWZQS. The flow chart for the study is shown in Figure 1.
We used dextran sulfate sodium salt (DSS) to cause colon injury in mice, developed a UC mouse model, and assessed the clinical symptoms of the mice. Mouse colons were collected, colonic lengths were measured, and the contents and expressions of TNF-α, IL-6, and IL-1β in tissues were measured. The phosphorylation of IκB-α and P65 in the NF-κB pathway was detected. Furthermore, the tissue was sliced to examine the damage to the colon. Immunohistochemistry (IHC) detected the expression of ZO-1, Occludin, and Claudin-1 in the mouse colon tissue.
## 2. Article Types
The article type was original research.
## 3.1. Collection of Components and Targets of JWZQS
We use the search terms “ Houpo, Chenpi, Banxia, Danggui, Chuanxiong, Huoxiang, Cangzhu, Xianhecao, and Shengma” in the TCM database of system pharmacology (TCMSP, http://tcmspw.com). Following the bioactive constituents criteria, active constituents with drug similarity (DL) not less than 0.18 and oral bioavailability (OB) not less than $30\%$ were screened [6]. Based on TCMSP, we identified potential targets of active JWZQS constituents. Using the UniProt database, species were restricted to “Homo sapiens,” and annotation normalization was applied to the target genes [7].
## 3.2. Identification and Collection of Disease Gene Targets
TGenecards (https://www.genecards.org), OMIM (https://www.omim.org), DrugBank (https://go.drugbank.com), and DisGeNet (https://www.disgenet.org) databases were searched for UC, and human genes were chosen [8–11]. After collecting disease-associated genes from the four databases, duplicate values were removed, yielding 2127 disease-related gene targets. Drugs and diseases may have similar targets for disease management [12]. String database (https://string-db.org) was used to screen out the common targets.
## 3.3. Construction of Drug-Components-Targets Network
Cytoscape software (3.10.1 version) is used to associate drugs, active ingredients, targets, and diseases and to build an interaction network diagram between components and targets. Edges represent node relationships, while nodes represent active ingredients and targets. The main active components of the drugs were then examined.
## 3.4. Construction of PPI Network and Screening of Core Targets
To establish the interaction between the two proteins, the common target was imported into the Cytoscape database (https://cytoscape.org), and restriction conditions were set for the human species [13]. The database calculated the interrelationship, closeness, degree, feature vector, average local connectivity, and network for network analysis. Finally, the traditional “degree” value was employed to assess the importance of PPI network nodes.
## 3.5. Analysis of Enrichment and Associated Signaling Pathways
To study JWZQS's biological process and signaling pathway in managing UC, we used the Metascape website to analyze molecular function (MF), biological process (BP), and cell component (CC) data. Their sequencing is determined by the number of targets involved in biological processes and signaling pathways.
## 3.6. Molecular Docking
We downloaded the protein crystal architectures used for docking from the PDB database (https://www.rcsb.org) in addition to the 3D structures of the small molecules irisolidone, naringenin, nobiletin, quercetin, and wogonin from the PubChem (https://pubchem.ncbi.nlm.nih.gov) database [14].
AutoDock Vina 1.1.2 was used in the current study to perform molecular docking [15]. PyMol 2.5.4 was used to remove small molecules, salt ions, and water molecules from the receptor proteins before docking [16]. A docking box was created to contain the entire protein architecture. The entire processed receptor proteins and small molecules were converted into the PDBQT format via ADFR suite 1.0 to prepare AutoDock Vina 1.1.2 docking [17]. While connecting, keep the default settings for interconnection parameters. The binding confirmation was determined to be the output docking confirmation with the highest score. The final step involved a visual analysis based on PyMol 2.5.4 docking results.
## 3.7. Animal Experiment
Animal experiments were carried out following relevant legal principles and were approved by the Institutional Animal Management and Use Committee of Jilin University (Changchun, China) according to the procedures (License No.: SYXK2018–0001). Suzhou Xishan Biotechnology Co. LTD. provided male Kunming mice (6~8 weeks old). Thirty Kunming mice were randomly divided into five groups: the blank group (NT), the model group (MOD), the sulfasalazine group (SASP), the JWZQS low-dose group (JWZQS-L), the JWZQS medium-dose group (JWZQS-M), and the JWZQS high-dose group (JWZQS-H). Each group had five mice. For UC model establishment, except for the control group, the other mice were given $3\%$ DSS solution for 7 days, according to the literature [18].The drugs studied in this experiment were obtained from the granule Pharmacy of the Affiliated Hospital of the Changchun University of Chinese Medicine. Houpo, Chenpi, Banxia, Danggui, Chuangxiong, Huoxiang, Cangzhu, Xianhecao, and Shengma were modulated in a ratio of 1.5: 1.5: 1.5: 2: 1: 1.5: 1: 3: 1 (batch numbers: 21097204, 21087554, 21076314, 21096634, 21106334, 21077654, 21097014, 21096254, and 21078184). According to the experimental drug dose recommended by FDA (Food and Drug Administration), the equivalent dose of mice (11.7 g/kg) was 9.0 times that of humans (93 g original drug/70 kg body weight = 1.3 g/kg). The equivalent dose was set as the JWZQS-M group (11.7 g/kg). The JWZQS-H group (23.4 g/kg) had twice the concentration as the JWZQS-M group, and the JWZQS-L group (5.85 g/kg) had 0.5 times the concentration as the JWZQS-M group. The SASP (Shanghai Xinyi Co. LTD.) group 100 mg/kg SASP suspension by gavage [19]. Gavage was used to administer the corresponding volume of distilled water to the NT group.
## 3.8. Clinical Scoring and Sample Collection
The disease activity index (DAI) scoring system (Table 1) was used to determine body weight (BW), fecal occult blood, fecal characteristics, and clinical symptoms. Finally, colons of mice were collected for relevant experimental analysis [20].
## 3.9. ELISA
TTNF-α, IL-1β, and IL-6 expression levels in mouse colon tissues were measured using the ELISA kit (ELISA MAX Deluxe Set Mouse IL-1β 432604, MAX Deluxe Set Mouse IL-6 431315, and ELISA MAX Deluxe Set Mouse TNF-α430915).
## 3.10. qRT-PCR Assay
Colon tissues from mice were collected and treated to extract mRNA, and the levels of IL-1β, TNF-α, and IL-6 were determined using qRT-PCR as previously described [21]. Colon tissues from mice were collected and treated to extract mRNA, and the levels of IL-1β, TNF-α, and IL-6 were determined using qRT-PCR as previously described [21]. Table 2 lists all of the substrates that were used. The arithmetic formula 2−ΔΔCT was used to obtain the relative quantitative results.
## 3.11. Western Blotting (WB) Assay
The total protein of colonic tissue was extracted in this study by grinding RIPA lysate. BCA (Thermo) reagent was used to measure colonic histamine levels in mice (Sevier Biotechnology, Wuhan, China). SDS-PAGE with $12\%$ was used to isolate the protein (15 μg), which was then transferred to a PVDF membrane (Millipore, Darmstadt, Germany). The PVDF membrane was then blocked for 2 h at room temperature with $5\%$ skim milk. PVDF membrane and primary antibody NF-κB p65 (Servicebio GB11997 1: 300), NF-κB p-p65 (Servicebio GB111421 1: 300), IκB-α (Abcam EPR20992 1: 2000), p-IκB-α (ser32 1: 2000), and GAPDH (Servicebio GB15002 1: 2000) were stored overnight at 4°C and washed with TBST lotion. The PVDF membrane was then washed in TBS-T washing solution and incubated for 2 h with goat anti-mouse/rabbit secondary antibody (Servicebio GB25301/GB23303 1: 5000). The trend of protein bands was then detected using an enhanced chemiluminescence solution.
## 3.12. Hematoxylin and Eosin (H&E) Staining
Tissues from mouse colons were fixed in $4\%$ formaldehyde and paraffin before being sliced into 5 μm sections for H&E staining. After staining, the histology was examined under a light microscope. Finally, histological scores were computed following Table 3 [22].
## 3.13. Immunohistochemistry (IHC)
Immunohistochemical analysis of mouse colon tissue was performed to observe the contents of ZO-1, Occludin, and Claudin-1 proteins in mouse colon tissue to further understand the protective mechanism of JWZQS on mouse colon. The sections were incubated in a citrate antigen repair solution for 20 min at 95°C. These sections were incubated overnight with the primary antibodies ZO-1 (D6L1E 1: 100), Occludin (ab216327 1: 100), and Claudin-1 (ab15098 1: 100) and then with the secondary antibodies for 50 min before being examined under a microscope.
## 3.14. Statistical Analysis
The data is presented as the mean ± SD. Prism 8.0 was used to compare two groups using unpaired Student's t-tests. Prism 8.0.20 was used to perform ANOVA (general linear model) comparisons of more than two groups. Nonparametric tests must be used to evaluate clinical and pathology scores.
## 4.1. Collection of Active Components of JWZQS and Study of Overlapping Targets
The following nine Chinese medicines were searched in the TCMSP database: Houpo, Chenpi, Banxia, Danggui, Chuangxiong, Huoxiang, Cangzhu, Xianhecao, and Shengma. These compounds' screening criteria were DL (≥0.18) and OB (≥$30\%$). The TCMSP database was used to screen the targets linked to active ingredients, after which the target information was compared, and the gene name was adjusted using the UniProt database [23]. After matching, 670 targets were discovered, and 201 targets were discovered after removing duplicate values, as shown in Table 4.
Using the search term “ulcerative colitis,” the human genome annotation databases GeneCards, OMIM, DisGeNet, and DrugBank were searched for disease-related genes. There were 2127 acquired disease targets. The acquired target genes were compared with the genes associated with the aforementioned drug active ingredients. The String website generated a Venn diagram to screen for common genes, as shown in Figure 2(a).
## 4.2. Construction of “Drug-Components-Targets,” “Drug-Components-Targets-Disease,” and “PPI” Network Maps and Screening of Core Targets
The obtained active ingredient and genetic data will be used to generate “Drug-Components-Targets” and “Drug-Components-Targets-Disease” network maps using Cytoscape software. Figures 2(b) and 2(c) depict the intuitive relationship between drugs, diseases, and targets. Thirteen components, including quercetin, naringenin, wogonin, nobiletin, and irisolidone, were linked to disease targets and may be useful in the treatment of UC.
With the common gene data imported, the String database was used to generate a protein-protein interaction network (PPI) interaction map with the species “Homo Sapiens” as the designation. The TSV file was exported with the lowest interaction score (0.7). The Cytoscape software was used to calculate the major protein interactions. We imported the 123 common targets in the intersection into the String database to generate the interaction map. We import the interaction diagram into Cytoscape software to generate the PPI network diagram (Figure 3(a)).
TAccording to the network, the core targets were AKT1, IL6, TP53, TNF, IL-1β, VEGFA, TPGS2, CASP3, HIF1A, MAPK3, and other genes, ranked by degree value (Figures 3(b) and 3(c)).
## 4.3. GO and KEGG Analyses
The GO function and KEGG pathway of JWZQS were investigated using Metascape to better understand the compound's treatment mechanism on UC. In terms of BP, the therapeutic effect of JWZQS on UC is mainly due to cellular responses to lipid, hormones, substances, lipid peptides, radiation, positive regulation of cell motility, and other factors. Protein homodimerization activity, cytokine activity, transcription factor binding, protein domain-specific binding, DNA binding, kinase binding, and other activities are all part of the MF project. In terms of CC, JQZQS may influence membrane rafts, transcription regulator complexes, membrane sides, vesicle lumens, and other cell structures. JWZQS may be used to treat UC by interfering with these BP (Figure 4(a)). Cancer, lipid and atherosclerosis, diabetic complications, chemogenic receptor activation, cellular senescence, platinum resistance, and diabetic cardiomyopathy were the most affected pathways in the KEGG analysis (Figure 4(b)). We discovered numerous associations between rich pathways and additional pathological effects, which could be attributed to shared molecular targets in various diseases (Figure 4(c)). JWZQS may regulate the expression of TNF-α, IL-6, and IL-1β in the NF-κB pathway. It may control the phosphorylation of IκB-α and P65 to prevent inflammation (Figure 5).
## 4.4. Molecular Docking
We discovered in the PPI network diagram that the core targets were AKT1, IL6, TP53, TNF, and IL-1β. In the “drug-components-targets-disease” network diagram, quercetin, naringenin, wogonin, nobiletin, and irisolidone are discovered to be the main active components of JWZQS. We used molecular docking to find their corresponding relationship in the drug-components-targets-disease network diagram. As shown in Figure 2(c), Vina 1.1.2 was used to investigate the docking of compounds irisolidone, naringenin, nobiletin, quercetin, and wogonin with IL-1β, AKT1, TNF, IL-6, and TP53 proteins, respectively (Figure 6).
TA negative binding affinity indicates the likelihood of binding, and when the affinity value is less than –6 kcal/mol, the binding is frequently thought to be highly likely. Docking scores for AKT1 and naringenin are –8.1 (kcal/mol), IL-1β and irisolidone are –6.4 (kcal/mol), IL-6 and quercetin are –7.4 (kcal/mol), TNF and nobiletin are –9.1 (kcal/mol), and TP53 and wogonin are –6.6 (kcal/mol). It is worth noting that AKT1 has a high affinity with naringenin and TNF with nobiletin. Then, we dock all ligands for each docked protein, to observe the interaction between the core targets and other active components (Table 5).
## 4.5. JWZQS Inhibited the Expression of the NF-κB Pathway
We examined the levels of inflammatory factors in the colons of mice in each group. We discovered that JWZQS could inhibit the expression of inflammatory factors, such as IL-6, IL-1β, and TNF-α in the MOD group, effectively controlling inflammation development (Figure 7). Protein bands from the NF-κB pathway were analyzed to investigate the protective mechanism of JWZQS in UC mice (Figure 8(b)). JWZQS effectively inhibited IκB-α and P65 phosphorylation in the NF-κB pathway (Figure 8(a)). Therefore, JWZQS can inhibit the NF-κB signaling pathway activation and reduce colon inflammation in mice.
## 4.6. JWZQS Showed a Good Protective Effect on DSS-Induced UC Mouse Model
Images of colon samples revealed that fecal urine appeared in the colons of MOD group mice, and the colon length was significantly reduced (Figures 9(a) and 9(b)). Every day, the clinical symptoms of mice were recorded (Figure 9(c)). Every day, except for the blank group, the weight of the mice was recorded (Figure 9(d)). The MOD group's average colon length was shorter than that of the other groups.
H&E staining was used to stain mouse colon specimens. After staining, the histology was examined under a light microscope, and histological evaluations were carried out. The MOD group mice had fewer colonic goblet cells, lost colonic crypts, and typical edematous infiltration of inflammatory tissue when compared to the other groups (Figure 10(a)). The results revealed that the content of MOD histone decreased significantly, and the SASP group clearly outperformed the MOD group, as did the Chinese medicine dose groups, which also outperformed the MOD group and demonstrated a concentration gradient advantage (Figure 10(b)).
## 5. Discussion
UC is currently classified as a refractory condition by the medical community. Mucous pus, diarrhea, hemafecia, and stomachache are some of the clinical symptoms. Its lesions mostly infiltrate the rectal and colonic mucosa/submucosa. Clinical characteristics include a long disease course, ease of recurrence, a difficult prognosis, and numerous complications [24]. Despite its low prevalence, UC is on the rise, and it is a refractory disease with a lengthy treatment cycle and a high fatality and disability rate [25]. The current oversimplified Western drug therapies are not only costly but also heavily dependent and drug-resistant. Therefore, it is critical to improve treatment methods and encourage the development of new, more effective therapeutic options with fewer side effects.
Along with the growth of network pharmacology, a new method for locating therapeutic agents has recently emerged. The multitarget activity in network pharmacology corresponds to the complex disease pathways and drug action. Understanding TCM through network pharmacology is a growing trend [26]. Network pharmacology is a cutting-edge research method that combines network science, bioinformatics, and systems biology to assess molecular associations between pharmaceuticals and therapeutic entities at the physiological network and system level facets and clarify drug methodological pharmacodynamics. Exploring such intricate TCM constituents is a natural fit for its multichannel and multitarget properties [27]. Therefore, the current study established the mechanism of action of JWZQS in methodologically managing UC.
The analysis of “drug-components-targets” and “drug-components-targets-disease” revealed 13 active ingredients, which are key ingredients in the treatment of UC. The PPI network discovered that AKT1, IL6, TP53, TNF, and IL-1β are core targets, and their interactions are critical to the completion of various biological processes. GO and KEGG analyses revealed that JWZQS's anti-UC effect is primarily derived from the cellular response to positive regulation of lipids, hormones, substances, lipid peptides, radiation, and cell movement. Some examples include protein homodimer activity, cytokine activity, transcription factor binding, protein domain-specific binding, DNA binding, kinase binding, and other MF items. JQZQS may influence membrane rafts, transcriptional regulatory complexes, membrane sides, vesicle lumens, and other cellular structures in CC. JWZQS may be used to treat UC by interfering with these biological processes. JWZQS are primarily involved in regulating cancer, lipid and atherosclerosis, diabetic complications, chemogenic receptor activation, cellular senescence, platinum resistance, and diabetic cardiomyopathy. However, TNF-α, IL-6, and IL-1β may be regulated in the NF-κB pathway. It also influences IκB-α and P65 phosphorylation. Animal studies will confirm this later. Irisolidone, naringenin, nobiletin, quercetin, and wogonin are some of the most promising UC treatments. Irisolidone, naringenin, nobiletin, quercetin, and wogonin were found to have affinity for IL-1β, AKT1, TNF, IL6, and TP53 proteins. However, the interaction of IL1β, IL6, and TP53 with the active constituents was weak in the docking results. There may be the following reasons. IL1β, IL6, and TP53 proteins were small and could not form obvious binding pockets. The active ingredient binds only to the surface of the protein. The molecular docking mode is semiflexible docking, which cannot completely simulate the induced fit in body. Molecular docking does not fully simulate the environment in body.
We found that the NF-κB pathway may be involved in the important pathway of JWZQS therapy for UC. The NF-κB pathway regulates the expression of numerous genes that regulate immune responses and inflammation, as well as cellular proliferation, apoptosis, and survival [28, 29]. Dimers are formed when NF-κB/Rel proteins, which are kept inactive by collaborating with IκB repressor proteins and p50 and p52's p105 and p100 precursors combined. The inducer of NF-κB is an IKK complex composed of the NEMO/IKKγ regulatory and IKKα and IKKβ catalytic subunits [30–32]. IKK compounds, particularly those with serine residues in protein phosphorylation IκB, predominate in response to various stimuli (e.g., TNF, IL-1β, and other proinflammatory cytokines) and bacterial processes, such as fat polysaccharide activation. These compounds target serine residues for significant degradation mediated by the ubiquitin-proteasome, resulting in the release of NF-κB, nuclear accumulation, and transcriptional activation of target genes [33, 34]. Because a variety of cell types are involved in the pathogenesis of the disease that resembles IBD, the intracellular functionalities of NF-κB differ by cell type. Increased proinflammatory mediator levels are most likely caused by increased NF-κB immunocyte activity, which worsens the inflammation. Therefore, we looked into how JWZQS affected the phosphorylation of P65 and IκB-α proteins in the NF-κB signaling pathway in animals. JWZQS inhibits the activation of the NF-κB signaling pathway, which inhibits the release of inflammatory cytokines IL-1β, IL-6, and TNF-α in a dose-dependent manner.
In this study, it was discovered that JWZQS not only had a protective effect on pathological injury and colon shortening throughout the DSS induced UC mouse model but also significantly improved body weight maintenance in mice. JWZQS was dose-dependent in UC mice, according to the findings. The breakdown of the intestinal immune barrier, which is formed by tight connections between cells, is a key factor in the progression of UC. Occludin, Claudin-1, and ZO-1 are intestinal immune barrier-tight junction proteins that are primarily responsible for maintaining the integrity of the intestinal epithelial barrier. Increased permeability of these junction proteins is associated with an increased risk of UC [35]. The study of these three tight junction proteins revealed that JWZQS had a protective effect on the intestinal immune barrier, indicating that the expression levels of Occludin, Claudin-1, and ZO-1 were lower in the DSS-induced UC mouse model than in the other groups. JWZQS could significantly correct the abnormal expression of three tight junction proteins in UC model mice, effectively reduce cell tissue permeability, repair intestinal mucosa, and alleviate UC symptoms.
These results suggest that JWZQS treat UC by inhibiting the expression of TNF-α, IL-6, and IL-1β in the NF-κB pathway. JWZQS protects DSS-induced UC mouse models by upregulating Occludin, Claudin-1, and ZO-1 in the colon of mice. In addition, many additional signaling pathways may be activated and their mechanisms remain to be investigated.
## 6. Conclusion
In this study, we used network pharmacology and related software system to study the potential mechanism of JWZQS against UC. The affinity between the core target and the active component was verified by molecular docking. Animal experiments revealed that JWZQS could inhibit the NF-κB pathway, reduce the expression of related inflammatory factors, and increase the expression of tight link protein. However, additional research is required to validate the reliability of the research findings. Finally, JWZQS can be used in the clinic. However, the mechanism of JWZQS in treating UC requires more research.
## Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
## Conflicts of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Authors' Contributions
Research design was done by Z-JH. Data collection and analysis were done by ZC. Manuscript writing was done by ZC. Manuscript revision was done by Z-CY and ZC. Graphic production was done by ZC.
## References
1. Ungaro R., Mehandru S., Allen P. B., Peyrin-Biroulet L., Colombel J. F.. **Ulcerative colitis**. (2017) **389** 1756-1770. DOI: 10.1016/s0140-6736(16)32126-2
2. Yu Y. R., Rodriguez J. R.. **Clinical presentation of Crohn's, ulcerative colitis, and indeterminate colitis: symptoms, extraintestinal manifestations, and disease phenotypes**. (2017) **26** 349-355. DOI: 10.1053/j.sempedsurg.2017.10.003
3. Kucharzik T., Koletzko S., Kannengiesser K., Dignass A.. **Ulcerative colitis-diagnostic and therapeutic algorithms**. (2020) **117** 564-574. DOI: 10.3238/arztebl.2020.0564
4. Hopkins A. L.. **Network pharmacology: the next paradigm in drug discovery**. (2008) **4** 682-690. DOI: 10.1038/nchembio.118
5. Li S., Niu M., Zhang S., Zhang B., Yang K.. **Interpretation of network pharmacology evaluation method guidance**. (2020) **52** 4119-4129
6. Ru J., Li P., Wang J.. **TCMSP: a database of systems pharmacology for drug discovery from herbal medicines**. (2014) **6** p. 13. DOI: 10.1186/1758-2946-6-13
7. Xu X., Fang C., Wang Y., Lu F., Liu S.. **Integrating network pharmacology and metabolomics to elucidate the mechanism of action of Huang Qin decoction for treament of diabetic liver injury**. (2022) **13, article 899043**. DOI: 10.3389/fphar.2022.899043
8. Amberger J. S., Hamosh A.. **Searching online Mendelian inheritance in man (OMIM): a knowledgebase of human genes and genetic phenotypes**. (2017) **58** 1.2.1-1.2.12. DOI: 10.1002/cpbi.27
9. Piñero J., Bravo À., Queralt-Rosinach N.. **DisGeNET: a comprehensive platform integrating information on human disease-associated genes and variants**. (2017) **45** D833-d839. DOI: 10.1093/nar/gkw943
10. Safran M., Dalah I., Alexander J.. **GeneCards version 3: the human gene integrator**. (2010) **2010, article baq020**. DOI: 10.1093/database/baq020
11. Wishart D. S., Feunang Y. D., Guo A. C.. **DrugBank 5.0: a major update to the DrugBank database for 2018**. (2018) **46** D1074-d1082. DOI: 10.1093/nar/gkx1037
12. Huang D., Wen X., Lu C.. **Investigating the molecular mechanism of compound Danshen dropping pills for the treatment of epilepsy by utilizing network pharmacology and molecular docking technology**. (2022) **10** p. 216. DOI: 10.21037/atm-22-195
13. Szklarczyk D., Gable A. L., Nastou K. C.. **The STRING database in 2021: customizable protein-protein networks, and functional characterization of user-uploaded gene/measurement sets**. (2021) **49** D605-d612. DOI: 10.1093/nar/gkaa1074
14. Berman H. M., Westbrook J., Feng Z.. **The Protein Data Bank**. (2000) **28** 235-242. DOI: 10.1093/nar/28.1.235
15. Trott O., Olson A. J.. **AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading**. (2010) **31** 455-461. DOI: 10.1002/jcc.21334
16. Seeliger D., De Groot B. L.. **Ligand docking and binding site analysis with PyMOL and Autodock/Vina**. (2010) **24** 417-422. DOI: 10.1007/s10822-010-9352-6
17. Ravindranath P. A., Forli S., Goodsell D. S., Olson A. J., Sanner M. F.. **AutoDockFR: advances in protein-ligand docking with explicitly specified binding site flexibility**. (2015) **11**. DOI: 10.1371/journal.pcbi.1004586
18. Kim J. J., Shajib M. S., Manocha M. M., Khan W. I.. **Investigating intestinal inflammation in DSS-induced model of IBD**. (2012) **60** p. 3678. DOI: 10.3791/3678
19. Lee S. H., Lee J. A., Shin M. R., Park H. J., Roh S. S.. (2022) **2022** 13. DOI: 10.1155/2022/4041402
20. Tian Z., Liu J., Liao M.. **Beneficial effects of fecal microbiota transplantation on ulcerative colitis in mice**. (2016) **61** 2262-2271. DOI: 10.1007/s10620-016-4060-2
21. Hao Y., Wang W., Wu D., Liu K., Sun Y.. **Retracted: Bilobalide alleviates tumor necrosis factor-alpha-induced pancreatic beta-cell MIN6 apoptosis and dysfunction through upregulation of miR-153**. (2020) **34** 409-417. DOI: 10.1002/ptr.6533
22. Zhang H., Wang Y., Su Y., Fang X., Guo W.. **The alleviating effect and mechanism of bilobalide on ulcerative colitis**. (2021) **12** 6226-6239. DOI: 10.1039/D1FO01266E
23. **UniProt: the universal protein knowledgebase**. (2017) **45** D158-d169. DOI: 10.1093/nar/gkw1099
24. Shen Z. H., Zhu C. X., Quan Y. S.. **Relationship between intestinal microbiota and ulcerative colitis: mechanisms and clinical application of probiotics and fecal microbiota transplantation**. (2018) **24** 5-14. DOI: 10.3748/wjg.v24.i1.5
25. Feuerstein J. D., Cheifetz A. S.. **Ulcerative colitis: epidemiology, diagnosis, and management**. (2014) **89** 1553-1563. DOI: 10.1016/j.mayocp.2014.07.002
26. Zhou Y., Dou F., Song H., Liu T.. **Anti-ulcerative effects of wogonin on ulcerative colitis induced by dextran sulfate sodium via Nrf2/TLR4/NF-**. (2022) **37** 954-963. DOI: 10.1002/tox.23457
27. Luo T. T., Lu Y., Yan S. K., Xiao X., Rong X. L., Guo J.. **Network pharmacology in research of Chinese medicine formula: methodology, application and prospective**. (2020) **26** 72-80. DOI: 10.1007/s11655-019-3064-0
28. Baldwin A. S.. **The NF-**. (1996) **14** 649-681. DOI: 10.1146/annurev.immunol.14.1.649
29. Ghosh S., May M. J., Kopp E. B.. **NF-kappa B and Rel proteins: evolutionarily conserved mediators of immune responses**. (1998) **16** 225-260. DOI: 10.1146/annurev.immunol.16.1.225
30. Israël A.. **The IKK complex: an integrator of all signals that activate NF-**. (2000) **10** 129-133. DOI: 10.1016/s0962-8924(00)01729-3
31. Karin M., Ben-Neriah Y.. **Phosphorylation meets ubiquitination: the control of NF-[kappa]B activity**. (2000) **18** 621-663. DOI: 10.1146/annurev.immunol.18.1.621
32. Zandi E., Karin M.. **Bridging the gap: composition, regulation, and physiological function of the IkappaB kinase complex**. (1999) **19** 4547-4551. DOI: 10.1128/mcb.19.7.4547
33. Hayden M. S., Ghosh S.. **Signaling to NF-**. (2004) **18** 2195-2224. DOI: 10.1101/gad.1228704
34. Verma I. M., Stevenson J. K., Schwarz E. M., Van Antwerp D., Miyamoto S.. **Rel/NF-**. (1995) **9** 2723-2735. DOI: 10.1101/gad.9.22.2723
35. Wang J., Zhang C., Guo C., Li X.. **Chitosan ameliorates DSS-induced ulcerative colitis mice by enhancing intestinal barrier function and improving microflora**. (2019) **20** p. 5751. DOI: 10.3390/ijms20225751
|
---
title: A phenome-wide comparative analysis of genetic discordance between obesity
and type 2 diabetes
authors:
- Daniel E. Coral
- Juan Fernandez-Tajes
- Neli Tsereteli
- Hugo Pomares-Millan
- Hugo Fitipaldi
- Pascal M. Mutie
- Naeimeh Atabaki-Pasdar
- Sebastian Kalamajski
- Alaitz Poveda
- Tyne W. Miller-Fleming
- Xue Zhong
- Giuseppe N. Giordano
- Ewan R. Pearson
- Nancy J. Cox
- Paul W. Franks
journal: Nature Metabolism
year: 2023
pmcid: PMC9970876
doi: 10.1038/s42255-022-00731-5
license: CC BY 4.0
---
# A phenome-wide comparative analysis of genetic discordance between obesity and type 2 diabetes
## Abstract
Obesity and type 2 diabetes are causally related, yet there is considerable heterogeneity in the consequences of both conditions and the mechanisms of action are poorly defined. Here we show a genetic-driven approach defining two obesity profiles that convey highly concordant and discordant diabetogenic effects. We annotate and then compare association signals for these profiles across clinical and molecular phenotypic layers. Key differences are identified in a wide range of traits, including cardiovascular mortality, fat distribution, liver metabolism, blood pressure, specific lipid fractions and blood levels of proteins involved in extracellular matrix remodelling. We find marginal differences in abundance of Bacteroidetes and *Firmicutes bacteria* in the gut. Instrumental analyses reveal prominent causal roles for waist-to-hip ratio, blood pressure and cholesterol content of high-density lipoprotein particles in the development of diabetes in obesity. We prioritize 17 genes from the discordant signature that convey protection against type 2 diabetes in obesity, which may represent logical targets for precision medicine approaches.
Coral et al. characterize genetically determined discordance between obesity and type 2 diabetes, identifying discordant genes that may convey protection against type 2 diabetes in obesity.
## Main
Cardiometabolic diseases are the leading cause of death globally, with obesity and type 2 diabetes mellitus (T2D) accounting for a large proportion of this burden1. The prevalences of obesity and T2D have risen sharply over the past decades worldwide2, corresponding with a shift to sedentary lifestyles and poor diet3. Even though obesity and T2D often coincide, their relationship is complex and remains incompletely understood. Indeed, while more than $80\%$ of people with T2D also have obesity, 10–$30\%$ of people with obesity appear metabolically healthy4–6. Conversely, metabolic abnormalities occur in ~$30\%$ of normal-weight individuals7–9. Likewise, despite weight loss improving glycaemic control in people with T2D10, when T2D occurs in people with normal weight, mortality rates are higher than those in people with overweight or obesity11,12. Here, we refer to these divergent features as ‘discordant diabesity’. We focus on this unusual phenotype because it helps leverage the independent roles of excess adiposity and T2D in life-threatening disease.
To some extent, this discordance can be attributed to the imprecision with which body mass index (BMI), the conventional metric used to define obesity, characterizes adiposity13,14. For instance, even when BMI is comparable, lean and fat mass distributions often vary from one person to the next15. Genetics has helped provide pathophysiological explanations for discordant diabesity, whereby, collectively, common variants affecting adipose distribution mimic monogenic syndromes such as familial lipodystrophies16–21. Expanding our knowledge of the phenotypic signature of discordant diabesity using the quantitative framework of genetics may help elucidate the mechanisms by which the broader health consequences of excess adiposity varies from one person to the next.
Here, we characterize genetically determined discordant diabesity through a comparative analysis with its concordant counterpart (that is, where higher genetic risk of obesity and T2D coincide). We used a range of machine learning methods to undertake phenome-wide scans to identify traits other than T2D that distinctively characterize these profiles. We concluded by undertaking robust causal inference analyses to determine the causal relationships underlying discordant diabesity with other features of health and disease.
## Assembly of concordant and discordant diabesity profiles
An analysis flowchart is presented in Extended Data Fig. 1a. We first identified genetic instruments for BMI22 and T2D23 by cross-referencing publicly accessible genome-wide association study (GWAS) summary statistics, finding 67 relatively independent single nucleotide polymorphisms (SNPs) strongly associated with both conditions ($P \leq 5$ × 10−8). After alignment to the BMI-increasing allele, these variants were labelled as ‘concordant’ (48 SNPs) or ‘discordant’ (19 SNPs) according to the positive or negative sign of their coefficients for T2D, respectively (Extended Data Fig. 1b and Supplementary Table 1; replication shown in Supplementary Table 2). Visual inspection of correlation patterns between BMI and T2D signals at each locus was undertaken using regional association plots (Supplementary Figs. 1 and 2).
## Phenome-wide scans
Among the clinical phenotypes, we found that concordant and discordant diabesity profiles differed predominantly in cardiometabolic features including high-density lipoprotein (HDL) cholesterol, waist-to-hip ratio (WHR), waist circumference, and blood pressure (Fig. 1 and Supplementary Table 3). Generally, the discordant profile was associated with a favourable phenotypic signature compared to the concordant profile. For example, systolic blood pressure (SBP) was lower in the discordant compared to the concordant profile (SBP: βC = 0.002 s.d. units per allele ($95\%$ confidence interval (CI): −0.001, 0.004), βD = −0.008 s.d. units per allele ($95\%$ CI: −0.012, −0.004), pδ = 1.39 × 10−4). We also found differences in risk of coronary heart disease (CHD) and stroke, which were lower in the discordant compared to the concordant profile (for example, CHD: odds ratio (OR)$c = 1.01$ per allele ($95\%$ CI: 1.01, 1.02), ORD = 0.98 per allele ($95\%$ CI: 0.97, 0.99), pδ = 1.3 × 10−6). The levels of biomarkers of liver function such as gamma-glutamyl transferase (GGT) and alanine aminotransferase (ALT) enzymes were lower in the discordant relative to the concordant profile (for example, ALT: βC = 0.008 s.d. units per allele ($95\%$ CI: 0.006, 0.011), βD = −0.011 ($95\%$ CI: −0.019, −0.003), pδ = 2.07 × 10−6). SHBG, a protein also produced in the liver, was higher in the discordant as opposed to the concordant profile (βC = −0.008 s.d. units per allele ($95\%$ CI: −0.012, −0.004), βD = 0.013 s.d. units per allele ($95\%$ CI: 0.007, 0.019), pδ = 1.94 × 10−8). Additionally, the discordant profile was associated with higher mean corpuscular volume (βC = −0.002 s.d. units per allele ($95\%$ CI: −0.005, 0), βD = 0.006 s.d. units per allele ($95\%$ CI: 0.002, 0.01), pδ = 8.76 × 10−4) and lower levels of urate (βC = 0.007 s.d. units per allele ($95\%$ CI: 0.004, 0.01), βD = −0.005 s.d. units per allele ($95\%$ CI: −0.01, −0.001), pδ = 3 × 10−6) compared to the concordant profile. The odds of receiving treatment with alendronate was higher in the discordant than in the concordant profile, a drug indicated for osteoporosis (ORC = 0.99 per allele ($95\%$ CI: 0.99, 0.99), ORD = 1.001 per allele ($95\%$ CI: 1.001, 1.001), pδ = 3.26 × 10−6).Fig. 1Summary-based comparison of concordant and discordant profiles. Concordant and discordant GRS coefficients for traits where we found differences between profiles using GWAS summary data. All are per-allele effect sizes, in s.d. units for continuous outcomes and ORs for binary traits (diseases and self-reported medication). Traits shown had at least one estimate significant after $5\%$ FDR correction and the difference between profiles was also significant after $5\%$ FDR. Statistical tests were based on a z-distribution and were two-sided. Bars show $95\%$ CIs. Sample sizes vary for every trait (N > 100,000 for all traits).
We extracted association data for concordant and discordant SNPs from a variety of sources. From the curated repository Open GWAS66, which we queried using the ‘ieugwasr’ package in R, we gathered data for >3,500 traits derived from UK Biobank and other consortia for a variety of traits; these traits are termed ‘clinical phenotypes’. In cases where the effect of a SNP on a trait was not found, we looked for the effect of the nearest proxy SNP up to an r2 of 0.5 over a 500-kb window. We only kept estimates obtained from European ancestry populations in order to be consistent with the GWAS used to identify the genetic profiles. To prevent inclusion of inflated signals due to low sample size, we only included studies of more than 500 individuals; for binary traits, we required at least 25 minor alleles in the smallest group67. We calculated z-scores by dividing the β coefficients by their corresponding standard errors, and then we computed standardized effect sizes as a function of MAF and sample size n using equations [1] and [2] (ref. 31):1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{SE} = \frac{1}{{\sqrt {2 \times \mathrm{MAF} \times \left({1 - \mathrm{MAF}} \right) \times (n + z^2)} }}$$\end{document}SE=12×MAF×1−MAF×(n+z2)2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta = z \times \mathrm{SE}$$\end{document}β=z×SE We aligned all the estimates from these scans to the BMI-increasing alleles, so that they represent phenotypic variations associated with higher BMI in both profiles.
We also obtained data for 657 blood metabolites68,69 and 3,282 proteins in plasma39. Associations with several bacterial taxa in the gut were obtained from the MiBioGen consortium70. Association with expression and splicing of nearby genes in multiple tissue samples and in whole blood were obtained from data generated by GTEx and eQTLGen consortia, respectively.
## Profile decomposition
Further exploration of the molecular features of the discordant and concordant profiles revealed that some variants used to characterize these profiles deviated from the overall pattern of trait association for their respective SNP set. Using single-linkage clustering on the SNP–trait matrix (Extended Data Fig. 1c), we identified two outliers in the concordant profile, one near GCKR, associated with higher SHBG and lower liver enzymes (SHBG: 0.07 ($95\%$ CI: 0.07, 0.08), $$P \leq 7.5$$ × 10−199) and a second near TOMM40 associated with higher HDL (0.07 s.d. units per allele ($95\%$ CI: 0.06, 0.08), $$P \leq 3.7$$ × 10−107). In the discordant profile, the last variant to be aggregated to the clustering tree (that is, the SNP most distal from the other SNPs within its set) is located near SLC2A2 and, in contrast to the overall discordant estimates, was associated with higher levels of AST and GGT (GGT: 0.02 s.d. units per allele ($95\%$ CI: 0.02, 0.03), $$P \leq 2.7$$ × 10−22).
## External validation in BioVU
We sought replication of the discoveries outlined above in an independent European-ancestry cohort from BioVU, a de-identified collection of electronic health records and a linked biobank including inpatient and outpatient data from Vanderbilt University Medical Center (VUMC), a tertiary-care centre in Nashville, Tennessee, USA24–26. We constructed separate genetic risk score (GRS) coefficients for concordant and discordant profiles and assessed their association with multiple phenotypes (Fig. 2 and Supplementary Table 4). We first confirmed that the concordant and discordant GRSs were associated with higher obesity risk, respectively (ORC = 1.03 per allele ($95\%$ CI: 1.03, 1.04), ORD = 1.02 per allele ($95\%$ CI: 1.01, 1.02), pδ = 1.6 × 10−3) and that the concordant and discordant profiles were positively and negatively associated with diabetes diagnosis, respectively (ORC = 1.03 per allele, ($95\%$ CI: 1.02, 1.04), ORD = 0.95 per allele ($95\%$ CI: 0.94, 0.96), pδ = 3.2 × 10−49). Both scores were associated with increased odds of bariatric surgery (ORC = 1.05 per allele ($95\%$ CI: 1.03, 1.06), ORD = 1.03 per allele ($95\%$ CI: 1.008, 1.06), pδ = 0.24). We found divergent associations in multiple diseases directly related to the main traits (for example, essential hypertension (HT): ORC = 1.014 per allele ($95\%$ CI: 1.009, 1.019), ORD = 0.99 per allele ($95\%$ CI: 0.98, 0.99), pδ = 1.2 × 10−6). We also observed differences for other disease outcomes such as chronic kidney disease (ORC = 1.02 per allele ($95\%$ CI: 1.01, 1.02), ORD = 0.98 per allele ($95\%$ CI: 0.97, 0.99), pδ = 2.9 × 10−6) and osteoarthrosis (ORC = 1.01 per allele ($95\%$ CI: 1, 1.01), ORD = 1.02 per allele ($95\%$ CI: 1.01, 1.03), pδ = 0.012). Because both scores were also strongly associated with type 1 diabetes (T1D; ORC = 1.05, ($95\%$ CI: 1.03, 1.05), ORD = 0.96, ($95\%$ CI: 0.94, 0.97), pδ = 4.8 × 10−17), we repeated the analyses excluding individuals with T1D. This attenuated the differences between concordant and discordant profiles for a small subset of traits including diabetic retinopathy and end-stage chronic kidney disease. Fig. 2Comparison of concordant and discordant profiles in BioVU.Concordant and discordant GRS coefficients for traits where we found differences between profiles in BioVU. Analyses of disease endpoints included data for up to 48,544 individuals. Continuous outcomes included data for up to 68,724 and 13,661 individuals of European and African descent, respectively. All are per-allele effect sizes, in s.d. units for continuous outcomes and ORs for disease endpoints. Traits shown had at least one estimate significant after $5\%$ FDR correction and the difference between profiles was also significant after $5\%$ FDR. Statistical tests were based on a z-distribution and were two-sided. Bars show $95\%$ CIs. CRP, C-reactive protein.
We also assessed the association of each GRS to multiple laboratory measurements in individuals of European (n > 68,000) and African American (n > 14,000) descent (Supplementary Table 5). The value per individual was computed as the median value over all measurements after a quality-control pipeline described in detail elsewhere27. Significant differences were found for several glycaemic traits consistent with the diabetes risk profiles (for example, HbA1c: s.d. difference per concordant allele: 0.05 ($95\%$ CI: 0.04, 0.06), s.d. difference per discordant allele: −0.06 ($95\%$ CI: −0.08, −0.05), pδ = 1.5 × 10−39). We confirmed the difference between the two profiles in HDL and observed differences in the other two main lipid fractions (for example, triglycerides: s.d. difference per concordant allele: 0.02 ($95\%$ CI: 0.01, 0.03), s.d. difference per discordant allele: −0.04 ($95\%$ CI: −0.05, −0.03), pδ = 1.5 × 10−14). The findings for red blood cell phenotypes were also replicated, and additional differences were found in leucocyte count, urea, creatinine, phosphate, C-reactive protein and parathyroid hormone (PTH), all of which were higher in carriers of concordant SNPs. Of the liver enzymes, only ALT values were available for comparison, whose levels were weakly associated with the concordant but not the discordant GRS (s.d. difference per concordant allele: 0.14 ($95\%$ CI: 0.02, 0.26), s.d. difference per discordant allele: 0.06 ($95\%$ CI: −0.08, 0.2), pδ = 0.2). In individuals of African American descent, significant differences were found in HbA1c, glucose and urea levels in urine.
## Differences in mortality in UK Biobank
We examined the relationship of GRSs to mortality owing to cardiovascular events in >337,000 participants of European descent from the UK Biobank (mean follow-up of 11.8 years). Around 35,000 deaths were reported, of which approximately $20\%$ were related to cardiovascular events. The concordant GRS was associated with higher mortality (hazard ratio (HR) per allele: 1.01 ($95\%$ CI: 1.01, 1.02)), whereas the discordant GRS was not (HR per allele: 0.99 ($95\%$ CI: 0.98, 1.01), pδ = 0.02). However, when assessing each SNP separately, we observed that the concordant variant near TOMM40 was associated with lower incidence of cardiovascular mortality (HR per allele: 0.85 ($95\%$ CI: 0.81, 0.90), $$P \leq 4.54$$ × 10−9 and Supplementary Table 6).
## Differences in serum metabolites
Of the metabolites available, those related to lipid subfractions were the strongest discriminators of concordant and discordant profiles (Fig. 3 and Supplementary Table 7). Discordant diabesity was associated with higher cholesterol in lipoprotein particles of all densities, while lower triglyceride content in lipoprotein particles of low densities, as opposed to concordant diabesity (for example, free cholesterol in HDL: βC = −0.008 s.d. units per allele ($95\%$ CI: −0.01, −0.005), βD = 0.008 s.d. units per allele ($95\%$ CI: 0.004, 0.013), pδ = 3.09 × 10−10). Discordant diabesity also correlated with lower levels of branched-chain amino acids and aromatic amino acids, whereas in concordant diabesity they tended to be higher (total concentration of branched-chain amino acids: βC = 0.004 s.d. units per allele ($95\%$ CI: 0.002, 0.008), βD = −0.008 s.d. units per allele ($95\%$ CI: −0.012, −0.003), pδ = 1.46 × 10−6).Fig. 3Comparison of concordant and discordant profiles in molecular phenotypes. Concordant and discordant GRS coefficients for traits where we found differences between profiles in molecular phenotypes. All are per-allele effect sizes, in s.d. units. Findings in metabolites shown here are derived from TwinsUK + KORA F4 ($$n = 7$$,824) and the UK Biobank ($$n = 115$$,078). Protein data were derived from the INTERVAL study ($$n = 3$$,301). Traits shown in these two domains had at least one estimate significant after $5\%$ FDR correction, and the difference between profiles was also significant after $5\%$ FDR. Statistical tests were based on a z-distribution and were two-sided. Bars show $95\%$ CIs. Microbiome data came from the MiBioGen consortium ($$n = 18$$,340); the genii shown here had at least one estimate nominally significant, and the difference between estimates was also nominally significant (two-sided $P \leq 0.05$).
## Differences in gut microbiota
There were no differences between pooled concordant and discordant estimates for bacterial abundance in the gut that were statistically significant after false discovery rate (FDR) correction. Across ten taxa, several were nominally associated ($P \leq 0.05$) within either the concordant or the discordant profiles (Fig. 3 and Supplementary Table 8). Four of these belonged to the phylum Bacteroidetes (family Bacteroidaceae and geni Bacteroides, Parabacteroides and Butyricimonas), all of which were less abundant in discordant relative to concordant diabesity (for example, family Bacteroidaceae: βC = 0.005 s.d. units per allele ($95\%$ CI: 0.001, 0.008), βD = −0.004 s.d. units per allele ($95\%$ CI: −0.004, −0.01), pδ = 0.004). The remaining taxa belonged to the phylum Firmicutes, most of them members of the obligately anaerobic class Clostridia, which tended to be more abundant in the discordant profile compared to the concordant profile (for example, genus Subdoligranulum: βC = −0.003 s.d. units per allele ($95\%$ CI: −0.006, 0.001), βD = 0.006 s.d. units per allele ($95\%$ CI: 0.007, 0.011), pδ = 0.006). The family Lactobacillaceae was also lower in the discordant compared to the concordant profile (βC = 0.006 s.d. units per allele ($95\%$ CI: 0.001, 0.01), βD = −0.006 s.d. units per allele ($95\%$ CI: −0.014, 0.003), pδ = 0.02).
## Differences in serum protein levels
We found a significant difference between concordant and discordant estimates after FDR correction in a single protein: heparan sulfate 6-O-sulfotransferase 2 (HS6ST2), which was higher in discordant relative to concordant diabesity (βC = −0.01 s.d. units per allele ($95\%$ CI: −0.017, 0), βD = 0.03 s.d. units per allele ($95\%$ CI: 0.02, 0.04), pδ = 7.52 × 10−7; Fig. 3). These analyses may be underpowered given that the effect of variants in trans is likely to be weaker than that of those in the gene encoding the protein. Thus, we also searched for strong cis effects ($P \leq 5$ × 10−8) in the discordant profile. We found one association between the discordant variant near PPARG and metalloproteinase inhibitor 4 (TIMP4; β = −0.28 s.d. units per allele ($95\%$ CI: −0.35, −0.2), $$P \leq 5$$ × 10−14).
## Functional annotation using DEPICT
We used the Data-driven Expression Prioritized Integration for Complex Traits (DEPICT)28 tool to compare the enrichment for tissues and biological pathways in each profile (Supplementary Figs. 3–5). The most notable difference was the significant enrichment ($P \leq 0.05$) for adipose tissue in the discordant profile, which was not found in the concordant profile. We also found significant enrichment for adrenal glands, ileum and kidney in the discordant but not in the concordant profile. Conversely, there was significant enrichment for endocrine tissue and retina in the concordant but not the discordant profile. Tissues for which there was significant enrichment in both profiles included pancreas and myometrium.
## Gene expression and splicing in discordant diabesity
We found 506 genes whose expression/splicing was significantly influenced by concordant SNPs and 76 which were influenced by discordant SNPs across multiple tissues in GTEx29. In eQTLGen30, we found significant associations of concordant SNPs with expression of 493 genes. Discordant SNPs were associated with 94 genes. Around $46\%$ of all the associations found in GTEx were replicated in eQTLGen ($47\%$ of the genes associated with concordant SNPs; $39\%$ of the genes associated with discordant SNPs).
To identify genes most likely involved in the molecular mechanisms leading to discordant diabesity, we chose genetic instruments for the 76 genes whose expression was influenced by discordant SNPs in the corresponding tissues in GTEx, and for the 94 genes in eQTLGen. A prerequisite for these instruments was that they are strongly associated with BMI ($P \leq 5$ × 10−8). We followed the SMR & HEIDI approach31, which utilizes the strongest instrument for gene expression/splicing within the cis region of the corresponding gene (±500 Mb from the transcription start site) to calculate an estimate of the pleiotropic association across gene expression, BMI and T2D risk. This approach then determines if the association found reflects true pleiotropy rather than mere linkage by testing for heterogeneity of the estimates of SNPs in linkage disequilibrium (LD) with the lead SNP. We found 17 genes with robust expression signals for obesity and T2D whose directions of effect were in contrast (FDR-corrected $P \leq 5$%, pHEIDI > 0.01; Fig. 4 and Supplementary Table 9). To locate the most likely tissue of action for these genes, we followed a scoring procedure32 through which a tissue specificity score is derived for each gene. This is calculated as (i) the proportion of median expression (in transcripts per million) across specific tissue types catalogued in GTEx and (ii) evidence of promoter/enhancer histone marks surrounding the genetic instruments, derived from multiple cell lines classified anatomically by the RoadMap Epigenomics Project33 that could be mapped to tissue samples in GTEx. For each gene, we sorted tissues where we found pleiotropic links according to its specificity score and presence of promoter/enhancer signals for the genetic instrument. This allowed us to prioritize potential main action sites of relevance to discordant diabesity, for example, LYPLAL1 in adipose tissue and JAZF1 in vasculature and pancreas, while confirming the widespread effects of SLC22A3 across multiple organs. Fig. 4Genes with likely discordant pleiotropic effects on BMI and T2D.Genes with likely pleiotropic, yet discordant, effects on BMI and T2D, as found in the SMR & HEIDI analysis. Genes were sorted by their chromosome location and tissue where the pleiotropic association was found, as well as the lead expression quantitative trait loci (eQTL). The first three panels comprise the effect sizes of the lead eQTL on BMI (s.d. units), gene expression/splicing (normalized effect size) and T2D risk (OR), respectively. Bars represent $95\%$ CIs. The right panel represents the logarithm of the tissue-of-action score. BMI data were derived from the GIANT + UK Biobank meta-analysis ($$n = 681$$,275). Gene expression data came from the GTEx ($$n = 838$$) and eQTLGen consortia ($$n = 31$$,684). T2D data came from the DIAGRAM meta-analysis dataset ($$n = 158$$,186).
## Discordant diabesity genes as therapeutic targets
We performed a lookup of the genes identified previously in the comprehensive public access databases DGIdb34 and PHAROS35. Notably, there was evidence of interaction between three of the genes with strong pleiotropic associations (SLC2A2, SLC22A3 and KCNJ11) and metformin in both databases. SLC22A3 interacted with various quinoline derivatives (decynium-22, disprocynium-24, found in both databases), SarCNU (an antineoplastic drug in phase 2 clinical trials), derivatives of the alpha blocker phenoxybenzamine, corticosterone and colchicine. There is also evidence of potent inhibition of GLUT2, the protein product of SLC2A2, by a specific class of pyrazolopyrimidines. SLC38A11, MAU2 and FBXO46 are classified in the ‘Tdark’ level of target development in the PHAROS database, composed of understudied targets, while the remaining genes fall under the ‘Tbio’ level, which includes targets with no known interactions yet satisfying other conditions, such as having functional annotations based on experimental evidence, repeated mentions in publications indexed in PubMed, and 50 or more available commercial antibodies.
## Instrumental variable analyses
To quantify the potential impact of traits that emerged from the previous steps on offsetting the diabetogenic effect of obesity, we derived genetic instruments for each of these traits using SNPs that were also robustly associated with BMI ($P \leq 5$ × 10−8) and decomposed these instruments into two groups based on their direction of effect on the trait of interest after alignment to the BMI-increasing allele. We then constructed two GRSs, one for each group of variants, and calculated the T2D risk conferred by each GRS using summary data from the DIAGRAM consortium; we focused on GRSs that confer protection from T2D.
From the clinical phenotypes, the GRSs that conveyed higher BMI but lower WHR and SBP were significantly associated with lower T2D risk (Extended Data Fig. 2a and Supplementary Table 10). For example, the estimate for the GRS associated with higher BMI but lower WHR had an OR of 0.96 per allele ($95\%$ CI: 0.94–0.98, $$P \leq 6.71$$ × 10−5). Some traits in the clinical phenotypes required instruments to be in cis with the gene encoding the corresponding protein (for example, SHBG), to prevent confounding due to pleiotropy. We found two such instruments for ApoA1 and SHBG, respectively, which were not associated with T2D risk ($$P \leq 0.17$$ and 0.84, respectively; Supplementary Table 11) despite their strong association with higher BMI. From the analysis of metabolites, we found two GRS coefficients associated with higher BMI and lower T2D risk. The strongest protection was found for the GRS conferring higher total concentration of lipoprotein particles (OR: 0.98, $95\%$ CI: 0.96, 0.99, $$P \leq 0.006$$; Supplementary Table 12), consistent with our findings in the phenome scans.
To test for the potential causal effect on diabesity discordance of HS6ST2 and TIMP4, the two proteins identified in the previous analysis, we searched for valid instruments (P value for both protein levels and BMI < 5 × 10−8) in the cis region of the corresponding genes. We could only derive a valid instrument for TIMP4. Using the SMR & HEIDI method, we found a significant pleiotropic effect ($$P \leq 3.8$$ × 10−7, pHEIDI = 0.4; Extended Data Fig. 2b). However, we noted that the lead instrument and its closest proxies were located within PPARG, which is proximal to TIMP4.
No instruments for the microbial taxa where we found nominally significant differences reached the significance threshold required for BMI. Extending the exploration to other taxa revealed a single significant association ($P \leq 5$ × 10−8) of the A allele of rs1530559 (a variant within the lactase persistence haploblock36 in LD with the lactase functional variant rs4988235 (r2 = 0.4)) with higher BMI and lower abundance of the order Bifidobacteriales. This variant was not associated with T2D risk ($$P \leq 0.76$$).
## Individuals within the top decile for each profile
To determine the relevance of concordant and discordant profiles in people with obesity (≥30 kg/m2), we focused on this subgroup in UK Biobank who localized to the top decile of one of the two profile GRSs37. Consistent with a binomial distribution, $18\%$ of individuals with obesity were present in the two groups of extreme GRSs. The health characteristics of these individuals differed from all others with obesity (Supplementary Table 13) in several ways: for example, HbA1c levels in individuals with obesity and the extreme concordant GRS were higher compared to all other individuals with obesity (Kruskal–Wallis $$P \leq 5.94$$ × 10−12). Conversely, individuals with obesity and an extreme discordant GRS had significantly lower HbA1c compared to the rest of individuals with obesity ($$P \leq 2.71$$ × 10−42). Persons with obesity at both extreme GRSs are also distinguished from the rest by the main clinical features identified in our previous analyses such as SBP, HDL and ALT. WHR did not adequately separate the concordant or discordant extreme GRS from the wider group of people with obesity. However, because the initial phenome scan revealed a gender-specific difference in WHR between concordant and discordant profiles, (Extended Data Fig. 1c), we performed an additional analysis for WHR stratified by sex, where we found that women with extreme discordant GRS had significantly lower WHR compared to other women with obesity ($$P \leq 6.05$$ × 10−10).
## Comparison with previous studies of discordant diabesity
We compared our results to those obtained in three previous investigations of discordant variants. For instance, Mahajan et al.23 calculated the change in estimates of SNPs associated with T2D before and after adjustment for BMI. They found 15 loci where signals were enhanced after adjustment, which was attributed to discordance. Consistently, the SNP effects in the diabesity discordant profile derived here were enhanced, while those of the concordant profile were attenuated after adjustment (as described in ref. 38; Supplementary Table 14). The change in SNP effect estimates was consistently associated with SNP effects on BMI (R2 = 0.8; Supplementary Fig. 6). However, we observed that for four of the 19 SNPs ($20\%$) from the discordant set near PPARG, JAZF1, KCNJ11 and LYPLAL1, the change in SNP effect estimates was less than predicted. These discordant variants are those most likely to directly alter the relationship between BMI and T2D. This is consistent with the known effect of PPARG on adipocyte differentiation, and with our findings linking adipose tissue-specific gene expression at the LYPLAL1 locus with higher BMI but lower T2D risk. Similarly, we found that KCNJ11 and JAZF1 had discordant effects on BMI and T2D, which is related to tissue-specific expression in heart and arteries; variants at both loci are known to influence insulin secretion.
We also sought replication of a finding from Pigeyre et al.27 linking discordance to levels of the protein IGFBP-3 in blood. We were not able to replicate this finding (Supplementary Table 15), possibly due to differences in the characteristics of the cohorts where this relationship was found. For our analysis, we used summary data from the INTERVAL study39, which includes predominantly healthy blood donors of European ancestry. In contrast, Pigeyre et al. used data from the ORIGIN trial, a cohort composed of individuals of European ($47\%$) and Latin American ($53\%$) ancestries, enriched for T2D cases (>$80\%$ had a prior diagnosis).
Finally, we searched for the SNPs comprising the concordant and discordant profiles described above in the cluster analysis of discordant SNPs performed by Huang et al.21. Fourteen of the 19 discordant SNPs identified in our analysis ($78\%$) are among or in LD with the 62 SNPs identified by Huang et al (r2 > 0.1 within a 1-Mb window, as specified in the publication; Supplementary Table 16). Two of the subclusters described by the authors were significantly overrepresented by these 14 SNPs: 5 (ARAP1, ADCY5, PPARG, TCF7L2, KCNJ11-NCR3LG1) were in the subcluster characterized mainly by higher BMI and lower fasting glucose and risk of T2D (enrichment $$P \leq 1.6$$ × 10−3) and 4 (GRB14, LYPLAL1, ADAMTS9 and VEGFA) in the subcluster that conveyed an apparent protective effect on multiple cardiometabolic traits via peripheral adipose distribution (higher BMI and body fat percentage, and lower WHR; enrichment $$P \leq 0.04$$). Four concordant variants (at GCKR, TOMM40, AKAP6 and PPP1R3B-TNKS-MSRA) were also among the 62 SNPs described by Huang et al. As opposed to other variants in the concordant set, the variant in AKAP6 was associated with lower SBP (in ICBP GWAS: β = −0.25 mm Hg ($95\%$ CI: −0.38, −0.12), $$P \leq 1.25$$ × 10−4) and the variant near PPP1R3B-TNKS-MSRA was associated with higher HDL (β = 0.02 s.d. units ($95\%$ CI: 0.012, 0.027), $$P \leq 1.72$$ × 10−6). As we found and discussed in our analyses, TOMM40 and GCKR deviate from the concordant set owing to their favourable associations with lipids and liver enzymes that resemble the discordant set, a pattern that was also reported by Huang et al.
## Discussion
Obesity conveys heterogenous effects in cardiometabolic health, making disease prevention and management challenging. Here we used genetics to deconstruct the obesity phenotype into concordant and discordant diabesity, with strikingly different health characteristics beyond diabetes and obesity. Through transcriptomic, metabolomic and metagenomic analyses, we identified biomarkers that shed light on mechanisms of action and may aid risk stratification. Further analyses identified potential targets for drug development and drug repurposing.
Obesity and T2D often coalesce, owing largely to the mediating effect of peripheral insulin resistance caused by excess adiposity. The trait discordances described here reflect mechanisms involved in uncoupling obesity risk from T2D risk, thereby exposing diabetes-independent pathways through which obesity affects disease risk, for example, through adipose distribution. It is likely that both a higher capacity to expand adipose tissue in the gluteo-femoral compartment40,41 and lower abdominal region around organs such as the liver, which might underlie the difference seen in biomarkers of liver failure42, play important and independent roles in genetically determined diabesity discordance.
Another key phenotypic distinction between concordant and discordant profiles concerns blood pressure. Although T2D often causes vascular dysfunction, changes in the vascular bed may also precede metabolic perturbations through nutrient and hormonal flux43,44, affecting pancreas, muscle and adipose tissue45. For instance, capillary recruitment and permeability are key determinants of whole-body glucose uptake and glycaemic variation46.
Our findings relating to lipid metabolites support the use of more refined profiling of lipid subfractions to help determine risk in people with obesity. The cholesterol content of HDL particles and BCAA levels appear especially informative biomarkers47, possibly because they enhance glucose homeostasis in obesity by improving cross-talk between peripheral tissue and the liver48.
Despite the contrasting health consequences of the two diabesity profiles, bariatric surgery was equally likely, which may predispose one group to health benefits following surgery, whereas the other may not benefit in this way.
We found a significant difference between concordant and discordant profiles in levels of HS6ST2, a protein expressed in brain, kidney and ovaries, which in animal knock-out models shows a strong association with increased body weight and insulin resistance, possibly owing to enhanced adipocyte differentiation49,50. We found only one robust pleiotropic effect for discordant diabesity at TIMP4, which is proximal to PPARG, the likely causal gene. Moreover, PPARG activator medication inhibits matrix metalloproteinases51,52. TIMP4 has been associated with adipogenesis, possibly through its effect on the adipose tissue extracellular matrix in obesity53.
The colocalization analyses underscore the importance of tissue pleiotropy and tissue cross-talk in the molecular mechanisms of diabesity discordance. This is especially evident for SLC22A3, but also for other potential targets such as LYPLAL1, whose differential expression in both adipose tissue and adrenal glands appears linked to discordant diabesity. Moreover, three of the genes with pleiotropic links to T2D risk (SLC2A2, SLC22A3 and KCNJ11) interact with metformin. This suggests a potential effect of metformin in shifting individuals with obesity from a concordant to a discordant phenotype. SLC2A2 encodes GLUT2, which is part of the glucose sensor apparatus in pancreas and liver and is involved in intestinal glucose absorption in the gut54. Variants in this gene have been associated with preference for sugary foods55 and modified response to metformin56,57. SLC22A3 encodes OCT3, a protein widely expressed across tissues that aids adipocyte beiging58 and perivascular adipose tissue remodelling59. KCNJ11 encodes the Kir6.2 subunit of the ATP-sensitive potassium channel. As this is the target of sulphonylureas, this group of drugs may also harbor potential candidates for the phenotype shift to discordance in diabesity. The other ligands identified in the lookup may also constitute potential therapeutic agents to prevent cardiometabolic complications in obesity. For the rest of the genes, especially those in the ‘Tdark’ level in PHAROS, follow-up functional experiments in the tissues indicated by the lead genetic instruments and its corresponding epigenetic annotations are warranted.
Certain SNPs deviate from the overall association pattern of the profile within which they reside. In the concordant profile, the BMI-increasing allele of the variant near TOMM40 increases T2D risk but, unlike other SNPs in the same profile, is associated with a better lipid profile and lower cardiovascular disease mortality. *This* gene and others in its proximity (APOC1 and APOE) have been consistently implicated in lipid metabolism60. In the discordant profile, the variant SLC2A2 conveys protection against T2D risk despite being associated with heavier weight and higher blood pressure, and worse liver function and dyslipidaemia. The opposite pattern was observed in the concordant variant in GCKR, which encodes a regulatory protein that inhibits glucokinase. This reflects disparate phenotypic effects of modulating the glucose sensor apparatus at different levels54. Deeper characterization of these mechanisms can further improve obesity stratification.
Although no statistically robust differences were observed in gut microbiota between the two diabesity profiles, possibly owing to low statistical power, nominal differences emerged in taxa belonging to the Bacteroidetes and Firmicutes phyla, which together constitute $90\%$ of the human intestinal flora61. Our results indicate higher Firmicutes and lower Bacteroidetes abundance in discordant diabesity, which may result in enhanced production of short-chain fatty acid species such as butyrate, which is involved in glucose-lowering and anti-inflammatory mechanisms62.
Previous strategies to characterize the discordance between BMI and metabolic risk have been based on predefined sets of phenotypes traditionally linked with metabolic status19,21. Our phenome-wide approach consisted of leveraging the wealth of genetic associations harvested to date to dissect the phenotypic structure relevant for discordant diabesity, having three main advantages: [1] variables defining the differential phenotypic structure of each profile are selected a in data-driven manner across many phenotypic layers; [2] leveraging genetic data across multiple datasets enhances power and minimizes cohort-specific biases that would be anticipated if analyses were performed in a single cohort; and [3] although concordant and discordant diabesity profiles may be driven by molecular mechanisms that are independent of DNA variation, using germline DNA variants helps mitigate reverse causality and other sources of confounding that hamper the interpretation of associations for most other types of biological variation and phenotypes. An example of this is the analysis of epigenetic factors, which has led to identification of obesity sub-phenotypes even in the context of genetic homogeneity, as found in monozygotic twins that are discordant for adiposity traits63. However, these findings might be driven by variations in environmental exposures and behaviours that exist within and between twin pairs, as well as confounded by factors such as age, which differed between twin pairs in the reported analyses.
In conclusion, obesity profiles with either diabetogenic or antidiabetogenic proclivities reveal distinctive aetiological subtypes, with key differences in fat distribution, blood pressure and cholesterol content in HDL particles. We identified 1 protein (TIMP4) and 17 genes potentially involved in the molecular mechanisms leading to diabesity discordance, involving pleiotropic effects across multiple tissues.
## BioVU
Collection of electronic health records in BioVU was established in 1990 and includes data on billing codes from the International Classification of Diseases, 9th and 10th editions (ICD-9 and ICD-10). Disease phenotypes (‘phecodes’) are derived from these billing codes as described previously24 and case, control and exclusion criteria are defined. Two codes on different visit days were required to instantiate a case for each phecode. The biobank was launched in 2007 and comprises excess blood samples that their donors had consented for use in biomedical research. Details of programme operations, ethical considerations, continuing oversight and patient engagement are published elsewhere25. DNA samples were analysed using genome-wide genotyping platforms including Illumina multi-ethnic genotyping array. After quality assessment, the genotype data were then imputed to the Haplotype Reference Consortium reference panel at the Michigan imputation server. Populations of African American and European descent were identified by projecting individuals onto the major principal-component space derived from 1000 Genomes reference panel.
## UK Biobank
The UK *Biobank is* an ongoing prospective study of approximately 500,000 adults. Initial enrolment took place from 2006 to 2010 and included individuals aged 40–69 years across the United Kingdom64. It has collected comprehensive genetic and phenotypic information, biochemical assays and longitudinal health outcomes through health records, such as hospitalization and mortality. The genotypes were assayed using the UK Biobank Lung Exome Variant Evaluation and the Applied Biosystems UK Biobank Axiom Array, and imputed to the Haplotype Reference Consortium panel. Population structure was also assessed using principal-component analysis. We excluded individuals with inconsistency between their reported and genetic sex, had sex chromosome aneuploidy or were outliers for heterozygosity or missingness. Only individuals who were included in the calculation of genetic principal components were included, which ensures minimal genetic kinship with other participants.
## Single-nucleotide polymorphism selection to construct concordant and discordant genetic profiles
We cross-referenced the largest GWAS for BMI and T2D and extracted common biallelic SNPs (minor allele frequency (MAF) > $1\%$). Insertions, deletions and potentially ambiguous palindromic SNPs (A/T or C/G alleles with MAF > $30\%$) were excluded. Because both GWAS were conducted predominantly in populations of European descent, we used 1000 Genomes EUR reference panel for clumping (r2 < 0.01 over a 500-kb window) to identify nearly independent SNPs that were strongly associated with both conditions ($P \leq 5$ × 10−8). The directions of the effect of these SNPs on T2D were consistent in a second independent set of GWAS summary statistics extracted from the FinnGen database65 (Supplementary Table 2).
## Profile comparison
We then compared the effects of discordant versus concordant SNPs for every trait in two stages: we first obtained the combined effect of concordant (βC) and discordant (βD) SNPs separately using a random-effects meta-analysis with the Paule–Mandel estimator of between-SNP variance τ2 (refs. 71,72). We then calculated their difference δ = |βC − βD| and computed its standard error as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{SE}_\delta ^2 = \mathrm{SE}_C^2 + \mathrm{SE}_D^2$$\end{document}SEδ2=SEC2+SED273. We excluded from these analyses T2D traits (Supplementary Table 17). Traits in which any of the combined estimates βC or βD and δ were statistically significant after $5\%$ FDR correction were taken forward to stage two, where we converted the effect estimates for each SNP and the selected traits to z-scores and placed them in a SNP–trait matrix, with SNPs coded as ‘0’ if concordant and ‘1’ if discordant. We then trained several Random Forest classifiers (1,000 iterations) to this matrix, which attempted to classify SNPs in their correct category, and used the Boruta algorithm74 to identify which traits were relevant to distinguish discordant from concordant SNPs. Briefly, this algorithm creates randomly shuffled copies of all traits in the SNP–trait matrix, and then evaluates for each trait if its contribution to the accuracy of decision trees in the Random *Forest is* higher than its corresponding random set.
## Genetic risk score analyses
Concordant and discordant GRSs for an individual i were calculated as:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{PRS}_{Pi} = \mathop {\sum }\limits_{j \in P}^{M_P} G_{ij}$$\end{document}PRSPi=∑j∈PMPGij Where P is the set of MP SNPs belonging to the concordant or discordant profiles and *Gij is* the genotype for SNP j in individual i. In BioVU, association analyses were carried out for each GRS using R package ‘PheWAS’ (v0.99.5-5)24. We kept phecodes with at least 200 cases67 and identified those associated with either of the GRS coefficients and a significant difference between the estimated effects after a $5\%$ FDR correction.
In the UK Biobank, we examined the relationship of GRSs to mortality due to cardiovascular events in individuals followed up to the latest censor date (30th September 2021) using Cox regression. Primary cause of death was ascertained using ICD-10 codes reported in death certificates (Supplementary Table 18). All association models were adjusted for age, sex and first ten genetic principal components.
## SMR & HEIDI
The SMR method consists of identifying for a protein or gene the strongest association signal, which is used as a genetic instrument to test for its pleiotropic effect on an outcome. The HEIDI method consists of calculating the heterogeneity in the estimates of SNPs in LD with the lead SNP used in SMR. A higher pHEIDI value means heterogeneity is less likely, which supports true pleiotropy across the gene/protein and outcome signal, while a lower pHEIDI value means there is heterogeneity in the estimates, and the SMR signal is probably due to linkage. We consider an association to be true pleiotropy if pHEIDI > 0.01 (ref. 75). We retained signals where we found evidence of true pleiotropy for both BMI and T2D.
## Scoring method using epigenetic annotation
The scoring method to identify the most likely tissue of action assumes that if a genetic instrument for the expression of a gene in a certain tissue where it is highly expressed (that is, high tissue specificity) is in or close (in LD) to a promoter/enhancer region in the same tissue, and this genetic instrument is also associated with an outcome, then it is likely that the pleiotropic association on the outcome is due to perturbation of gene activity in that tissue. Promoter/enhancer signals were obtained by querying the RoadMap Epigenomics Project through the ‘haploR’ package in R.
## Genes as therapeutic targets
The lookups in DGIdb and PHAROS were performed using the web-based tool. DGIdb assigns an interaction score to the drug–gene interactions, which is the result of combining publication count, source count, relative drug specificity and relative gene specificity. The PHAROS database classifies targets into four ‘Target Development Levels’, according to the evidence of drug interactions available: ‘Tdark’ contains understudied targets, ‘Tbio’ contains highly studied targets but without interaction with compounds, ‘Tchem’ includes targets that bind to small molecules, and ‘Tclin’ interact with approved drugs.
All analyses were done using packages within the R environment (v4.1.2)76.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary InformationSupplementary Figs. 1–6 Reporting Summary Supplementary Tables 1–18 The online version contains supplementary material available at 10.1038/s42255-022-00731-5.
## Extended data
Extended Data Fig. 1Analysis flowchart and profile identification. Panel A: Analysis flowchart. Panel B: BMI and T2D risk estimates of concordant and discordant SNPs after alignment to the BMI increasing allele. Panel C: summary-based concordant and discordant GRS coefficients (standard deviation units for continuous traits, log OR for binary traits). Traits shown have at least 1 estimate significant after $5\%$ FDR correction and the difference between profiles is also significant after $5\%$ FDR. Statistical tests were based on a Z-distribution and were two-sided. Bars show $95\%$ confidence intervals. Sample sizes vary for every trait (> 100.000 for all traits). The heatmap shows the Z-scores of the SNPs in every trait, with the single-linkage tree at the bottom, separately for concordant and discordant SNPs.
Extended Data Fig. 2Traits with potential causal effect on diabesity discordance. Panel A: Traits where a difference was found in the comparison of profiles and one of the two direction-specific GRS associated with BMI was associated with lower risk of T2D (two-sided Z-statistic $P \leq 0.05$). To derive the GRS we used BMI data from the GIANT + UK Biobank meta-analysis ($$n = 681$$,275). WHR data came from the GIANT consortium ($$n = 212$$,244). SBP data came from the meta-analysis performed by the ICBP ($$n = 757$$,601). Metabolite data came from the UK Biobank ($$n = 115$$,078). Estimates represent T2D OR, bars represent $95\%$ confidence intervals, which are derived from the DIAGRAM meta-analysis ($$n = 158$$,186). Panel C: Regional association plot showing the pleiotropic effect of genetic instruments for blood levels of TIMP4 protein and high BMI and lower T2D risk. Protein data was derived from the INTERVAL study ($$n = 3$$,301).
is available for this paper at 10.1038/s42255-022-00731-5.
## Peer review information
Nature Metabolism thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Isabella Samuelson, in collaboration with the Nature Metabolism team.
## References
1. 1.World Health Organization. Cardiovascular diseases (CVDs). Available from: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds). Accessed October 2022.
2. Magliano DJ. **Trends in incidence of total or type 2 diabetes: systematic review**. *BMJ.* (2019.0) **366** l5003. DOI: 10.1136/bmj.l5003
3. Blüher M. **Obesity: global epidemiology and pathogenesis**. *Nat. Rev. Endocrinol.* (2019.0) **15** 288-298. DOI: 10.1038/s41574-019-0176-8
4. van Vliet-Ostaptchouk JV. **The prevalence of metabolic syndrome and metabolically healthy obesity in Europe: a collaborative analysis of ten large cohort studies**. *BMC Endocr. Disord.* (2014.0) **14** 9. DOI: 10.1186/1472-6823-14-9
5. Blüher M. **Metabolically healthy obesity**. *Endocr. Rev.* (2020.0) **41** 405-420. DOI: 10.1210/endrev/bnaa004
6. Vecchié A. **Obesity phenotypes and their paradoxical association with cardiovascular diseases**. *Eur. J. Intern Med* (2018.0) **48** 6-17. DOI: 10.1016/j.ejim.2017.10.020
7. Ding C, Chan Z, Magkos F. **Lean, but not healthy: the ‘metabolically obese, normal-weight’ phenotype**. *Curr. Opin. Clin. Nutr. Metab. Care* (2016.0) **19** 408-417. DOI: 10.1097/MCO.0000000000000317
8. Wang B. **Prevalence of metabolically healthy obese and metabolically obese but normal weight in adults worldwide: a meta-analysis**. *Horm. Metab. Res* (2015.0) **47** 839-845. DOI: 10.1055/s-0035-1559767
9. Loos RJF, Kilpeläinen TO. **Genes that make you fat, but keep you healthy**. *J. Intern Med.* (2018.0) **284** 450-463. DOI: 10.1111/joim.12827
10. Schwartz SS, Kohl BA. **Glycemic control and weight reduction without causing hypoglycemia: the case for continued safe aggressive care of patients with type 2 diabetes mellitus and avoidance of therapeutic inertia**. *Mayo Clin. Proc.* (2010.0) **85** S15-S26. DOI: 10.4065/mcp.2010.0468
11. Carnethon MR. **Association of weight status with mortality in adults with incident diabetes**. *JAMA* (2012.0) **308** 581-590. PMID: 22871870
12. Badrick E, Sperrin M, Buchan IE, Renehan AG. **Obesity paradox and mortality in adults with and without incident type 2 diabetes: a matched population-level cohort study**. *BMJ Open Diabetes Res. Care* (2017.0) **5** e000369. DOI: 10.1136/bmjdrc-2016-000369
13. Shah NR, Braverman ER. **Measuring adiposity in patients: the utility of body mass index (BMI), percent body fat and leptin**. *PLoS ONE* (2012.0) **7** e33308. DOI: 10.1371/journal.pone.0033308
14. Müller MJ. **The case of GWAS of obesity: does body weight control play by the rules?**. *Int. J. Obes.* (2018.0) **42** 1395-1405. DOI: 10.1038/s41366-018-0081-6
15. Goossens GH. **The metabolic phenotype in obesity: fat mass, body fat distribution, and adipose tissue function**. *Obes. Facts* (2017.0) **10** 207-215. DOI: 10.1159/000471488
16. Shungin D. **New genetic loci link adipose and insulin biology to body fat distribution**. *Nature* (2015.0) **518** 187-196. DOI: 10.1038/nature14132
17. Rask-Andersen M, Karlsson T, Ek WE, Johansson Å. **Genome-wide association study of body fat distribution identifies adiposity loci and sex-specific genetic effects**. *Nat. Commun.* (2019.0) **10** 339. DOI: 10.1038/s41467-018-08000-4
18. Lotta LA. **Integrative genomic analysis implicates limited peripheral adipose storage capacity in the pathogenesis of human insulin resistance**. *Nat. Genet.* (2017.0) **49** 17-26. DOI: 10.1038/ng.3714
19. Ji Y. **Genome-wide and abdominal MRI data provide evidence that a genetically determined favorable adiposity phenotype is characterized by lower ectopic liver fat and lower risk of type 2 diabetes, heart disease, and hypertension**. *Diabetes* (2019.0) **68** 207-219. DOI: 10.2337/db18-0708
20. Kilpeläinen TO. **Genetic variation near**. *Nat. Genet.* (2011.0) **43** 753-760. DOI: 10.1038/ng.866
21. Huang LO. **Genome-wide discovery of genetic loci that uncouple excess adiposity from its comorbidities**. *Nat. Metab.* (2021.0) **3** 228-243. DOI: 10.1038/s42255-021-00346-2
22. Yengo L. **Meta-analysis of genome-wide association studies for height and body mass index in ∼700,000 individuals of European ancestry**. *Hum. Mol. Genet.* (2018.0) **27** 3641-3649. DOI: 10.1093/hmg/ddy271
23. Mahajan A. **Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps**. *Nat. Genet.* (2018.0) **50** 1505-1513. DOI: 10.1038/s41588-018-0241-6
24. Denny JC. **PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene–disease associations**. *Bioinformatics* (2010.0) **26** 1205-1210. DOI: 10.1093/bioinformatics/btq126
25. Roden D. **Development of a large-scale de-identified DNA biobank to enable personalized medicine**. *Clin. Pharmacol. Ther.* (2008.0) **84** 362-369. DOI: 10.1038/clpt.2008.89
26. Dennis JK. **Clinical laboratory test-wide association scan of polygenic scores identifies biomarkers of complex disease**. *Genome Med.* (2021.0) **13** 6. DOI: 10.1186/s13073-020-00820-8
27. Pigeyre M. **Identification of novel causal blood biomarkers linking metabolically favorable adiposity with type 2 diabetes risk**. *Diabetes Care* (2019.0) **42** 1800-1808. DOI: 10.2337/dc18-2444
28. Pers T. **Biological interpretation of genome-wide association studies using predicted gene functions**. *Nat. Commun.* (2015.0) **6** 5890. DOI: 10.1038/ncomms6890
29. **The GTEx Consortium atlas of genetic regulatory effects across human tissues**. *Science* (2020.0) **369** 1318-1330. DOI: 10.1126/science.aaz1776
30. Võsa U. **Large-scale**. *Nat. Genet.* (2021.0) **53** 1300-1310. DOI: 10.1038/s41588-021-00913-z
31. Zhu Z. **Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets**. *Nat. Genet.* (2016.0) **48** 481-487. DOI: 10.1038/ng.3538
32. Torres JM. **A multi-omic integrative scheme characterizes tissues of action at loci associated with type 2 diabetes**. *Am. J. Hum. Genet.* (2020.0) **107** 1011-1028. DOI: 10.1016/j.ajhg.2020.10.009
33. Kundaje A. **Integrative analysis of 111 reference human epigenomes**. *Nature* (2015.0) **518** 317-330. DOI: 10.1038/nature14248
34. Freshour SL. **Integration of the drug–gene interaction database (DGIdb 4.0) with open crowdsource efforts**. *Nucleic Acids Res.* (2021.0) **49** D1144-D1151. DOI: 10.1093/nar/gkaa1084
35. Sheils TK. **TCRD and Pharos 2021: mining the human proteome for disease biology**. *Nucleic Acids Res.* (2021.0) **49** D1334-D1346. DOI: 10.1093/nar/gkaa993
36. Enattah N. **Identification of a variant associated with adult-type hypolactasia**. *Nat. Genet.* (2002.0) **30** 233-237. DOI: 10.1038/ng826
37. Udler MS. **Type 2 diabetes genetic loci informed by multi-trait associations point to disease mechanisms and subtypes: a soft clustering analysis**. *PLoS Med.* (2018.0) **15** e1002654. DOI: 10.1371/journal.pmed.1002654
38. Aschard H, Vilhjálmsson BJ, Joshi AD, Price AL, Kraft P. **Adjusting for heritable covariates can bias effect estimates in genome-wide association studies**. *Am. J. Hum. Genet.* (2015.0) **96** 329-339. DOI: 10.1016/j.ajhg.2014.12.021
39. Sun BB. **Genomic atlas of the human plasma proteome**. *Nature* (2018.0) **558** 73-79. DOI: 10.1038/s41586-018-0175-2
40. Emdin CA. **Genetic association of waist-to-hip ratio with cardiometabolic traits, type 2 diabetes, and coronary heart disease**. *JAMA* (2017.0) **317** 626-634. DOI: 10.1001/jama.2016.21042
41. Lotta LA. **Association of genetic variants related to gluteofemoral vs abdominal fat distribution with type 2 diabetes, coronary disease, and cardiovascular risk factors**. *JAMA* (2018.0) **320** 2553-2563. DOI: 10.1001/jama.2018.19329
42. Silva NMGD. **Liver function and risk of type 2 diabetes: bidirectional Mendelian randomization study**. *Diabetes* (2019.0) **68** 1681-1691. DOI: 10.2337/db18-1048
43. Karaca Ü, Schram MT, Houben AJHM, Muris DMJ, Stehouwer CDA. **Microvascular dysfunction as a link between obesity, insulin resistance and hypertension**. *Diabetes Res. Clin. Pract.* (2014.0) **103** 382-387. DOI: 10.1016/j.diabres.2013.12.012
44. Stehouwer CDA. **Microvascular dysfunction and hyperglycemia: a vicious cycle with widespread consequences**. *Diabetes* (2018.0) **67** 1729-1741. DOI: 10.2337/dbi17-0044
45. Barrett EJ. **The vascular actions of insulin control its delivery to muscle and regulate the rate-limiting step in skeletal muscle insulin action**. *Diabetologia* (2009.0) **52** 752–764. DOI: 10.1007/s00125-009-1313-z
46. Meijer RI. **Insulin-induced microvascular recruitment in skin and muscle are related and both are associated with whole-body glucose uptake**. *Microcirculation* (2012.0) **19** 494-500. DOI: 10.1111/j.1549-8719.2012.00174.x
47. Holeček M. **Branched-chain amino acids in health and disease: metabolism, alterations in blood plasma, and as supplements**. *Nutr. Metab.* (2018.0) **15** 33. DOI: 10.1186/s12986-018-0271-1
48. Guasch-Ferré M. **Metabolomics in prediabetes and diabetes: a systematic review and meta-analysis**. *Diabetes Care* (2016.0) **39** 833-846. DOI: 10.2337/dc15-2251
49. Nagai N. **Involvement of heparan sulfate 6-**. *Glycobiology* (2013.0) **23** 980-992. DOI: 10.1093/glycob/cwt037
50. 50.Matsuzawa, T. et al. Heparan sulfate promotes differentiation of white adipocytes to maintain insulin sensitivity and glucose homeostasis. J. Biol. Chem.297, 101006 (2021).
51. Sakamuri SSVP. **Absence of tissue inhibitor of metalloproteinase-4 (TIMP4) ameliorates high fat diet-induced obesity in mice due to defective lipid absorption**. *Sci. Rep.* (2017.0) **7** 6210. DOI: 10.1038/s41598-017-05951-4
52. Marx N. **Antidiabetic PPAR gamma-activator rosiglitazone reduces MMP-9 serum levels in type 2 diabetic patients with coronary artery disease**. *Arterioscler Thromb. Vasc. Biol.* (2003.0) **23** 283-288. DOI: 10.1161/01.ATV.0000054195.35121.5E
53. Maquoi E, Munaut C, Colige A, Collen D, Lijnen HR. **Modulation of adipose tissue expression of murine matrix metalloproteinases and their tissue inhibitors with obesity**. *Diabetes* (2002.0) **51** 1093-1101. DOI: 10.2337/diabetes.51.4.1093
54. Thorens B. **GLUT2, glucose sensing and glucose homeostasis**. *Diabetologia* (2015.0) **58** 221-232. DOI: 10.1007/s00125-014-3451-1
55. Eny KM, Wolever TMS, Fontaine-Bisson B, El-Sohemy A. **Genetic variant in the glucose transporter type 2 is associated with higher intakes of sugars in two distinct populations**. *Physiol. Genomics* (2008.0) **33** 355-360. DOI: 10.1152/physiolgenomics.00148.2007
56. Zhou K. **Variation in the glucose transporter gene**. *Nat. Genet.* (2016.0) **48** 1055-1059. DOI: 10.1038/ng.3632
57. Rathmann W. **A variant of the glucose transporter gene SLC2A2 modifies the glycaemic response to metformin therapy in recently diagnosed type 2 diabetes**. *Diabetologia* (2019.0) **62** 286-291. DOI: 10.1007/s00125-018-4759-z
58. Song W. **Organic cation transporter 3 (Oct3) is a distinct catecholamines clearance route in adipocytes mediating the beiging of white adipose tissue**. *PLoS Biol.* (2019.0) **17** e2006571. DOI: 10.1371/journal.pbio.2006571
59. Saxton SN. **Restoring perivascular adipose tissue function in obesity using exercise**. *Cardiovasc. Drugs Ther.* (2021.0) **35** 1291-1304. DOI: 10.1007/s10557-020-07136-0
60. Middelberg RP. **Genetic variants in**. *BMC Med. Genet.* (2011.0) **12** 123. DOI: 10.1186/1471-2350-12-123
61. Qin J. **A human gut microbial gene catalogue established by metagenomic sequencing**. *Nature* (2010.0) **464** 59-65. DOI: 10.1038/nature08821
62. 62.den Besten, G. et al. The role of short-chain fatty acids in the interplay between diet, gut microbiota, and host energy metabolism. J. Lipid Res. 54, 2325–2340 (2013).
63. Yang CH. **Independent phenotypic plasticity axes define distinct obesity sub-types**. *Nat. Metab.* (2022.0) **4** 1150-1165. DOI: 10.1038/s42255-022-00629-2
64. Bycroft C. **The UK Biobank resource with deep phenotyping and genomic data**. *Nature* (2018.0) **562** 203-209. DOI: 10.1038/s41586-018-0579-z
65. 65.Kurki, M. I. et al. FinnGen: unique genetic insights from combining isolated population and national health register data. Preprint at https://www.medrxiv.org/content/10.1101/2022.03.03.22271360v1 (2022).
66. 66.Elsworth B, et al. The MRC IEU OpenGWAS data infrastructure. Preprint at bioRxiv10.1101/2020.08.10.244293 (2020).
67. Verma A. **PheWAS and beyond: the landscape of associations with medical diagnoses and clinical measures across 38,662 individuals from Geisinger**. *Am. J. Hum. Genet* (2018.0) **102** 592-608. DOI: 10.1016/j.ajhg.2018.02.017
68. Shin S-Y. **An atlas of genetic influences on human blood metabolites**. *Nat. Genet.* (2014.0) **46** 543-550. DOI: 10.1038/ng.2982
69. Kettunen J. **Genome-wide study for circulating metabolites identifies 62 loci and reveals novel systemic effects of LPA**. *Nat. Commun.* (2016.0) **7** 11122. DOI: 10.1038/ncomms11122
70. Kurilshikov A. **Large-scale association analyses identify host factors influencing human gut microbiome composition**. *Nat. Genet.* (2021.0) **53** 156-165. DOI: 10.1038/s41588-020-00763-1
71. Dastani Z. **Novel loci for adiponectin levels and their influence on type 2 diabetes and metabolic traits: a multi-ethnic meta-analysis of 45,891 individuals**. *PLoS Genet.* (2012.0) **8** e1002607. DOI: 10.1371/journal.pgen.1002607
72. Veroniki AA. **Methods to estimate the between-study variance and its uncertainty in meta-analysis**. *Res. Synth. Methods* (2016.0) **7** 55-79. DOI: 10.1002/jrsm.1164
73. 73.Borenstein, M., Hedges, L. V., Higgins, J. P. T. & Rothstein, H. R. Multiple outcomes or time-points within a study. In Introduction to Meta-Analysis, 225–238 (John Wiley & Sons, 2009).
74. Kursa MB, Rudnicki WR. **Feature selection with the Boruta package**. *J. Stat. Softw.* (2010.0) **36** 1-13. DOI: 10.18637/jss.v036.i11
75. Wu Y. **Integrative analysis of omics summary data reveals putative mechanisms underlying complex traits**. *Nat. Commun.* (2018.0) **9** 918. DOI: 10.1038/s41467-018-03371-0
76. 76.R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Accessed at https://www.R-project.org/ (2022).
|
---
title: Delta (B1.617.2) variant of SARS-CoV-2 induces severe neurotropic patterns
in K18-hACE2 mice
authors:
- Ju-Hee Yang
- Myeon-Sik Yang
- Dae-Min Kim
- Bumseok Kim
- Dongseob Tark
- Sang-Min Kang
- Gun-Hee Lee
journal: Scientific Reports
year: 2023
pmcid: PMC9970970
doi: 10.1038/s41598-023-29909-x
license: CC BY 4.0
---
# Delta (B1.617.2) variant of SARS-CoV-2 induces severe neurotropic patterns in K18-hACE2 mice
## Abstract
A highly contagious virus, severe acute respiratory syndrome coronavirus 2, caused the coronavirus disease 19 (COVID-19) pandemic (SARS-CoV-2). SARS-CoV-2 genetic variants have been reported to circulate throughout the COVID-19 pandemic. COVID-19 symptoms include respiratory symptoms, fever, muscle pain, and breathing difficulty. In addition, up to $30\%$ of COVID-19 patients experience neurological complications such as headaches, nausea, stroke, and anosmia. However, the neurotropism of SARS-CoV-2 infection remains largely unknown. This study investigated the neurotropic patterns between the B1.617.2 (Delta) and Hu-1 variants (Wuhan, early strain) in K18-hACE2 mice. Despite both the variants inducing similar pathogenic patterns in various organs, B1.617.2-infected K18-hACE2 mice demonstrated a higher range of disease phenotypes such as weight loss, lethality, and conjunctivitis when compared to those in Hu-1-infected mice. In addition, histopathological analysis revealed that B1.617.2 infects the brain of K18-hACE2 mice more rapidly and effectively than Hu-1. Finally, we discovered that, in B1.617.2-infected mice, the early activation of various signature genes involved innate cytokines and that the necrosis-related response was most pronounced than that in Hu-1-infected mice. The present findings indicate the neuroinvasive properties of SARS-CoV-2 variants in K18-hACE2 mice and link them to fatal neuro-dissemination during the disease onset.
## Introduction
A novel, highly contagious virus termed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emerged in December 2019 and swiftly expanded globally1. Coronavirus disease 19 (COVID-19) primarily affects the respiratory system and various organs such as the brain2. In addition, a high proportion of COVID-19 patients exhibit neurological manifestations such as hypogeusia, dizziness, headaches, myalgia, impaired consciousness, hyposmia, and ataxia3,4. In the early pandemic, $36.4\%$ of patients in Wuhan, China exhibited neurological symptoms, with $8.9\%$ having peripheral nervous system symptoms, including anosmia ($5.1\%$)5. Multiple lines of evidence demonstrated that SARS-CoV-2-induced brain damage-induced neurological symptoms are frequent and disabling events6–8. These findings suggested that SARS-CoV-2 infection can infiltrate the central nervous system (CNS) and the respiratory system. However, the mechanism by which SARS-CoV-2 invades the CNS is unclear. ACE2 and TMPRSS2, which are recognized as SARS-CoV-2 receptors for infection and dissemination in the body, are expressed at limited levels in the brain relative to that in the lungs9. However, early studies reported that neuropilin-1 (NRP-1), basigin, cathepsin L, and furin can enhance SARS-CoV-2 invasiveness and are more highly expressed in the brain than ACE2 or TMPRSS210–12. Several studies have revealed that the SARS-CoV-2 neuroinvasive mechanism involves hematological dissemination across a disrupted blood–brain barrier (BBB) and direct CNS infection13–15. In addition, SARS-CoV-2 infection-induced inflammatory responses cause BBB dysfunction, facilitating viral entry into the CNS16. Although the mechanism of SARS-CoV-2 neuroinvasion is unclear, considering the similar genetic characteristics of SARS-CoV, a similar neuroinvasion mechanism for SARS-CoV-2 may exist17.
SARS-CoV-2, like other RNA viruses, has evolved rapidly and accumulated mutations in its viral genome through frequent recombination and evolutionary adaptation, giving rise to multiple variants of concern18. The B1.617.2 (delta) variant emerged in India in October 2020, and it exhibited nearly twofold higher infectivity than the early strain (Wuhan, Hu-1) during its predominant circulation period of June–December 202119. Despite the unknown etiology, in-hospital death rate was higher for patients infected with B1.617.2 (vaccinated or unvaccinated cases) than for those infected with the other strains20. Furthermore, immune escape mutations of 13 amino acids in the spike (S) protein, including the receptor-binding domain, enhance viral infectivity, viral particle production21, and stability22. We also expected to observe pathogenic differences between patients infected with B1.617.2 and the early strain in the brain and respiratory system. In this study, we assessed the Hu-1 and B1.617.2 respiratory symptoms and the neurotropic patterns in K18-hACE2 mice, which are susceptible to SARS-CoV-2 infection, and often demonstrate COVID-19–like disease symptoms23. Our studies clarified the variant’s lethality and histopathology, as well as changes in various cellular response genes after infection of the brains or lungs of the K18-hACE2 mice. These observations can help clarify the pathogenic characteristics and the neurotropic patterns of B1.617.2 and provide insights into the potential treatment responses for COVID-19 patients.
## The B1.617.2 variant is more virulent than the early strain in the brains of K18-hACE2 mice
Intranasally, we administered 2.5 × 104 of $50\%$ tissue culture-infectious dose (TCID50)/mL SARS-CoV-2, Hu-1 (early strain), or B1.617.2 (delta variant) into 7-week-old heterozygous K18-hACE2 mice (Fig. 1A). The resultant clinical symptoms were monitored for 8 days post-infection (dpi). K18-hACE2 mice infected with B1.617.2 exhibited more severe weight loss (> $20\%$) and an earlier onset of lethality in mice (death at day 5) than those infected with the Hu-1 strain (death at day 6). Both viruses infected mice showed distinguished $100\%$ lethality by day 7 in the B1.617.2 variant and, by day 9, in the Hu-1 strain (Fig. 1B). Furthermore, K18-hACE2 mice infected with B1.617.2 frequently exhibited inflammation throughout the eye, but in the Hu-1 infection group, inflammation was only observed in the corners of the eyes (Sup. Fig. S1A). The autopsy revealed damage in several organs including severe hemorrhage in the brain, lungs, and spleen (Sup. Fig. S1B–D), albeit no discernable injury in the kidneys (Sup. Fig. S1E) or other organs. These clinical lesions of the brain and lungs differed between mice inoculated with B1.617.2 and Hu-1; specifically, B1.617.2-infected mice showed severe brain and mild lung damage, whereas the Hu-1–infected mice showed severe lung and moderate brain damage. At 3–6 dpi, we assessed the viral burden in the lung and brain homogenates. B1.617.2 revealed similar trends in reducing viral RNA copies and subgenomic RNA when compared to Hu-1. The lower levels of viral RNA (Fig. 1C), subgenomic RNA (Fig. 1D), and infectious SARS-CoV-2 (Fig. 1E) were detected in B1.617.2-infected lungs over time, whereas a significant difference showed at only 4 dpi in the B1.617.2-infected brains. At 6 dpi, the viral RNA levels in other tissues, including the heart, kidneys, and spleen, were similar between B1.617.2- and Hu-1–infected lungs. By contrast, no viral RNA was detected in the liver or trachea (Sup. Fig. S2). The expression of hACE2, a SARS-CoV-2 receptor, was constant, supporting SARS-CoV-2 infection in the brain and other tissues (Sup. Fig. S3). During SARS-CoV-2 infection, hACE2 expression was lower in the lungs than in the brains (Sup. Fig. S4). Similarly, the expression of the nucleocapsid (N) protein declined in both Hu-1–and B1.617.2-infected lungs with time. At 4 dpi, viral N protein was detected in the B1.617.2-infected brain (Fig. 1F). These data suggested that the B1.617.2 variant infects the brain earlier than the Hu-1 strain and disseminates more rapidly, which may be associated with the early onset of clinical symptoms. Figure 1The B1.617.2 variant is more virulent than the Hu-1 strain in K18-hACE2 mice. The schematic illustration of the animal experiment. K18-hACE2 mice were intranasally inoculated with the Hu-1 strain or B1.617.2 variant (A). Weight loss and mortality were monitored in K18-hACE2 mice after inoculation with the indicated strains (B). Representative data were analyzed from three independent experiments and presented as the mean ± SEM value ($$n = 12$$). The viral burden (C) and negative-sense strand (D) in the lungs and brains were quantified by qRT-PCR at the indicated times after infection. The titration of SARS-CoV-2 in the lungs and brains was performed by TCID50 for the infectious virus (E). Western blotting detected the N protein of SARS-CoV-2 in the lung and brain homogenates at the indicated times after infection (F). Two-way ANOVA was performed, followed by Dunnett’s multiple-comparison tests. Statistical significance is indicated by asterisks ($$n = 5$$, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 001$, ****$p \leq 0.0001$, ns not significant). C, Control; H, Hu-1; B, B1.617.2. The viral RNA copies and infectious SARS-CoV-2 limit of detection were log10/copies and 2 log10 TCID50/mL, respectively (dot line). The image J software (version 1.53 k, http://imagej.nih.gov/ij/) normalized the relative viral N protein expression with β-actin.
## Differences of neuropathological complications in the brains of K18-hACE2 mice after SARS-CoV-2 infection
We assessed histopathological changes in hematoxylin and eosin-stained lung (Sup. Fig. S5) and brain sections (Fig. 2) from Hu-1–or B1.617.2-infected K18-hACE2 mice. The histopathological score was calculated using a microscopic grading system as the average of the representative severity of the lungs (Sup. Table S1). At 3 dpi, we observed similar lung pathology in both variants, including moderate pulmonary edema and infiltrating inflammatory cells in the perivascular and peribronchial regions with progressive inflammation. At 4 dpi, these patterns accelerated lung consolidation, infiltrating inflammatory cells, and a partial loss of bronchiole epithelial cilia in the Hu-1–infected lung sections. At 6 dpi, Hu-1 and B1.617.2-infected K18-hACE2 mice displayed consolidation in $35\%$–$50\%$ of the lungs and blood leakage from vessels into the adjacent alveolar space and alveolar wall thickening (Sup. Fig. S5). Furthermore, we observed the brain pathology in both SARS-CoV-2-infected brain tissue, including the meninges and cerebrum. At 4 dpi, the B1.617.2-infected brain sections displayed an increase in microglia and radial glia cell counts (Sup. Fig. S6) in the adjacent meningeal vessels and the perivascular region of the cerebrum, as well as a partial detachment of the meninges. Nonetheless, these changes were not present in the Hu-1–infected brain sections. B1.617.2 infection continuously increased the infiltrating glial cell numbers, culminating in meningeal disruption at 6 dpi (Fig. 2A and Sup. Fig. S6). Conversely, the predominant histopathological changes of Hu-1–infected brain sections were generally weak inflammatory responses that developed as late-onset symptoms when compared to the findings in B1.617.2-infected brain sections. To determine whether brain damage was correlated with the severity of SARS-CoV-2 infection, we stained the brain sections with immunohistochemistry for the SARS-CoV-2 N protein. At 4 dpi, N protein was distributed predominantly in perivascular neuronal cells in the B1.617.2-infected cerebrum sections, but not detected in the Hu-1–infected cerebrum sections (Fig. 2B). At 5 and 6 dpi, SARS-CoV-2 infection of neuronal cells widely spread throughout the cerebrum (Fig. 2B). The cells infected with SARS-CoV-2 were then stained with a neuron-specific marker known as microtubule-associated protein 2 (MAP2) to demonstrate that they were neuronal cells. At 4 dpi, the brain section infected with B1.617.2 demonstrated the co-localization of N protein (SARS-CoV-2) with MAP2 (neuronal cell) compared to the Hu-1-infected brain section (Fig. 2C). Furthermore, the same pattern was detected in the cerebrum sections stained with GFAP, an astrocyte marker, and glial fibrillary acidic protein (GFAP). At 4 dpi, the viral RNA and protein levels were significantly increased in B1.617.2-infected brains (Fig. 2D). These data suggested that brain damage caused by SARS-CoV-2 infection recruited multinucleated cells including microglia, radial glia, and astrocytes. In addition, the neuronal cells were more infected and sensitive to the B1.617.2 variant than the Hu-1 strain. Figure 2Histopathological analysis of SARS-CoV-2–infected K18-hACE2 mice. The abnormalities of the brain sections stained with hematoxylin and eosin were observed after the SARS-CoV-2 infection (A). The N protein level was detected in the brain sections and cerebrum at the indicated times after infection (left panel), and the N protein-positive area was quantified (right panel) (B). Representative immunofluorescence images of 4 dpi brain sections infected with SARS-CoV-2 display the localization of N protein (SARS-CoV-2, red) and MAP2 (neuronal cells, green) (C). GFAP, a marker of astrocytes, was stained in the brain sections at 4 dpi (left panel), and its levels were quantified by qRT-PCR (middle panel, $$n = 5$$) and Western blotting (right panel) at the indicated time after infection (D). Two-way ANOVA with Dunnett’s multiple-comparison test was performed to determine the significance (**$p \leq 0.01$, ns; not significant). Images are representative of four images per group. Magnification, × 200. C, Control; H, Hu-1; B, B1.617.2. The relative GFAP expression was normalized with β-actin by using the Image J software (version 1.53 k, http://imagej.nih.gov/ij/).
## The early cellular responses to SARS-CoV-2 infection
To assess how the kinetics of infection and the subsequent processes modulate the early cellular response to SARS-CoV-2 in the brain, we performed the *Next* generation sequencing (NGS) of SARS-CoV-2-infected brain homogenates at 0 (control), 3, and 4 dpi. The Venn diagram in Fig. 3A depicts the upregulated and downregulated genes in SARS-CoV-2–infected brain homogenates at 3 and 4 dpi as compared to that in the control animals. The enrichment of gene signatures in Hu-1–infected brain homogenates demonstrated an increase in the number of upregulated (from 47 to 305 genes) and downregulated genes (from 37 to 68 genes), with only 49 genes overlapping at 3 and 4 dpi. Conversely, upregulated (from 62 to 586 genes), downregulated (from 31 to 52 genes), and overlapped (18 genes) gene signatures were increased in B1.617.2-infected homogenates. Gene *Ontology analysis* of the top-upregulated genes identified various cellular responses, such as the immune system processes, innate immune responses, inflammation responses, viral defense responses, and interferon (IFN) responses (Fig. 3B). These findings highlighted the regulation of gene sets involved in type-I IFN signaling, inflammatory cytokine signaling, and glial cell migration. An hiPSC-derived neuronal organoid study reported dysregulated inflammatory and innate immune responses coupled with cell-death regulation24. At 4 dpi, we revealed that inflammatory cytokine-associated genes (Ccl5, Ccl7, Cxcl1, Il1b, and Csf3), Type-I IFN-associated genes (Irf7, Stat1, and Oas2), and certain IFN-stimulated genes (Ifit1, Oas2, Mx2, and Irf7) were upregulated in the B1.617.2-infected tissues compared to those in the Hu-1–infected tissues (Fig. 3C). In addition, the cell-death process was activated earlier in the B1.617.2-infected brains than in the Hu-1–infected brains. SARS-CoV-2 infection upregulated the expression of genes involved in apoptotic (Casp8, Casp7, and Nod1) and necrotic (Tnf, Ngfr, Ripk1, Ripk3, and Pygl) processes. An in vitro study reported that SARS-CoV-2 limits autophagy signaling and inhibits autophagic flux25. Similarly, the expression of autophagy-associated genes (Becn1 and Atg7) was neither altered nor reduced, indicating that an autophagy-independent cell-death program was activated in the SARS-CoV-2–infected brain (Fig. 3C). These distinct transcriptional changes were determined as temporary occurrences in the early stage of SARS-CoV-2-infected brain, indicating immune-associated or cellular-regulated characteristics. Figure 3SARS-CoV-2 infection-induced multi-transcriptional signatures associated with various cellular responses. The NGS data of the brain homogenates of 3 and 4 dpi. Venn diagram indicated the overlapping genes with differential expression. The total number of significantly upregulated and downregulated genes when compared to the corresponding findings in the non-infected mice (A). Biological responses enriched in differentially expressed genes were analyzed by Gene Ontology at 4 dpi when compared to the corresponding findings in the non-infected mice. The false discovery process (q value) determined the rank, and the indicated responses are listed after eliminating the redundant genes (B). Heat maps indicating upregulated genes at 3 and 4 days after Hu-1 or B1.617.2 infection. The cellular signal pathways with enriched inflammatory cytokines, type-I IFN responses, apoptosis, autophagy, and necrosis signaling were identified by Gene Ontology (C). The differentially expressed genes presented in each cellular pathway are the combinations of differential expressed genes. Based on the RNA-seq database, the gene ontology and Heat maps were generated using GraphPad Prism 7.0 (https://www.graphpad.com/).
## The necrotic pathway activated during the early brain response to SARS-CoV-2 infection
We investigated the commitment of the necrotic process in the early brain response after SARS-CoV-2 infection. The mouse necrosis RT2-profiler PCR array assessed the expression of necrosis-related genes in the brain homogenates infected with SARS-CoV-2 at 3 and 4 dpi. At 4 dpi, unsupervised clustering analysis illustrated that most necrosis-related genes were expressed at higher levels in B1.617.2-infected brains than in Hu-1–infected brains (Fig. 4A, left). Necrosis-related genes were upregulated (Hu-1: 4 and 4; B1.617.2: 4 and 62) and downregulated (Hu-1: 1 and 2; B1.617.2: 1 and 6) after 3 and 4 days of infection, respectively. The indicated genes included necrotic markers such as receptor-interacting kinase 1 (Ripk1), Ripk3, glycogen phosphorylase L (Pygl), Poly (ADP-ribose) polymerase 1 (Parp1), and a calpain-1 catalytic subunit (Capn1) (Fig. 4A, right). In addition, we identified a significant difference in the Ripk3 expression across SARS-CoV-2 infection groups. However, Bcl2 (pro-apoptotic), Casp3 (apoptotic), and Becn1 (autophagy) expression did not differ between the groups (Fig. 4B). These findings suggest that SARS-CoV-2–infected neuronal cells can promote a Ripk3-dependent cell-death program in the early brain response. Figure 4Necroptosis-associated genes were induced in early-stage brain homogenates during B1.617.2 infection. The brain homogenates obtained at 3 and 4 dpi were analyzed simultaneously to profile 84 necrosis genes by using the RT2 profiler PCR assay. The heat maps present the upregulated genes in the brain homogenates (left panel), and a scatter plot reveals the upregulated (red), downregulated (blue), and unchanged (black) genes (A). The plot presents the log tenfold changes in gene expression in the two groups. The quantification of the cell death-related genes Bcl2 (pro-apoptotic), Ripk3 (necroptosis), Casp3 (apoptotic), and Becn1 (autophagic) in the brain homogenates at the indicated time points after infection (B). Based on the GeneGlobe database (https://dataanalysis2.qiagen.com/pcr), the heat map for the RT2-profiler PCR array was generated using GraphPad Prism 7.0 (https://www.graphpad.com/).
## Discussion
The B1.617.2 variant was first identified in Maharashtra, India, and it carries three key mutations (i.e., L452R, T478K, and P681R) in the receptor-binding motif of S protein. These alterations rapidly became dominant globally because of the increased virus infectivity and the reported evasion of neutralizing antibodies, which is associated with its transmissibility26–29. Moreover, the epidemiological characteristics of the B1.617.2 variant included higher risks of hospitalization, intensive care unit admission, and mortality when compared to that observed for N501Y-positive variants and the early strain30–32. K18-hACE2 mice consistently expressed hACE2, thereby promoting systemic virus dissemination in most of the tissues and enhancing the infectivity of SARS-CoV-2. Several infected mice had a significant error value of the viral burden and titration following the change of hACE2 expression because of SARS-CoV-2–induced hACE2 downregulation or cell death in the infected tissues33,34. Since K18-hACE2 mice exhibited enhanced neurotropism, which was not detected in patients due to the overexpression of hACE2 in all epithelial tissues, they have limitations in that they do not accurately represent the disease phenotype observed in humans. Nevertheless, K18-hACE2 mice were considered an appropriate model for studying the lethal cases of COVID-19. In addition, SARS-CoV-2–infected K18-hACE2 mice exhibited neurological signs such as circling, rolling, and flaccid paralysis of the hind legs, which eventually led to death23. Moreover, it was previously reported that the potential replication of SASR-CoV-2 in neuronal cells could have lethal consequences in the CNS of K18-hACE2 mice35. In this study, we investigated the pathological changes and temporal changes of host factors involved in cellular and inflammatory responses during SARS-CoV-2 (Hu-1 or B1.617.2) infection in K18-hACE2 mice. As the B1.617.2 variant was associated with high infectivity and mortality, we expected this variant to cause more extensive changes in the clinical indices of the virus-infected brain relative to that by Hu-1. As expected, B1.617.2-infected mice displayed high lethality, neurological signs, severe hemorrhage, and weight loss when compared to the corresponding findings in Hu-1-infected mice. However, the B1.617.2 variant tends to feature the opposite patterns in terms of the viral burden, infectious virus titer, and viral replication, as well as the N protein levels in the infected lungs and brains. SARS-CoV-2 infection in the lungs leads to hACE2 downregulation, hACE2 shedding, or death in the hACE2-expressing cells36,37. We also demonstrated that the hACE2 levels were decreased in the lungs following SARS-CoV-2 infection, which suggests the high virulence of B1.617.2 in the pneumocytes and the possibility that the variant promotes inflammatory processes associated with hACE2 imbalance on the cell surface. Histopathological analysis revealed that SARS-CoV-2 infection caused the disruption of meninges and the infiltration of inflammatory cells, but no ischemia in the brain and lungs35,38,39.
The distinct mechanism of SARS-CoV-2 infection in the brain remains unclarified. The possible mechanisms include SARS-CoV-2 infection in the olfactory nerve, vascular endothelial cell infection, and invasion through inflammation-induced disruption of the BBB40. A few studies have reported that SARS-CoV-2 targets neurons and neuronal progenitors for subsequent replications14,41. Our results indicated that the patterns of neuron and reactive astrocyte infection, known as gliosis, caused by Hu-1 or B1.617.2 differed among the K18-hACE2 mice. We attributed this difference to the possible susceptibility of the brain by neuropilin-1 (NRP1), which enhances the SARS-CoV-2 entry10, resulting in sustained brain damage and enhanced severity.
Cytokine analysis of the brains of K18-hACE2 mice infected with Hu-1 or B1.617.2 revealed markedly different cell-death profiles, including apoptosis and necrosis, whereas the inflammatory and innate immune responses were similar between the groups. SARS-CoV-2–mediated regulation of the cell-death pathways has been reported in several cell types and neuronal cells40,42. The cell death processes can cooperate and they are often accompanied by the activation of multicellular factors, which implies the activation of defenses against intracellular infection. In addition, these processes can promote innate and adaptive immune responses and inflammatory responses, which act synergistically to regulate cell fate43. Ripk1 and Ripk3 are the key factors that regulate necrotic cell death. Under virus infection, which promotes Ripk1-dependent necrosis by stimulating tumor necrosis factor (TNF), Ripk3 subsequently induced the phosphorylation of Ripk1, which resulted in the formation of a pro-necrotic necrosome complex. In addition, Ripk3 phosphorylates mixed lineage kinase domain-like pseudokinase (MLKL), which is distributed on the plasma membrane, demonstrating necroptosis activation, and it also interacts with metabolic enzymes, such as Pygl, which contributes to ROS production and necroptosis44. Our results revealed that B1.617.2 induced the expression of necrosis- and apoptosis-related genes including Ripk1, Ripk3, Tradd, Pygl, Fadd, Il-1β, and Casp3 when compared to the corresponding findings in Hu-1-infected mice. Necroptosis and apoptosis are programmed forms of cell death that are activated by SARS-CoV-2 infection45,46. Our results demonstrated that the expression of the necroptosis marker RIPK3 was increased by B1.617.2 at the early time when compared with that after Hu-1 infection; moreover, the mortality was earlier in mice infected with B1.617.2. In addition, the apoptosis marker caspase-3 was increased after 6 days of infection with B1.617.2 when compared to that in the Hu-1-infected mice. These results can serve as evidence for analyzing the causes of why B1.617.2 is classified as variants of concern and fatal symptoms in people. In our subsequent studies, we aim to clarify the mechanism by which SARS-CoV-2 induces necroptosis and cell-death responses in the brain. Furthermore, whether necrotic factors are conserved in the brains of patients infected with SARS-CoV-2 variants is an interesting question for future studies. Finally, our study findings can facilitate the clarification of the pathogenic characteristics of the B1.617.2 variant and identify the potential factors that control brain damage and improve the survival outcomes of patients with COVID-19.
## Virus and cells
The Hu-1 (BetaCoV/Korea/KCDC$\frac{03}{2020}$, NCCP43326) and B1.617.2 strains (hCoV-19/Korea/KDCA$\frac{119861}{2021}$, NCCP43390) were obtained from the Korea Disease Control and Prevention Agency. African green monkey kidney epithelial cells (Vero E6, ATCC CRL-1586) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with $10\%$ fetal bovine serum (FBS, Thermo Fisher Scientific), 100 U/mL penicillin, and 100 µg/mL streptomycin (Thermo Fisher Scientific). The propagation and titration of SARS-CoV-2 in the Vero E6 cells were calculated using TCID50, as described previously1. Briefly, Vero E6 cells were infected with the B1.617.2 (2.9 × 106 TCID50/mL) or Wuhan strain (1 × 106 TCID50/mL), and the cytopathic effect (CPE) was monitored at 3–5 dpi. The experiments associated with SARS-CoV-2 infection were performed at a biosafety level 3 laboratory with the use of personal protection equipment in accordance with the biosafety manual instructions issued by the Korea Zoonosis Research Institute of Jeonbuk National University.
## Mouse experiment
Heterozygous K18-hACE2 mice [strain: JAX 034,860 B6.Cg-Tg(K18-hACE2)2Prlmn/J] were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Seven-week-old male K18-hACE2 mice were administrated 2.5 × 104 TCID50/mL SARS-CoV-2 via the intranasal route. The K18-hACE2 mice were monitored for changes in weight loss, lethality, and clinical symptoms every day after inoculation. At 6 dpi, SARS-CoV-2–infected K18-hACE2 mice were sacrificed by isoflurane and autopsied to assess the clinical lesions in several organs such as the brain, heart, lungs, spleen, and kidneys. The animal experiments were approved by the Institutional Animal Committee of the Jeonbuk National University (JBNU 2020-11-001) and performed in accordance with the guidelines of the Institutional Biosafety Committee. The study was conducted in compliance with the ARRIVE guidelines.
## Measurement of the viral burden and viral protein levels in K18-hACE2 mouse tissues
Different tissues of SARS-CoV-2–infected K18-hACE2 mice were obtained by autopsy and then homogenized in a 2-mL homogenous tube (Bertin Technologies SAS, France) with RIPA lysis buffer or TRIzol (Invitrogen, Carlsbad, CA, USA) after treatment with RNA protect-tissue reagent (Qiagen, Venlo, Netherlands). Total RNA was purified as per a commercial manual, and, subsequently, cDNA was synthesized using an all-in-one master mix (Cellsafe, Yongin, South Korea) for 5 min at 25 °C, for 60 min at 42 °C, and 5 s at 85 °C. The target genes were quantified by qRT-PCR with the IQ SYBR Green (Bio-Rad, Seoul, South Korea) using target-specific primer sets. The supernatants were collected from homogenized tissues in the RIPA lysis buffer. The quantification of total protein was performed by using a BCA protein assay kit (Thermo Fisher Scientific) and subjected to SDS-PAGE, followed by Western blotting with specific antibodies. The protein was detected by developing (Poohung, Kyunggi, South Korea) into X-ray films (AGFA, Mortsel, Belgium) using an ECL kit (ELPIS, Daejeon, South Korea). The western blot images comply with the digital and integrity policy (the full, unprocessed images are included in the supplementary information file S1). The target-specific primer sets and primary antibodies are described in Supplementary Table S2.
## Histopathological and immunohistochemical analyses
SARS-CoV-2–infected brain and lung tissues were fixed in $10\%$ formalin solution (Sigma–Aldrich, St. Louis, MO, USA) and then embedded in paraffin wax (Leica Biosystems, Wetzlar, Germany). Formalin-fixed, paraffin-embedded tissue blocks were sectioned at a thickness of 4 μm with an HM 340 electronic rotary microtome (Thermo Fisher Scientific). The tissue sections were then stained with hematoxylin and eosin as per the standard laboratory protocol47, and the pulmonary abnormalities were scored based on the representative microscopic lesions. The severity of each criterion was scored 0–3 as described in Supplementary Table S1. As the lesions were not uniformly distributed and different patterns were detected in the tissues, caution was practiced when scoring. The scores for each criterion were summed, with the higher scores indicating more severe damage. For immunohistochemistry, the sections were mounted onto silane-coated slides and treated with citrate buffer (pH 6.0) at 95 °C for 30 min and room temperature for 20 min. The sections were incubated overnight with SARS-CoV-2 nucleocapsid protein (Sino Biological, China), MAP2 (Invitrogen, Carlsbad, CA, USA), and GFAP antibody (Cell Signaling Technology, CA, USA) at 4 °C. Each slide was washed thrice for 15 min each with the wash buffer (0.145 M NaCl, 0.0027 M KCl, 0.0081 M Na2HPO4, 0.0015 M KH2PO4, pH 7.4 in PBS). The sections were incubated with horseradish peroxidase-conjugated anti-rabbit IgG (Vector Laboratories, CA, USA). The antibodies were visualized with 3,3′-diaminobenzidine (Vector Laboratories) in accordance with the manufacturer’s instructions. The histopathological examinations were performed in a double-blinded manner by trained pathologists. To quantify the immunohistochemistry outcomes, the images were randomly captured from each stained tissue and analyzed by the TS Auto 5.1 (Olympus, Tokyo, Japan). The percent immunohistochemistry-positive area was analyzed in a defined magnification field and area (magnification, × 200; field, 0.144 mm2).
## TCID50 assay
SARS-CoV-2–infected K18-hACE2 mouse lungs or brains were homogenized with PBS. The clarified supernatants were collected via centrifugation and then serially diluted with the DMEM without serum. Vero E6 cells (3 × 104 cells/well) were inoculated with four replicates from 1 × 10−8 to 1 × 10−1 diluents. The diluents were then removed, and the medium was sequentially replaced with DMEM supplemented with $2\%$ FBS. At 3–5 dpi, CPE was monitored, and TCID50 of SARS-CoV-2 was calculated by the Reed and Muench method48.
## RNA-seq analysis
Total RNA was isolated and quantified with the Bioanalyzer 2100 (Agilent Technologies, CA, USA). Redundant ribosomal RNA (rRNA) was eliminated from total RNA using the RiboCop rRNA Depletion Kit (Lexogen, Vienna, Australia). The RNA-seq libraries were prepared using the Next Ultra II Directional RNA kit (NEB, MA, USA) according to the manufacturer’s protocol. The libraries were pooled and analyzed as paired-end sequenced on the NovaSeq 6000 (Illumina, CA, USA) targeting 40 million read pairs and extended. The RNA-seq reads were then aligned to the mouse reference genome (mm10) with the TopHat. *All* gene counts were preprocessed with the EdgeR to adjust the samples for differences in the library size using the trimmed mean of M values. The results of the differential signature genes were analyzed with the ExDEGA (eBiogen, Seoul, South Korea). Gene *Ontology analysis* and classification were performed using the data from the Database for Annotation, Visualization, and Integrated Discovery.
## RT2-profiler PCR array
Total RNA extracted from SARS-CoV-2–infected K18-hACE2 mouse brains at 3 and 4 dpi was synthesized using the RT2 First Strand Kit (Qiagen, Hilden, Germany). The synthesized cDNA was mixed with RT2 SYBR Green Mastermix, and the RT2 profiler™ PCR array mouse necrosis pathway (Qiagen, PAMM-141ZA/330231). The qPCR array was performed by holding for 10 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 60 s at 60 °C. The result was then analyzed with GeneGlobe (https://dataanalysis2.qiagen.com/pcr), and the Ct values were normalized with the supplied internal housekeeping gene.
## Supplementary Information
Supplementary Information 1.Supplementary Information 2.Supplementary Information 3. The online version contains supplementary material available at 10.1038/s41598-023-29909-x.
## References
1. 1.Roberts, D. L., Rossman, J. S. & Jarić, I. Dating first cases of COVID-19. PLoS Pathog.17, e1009620. 10.1371/journal.ppat.1009620 (2021).
2. Puelles VG. **Multiorgan and renal tropism of SARS-CoV-2**. *N. Engl. J. Med.* (2020.0) **383** 590-592. DOI: 10.1056/NEJMc2011400
3. Nuzzo D. **Post-acute COVID-19 neurological syndrome: A new medical challenge**. *J. Clin. Med.* (2021.0) **10** 1947. DOI: 10.3390/jcm10091947
4. Acharya A. **SARS-CoV-2 infection leads to neurological dysfunction**. *J. Neuroimmune Pharmacol.* (2020.0) **15** 167-173. DOI: 10.1007/s11481-020-09924-9
5. Mao L. **Neurologic manifestations of hospitalized patients with coronavirus disease 2019 in Wuhan China**. *JAMA. Neurol.* (2020.0) **77** 683-690. DOI: 10.1001/jamaneurol.2020.1127
6. Marshall M. **How COVID-19 can damage the brain**. *Nature* (2020.0) **585** 342-343. DOI: 10.1038/d41586-020-02599-5
7. Pleasure SJ, Green AJ, Josephson SA. **The spectrum of neurological disease in the severe acute respiratory syndrome coronavirus 2 pandemic infection: Neurologists move to the frontlines**. *JAMA Neurol.* (2020.0) **77** 679-680. DOI: 10.1001/jamaneurol.2020.1065
8. 8.Chen, T., et al. Clinical characteristics of 113 deceased patients with coronavirus disease 2019: Retrospective study. BMJ368, m1091. 10.1136/bmj.m1091 (2020).
9. 9.Hammin, I., et al. Tissue distribution of ACE2 protein, the functional receptor for SARS coronavirus. A first step in understanding SARS pathogenesis. J. Pathol.203, 631–637. 10.1002/path.1570 (2004).
10. Cantuti-Castelvetri L. **Neuropilin-1 facilitates SARS-CoV-2 cell entry and infectivity**. *Science* (2020.0) **370** 856-860. DOI: 10.1126/science.abd2985
11. 11.Coutard, B., et al. The spike glycoprotein of the new coronavirus 2019-nCoV contains a furin-like cleavage site absent in CoV of the same clade. Antiviral Res.176, 104742. 10.1016/j.antiviral.2020.104742 (2020).
12. Davies J. **Neuropilin-1 as a new potential SARS-CoV-2 infection mediator implicated in the neurologic features and central nervous system involvement of COVID-19**. *Mol. Med. Rep.* (2020.0) **22** 4221-4226. DOI: 10.3892/mmr.2020.11510
13. Krasemann S. **The blood-brain barrier is dysregulated in COVID-19 and serve as a CNS entry route for SARS-CoV-2**. *Stem Cell Rep.* (2020.0) **17** 307-320. DOI: 10.1016/j.stemcr.2021.12.011
14. 14.Ramani, A., et al. SARS-CoV-2 targets neurons of 3D human brain organoids. EMBO J.39, e106230. 10.15252/embj.2020106230 (2020).
15. Pennisi M. **SARS-CoV-2 and the nervous system: from clinical features to molecular mechanisms**. *Int. J. Mol. Sci.* (2020.0) **21** 5475. DOI: 10.3390/ijms21155475
16. Guadarrama-Ortiz P. **Neurological aspects of SARS-CoV-2 infection: Mechanisms and manifestations**. *Front. Neurol.* (2020.0) **11** 1039. DOI: 10.3389/fneur.2020.01039
17. Li Y-C, Bai W-Z, Hashikawa T. **The neuroinvasive potential of SARS-CoV-2 may play a role in the respiratory failure of COVID-19 patients**. *J. Med. Virol.* (2020.0) **92** 552-555. DOI: 10.1002/jmv.25728
18. Burki T. **The origin of SARS-CoV-2 variants of concern**. *Lancet Infect. Dis.* (2022.0) **22** 174-175. DOI: 10.1016/S1473-3099(22)00015-9
19. 19.Liu, Y., & Rocklöv, J. The reproductive number of the Delta variant of SARS-CoV-2 is far higher compared to the ancestral SARS-CoV-2 virus. J. Travel. Med.28, taab124. 10.1093/jtm/taab124 (2021).
20. 20.Lauring, A. S., et al. Clinical severity of, and effectiveness of mRNA vaccines against, covid-19 from omicron, delta, and alpha SARS-CoV-2 variants in the United States: prospective observational study. BMJ. 376, e069761. 10.1136/bmj-2021-069761 (2022).
21. Callaway E. **The mutation that helps Delta spread like wildfire**. *Nature* (2021.0) **596** 472-473. DOI: 10.1038/d41586-021-02275-2
22. Liu Z. **Identification of SARS-CoV-2 spike mutations that attenuate monoclonal and serum antibody neutralization**. *Cell Host Microbe.* (2021.0) **29** 477-488. DOI: 10.1016/j.chom.2021.01.014
23. 23.Yinda, C. K., et al. K18-hACE2 mice develop respiratory disease resembling severe COVID-19. PLoS Pathog.17, e1009195 (2021). 10.1371/journal.ppat.1009195
24. Jacob F. **Human pluripotent stem cell-derived neural cells and brain organoids reveal SARS-CoV-2 neurotropism predominates in choroid plexus epithelium**. *Cell Stem Cell* (2020.0) **27** 937-950. DOI: 10.1016/j.stem.2020.09.016
25. Gassen NC. **SARS-CoV-2-mediated dysregulation of metabolism and autophagy uncovers host-targeting antivirals**. *Nat. Commun.* (2021.0) **12** 3818. DOI: 10.1038/s41467-021-24007-w
26. Seeley JJ. **Induction of innate immune memory via microRNA targeting of chromatin remodelling factors**. *Nature* (2018.0) **559** 114-119. DOI: 10.1038/s41586-018-0253-5
27. Harvey WT. **SARS-CoV-2 variants, spike mutations and immune escape**. *Nat. Rev. Microbial.* (2021.0) **19** 409-424. DOI: 10.1038/s41579-021-00573-0
28. Deng X. **Transmission, infectivity, and neutralization of a spike L452R SARS-CoV-2 variant**. *Cell* (2021.0) **184** 3426-3437. DOI: 10.1016/j.cell.2021.04.025
29. Giacomo SD. **Preliminary report on severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike mutation T478K**. *J. Med. Virol.* (2021.0) **93** 5638-5643. DOI: 10.1002/jmv.27062
30. Fisman DN, Tuite AR. **Evaluation of the relative virulence of novel SARS-CoV-2 variants: A retrospective cohort study in Ontario Canada**. *CMAJ* (2021.0) **193** E1619-1625. DOI: 10.1503/cmaj.211248
31. Callaway E. **Delta coronavirus variant: Scientists brace for impact**. *Nature* (2021.0) **595** 17-18. DOI: 10.1038/d41586-021-01696-3
32. Sheikh A. **SARS-CoV-2 Delta VOC in Scotland: Demographics, risk of hospital admission, and vaccine effectiveness**. *Lancet* (2021.0) **397** 2461-2462. DOI: 10.1016/S0140-6736(21)01358-1
33. Winkler ES. **SARS-CoV-2 infection of human ACE2-transgenic mice causes severe lung inflammation and impaired function**. *Nat. Immunol.* (2020.0) **21** 1327-1335. DOI: 10.1038/s41590-020-0778-2
34. Kuba K. **A crucial role of angiotensin converting enzyme 2 (ACE2) in SARS coronavirus-induced lung injury**. *Nat. Med.* (2005.0) **11** 875-879. DOI: 10.1038/nm1267
35. 35.Song, E., et al. Neuroinvasion of SARS-CoV-2 in human and mouse brain. J. Exp. Med.218, e20202135 (2021). 10.1084/jem.20202135
36. 36.Wang, J., Zhao, H. & An, Y. ACE2 shedding and the role in COVID-19. Front. Cell Infect. Microbiol.11, 789180 (2021). 10.3389/fcimb.2021.789180
37. Vaduganathan M. **Renin-Angiotensin-Aldosterone system inhibitors in patients with Covid-19**. *N. Engl. J. Med.* (2020.0) **382** 1653-1659. DOI: 10.1056/NEJMsr2005760
38. Vidal E. **Chronological brain lesions after SARS-CoV-2 infection in hACE2-transgenic mice**. *Vet. Pathol.* (2022.0) **59** 613-626. DOI: 10.1177/03009858211066841
39. Chavda V. **Ischemic stroke and SARS-CoV-2 infection: The bidirectional pathology and risk morbidities**. *Neurol. Int.* (2022.0) **14** 391-405. DOI: 10.3390/neurolint14020032
40. Wan D. **Neurological complications and infection mechanism of SARS-CoV-2**. *Signal Transduct. Target Ther.* (2021.0) **6** 406. DOI: 10.1038/s41392-021-00818-7
41. Zhang B-Z. **SARS-CoV-2 infects human neural progenitor cells and brain organoids**. *Cell Res.* (2020.0) **30** 928-931. DOI: 10.1038/s41422-020-0390-x
42. Bader SM. **Programmed cell death: The pathways to severe COVID-19?**. *Biochem. J.* (2022.0) **479** 609-628. DOI: 10.1042/BCJ20210602
43. Bedoui S, Herold MJ, Strasser A. **Emerging connectivity of programmed cell death pathways and its physiological implications**. *Nat. Rev. Mol. Cell Biol.* (2020.0) **21** 678-695. DOI: 10.1038/s41580-020-0270-8
44. Humphries F. **RIP kinases: Key decision makers in cell death and innate immunity**. *Cell Death Differ.* (2015.0) **22** 225-236. DOI: 10.1038/cdd.2014.126
45. Fritsch M. **Caspase-8 is the molecular switch for apoptosis, necroptosis and pyroptosis**. *Nature* (2019.0) **575** 683-687. DOI: 10.1038/s41586-019-1770-6
46. Li S. **SARS-CoV-2 triggers inflammatory responses and cell death through caspase-8 activation**. *Signal Transduct. Target Ther.* (2020.0) **5** 235. DOI: 10.1038/s41392-020-00334-0
47. Yang M-S. **Ultra- and micro-structural changes of respiratory tracts in SARS-CoV-2 infected Syrian hamsters**. *Vet. Res.* (2021.0) **52** 121. DOI: 10.1186/s13567-021-00988-w
48. Reed LJ, Muench H. **A simple method of estimating fifty percent endpoints**. *Am. J. Epidemiol.* (1938.0) **27** 493-497. DOI: 10.1093/oxfordjournals.aje.a118408
|
---
title: Itaconate ameliorates autoimmunity by modulating T cell imbalance via metabolic
and epigenetic reprogramming
authors:
- Kuniyuki Aso
- Michihito Kono
- Masatoshi Kanda
- Yuki Kudo
- Kodai Sakiyama
- Ryo Hisada
- Kohei Karino
- Yusho Ueda
- Daigo Nakazawa
- Yuichiro Fujieda
- Masaru Kato
- Olga Amengual
- Tatsuya Atsumi
journal: Nature Communications
year: 2023
pmcid: PMC9970976
doi: 10.1038/s41467-023-36594-x
license: CC BY 4.0
---
# Itaconate ameliorates autoimmunity by modulating T cell imbalance via metabolic and epigenetic reprogramming
## Abstract
Dysregulation of Th17 and Treg cells contributes to the pathophysiology of many autoimmune diseases. Herein, we show that itaconate, an immunomodulatory metabolite, inhibits Th17 cell differentiation and promotes Treg cell differentiation by orchestrating metabolic and epigenetic reprogramming. Mechanistically, itaconate suppresses glycolysis and oxidative phosphorylation in Th17- and Treg-polarizing T cells. Following treatment with itaconate, the S-adenosyl-L-methionine/S-adenosylhomocysteine ratio and 2-hydroxyglutarate levels are decreased by inhibiting the synthetic enzyme activities in Th17 and Treg cells, respectively. Consequently, these metabolic changes are associated with altered chromatin accessibility of essential transcription factors and key gene expression in Th17 and Treg cell differentiation, including decreased RORγt binding at the Il17a promoter. The adoptive transfer of itaconate-treated Th17-polarizing T cells ameliorates experimental autoimmune encephalomyelitis. These results indicate that itaconate is a crucial metabolic regulator for Th17/Treg cell balance and could be a potential therapeutic agent for autoimmune diseases.
Dysregulation of T cell homeostasis is known to contribute to the immunopathology of autoimmune diseases. Here the authors show that itaconate impacts autoimmune pathology by altering T cells via modulation of metabolic and epigenetic programs.
## Introduction
Autoimmune diseases are characterized by the loss of self-tolerance and systemic inflammation that targets vital organs1. Immunosuppressive agents play a central role in the treatment of autoimmune diseases. Conventional drugs, including corticosteroids, are broad-acting and increase the risk of severe infection, which is a leading cause of death1,2. Although more targeted drugs against distinct immune cells or cytokines have been developed, the balance between their efficacy and side effects is still challenging2. Considering the unsatisfactory remission rate in the treatment with these drugs3, more specific treatments targeting the pathogenic mechanisms underlying autoimmune diseases are required.
Upon antigen stimulation in the presence of unique cytokine signals and microenvironment, distinct T cell subsets differentiate from naive CD4+ T cells4. Although T helper 17 (Th17) and regulatory T (Treg) cells require a common tumor growth factor (TGF)-β signal for their differentiation5, these cells fulfill opposite functions. Th17 cells play a pathogenic role in several autoimmune diseases, while Treg cells maintain immune homeostasis and inhibit autoimmunity6. Dysregulation of Th17 and Treg cells contributes to the pathophysiology of many autoimmune diseases7, including multiple sclerosis (MS)8, systemic lupus erythematosus9, and rheumatoid arthritis10. However, therapy targeting Th17/Treg cell imbalance has not been established in clinical settings11. The balance between these T cell subsets depends on cellular metabolism, which alters cellular epigenetics and transcription, and modulates their effector functions4. Effector T cells, such as T helper 1 (Th1) and Th17 cells, depend primarily on glycolysis, whereas Treg cells utilize oxidative phosphorylation (OXPHOS) and fatty acid oxidation (FAO) for survival, differentiation, and effector function. Failure to induce appropriate metabolic pathways impairs differentiation and effector function in activating CD4+ Th cells from the naïve T cells4. Pharmacological inhibition of pyruvate kinase muscle isozyme 2, which is the final rate-limiting enzyme in glycolysis, reduces glycolysis and limited Th1 and Th17 cell differentiation in vitro12. Furthermore, recent studies have shown that some metabolites regulate Th17 and Treg cell differentiation. Reduced universal methyl donor S-adenosyl-L-methionine (SAM) levels, following restriction of methionine metabolism, decreases histone methylation at the Il17a promoter and restricts Th17 cell differentiation13. During Treg cell differentiation, the reduction of 2-hydroxyglutarate (2-HG) levels, following inhibition of glutaminolysis, induced demethylation of Foxp3 and promoted Treg cell differentiation14. Previous results have suggested that T cell function and differentiation are determined by the changes of crucial “metabolites” due to the altered balance of the distinct metabolic pathways.
Itaconate (ITA) is an endogenous metabolite derived from the mitochondrial tricarboxylic acid (TCA) cycle. In recent years, ITA has gained attention owing to its anti-inflammatory, antiviral, and antimicrobial effects15–17. ITA inhibits the production of proinflammatory cytokines, including interleukin-1β (IL-1β) and IL-6, through various mechanisms in macrophages18–20. ITA has an antimicrobial effect with direct inhibition of the bacterial isocitrate lyase21 and potently inhibits viral replication and excessive inflammatory host responses to human pathogenic viruses, including severe acute respiratory syndrome coronavirus 222. Furthermore, ITA inhibits key glycolytic enzymes and impairs glycolysis in macrophages20. However, its immunomodulatory role in T cell differentiation and function has not been elucidated thoroughly.
Here, in this study, we aim to identify the role of ITA in regulating T cell differentiation and its potential as a candidate to treat T cell-mediated autoimmune diseases. Consequently, we identify ITA supplementation inhibits Th17 cell differentiation and promotes Treg cell differentiation through metabolic and epigenetic reprogramming. Furthermore, we demonstrate that ITA ameliorates the disease activity of T cell-driven autoimmune disorder model.
## Itaconate modulates Th17 and Treg cell differentiation
To determine if ITA influences T cell differentiation in vitro in naive CD4+ T cells isolated from the spleen of C57BL/6 J mice were cultured under Th1, Th2, Th17, and Treg-polarizing conditions, with or without ITA (Supplementary Fig. 1a). ITA inhibited Th17 and promoted Treg cell differentiation in a dose-dependent manner (Fig. 1a, b), without affecting their cell viability, absolute number, and proliferation (Supplementary Fig. 1b−d). In contrast, ITA did not impact the differentiation of Th1 and Th2 cells. ( Fig. 1a, b). ITA reduced the expression of Th17-related genes, including Il17a and Il17f, while ITA enhanced the expression of Rorc, which encodes a master Th17 lineage transcription factor RORγt in Th17-polarizing T cells (Fig. 1c). Additionally, ITA enhanced the expression of Foxp3 in Treg-polarizing T cells (Fig. 1c). Protein levels of RORγt were also elevated in Th17-polarizing T cells (Fig. 1d). In pathogenic Th17 cells induced by IL-6, IL-1β, and IL-23, ITA inhibited the production of IL-17 and granulocyte-macrophage colony-stimulating factor (GM-CSF) (Supplementary Fig. 1e). These results indicated that ITA regulated Th17 and Treg cell differentiation. Fig. 1Itaconate inhibits Th17 differentiation and enhances Treg differentiation. Representative flow plots (a) and cumulative data (b) of the differentiation of murine naive CD4+ T cells from wild-type B6 mice activated under Th1, Th2, Th17, and Treg cell conditions in the presence or absence of itaconate (ITA; 0, 3, and 6 mM) after 3 days culture (Th1 and Th2, $$n = 5$$; Th17, $$n = 6$$; Treg, $$n = 4$$). c Expression of Th17- and Treg-related genes in the presence or absence of ITA (0 and 6 mM) ($$n = 4$$, each condition). d Mean fluorescence intensity (MFI) of RORγt expression under Th17-polarizing condition in the presence or absence of ITA (6 mM) ($$n = 5$$). P values are calculated using one-way ANOVA with Bonferroni post hoc test for (b) and two-tailed unpaired Student’s t-test for (c, d). Data are representative of mean ± s.e.m. Source data are provided as a Source Data file.
ITA is a metabolite produced by macrophages in which Irg1 is highly expressed. In contrast, our study demonstrated no upregulation of Irg1 in T cells (Supplementary Fig. 1f). Moreover, a previous study has shown that ITA could be released from activated macrophages at an extracellular concentration of 1–5 μM23. However, here, we demonstrated that the effect of ITA on Th17 and Treg differentiation was not statistically significant at a lower concentration than 1 mM (Supplementary Fig. 1g, h). Altogether, these results suggested that T cell differentiation was not affected by macrophage-derived ITA in vivo.
## Itaconate ameliorated the EAE model
To investigate the potential of ITA as a candidate to treat autoimmune diseases, we used an experimental autoimmune encephalomyelitis (EAE) model in vivo. EAE mice were intraperitoneally injected with 50 mg kg−1 ITA every other day from day 0 to day 14 post-immunization with myelin oligodendrocyte glycoprotein (MOG)35-55 and complete Freund’s adjuvant (CFA). ITA treatment significantly reduced the clinical scores and loss of body weight in EAE mice compared to PBS treatment (Fig. 2a, b). These results indicated the potential of ITA to reduce the disease activity of EAE.Fig. 2Itaconate ameliorates experimental autoimmune encephalomyelitis.a, b C57BL/6J mice were immunized with myelin oligodendrocyte glycoprotein (MOG)35-55 and complete Freund’s adjuvant (CFA). Mice were intraperitoneally injected with 50 mg kg−1 itaconate (ITA) every other day from day 0 to day 14. Clinical scores (a) and body weight (b) in EAE mice treated with PBS (Ctrl, $$n = 11$$) or ITA ($$n = 11$$). c–j For adoptive transfer EAE (tEAE), pathogenic Th17-polarizing CD4+ T cells from 2D2 mice were cultured with or without ITA ex vivo for 3 days. Then, the harvested cells were transferred to recipient Rag1-deficient mice intravenously. Clinical scores (c) and body weight (d) of control (Ctrl, $$n = 10$$) or ITA ($$n = 13$$) recipient mice in tEAE models. e Representative histology of spinal cord stained with hematoxylin and eosin (H&E) and luxol fast blue (LFB). Scale bar, 500 μm. f Inflammation scores of spinal cords are shown (Ctrl, $$n = 6$$; ITA, $$n = 8$$, biologically independent samples). Absolute numbers (left) and frequency (right) of IL-17A+ (g) and IL-17A+GM-CSF+ (h) CD4+ T cells in the spinal cord of recipient mice, as assessed using flow cytometry ($$n = 4$$ in each condition, biologically independent samples). i Absolute number of macrophages, Ly6Chi monocytes, and neutrophils in the spinal cord of Rag1-deficient mice on day 14 after induction of EAE (Ctrl, $$n = 11$$; ITA, $$n = 10$$, biologically independent samples). j Absolute number of IL-1β-producing cells in three cell types as described in (i) (Ctrl, $$n = 11$$; ITA, $$n = 10$$, biologically independent samples). Data of e are representative of four independent experiments with similar results. P values are calculated using two-way ANOVA for (a–d) and two-tailed unpaired Student’s t-test for (f–j). Data are representative of mean ± s.e.m. Source data are provided as a Source Data file.
The MOG-induced EAE model could not strictly exclude the possible effects of ITA on cell types other than T cells, such as the anti-inflammatory effect on macrophages. To further investigate the role of ITA on in vivo functionality of Th17, pathogenic Th17-polarizing CD4+ T cells from 2D2 mice were cultured ex vivo for 3 days following adoptive transfer into Rag1-deficient mice (Supplementary Fig. 2a). ITA treatment reduced the percentage of IL-17A and GM-CSF-producing CD4+ T cells in ex vivo culture (Supplementary Fig. 2b). Clinical scores and loss of body weight were significantly reduced in Rag1−/− mice administered ITA-treated cells compared to those in mice administered with control cells (Fig. 2c, d). Histological sections of spinal cords showed significantly reduced cell infiltration and demyelination in mice administered ITA-treated cells (Fig. 2e, f). Furthermore, mice administered with ITA-treated Th17 cells showed a significantly reduced number and percentage of IL-17A-producing CD4+ T cells, as well as IL-17A and GM-CSF-producing CD4+ T cells, in the spinal cord 15 days after EAE induction compared to that in control mice (Fig. 2g, h). We next evaluated the infiltrating macrophages, inflammatory Ly6Chi monocytes, and neutrophils, which drive T cells that mediate pathology in EAE24, in the spinal cord of the adoptive transfer EAE (tEAE) model. The absolute number of these infiltrated cells or IL-1β-producing cells in recipient mice administered ITA-treated Th17 cells did not differ significantly from that in their counterparts (Supplementary Fig. 2e and Fig. 2i, j). These data suggest that in vivo induction of neutrophils, inflammatory monocytes, and macrophages is unlikely to be the main pathogenicity of the tEAE attenuation by ITA-treated Th17 cells. Additionally, the intraperitoneal injection of ITA to Rag1-deficient recipient mice following the adoptive transfer of Th17-polarizing CD4+ T cells from 2D2 mice significantly attenuated the severity of the adoptive tEAE model (Supplementary Fig. 2c, d). Collectively, these findings suggest that ITA is a potential therapeutic agent to treat T cell-driven autoimmune disorders.
## Itaconate inhibits glycolysis and OXPHOS under Th17- and Treg-polarizing conditions
To understand the mechanisms underlying ITA-dependent modulation of T cell differentiation, we performed RNA sequencing (RNA-seq) using Th17- and Treg-polarizing T cells with or without ITA treatment (Fig. 3a, b). A total of 1072 differentially expressed genes (DEGs) were identified between ITA-treated and control Th17-polarizing T cells (Fig. 3c). In ITA-treated Treg-polarizing T cells, 1554 genes were identified as DEGs, of which 610 DEGs were identified in Th17-polarizing T cells as well (Fig. 3d). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis identified that some metabolic pathways were enriched in Th17- and Treg-polarizing T cells (Supplementary Fig. 3a, b). Among Th17 signature genes25, the relative expression of genes that encode the effector cytokine IL-17A (Il17a) and F (Il17f) and signal transducer and activator of transcription 3 (Stat3) were downregulated by ITA under Th17 condition compared to that in control (Fig. 3c and Supplementary Fig. 3c). Interestingly, the relative expression of gene encoding important transcription factor for Th17 polarization, including RORγt (Rorc) was not downregulated by ITA (Supplementary Fig. 3c). KEGG pathway analysis of the overlapped DEGs between ITA-treated and control T cells under Th17 and Treg conditions revealed significant enrichment of ‘glycolysis’ and ‘HIF-1 signaling pathway’ (Fig. 3e). The results showed that several glycolysis-related genes, including the glucose transporter Slc2a1 (encoding Glut-1), were downregulated in ITA-treated T cells under Th17- and Treg-polarizing conditions (Fig. 3f). Glut-1 regulates glucose uptake and glycolysis upon naive CD4+ T cell activation and differentiation26. These results indicated that ITA broadly inhibited glycolysis-related gene expression. Fig. 3Itaconate decreases glycolysis-related genes under Th17- and Treg-polarizing conditions.a–f Th17- and Treg-polarizing T cells from B6 mice in the presence or absence of itaconate (ITA) after 2 days of culture were subjected to RNA sequencing ($$n = 3$$ in each condition, independent experiments). Heatmap (a) shows the top 100 and bottom 100 modified genes from RNA-seq in Th17- (left) and Treg- (right) polarizing T cells in the presence or absence of ITA. Principle component analysis of global gene expression from RNA-seq in four T cell populations (b). Volcano plot of differential gene expression in ITA-treated Th17- (c) or Treg- (d) polarizing T cells compared to Control (Ctrl). Venn diagram (e) displays the overlapping DEGs between ITA-treated Th17- and Treg-polarizing T cells. The top gene ontology pathways in the overlapping group are shown. P value for (e) indicates gene enrichment analysis test implemented in Metascape without adjustment for multiple comparisons. Heatmap (f) shows relative expression (z-score) of glycolysis and HIF-1α-related genes according to RNA-seq. Source data are provided as a Source Data file.
Hypoxia-inducible factor 1 alpha (HIF-1α) is an important regulator of glycolysis4. HIF-1α-deficient T cells demonstrate suppressed glycolysis, resulting in reduced Th17 cell differentiation and increased Treg cell differentiation27. ITA modestly increased Hif1a gene expression (Fig. 3f) and HIF-1α protein levels under Th17- and Treg-polarizing conditions (Fig. 4a, b). We next performed glycolytic rate assays using an extracellular flux analyzer to assess the metabolic function of T cell differentiation in the presence or absence of ITA. Basal glycolysis and compensatory glycolysis were inhibited in ITA-treated Th17- and Treg-polarizing T cells, as demonstrated by extracellular acidification rate (ECAR) (Fig.4c, d). Given that ITA inhibits the enzymatic activity of succinate dehydrogenase (SDH) in macrophages28, we also evaluated mitochondrial OXPHOS. The analysis of mitochondrial oxygen consumption rate (OCR) showed decreased basal and maximal respiration and ATP production in ITA-treated Th17- and Treg-polarizing T cells (Fig. 4e, f). Overall, these results indicate that ITA inhibits glycolysis and OXPHOS in Th17- and Treg-polarizing T cells. Fig. 4Itaconate inhibits glycolysis without suppressing HIF-1α. Intracellular HIF-1α expression in Th17- and Treg-polarizing T cells with or without itaconate (ITA) (a), percentage of HIF-1α+ cells under each growth condition (b) (Th17, $$n = 3$$; Treg, $$n = 4$$). Extracellular acidification rate (ECAR) of Th17- (c) and Treg-polarizing T cells (d) with or without ITA measured using glycolytic rate assay (Th17, $$n = 4$$; Treg, $$n = 3$$). Basal glycolysis and compensatory glycolysis were calculated. Mitochobdrial oxygen consumption rate (OCR) using an extracellular flux analyzer in Th17- (e) and Treg-polarizing T cells (f) with or without ITA ($$n = 4$$, each condition). P values are calculated using two-tailed unpaired Student’s t-test for (b–f). Data are representative of mean ± s.e.m. Source data are provided as a Source Data file.
Unique cytokine signals direct metabolic changes and distinct CD4+ T cell subsets differentiation4. To investigate whether the effect of ITA on T cell differentiation is dependent on the cytokine signals, we performed RNA-seq using Th0 cells with or without ITA treatment. We identified 1461 DEGs between ITA-treated and control T cells under Th0 condition (Supplementary Fig. 4a). In KEGG pathway analysis, these genes presented main pathways different from those presented by Th17 and Treg RNA-seq data (Supplementary Fig. 4b). The absence of glycolysis and HIF-1 signaling pathway in the overlap between DEGs between ITA-treated and control T cells under Th0, Th17, and Treg conditions suggested that ITA strongly induced these metabolic changes dependent on unique cytokine signals for Th17 or Treg differentiation (Supplementary Fig. 4c). Furthermore, pre-treatment of Th0 cells with ITA followed by polarization in Th17 or Treg conditions induced no significant change in Th17 differentiation and a slight increase in Treg differentiation (Supplementary Fig. 4d, e). These findings indicate that the coordination between ITA treatment and unique cytokine signals is required to regulate Th17/Treg differentiation in ITA-treated T cells.
## Itaconate induces key metabolic changes by inhibiting MAT and IDH1/2 enzymatic activity
To further unravel metabolic profiling, we performed a metabolomic analysis using capillary electrophoresis time-of-flight mass spectrometry (CE-TOF MS) to evaluate intracellular metabolites in ITA-treated Th17- and Treg-polarizing T cells (Fig. 5a and Supplementary Fig. 5a, b). ITA treatment in vitro increased intracellular levels of ITA in activated naïve CD4+ T cells after CD3/CD28 stimulation (Supplementary Fig. 5c). Consistent with the RNA-seq results, the levels of glycolysis pathway metabolites were markedly decreased in ITA-treated Th17- and Treg-polarizing cells (Fig. 5b).Fig. 5Metabolic reprogramming and enzymatic inhibition in Itaconate-treated Th17- and Treg-polarizing T cells.a Intracellular levels of metabolites implicated in TCA cycle, glutaminolysis, and methionine metabolism from Th17- and Treg-polarizing T cells in the presence or absence of itaconate (ITA) after 2 days of culture ($$n = 3$$ independent experiments). b Heatmap of glycolytic metabolites from Th17- and Treg-polarizing T cells in the presence or absence of ITA. c Methylation index (SAM/SAH ratio) of Th17- and Treg-polarizing T cells in the presence or absence of ITA ($$n = 3$$, each condition). Enzymatic activity of methionine adenosyltransferase (MAT) (d) (Th17, $$n = 3$$; Treg, $$n = 4$$) and isocitrate dehydrogenase (IDH)1 and 2 (e) in Th17- and Treg-polarizing T cells with or without ITA ($$n = 4$$, each condition). IDH1 (f) and 2 (g) activity were measured in the presence of a variable concentration of ITA. IDH activity was calculated as the ratio of measured data to the average control value. Cumulative data of the differentiation of murine naive CD4+ T cells from Nrf2-knockout mice activated under Th17- (h), and Treg cell conditions (i) in the presence or absence of ITA ($$n = 4$$, each condition). P values are calculated using two-tailed unpaired Student’s t-test for (a, c–e) and one-way ANOVA with Bonferroni post hoc test for (h, i). arb.units, arbitrary unit. Data are representative of mean ± s.e.m. Source data are provided as a Source Data file.
The universal methyl donor SAM, which is synthesized from methionine by an enzyme methionine adenosyltransferase (MAT), regulates gene expression by modulating histone methylation via histone methyltransferases29. Decreased SAM levels induced the demethylation of histone H3K4 trimethylation (H3K4me3) in the Il17a promoter, resulting in the downregulation of Il17a gene expression in Th17 cells13. We identified that ITA treatment significantly decreased the level of SAM (Fig. 5a). Interestingly, the methylation index, calculated as the SAM to S-adenosylhomocysteine (SAH) ratio, which is an indicator of cellular methylation potential13, was reduced in ITA-treated Th17-polarizing T cells, but not in ITA-treated Treg-polarizing T cells (Fig. 5c). Further supporting our results, ITA-treated Th17 cells exhibited a similar trend towards the change of Th17 signature genes (Supplementary Fig. 3c), which were observed in Th17 cells cultured under methionine restriction13. The accumulation of 2-HG, synthesized by wild-type isocitrate dehydrogenase (IDH)1 and IDH2, reduces FOXP3 levels in T cells14. Similarly, the knockdown of both IDH1 and IDH2 reduces the production of 2-HG and increases the expression of FOXP314. We found that ITA reduced the level of 2-HG in Treg-polarizing T cells but not in Th17-polarizing T cells (Fig. 5a). We examined whether increasing intracellular SAM or 2-HG restored the effect of ITA on Th17 or Treg cell differentiation, respectively. Intracellular SAM is derived from extracellular methionine in Th17-polarizing T cells13. Treatment with escalating doses of methionine gradually promoted Th17 differentiation in ITA-treated Th17-polarizing T cells (Supplementary Fig. 5g). Additionally, increasing doses of cell-permeable 2-HG impaired Treg differentiation in ITA-treated Treg-polarizing T cells (Supplementary Fig. 5h). These data indicate that the SAM or 2-HG levels affect the polarization programs of ITA-treated Th17 or Treg cells, respectively.
To understand the functional relevance of the altered metabolic intermediates, including SAM and 2-HG, the activities of the synthetic enzymes MAT and IDH$\frac{1}{2}$ were evaluated. In line with the metabolic profiling analysis, the enzymatic activity of MAT was inhibited in ITA-treated Th17- and Treg-polarizing cells (Fig. 5d and Supplementary Fig. 5d). Besides, the enzymatic activities of IDH$\frac{1}{2}$ were inhibited in Treg-polarizing T cells, but not in ITA-treated Th17-polarizing T cells upon ITA treatment (Fig. 5e and Supplementary Fig. 5e). A previous study has shown that ITA binds directly to TET-family DNA dioxygenases like the co-substrate α-ketoglutarate, and inhibits its catalytic activity30. Since isocitrate, a co-substrate of IDH, has a similar structure to ITA, we speculated that ITA directly inhibits IDH activity. As shown in Fig. 5f, g, our results supported this speculation and showed that ITA directly inhibited the activity of purified wild-type IDH1 and 2. Further, we prepared whole-cell extract including IDH from Th17- and Treg-polarizing T cells and demonstrated that ITA inhibits the IDH activity of the extract derived from Treg but not Th17 (Supplementary Fig. 5f). Our metabolomics data revealed that ITA treatment led to increased succinate and decreased fumarate levels in T cells, suggesting an inhibitory effect on SDH (Fig. 5a). In the TCA cycle, the metabolite gradients generated by SDH inhibition were preserved and led to decreased isocitrate levels in ITA-treated Treg cells alone after the influx of metabolic components such as pyruvate (Fig. 5a). Based on these data, we inferred that the difference in the co-substrate levels of each cell may affect the inhibitory effect of ITA in IDH enzyme-catalyzed reaction. Altogether we demonstrated the functional link between the metabolic profile and enzymatic activities, indicating that ITA modulates the balance between Th17 and Treg cell differentiation via interaction with key enzymes and metabolites.
ITA exerts anti-inflammatory effects via activation of Nrf2 in macrophages18. We quantified the expression of Nrf2 (Nfe2l2) and Hmox1, a prototypical Nrf2 target gene31, in Th17- and Treg-polarizing T cells with or without ITA. ITA showed a trend toward increased gene expression of Nfe2l2; however, the expression of the downstream target gene transcripts was not significantly increased (Supplementary Fig. 5i). Immunoblotting exhibited that the protein levels of NRF2 and heme oxygenase 1 (HMOX1) did not increase in ITA-treated Th17- and Treg- polarizing T cells compared to those in the control (Supplementary Fig. 5j−l). Nrf2 is also known to facilitate glutaminolysis and redirect glutamate into anabolic pathways32. We assessed whether Nrf2 contributes to ITA-mediated regulation of Th17/Treg cell differentiation using Nrf2-deficient mice. We demonstrated that ITA inhibited Th17 cell differentiation and promoted Treg cell differentiation in Nrf2-deficient T cells (Fig. 5h, i). These results suggest that Nrf2 activation is unlikely to be the main mechanism of the regulation of Th17/Treg cell differentiation by ITA.
## Metabolic changes were associated with altered chromatin accessibility of essential transcription factors in Th17 and Treg cell differentiation
Histone modifications can alter the accessibility of transcription factors to certain genomic regions33. Although ITA increased RORγt and HIF-1α expression, a significant decrease in IL-17A expression was observed (Figs. 1b, d and 4a, b). Therefore, we hypothesized that ITA alters chromatin accessibility at key gene loci. To test this hypothesis, we examined RORγt-binding to the Il17a promoter in ITA-treated Th17-polarizing T cells using chromatin immunoprecipitation (ChIP) analysis. Naive CD4+ T cells were analyzed as a negative control. Our results showed that ITA suppressed RORγt binding to the Il17a promoter (Fig. 6a). During Th17 differentiation, multiple transcription factors are required for the induction of RORγt and IL-17A. In TCR-activated CD4+ T cells, BATF and IRF4 bind cooperatively to the regions of Il17a loci. Following Th17-polarizing cytokine activation, HIF-1α, STAT3, and Runx1 are recruited to the same regions. Finally, the RORγt binding to the regions determines Il17a gene expression34. Because multiple transcription factors are involved in T cell differentiation, we performed an assay for transposase-accessible chromatin sequencing (ATAC-seq) to reveal whether ITA changes chromatin accessibility of the transcription factors in Th17- and Treg-polarizing T cells. As RNA expression depends on DNA accessibility, we integrated the ATAC-seq and RNA-seq datasets (Fig. 6b). Overall, 703 DEGs showed differential accessibility between ITA-treated and control T cells under Th17 or Treg conditions. *These* genes included Il17a, glycolysis-related genes, and foxp3, but not csf2, which encodes GM-CSF (Fig. 6c, d). We next focused on two groups: chromatin more closed in ITA-treated Th17 cells, and chromatin more open in ITA-treated Treg cells. The analysis of transcription factor binding motifs revealed that the peaks of Th17 group were enriched for the motifs regulating Il17a gene expression, including BATF, STAT3, IRF4, RUNX1, and HIF-1α (Fig. 6e). In addition, motif discovery also indicated that the peaks of Treg group were enriched for ETS1, RUNX1, SMAD3, STAT5, and AP-1 (Fig. 6f), suggesting that ITA mainly altered the accessibility of conserved non-cording sequence 2 (CNS2)-targeting transcription factors which maintain Foxp3 expression35. Collectively, these results demonstrated that ITA treatment leads the chromatin accessibility to closed in Il17a loci and open in Foxp3 loci for key transcription factors. Fig. 6Itaconate altered the chromatin accessibility of essential transcription factors in Th17 and Treg cell differentiation.a Chromatin immunoprecipitation (ChIP) analysis of RORγt at the Il17a promoter regions in naive CD4+ and Th17-polarizing T cells in the presence or absence of itaconate (ITA) (Naive CD4+, $$n = 3$$; Th17, $$n = 3$$). b–f Assay for transposase-accessible chromatin sequencing (ATAC-seq) was performed using Th17- and Treg-polarizing T cells from B6 mice in the presence or absence of ITA after 2 days of culture. Two replicates ($$n = 2$$, each group) were used for ATAC-seq. Venn diagram displaying the overlap between DEGs between ITA-treated and control T cells under Th17 or Treg conditions and between genes that show differential chromatin accessibility (DA) (b). Heatmap displaying relative expression (z-score) of genes that show DA and DE between ITA-treated and control T cells under Th17 or Treg conditions (c). d Representative ATAC-seq tracks in Th17- and Treg-polarizing T cells. e Motif discovery of the peaks which significantly changed to be more closed in ITA-treated Th17 cells compared to Ctrl. f Motif discovery of the peaks which significantly changed to be more open in ITA-treated Treg cells compared to Ctrl. P values are calculated using one-way ANOVA with Bonferroni post hoc test for (a) and two-sided Wald test with Benjamini and Hochberg method for (e, f). Data are representative of mean ± s.e.m. Source data are provided as a Source Data file.
## Discussion
Our study revealed that ITA is a key metabolite in metabolic and epigenetic reprogramming to suppress Th17 and promote Treg cell differentiation. Consistently, the adoptive transfer of ITA-treated Th17-polarizing T cells ameliorated EAE.
Activated CD4+ T cells require metabolic and epigenetic reprogramming for proliferation and differentiation from the naive state4. Several studies have shown that multiple metabolic pathways, including glycolysis, glutaminolysis, and one-carbon metabolism, induce the reprogramming and control of Th17/Treg cell differentiation13,14,36. However, it remains unclear whether these processes are regulated independently of one another. Although the mammalian target of rapamycin (mTOR)-HIF-1α pathway is a crucial process for integrating glycolysis and Th17/Treg cell differentiation27, our results showed that ITA inhibits glycolysis without suppressing HIF-1α. Following inhibition of discrete metabolism, the intracellular metabolites, such as SAM and 2-HG, influence Th17 or Treg cell differentiation by epigenetic reprogramming13,14. Mechanistically, the reduced SAM/SAH ratio in Th17 cells induced histone demethylation not only H3K4me3 but also histone marks for transcriptional repression, including H3K27me313. Decreased 2-HG induced hypomethylation at the Foxp3 promoter and CNS214. Given that individual histone marks have discrete regulatory roles33, it is complicated how histone modification by altering the metabolites contributes to gene expression. Our study revealed that ITA decreased SAM/SAH ratio and 2-HG by inhibiting MAT and IDH$\frac{1}{2}$, respectively. Consequently, ITA altered the chromatin accessibility of essential transcription factors at the Il17a and Foxp3 loci, resulting in suppressed IL-17A and increased FOXP3 expression.
In this study, we showed the modulatory role of ITA in mediating the balance of Th17/Treg cell differentiation. The high polarity and low electrophilicity of ITA result in its low cell permeability37. To further confirm the mechanism of ITA uptake in T cells, a transport-mediated process such as mitochondrial oxoglutarate, dicarboxylate, and citrate carriers18 should be assessed. We focused on the discrepancy that RORγt expression was increased with ITA, but IL-17A expression was decreased. Concordant with our study, a previous study has reported that Th17 cells cultured under methionine restriction, which induces SAM reduction, showed reduced IL-17 production but stable RORγt expression13. Decreased SAM levels induced demethylation of histone H3K4me3 at the Il17a promoter, resulting in the downregulation of Il17a gene expression in Th17 cells. In the same study, SAM reduction also resulted in the demethylation of H3Kme3 around the Rorc transcription start site, but this did not decrease *Rorc* gene expression. Thus, the influence of SAM reduction on the histone modification may not be specific to Il17a gene loci, and additional mechanisms may affect Rorc expression. Further studies are required to reveal this selectivity of gene expression. In the motif analysis, RORγt was not detected in the Th17 group. Reportedly, RORγt possesses some binding sites other than at the promoter region of Il17a38, and the promoter region is marked by H3K4me339, the most labile mark affected by the restriction of the methionine cycle13. Therefore, we speculated that ITA might not change the accessibility of RORγt at the Il17a locus other than at the promoter region, and the accessibility in these RORγt-binding sites does not appear to change. Although our work focused on key metabolites and the binding of essential transcription factors at the Il17a and Foxp3 loci, the findings indicate that ITA may regulate other metabolites or transcription factors that impact T cell differentiation. As multiple pathways are involved in the anti-inflammatory effect of ITA on macrophages18,19, unknown mechanisms may contribute to the regulation of ITA-mediated T cell differentiation. In addition, future studies should address the reliable delivery of ITA to T cells for translation of its use in a clinical context.
In summary, we identified ITA as a key regulator reducing Th17 cell differentiation and promoting Treg cell differentiation through metabolic and epigenetic reprogramming. Our results could integrate previous knowledge of key metabolites and epigenetics in T cells and offer mechanisms and options for the modulation of T cell differentiation. Given the pathogenic roles of Th17/Treg imbalance in a wide variety of autoimmune diseases, our study makes a worthwhile contribution to suggesting simple therapeutic approaches which regulate T cell differentiation.
## Mice
C57BL/6J mice were purchased from Charles River Laboratories (Wilmington, MA). 2D2 (C57BL/6-Tg(Tcra2D2, Tcrb2D2)1Kuch/J) and Rag1−/− (B6.129S7-Rag1tm1Mom/J) were purchased from Jackson Laboratories (Bar Harbor, ME). Nrf2−/− (B6.129P2-Nfe2l2tm1Mym/MymRbrc) mice were provided by the RIKEN BRC through the National BioResource Project of the MEXT/AMED, Japan. All mice were bred in house and maintained in temperature- and humidity-controlled facilities under pathogen-free conditions at the Hokkaido University, group housed with free access to food and water and 12 h light/dark cycles. Both male and female mice were used at 8–10 weeks old with age- and sex-matched controls. All animal experiments were approved by the Institutional Animal Care and Use Committee of Hokkaido University (permission number: 19-0147).
## T cell in vitro activation and culture
Naive CD4+ T cells were isolated from the murine spleen by magnetic cell sorting with the naive CD4+ T Cell Isolation kit (Miltenyi Biotec). Approximately 0.3 million naive CD4+ T cells were plated into 48-well-plate pre-coated with goat anti-hamster IgG (MP Biomedicals) and stimulated for 2–3 days with anti-CD3 (0.25 μg mL−1, clone 145-2C11, Biolegend, 100340) and anti-CD28 (0.5 μg mL−1, clone 37.51, Biolegend, 102116) antibodies40. For each T cell differentiation, subset-specific antibodies and cytokines were further supplemented. For Th1 differentiation, cells were cultured with anti-IL-4 antibody (3 μg mL−1, clone 11B11, Biolegend, 504122) and IL-12 (20 ng mL−1, Biolegend, 577004). For Th2 differentiation, cells were cultured with anti-IFNγ antibody (3 μg mL−1, clone AN-18, Biolegend, 517906) and IL-4 (100 ng mL−1, Biolegend, 574304). For Th17 differentiation, cells were cultured with anti-IL-4 antibody (2 μg mL−1, clone 11B11, Biolegend, 504122), anti-IFNγ antibody (2 μg mL−1, clone AN-18, Biolegend, 517906), IL-6 (30 ng mL−1, Biolegend, 575704), and TGF-β (0.3 ng mL−1, Miltenyi Biotec, 130-095-066). For Treg differentiation, cells were cultured with anti-IL-4 antibody (2 μg mL−1, clone 11B11, Biolegend, 504122), anti- IFNγ antibody (2 μg mL−1, clone AN-18, Biolegend, 517906), IL-2 (20 ng mL−1, Biolegend, 575404), and TGF-β (1 ng mL−1, Miltenyi Biotec, 130-095-066). All cells were cultured in RPMI 1640 medium containing $10\%$ FBS, $0.1\%$ 2-mercaptoethanol, and penicillin-streptomycin at 37 °C under $5\%$ CO2. Itaconate (ITA, Sigma-Aldrich, I29204), L-Methionine (Sigma-Aldrich, M5308), and disodium (R)−2-hydroxtglutarate (2-HG, Selleck, S7873) were prepared as 750, 250, 100 mM stock solutions in PBS, respectively, and diluted directly into culture media at various concentrations. The medium, including ITA, L-Methionine, and 2-HG, was adjusted to pH 7.4 with 1 N NaOH at 37 °C.
## BMDMs culture
Bone marrow cells were harvested from the femur and tibia of C57BL/6 J mice and differentiated in the presence of M-CSF (20 ng mL−1, R&D Systems, 416-ML-010) in RPMI 1640 medium containing $10\%$ FBS, $0.1\%$ 2-mercaptoethanol, and penicillin-streptomycin at 37 °C under $5\%$ CO2 for 8 days. On day 8, the bone-marrow-derived macrophage (BMDMs) were washed and stimulated with lipopolysaccharide (LPS, 100 ng mL−1, Sigma-Aldrich, L2880) for 6 h.
## Immunoblotting
Cultured T cells were dissolved in lysis buffer (FUJIFILM Wako, 038-21141) supplemented with protease inhibitor (Sigma-Aldrich, P8340). The lysates were boiled at 95 °C for 5 min in Laemmli sample buffer. They were then resolved on 4−$15\%$ SDS-PAGE gel electrophoresis and transferred to a polyvinylidene difluoride membrane. The membranes were then blocked for 30 min with $2.5\%$ nonfat milk in TBS-Tween for 30 min at 37 °C. After blocking, the membranes were incubated overnight at 4 °C with the following primary antibodies: anti-mouse NRF2 (clone D1Z9C, Cell Signaling Technology, 12721S; 1:500), anti-mouse HO-1 (HMOX1, clone E6Z5G, Cell Signaling Technology, 82206; 1:500), and anti-β-actin (clone AC-15, Sigma-Aldrich, A3854; 1:50,000); β-actin was used as an endogenous control. Afterward, horseradish peroxidase-conjugated anti-rabbit (Sigma-Aldrich, A0545; 1:80,000) and anti-mouse (Sigma-Aldrich, A9044; 1:120,000) antibodies were used as secondary reagents. Signals were detected using enhanced chemiluminescence western blot detection reagents (Cytiva, RPN2209) and the LAS-4000 imaging system (FUJIFILM, Japan).
## Flow cytometry
Cells were collected from in vitro culture after 3 days post-activation or in vivo mice experiments. For intracellular staining, cells were re-stimulated with phorbol 12-myristate 13-acetate (PMA: 500 ng mL−1, Sigma-Aldrich, P1585) and ionomycin (1 μg mL−1, Sigma-Aldrich, I19657) under GolgiStop (BD Biosciences, 554724), except for Foxp3 detection. After 4 h, cells were stained with Zombie Aqua Fixable Viability Kit (Biolegend, 423102) to eliminate dead cells and anti-CD16/CD32 mix (0.01 mg mL−1, clone 2.4G2, BD Biosciences, 553141) to prevent nonspecific binding. Cells were fixed and permeabilized with fixation and permeabilization solution (BD Biosciences, 554715), followed by staining with cytokine- and/or transcription factor-specific antibodies as follows: anti-mouse CD4 (clone H129.19, Biolegend, 130308; 1:100), CD3 (clone 17A2, Biolegend, 100217; 1:100), CD90.2 (clone 53-2.1, Biolegend, 140310; 1:100), CD11b (clone M$\frac{1}{70}$, BD Biosciences, 557657; 1:100), Ly6C (clone HK1.4, Biolegend, 128006; 1:100), Ly6G (clone 1A8, Biolegend, 127613; 1:100), B220/CD45R (clone RA3-6B2, Biolegend, 563103; 1:100), F$\frac{4}{80}$ (clone BM8, Biolegend, 123113; 1:100) antibodies, for surface, and anti-mouse IFN-γ (clone XMG1.2, Biolegend, 505830; 1:50), IL-4 (clone 11B11, Biolegend, 504104; 1:50), IL-17A (clone TC11-18H10.1, Biolegend, 506904; 1:50), RORγt (clone B2D, eBioscience, 17-6981-80; 1:50), GM-CSF (clone MP1-22E9, eBioscience, 17-7331-82; 1:50), HIF-1α (clone 241812, R&D Systems, IC1935A; 1:50), and pro-IL-1β (clone NJTEN3, eBioscience, 12-7114-80; 1:50) antibodies, for intracellular. For Treg staining, cells were stained with anti-mouse CD25 (for surface, clone 3C7, BD Biosciences, 564370; 1:100), Foxp3 (for intracellular, clone FJK-16s, eBioscience, 12-5773-82; 1:50), and HIF-1α (for intracellular, clone 241812, R&D Systems, IC1935A) antibodies. All flow cytometry data were acquired on a BD FACSAria III cell sorter (BD Biosciences) and analyzed using the Flowjo software.
## Cell proliferation assay
Naive CD4+ T cells were labeled with CSFE (1 μM, BD Biosciences, 565082) and then cultured under Th17 or Treg cell conditions for 3 days before FACS analysis.
## Active EAE model
EAE was induced by immunizing C57BL/6J mice with myelin oligodendrocyte glycoprotein (MOG)35-55 peptide (100 μg per mouse, ANASPEC, 60130-5) and complete Freund’s adjuvant (CFA) containing 4 mg ml−1 (0.4 mg per mouse) of heat-killed *Mycobacterium tuberculosis* (Chondrex, 7001)24. Mice were intraperitoneally injected with pertussis toxin (PTX, 250 ng/mouse, List Biological Laboratories, 180) on days 0 and 2. PBS or ITA (50 mg kg−1) was intraperitoneally injected every other day from day 0 to day 14. An ITA solution was adjusted to pH 7.4 with 1 N NaOH at 37 °C. The disease scores were assigned according to the following scale: 0, no clinical signs; 0.5, partially limp tail;1, paralyzed tail; 2, loss in coordinated movement, hind limb paresis; 2.5; one hind limb paralyzed; 3, both hind limbs paralyzed; 3.5, hind limbs paralyzed, weakness in forelimbs; 4, forelimbs paralyzed; 5, moribund or death40.
## Passive transfer EAE model
On day 0, naive CD4+ T cells were isolated from 8–10-week-old 2D2 mice, as described above. Approximately 0.2 million naive CD4+ T cells were plated into 48-well-plate pre-coated with goat anti-hamster IgG (MP Biomedicals) and stimulated for 3 days with anti-CD3 (0.25 μg mL−1, clone 145-2C11, Biolegend) and anti-CD28 (0.5 μg mL−1, clone 37.51, Biolegend) antibodies12. Cells were cultured under Th17-polarizing conditions with anti-IL-4 (2 μg mL−1, clone 11B11, Biolegend), anti-IFNγ (2 μg mL−1, clone AN-18, Biolegend), IL-6 (30 ng mL−1, Biolegend), IL-23 (10 ng mL−1, Biolegend, 589004), IL-1β (10 ng mL−1, Biolegend, 575104), and TGF-β (0.3 ng mL−1, Miltenyi Biotec) with or without ITA (3 mM, Sigma-Aldrich). Cultured cells were purified on day 3 of culture, and 1 × 107 cells were injected intravenously into each Rag1-deficient recipient mouse (8–10-week-old). Pertussis toxin (400 ng per mouse, List Biological Laboratories, 180) was intraperitoneally injected on the day of transfer and two days later. After 14 days of EAE induction, the mice were anesthetized, and the spinal cords were collected for analysis.
For histological staining, sections from $10\%$ formalin-fixed spinal cords were stained using H&E and luxol fast blue. The specimens were examined using the Keyence BZ-X$\frac{700}{710}$ microscopy. Images were analyzed with the Keyence BZ-X Analyzer software. Histology was scored by an investigator blinded to the experimental group. Spinal cord sections were scored as follows: 0, no infiltration (<50 cells); 1, mild infiltration (50–100 cells); 2, moderate infiltration (100–150 cells); 3, severe infiltration (150–200 cells); and 4, massive infiltration (>200 cells)40.
## RNA isolation and quantitative PCR
Total RNA was extracted from BMDMs on day 8 and from Th0, Th1, Th2, Th17, and Treg cells on day 2 using RNeasy Plus Micro Kit (QIAGEN) according to the manufacturer’s instructions. For reverse transcription quantitative PCR (qPCR), isolated RNA was reverse transcribed, and gene expression was quantified using predesigned TaqMan Gene Expression Assays. The primers used were as follows: TaqMan probe Foxp3 (Mm00475162_m1), TaqMan probe Rorc (Mm01261022_m1), TaqMan probe Il17a (Mm00446973_m1), TaqMan probe Il17f (Mm00521423_m1), TaqMan probe Irg1 (Mm01224532_m1), TaqMan probe Nfe2l2 (Mm00477784_m1), TaqMan probe Hmox1 (Mm00516005_m1), TaqMan probe Tbp (Mm00446973_m1), and TaqMan probe Gusb (Mm1197698_m1). Gene expression was normalized to the reference gene Tbp and Gusb.
## RNA-seq analysis
Total RNA freshly isolated from Th0, Th17, and Treg cells with or without ITA treatment on day 2 was extracted using the RNeasy Plus Micro Kit, and subjected to RNA-seq analysis. Total RNA was quantified and qualified using NanoDrop and Qubit RNA Assay (Thermo Fisher Scientific) and RNA ScreenTape on TapeStation 4200 (Agilent Technologies, Palo Alto, CA, USA). Approximately 500 ng total RNA with RNA integrity number 7 or higher was used for poly-A mRNA enrichment (NEBNext Poly(A) mRNA Magnetic Isolation Module). The library was prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England Biolabs), according to the manufacturer’s protocol. The resultant libraries were then quantified and sequenced by Illumina NovaSeq according to the manufacturer’s instructions (Illumina, San Diego, CA, USA) with a 150 bp paired-end configuration. The library preparation and sequencing were performed by GENEWIZ. The raw sequencing reads with low quality and adapter sequences were removed using Cutadapt v 2.1. The trimmed reads were mapped to the mm10 (Ensembl 99) and quantified using STAR v 2.7.1a41. Data analysis was performed using R platform v 3.6.1. DEGs were identified based on differences in expression levels (log2 fold-change > 0.5 and adjusted p value < 0.01) between samples after removing genes with zero read count using DESeq2 v 1.24.042. The Benjamini-Hochberg method was used to adjust the p value for multiple hypothesis testing. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of DEGs was performed using Metascape v 3.5 web-based platform43.
## Extracellular flux analysis
The Seahorse XFp Extracellular Flux Analyzer (Agilent Technologies) was used to measure ECAR and OCR36. Briefly, naive CD4+ T cells isolated from WT mice were cultured under Th17- and Treg-polarizing conditions, with or without ITA. After 2 days, 1.5 × 105 cells were plated in an XFp microplate coated with cell-tak (Corning). The Glycolytic Rate Assay (Agilent Technologies) was performed to measure ECAR, followed by sequential injections of rotenone/antimycin A (Rot/AA: 0.5 μM) and 2-deoxyglucose (50 mM). The Cell Mito Stress test (Agilent Technologies) was performed to measure OCR, followed by sequential injections of Oligomycin (1.5 μM), FCCP (0.5 μM), and Rot/AA (0.5 μM). The obtained data were analyzed by Seahorse Wave v 2.6.0 (Agilent technologies).
## Profiling of intracellular metabolites
Intracellular metabolites were extracted from 14 × 106 cells for each sample. Cells were collected by centrifugation (800 × g and 4 °C for 2 min) and washed with 10 mL $5\%$ mannitol solution, then treated with 800 μL methanol and vortexed for 30 s to inactivate enzymes. Next, the cell extract was treated with 550 μL Milli-Q water containing internal standards (Human Metabolome Technologies) and left to rest for another 30 s. The extract was obtained and centrifuged at 2300 × g and 4 °C for 5 min. Then, 800 μL upper aqueous layer was centrifugally filtered through a Milli-pore 5-kDa cutoff filter at 9100 × g and 4 °C for 120 min to remove proteins. The filtrate was centrifugally concentrated and resuspended in 50 μL Milli-Q water for CE-TOF MS analysis44. Samples were measured and analyzed by Human Metabolome Technologies.
## Enzyme activity assay
Enzymes, including MAT and IDH$\frac{1}{2}$, were purified from Th17- and Treg-polarizing T cells with or without ITA on day 2 and the activities were assessed by an enzyme activity assay kit (all from Biovision) according to the manufacturer’s instructions.
To evaluate the effect of ITA on IDH1 and 2 activities, ten-dose IDH profiling was performed in Reaction Biology Corporation. IDH1 or 2 wild-type enzymes and NADP+ (for IDH1; 50 μM NADP+, for IDH2; 25 μM NADP+) in the reaction buffer (50 mM KH2PO4, pH 7.5, 10 mM MgCl2, $2.5\%$ glycerol, 150 mM NaCl, $0.05\%$ BSA, 2 mM β-ME, $0.003\%$ Brij35) were plated into wells of the reaction plate. ITA in PBS (pH was adjusted to 7.4 with 5 N NaOH) or control compounds were delivered into the enzyme mixture and pre-incubated for 60 min at room temperature. To initiate the enzyme-catalyzed reaction, a substrate mixture (for IDH1, 65 μM Isocitrate; for IDH2, 150 μM Isocitrate) was added and incubated for 60 min at room temperature. A detection mixture was added before measuring the IDH activity by EnVision (Ex/Em = $\frac{535}{590}$ nm). An ITA 100 mM stock solution was prepared in PBS and adjusted to pH 7.4 with 5 N NaOH at room temperature.
## ATAC-seq analysis
Cells were harvested and frozen in culture media containing FBS and $5\%$ DMSO. Cryopreserved cells were sent to Active Motif to perform the ATAC-seq assay. The cells were then thawed in a 37 °C water bath, pelleted, washed with cold PBS, and tagmented as previously described45, with some modifications following a previous study46. Briefly, cell pellets were resuspended in lysis buffer, pelleted, and tagmented using the enzyme and buffer provided in the Nextera Library Prep Kit (Illumina). Tagmented DNA was then purified using the MinElute PCR purification kit (Qiagen), amplified with 10 cycles of PCR, and purified using Agencourt AMPure SPRI beads (Beckman Coulter). The resulting material was quantified using the KAPA Library Quantification Kit for Illumina platforms (KAPA Biosystems) and sequenced with PE42 sequencing on the NextSeq 500 sequencer (Illumina).
Reads were aligned using the BWA algorithm (mem mode; default settings). Duplicate reads were removed, the reads mapping as matched pairs and the uniquely mapped reads (mapping quality ≥ 1) were used for further analysis. Alignments were extended in silico at their 3′-ends to a length of 200 bp and assigned to 32-nt bins along the genome. The resulting histograms (genomic “signal maps”) were stored in bigWig files. Peaks were identified using the MACS 2.1.0 algorithm at a cutoff of p value 1e-7, without a control file, and with the ‘–nomodel’ option. Peaks that were on the ENCODE blacklist of known false peaks were removed. Signal maps and peak locations were used as input data to Active Motifs proprietary analysis program, which creates Excel tables containing detailed information on sample comparison, peak metrics, peak locations, and gene annotations. Chromatin accessibility was accessed based on the output table. Differently accessible (DA) peaks were selected with the peak metrics of |shrunkenLog2FC| > 0.3 and adjusted p value < 0.1. DA genes were defined as the genes with DA peaks. The DA genes were classified into “open” or “close” when all of the DA peaks on the annotated genes were in the same direction (i.e., All DA peaks increased on the gene “open”).
## ChIP analysis
Freshly isolated naive CD4+ T cells from wild-type mice were cultured under Th17- polarizing conditions, with or without ITA treatment. After 2 days, 4.5 × 106 cells were harvested, and ChIP was performed using the ChIP system (Thermo Fisher Scientific) according to the manufacturer’s instructions. The soluble chromatin supernatant was immunoprecipitated with an anti-RORγt antibody (40 μg mL−1, clone AFKJS-9, eBioscience, 14-6988-82) and IgG control (40 μg mL−1, clone eBR2a, eBioscience, 14-4321-82). Immunoprecipitated DNA and input DNA were measured by qPCR using a SYBR green reagent. Data were expressed as the percent input for each ChIP fraction. The primers used for amplification of the promoter of Il17a containing the RORγt-binding site were 5′ CAGCTCCCAAGAAGTCATGC 3′ (forward) and 5′ GCAACATCTGTCTCGAAGGTAG 3′ (reverse)47.
## Statistics and reproducibility
Except for RNA-seq and ATAC-seq experiments, statistical analyses were performed with Prism software version 8.0 (GraphPad Prism software) using a two-tailed unpaired Student’s t-test for pairwise comparison of variables and one-way analysis of variance (ANOVA) with Bonferroni post hoc test for multiple comparisons of variables. For the EAE model, clinical scores and body weight changes of each group were compared using two-way ANOVA. A P value of <0.05 was considered statistically significant. Error bars present the mean ± standard error of the mean (s.e.m.). No statistical methods were used to predetermine sample sizes. Sample sizes were based on pilot experiments conducted in the same laboratory and comparable to similar studies in the field. No data were excluded from the analyses. The evaluation and scoring of histopathology of HE-stained tissue sections was performed in a blinded fashion. In the other experiments, no blinding was used during allocation of experimental groups, because all data collection and analysis is quantitative and not qualitative in nature. To avoid introducing bias, samples were measured in a standardized way.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Peer Review File Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36594-x.
## Source data
Source Data
## Peer review information
Nature Communications thanks Dan Littman, Evanna Mills and Luke O’Neill for their contribution to the peer review of this work. Peer reviewer reports are available.
## References
1. Tsokos GC. **Autoimmunity and organ damage in systemic lupus erythematosus**. *Nat. Immunol.* (2020) **21** 605-614. DOI: 10.1038/s41590-020-0677-6
2. Yasuda S. **Emerging targets for the treatment of lupus erythematosus: there is no royal road to treating lupus**. *Mod. Rheumatol.* (2019) **29** 60-69. DOI: 10.1080/14397595.2018.1493909
3. Durcan L, O’Dwyer T, Petri M. **Management strategies and future directions for systemic lupus erythematosus in adults**. *Lancet* (2019) **393** 2332-2343. DOI: 10.1016/S0140-6736(19)30237-5
4. Geltink RIK, Kyle RL, Pearce EL. **Unraveling the complex interplay between T cell metabolism and function**. *Annu Rev. Immunol.* (2018) **36** 461-488. DOI: 10.1146/annurev-immunol-042617-053019
5. Zhou L. **TGF-beta-induced Foxp3 inhibits T(H)17 cell differentiation by antagonizing RORgammat function**. *Nature* (2008) **453** 236-240. DOI: 10.1038/nature06878
6. Zhu J, Yamane H, Paul WE. **Differentiation of effector CD4 T cell populations**. *Annu. Rev. Immunol.* (2010) **28** 445-489. DOI: 10.1146/annurev-immunol-030409-101212
7. Wagner A. **Metabolic modeling of single Th17 cells reveals regulators of autoimmunity**. *Cell* (2021) **184** 4168-4185.e21. DOI: 10.1016/j.cell.2021.05.045
8. Dendrou CA, Fugger L, Friese MA. **Immunopathology of multiple sclerosis**. *Nat. Rev. Immunol.* (2015) **15** 545-558. DOI: 10.1038/nri3871
9. Sharabi A, Tsokos GC. **T cell metabolism: new insights in systemic lupus erythematosus pathogenesis and therapy**. *Nat. Rev. Rheumatol.* (2020) **16** 100-112. DOI: 10.1038/s41584-019-0356-x
10. Samson M. **Brief report: inhibition of interleukin-6 function corrects Th17/Treg cell imbalance in patients with rheumatoid arthritis**. *Arthritis Rheum.* (2012) **64** 2499-2503. DOI: 10.1002/art.34477
11. Sakaguchi S. **Taking regulatory T cells into medicine**. *J. Exp. Med.* (2021) **218** e20210831. DOI: 10.1084/jem.20210831
12. Kono M. **Pyruvate kinase M2 is requisite for Th1 and Th17 differentiation**. *JCI Insight* (2019) **4** e127395. DOI: 10.1172/jci.insight.127395
13. Roy DG. **Methionine metabolism shapes T helper cell responses through regulation of epigenetic reprogramming**. *Cell Metab.* (2020) **31** 250-266.e259. DOI: 10.1016/j.cmet.2020.01.006
14. Xu T. **Metabolic control of T(H)17 and induced T(reg) cell balance by an epigenetic mechanism**. *Nature* (2017) **548** 228-233. DOI: 10.1038/nature23475
15. O’Neill LAJ, Artyomov MN. **Itaconate: the poster child of metabolic reprogramming in macrophage function**. *Nat. Rev. Immunol.* (2019) **19** 273-281. DOI: 10.1038/s41577-019-0128-5
16. Swain A. **Comparative evaluation of itaconate and its derivatives reveals divergent inflammasome and type I interferon regulation in macrophages**. *Nat. Metab.* (2020) **2** 594-602. DOI: 10.1038/s42255-020-0210-0
17. Hooftman A, O’Neill LAJ. **The immunomodulatory potential of the metabolite itaconate**. *Trends Immunol.* (2019) **40** 687-698. DOI: 10.1016/j.it.2019.05.007
18. Mills EL. **Itaconate is an anti-inflammatory metabolite that activates Nrf2 via alkylation of KEAP1**. *Nature* (2018) **556** 113-117. DOI: 10.1038/nature25986
19. Bambouskova M. **Electrophilic properties of itaconate and derivatives regulate the IκBζ-ATF3 inflammatory axis**. *Nature* (2018) **556** 501-504. DOI: 10.1038/s41586-018-0052-z
20. Qin W. **S-glycosylation-based cysteine profiling reveals regulation of glycolysis by itaconate**. *Nat. Chem. Biol.* (2019) **15** 983-991. DOI: 10.1038/s41589-019-0323-5
21. McFadden BA, Purohit S. **Itaconate, an isocitrate lyase-directed inhibitor in Pseudomonas indigofera**. *J. Bacteriol.* (1977) **131** 136-144. DOI: 10.1128/jb.131.1.136-144.1977
22. Olagnier D. **SARS-CoV2-mediated suppression of NRF2-signaling reveals potent antiviral and anti-inflammatory activity of 4-octyl-itaconate and dimethyl fumarate**. *Nat. Commun.* (2020) **11** 4938. DOI: 10.1038/s41467-020-18764-3
23. Albuquerque EX, Pereira EF, Alkondon M, Rogers SW. **Mammalian nicotinic acetylcholine receptors: from structure to function**. *Physiol. Rev.* (2009) **89** 73-120. DOI: 10.1152/physrev.00015.2008
24. McGinley AM. **Interleukin-17A serves a priming role in autoimmunity by recruiting IL-1β-producing myeloid cells that promote pathogenic T cells**. *Immunity* (2020) **52** 342-356.e346. DOI: 10.1016/j.immuni.2020.01.002
25. Ciofani M. **A validated regulatory network for Th17 cell specification**. *Cell* (2012) **151** 289-303. DOI: 10.1016/j.cell.2012.09.016
26. Macintyre AN. **The glucose transporter Glut1 is selectively essential for CD4 T cell activation and effector function**. *Cell Metab.* (2014) **20** 61-72. DOI: 10.1016/j.cmet.2014.05.004
27. Dang EV. **Control of T(H)17/T(reg) balance by hypoxia-inducible factor 1**. *Cell* (2011) **146** 772-784. DOI: 10.1016/j.cell.2011.07.033
28. Lampropoulou V. **Itaconate links inhibition of succinate dehydrogenase with macrophage metabolic remodeling and regulation of inflammation**. *Cell Metab.* (2016) **24** 158-166. DOI: 10.1016/j.cmet.2016.06.004
29. Schvartzman JM, Thompson CB, Finley LWS. **Metabolic regulation of chromatin modifications and gene expression**. *J. Cell Biol.* (2018) **217** 2247-2259. DOI: 10.1083/jcb.201803061
30. Chen LL. **Itaconate inhibits TET DNA dioxygenases to dampen inflammatory responses**. *Nat. Cell Biol.* (2022) **24** 353-363. DOI: 10.1038/s41556-022-00853-8
31. Piantadosi CA. **Heme oxygenase-1 couples activation of mitochondrial biogenesis to anti-inflammatory cytokine expression**. *J. Biol. Chem.* (2011) **286** 16374-16385. DOI: 10.1074/jbc.M110.207738
32. Mitsuishi Y. **Nrf2 redirects glucose and glutamine into anabolic pathways in metabolic reprogramming**. *Cancer Cell* (2012) **22** 66-79. DOI: 10.1016/j.ccr.2012.05.016
33. Kouzarides T. **Chromatin modifications and their function**. *Cell* (2007) **128** 693-705. DOI: 10.1016/j.cell.2007.02.005
34. Kurebayashi Y, Nagai S, Ikejiri A, Koyasu S. **Recent advances in understanding the molecular mechanisms of the development and function of Th17 cells**. *Genes Cells* (2013) **18** 247-265. DOI: 10.1111/gtc.12039
35. Lu L, Barbi J, Pan F. **The regulation of immune tolerance by FOXP3**. *Nat. Rev. Immunol.* (2017) **17** 703-717. DOI: 10.1038/nri.2017.75
36. Kono M. **Pyruvate dehydrogenase phosphatase catalytic subunit 2 limits Th17 differentiation**. *Proc. Natl Acad. Sci. USA* (2018) **115** 9288-9293. DOI: 10.1073/pnas.1805717115
37. Lin J, Ren J, Gao DS, Dai Y, Yu L. **The emerging application of itaconate: promising molecular targets and therapeutic opportunities**. *Front. Chem.* (2021) **9** 669308. DOI: 10.3389/fchem.2021.669308
38. Wang X. **Transcription of Il17 and Il17f is controlled by conserved noncoding sequence 2**. *Immunity* (2012) **36** 23-31. DOI: 10.1016/j.immuni.2011.10.019
39. Tripathi SK, Lahesmaa R. **Transcriptional and epigenetic regulation of T-helper lineage specification**. *Immunol. Rev.* (2014) **261** 62-83. DOI: 10.1111/imr.12204
40. Yoshida N. **ICER is requisite for Th17 differentiation**. *Nat. Commun.* (2016) **7** 12993. DOI: 10.1038/ncomms12993
41. Dobin A. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013) **29** 15-21. DOI: 10.1093/bioinformatics/bts635
42. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014) **15** 550. DOI: 10.1186/s13059-014-0550-8
43. Zhou Y. **Metascape provides a biologist-oriented resource for the analysis of systems-level datasets**. *Nat. Commun.* (2019) **10** 1523. DOI: 10.1038/s41467-019-09234-6
44. Kono M, Yoshida N, Maeda K, Tsokos GC. **Transcriptional factor ICER promotes glutaminolysis and the generation of Th17 cells**. *Proc. Natl Acad. Sci. USA* (2018) **115** 2478-2483. DOI: 10.1073/pnas.1714717115
45. Buenrostro JD, Giresi PG, Zaba LC, Chang HY, Greenleaf WJ. **Transposition of native chromatin for fast and sensitive epigenomic profiling of open chromatin, DNA-binding proteins and nucleosome position**. *Nat. Methods* (2013) **10** 1213-1218. DOI: 10.1038/nmeth.2688
46. Corces MR. **An improved ATAC-seq protocol reduces background and enables interrogation of frozen tissues**. *Nat. Methods* (2017) **14** 959-962. DOI: 10.1038/nmeth.4396
47. Endo Y. **Obesity drives Th17 cell differentiation by inducing the lipid metabolic kinase, ACC1**. *Cell Rep.* (2015) **12** 1042-1055. DOI: 10.1016/j.celrep.2015.07.014
|
---
title: KAT8 acetylation-controlled lipolysis affects the invasive and migratory potential
of colorectal cancer cells
authors:
- Bingquan Qiu
- Shen Li
- Meiting Li
- Shuo Wang
- Guanqun Mu
- Keyu Chen
- Meng Wang
- Wei-guo Zhu
- Weibin Wang
- Jiadong Wang
- Ziyu Li
- Jichun Yang
- Yang Yang
journal: Cell Death & Disease
year: 2023
pmcid: PMC9970984
doi: 10.1038/s41419-023-05582-w
license: CC BY 4.0
---
# KAT8 acetylation-controlled lipolysis affects the invasive and migratory potential of colorectal cancer cells
## Abstract
Epigenetic mechanisms involved in gene expression play an essential role in various cellular processes, including lipid metabolism. Lysine acetyltransferase 8 (KAT8), a histone acetyltransferase, has been reported to mediate de novo lipogenesis by acetylating fatty acid synthase. However, the effect of KAT8 on lipolysis is unclear. Here, we report a novel mechanism of KAT8 on lipolysis involving in its acetylation by general control non-repressed protein 5 (GCN5) and its deacetylation by Sirtuin 6 (SIRT6). KAT8 acetylation at K$\frac{168}{175}$ residues attenuates the binding activity of KAT8 and inhibits the recruitment of RNA pol II to the promoter region of the lipolysis-related genes adipose triglyceride lipase (ATGL) and hormone-sensitive lipase (HSL), subsequently down-regulating lipolysis to affect the invasive and migratory potential of colorectal cancer cells. Our findings uncover a novel mechanism that KAT8 acetylation-controlled lipolysis affects invasive and migratory potential in colorectal cancer cells.
## Introduction
Emerging evidence shows the wide interplays between epigenetic regulation and cell metabolism [1, 2]. Colorectal cancer (CRC) is one of the most common malignant diseases with a high mortality rate characteristic of metabolism disorder, including lipid metabolism [3–6]. Recently, some studies revealed that histone deacetylases played key roles in lipid metabolism of colorectal cancer [7, 8]. For example, bouchardatine was reported to suppress rectal cancer by disrupting its metabolic pathways via activating the SIRT1-PGC-1α-UCP2 axis [8]. In addition, our previous data showed that SIRT6 is phosphorylated by PKCζ at threonine 294 residue, which promotes SIRT6 enrichment on the chromatin to regulate the expression of fatty acid β-oxidation-related genes [9]. Also, we found that p53 physically interacts with histone deacetylase SIRT6 in vitro and in vivo, and cooperates with SIRT6 to regulate cardiolipin de novo biosynthesis [10]. However, the function of histone acetyltransferase in lipid metabolism of colorectal cancer is seldom mentioned and needs to further explore.
Lysine acetyltransferase 8 (KAT8), also called MOF (Males Absent on the First), is a histone acetyltransferase that belongs to MYST family. KAT8 is a chromatin regulatory protein that mainly functions by acetylating H4K16 [11–13]. KAT8 participates in cell proliferation, tumor invasion, DNA repair, autophagy and transcription regulation [13–18]. The effects of post-translational modification (PTM) of KAT8 on cellular processes and tumor progression have received increasing attention [19–22]. MOF-T392 phosphorylation mediated by ataxia telangiectasia-mutated (ATM) was reported to regulate 53BP1-mediated double-strand break repair pathway choice [19]. Mass spectrometry was used to map the lysines in KAT8, which are ubiquitylated by MSL2 in vitro and identified ubiquitylation sites of KAT8 in male and female cells of Drosophila melanogaster in vivo [21]. Also, KAT8 autoacetylation was shown in few studies [11, 13, 23]. For example, Lu Lu and his colleagues reported that KAT8 auto-acetylated at K274 residue in vitro and in vivo, and SIRT1 negatively modulates this process through regulating KAT8 recruitment to the chromatin [11]. However, little is known about the effects of KAT8 or KAT8 acetylation besides autoacetylation on lipolysis in colorectal cancer cells.
Lipolysis is a catabolic branch of the fatty acids (FA) cycle that provides FAs to meet metabolic need and removes them when in excess. Lipolysis hydrolyses the ester bonds of the triglyceride (TG) and plays a critical role in cell metabolism homoeostasis. Neutral hydrolysis of TGs to FAs and glycerol requires three consecutive steps and at least three different enzymes are involve, including adipose triglyceride lipase (ATGL), hormone-sensitive lipase (HSL) and monoacylglycerol lipase (MGL) [24–26]. ATGL catalyzes the initial step of lipolysis, catabolizing TGs to diacylglycerols (DGs). HSL, the main DG lipase, is rate-limiting for catabolism in adipose and non-adipose tissue and coordinates with ATGL to affect both the first two steps of TG breaking down, which mainly hydrolyze DGs to generate monoglycerides (MGs) and free FAs [27–29]. MGL is selectively responsible for the irreversible hydrolysis of monoglyceride (MGs) deriving from both extracellular and intracellular TG hydrolysis [30]. Lipolysis produces free fatty acid and then uses for β-oxidation and ATP production, which involves in various diseases such as cancer, type 2 diabetes and fatty liver disease [31].
Our previous study has reported that histone deacetylase SIRT6 plays a key role to regulate Cardiolipin de novo biosynthesis and fatty acid β-oxidation [9, 10]. Here, we are interested in the effect of histone acetyltransferase KAT8 on lipid metabolism. We identified KAT8 as an acetylated protein and KAT8 acetylation can be dynamically regulated by deacetylase SIRT6 and acetyltransferase GCN5, which is independent of its auto-acetylation. The K$\frac{168}{175}$ acetylation of KAT8 attenuates the binding activity of KAT8 and inhibits the recruitment of RNA pol II to the promoter region of the lipolysis-related genes ATGL and HSL, subsequently downregulating their expression to mediate lipolysis. Taken together, our data reveals that KAT8 acetylation plays a critical role in lipolysis to further affect invasive and migratory potential of CRC cells, and it may serve as a potential target for colorectal cancer therapy.
## KAT8 regulates lipolysis accompanied by enhanced KAT8 acetylation after PA treatment in human CRC cells
KAT8 is essential for cell proliferation and plays a critical role in multiple physiological process [32, 33]. However, its role on lipolysis is still unclear. Several stimuli were used to detect their function on lipolysis in colon cancer HCT116 cells. As shown in Fig. 1A, only palmitic acid (PA) can effectively increase the release of glycerol showing that PA involves in lipolysis of colon cancer cells. Next, we explored the relationship between KAT8 and lipolysis of CRC cells during PA stimulation. The lipid accumulation significantly reduced in KAT8 siRNA cells compared with control cells after PA treatment (Fig. 1B, C). The glycerol release experiments also showed that KAT8 may play a role in lipolysis induced by PA stimulation in HCT116 cells (Fig. 1D). In the meanwhile, opposite results were shown in PA-treated KAT8 overexpression cells, indicating that KAT8 overexpression inhibits lipolysis by PA stimulation (Fig. 1E–G). Also, the key regulation effect of KAT8 in lipolysis was found in colon cancer RKO and sw480 cells (Fig. 1H, I).Fig. 1KAT8 regulates lipolysis accompanied by enhanced KAT8 acetylation after PA treatment in human CRC cells. A HCT116 cells were treated with PA (0.2 mM) for 24 h, glucose starvation (GS) for 18 h, etoposide (40 μM) for 8 h, HU (2 mM) for 2 h and CPT (1 μM) for 1 h. Cells were lysed and subjected to quantify the amount of glycerol inside cells by glycerol colorimetric assay kit. B, E HCT116 cells were transfected with KAT8 siRNA (B) or KAT8 plasmid (E), and then treated with or without PA at 0.2 mM for 24 h. Oil red O staining was used to detect the accumulation of lipid droplets. C The siRNA efficiency of KAT8 was detected by real-time PCR. D, G HCT116 cells were treated as outlined above, cells were lysed and subjected to quantify the amount of glycerol inside cells by glycerol colorimetric assay kit. F The transfected efficiency of KAT8 was detected by Western blotting. H, I SW480 (H) or RKO (I) cells were treated as outlined above, cells were lysed and subjected to quantify the amount of glycerol inside cells by glycerol colorimetric assay kit. J, K HCT116 cells were treated with 0.2 mM PA for 0, 18 h and 24 h (J) or PA (0.1 mM and 0.2 mM) for 24 h (K). The mRNA expression of KAT8 was analyzed by real-time PCR. mRNA levels of the control sample were set as 1, and relative mRNA levels of the experimental samples were normalized to this control. L, M HCT116 cells were stimulated under the same condition above. The total protein expression of KAT8 was detected by Western blotting. β-actin was used as a loading control. N, O HCT116 (N), SW480 and RKO (O) cells were treated with or without 0.2 mM PA for 24 h, cell lysates were then extracted for a Co-IP assay to detect endogenous acetylation levels of KAT8 in these cells. The bar (−) represents the means ($$n = 3$$).
Given that KAT8 involves in the process of lipolysis, we next detected the mRNA and protein levels of KAT8 after PA treatment. No significant difference was found in both KAT8 mRNA and protein levels after PA treatment (Fig. 1J–M). Remarkably, the endogenous acetylation of KAT8 increased sharply after PA treatment in these CRC cells, suggesting that KAT8 acetylation might be play a critical role in regulating lipolysis in CRC cells (Fig. 1N, O). Meanwhile, the familiar phenomena were shown in other cancer cells, such as hepatocellular carcinoma HepG2 cells, gastric carcinoma MGC-803 cells and pancreatic cancer PANC1 cells (Supplementary Fig. S1A, 1C–E). However, KAT8 acetylation was shown not significantly increased in human embryonic lung diploid fibroblasts 2BS cells (Supplementary Fig. S1B). These data suggest that KAT8 acetylation involved in the process of lipolysis is universal in cancer cells.
## GCN5 is an acetyltransferase of KAT8
Since KAT8 acetylation might play a critical role in regulating lipolysis in CRC cells, we next aimed to identify the specific acetyltransferase by screening a series of acetyltransferases. We co-transfected KAT8 with different histone acetyltransferases (HATs) and found that only GCN5 dramatically increased the acetylation level of KAT8 (Fig. 2A, B). KAT8 acetylation was also dramatically decreased when the cells were treated with GCN5 siRNA or GCN5 enzymatic inhibitor MB-3 (Fig. 2C). In vitro acetylation assay further confirmed that GCN5 is the acetyltransferase of KAT8 (Fig. 2D).Fig. 2GCN5 is an acetyltransferase of KAT8.A, B KAT8 was co-transfected with several critical acetyltransferases in HCT116 cells and then Co-IP was performed to detect the level of KAT8 acetylation. C HCT116 cells were transfected with GCN5 siRNA or treated with GCN5 inhibitor MB-3, and then Co-IP was performed to detect the level of KAT8 acetylation. D Flag-GCN5 was transfected into HCT116 cells and then KAT8 and Flag-GCN5 proteins were purified to perform in vitro acetylation assay. The level of KAT8 acetylation was detected by Western blotting. E HCT116 cells were co-transfected with Flag-GCN5 and myc-KAT8 and then proteins were extracted for Co-IP to detect the exogenous interaction between KAT8 and GCN5. F HCT116 cells were transfected with Flag-GCN5 and then proteins were extracted for Co-IP to detect the semi-exogenous interaction between KAT8 and GCN5. G Protein of HCT116 cells was extracted for Co-IP to detect the endogenous interaction between KAT8 and GCN5. H His-GCN5, GST-KAT8 and GST were expressed, and purified in bacteria. GST-pull down assays were performed to show the direct interaction between GCN5 and KAT8 in vitro. I Schematic of plasmids encoding KAT8. J GST-KAT8 FL or fragments were incubated with His-GCN5, and Western blotting or Coomassie staining was performed to detect the direct binding of KAT8 and GCN5 in vitro. K Schematic of plasmids encoding GCN5. L His-GCN5 FL or fragments were incubated with GST-KAT8, and Western blotting or Coomassie staining was performed to detect the direct binding of GCN5 and KAT8 in vitro. # indicates the specific bands.
Next, we detected whether there is a molecular link between GCN5 and KAT8. Exogenous, semi-exogenous and endogenous co-immunoprecipitation showed the interaction of GCN5 with KAT8 (Fig. 2E–G). We then performed a GST pull-down assay to investigate whether KAT8 directly interacts with GCN5. We found that GCN5 directly interacts with GST-KAT8 but not GST alone, showing that GCN5 can interact with KAT8 in vitro (Fig. 2H). In the process of mapping the regions of KAT8 involves in GCN5 binding, we found that the N-terminus domain of KAT8 is responsible for the interaction with GCN5 (Fig. 2I, J). Also, N-terminus fragment of GCN5 was confirmed to be responsible for this interaction (Fig. 2K, L). Collectively, these experiments indicate that GCN5 can interact with KAT8 and acetylate KAT8 in vivo and in vitro.
## SIRT6 is a deacetylase of KAT8 and is responsible for KAT8 acetylation
We next sought to identify the deacetylase of KAT8 acetylation. HDACI, II family inhibitor TSA and the Sirtuin family inhibitor nicotinamide were used and we found that nicotinamide significantly increased the acetylation level of KAT8, suggesting that Sirtuins might be the major deacetylase of KAT8 (Fig. 3A). Subcellular location of KAT8 showed that KAT8 is mainly distributed in nucleus (data not shown). Thus, we co-transfected myc-tagged KAT8 together with Flag-tagged SIRT$\frac{1}{6}$/7 into HCT116 cells. As shown in Fig. 3B, the deacetylase activity of SIRT1 or SIRT6 on KAT8 acetylation was stronger than SIRT7 in transfected HCT116 cells. Interestingly, KAT8 acetylation was only recovered in SIRT6 transfected HCT116 cells after PA treatment, implying that SIRT6 may have certain relationship with KAT8 acetylation after PA treatment (Fig. 3C). When SIRT6 siRNA or Flag-SIRT6 plasmid was transfected into HCT116 cells, KAT8 acetylation was changed sharply compared with control group (Fig. 3D, E). Also, the data of in vitro deacetylation assay suggested that SIRT6 significantly deacetylates KAT8 in vitro (Fig. 3F). These data indicated that SIRT6 is the deacetylase of KAT8.Fig. 3SIRT6 is a deacetylase of KAT8 and is responsible for KAT8 acetylation. A HCT116 cells were treated with 1 μM TSA or 5 mM NAM for 8 h and then proteins were extracted for Co-IP assay to detect the level of KAT8 acetylation. B HCT116 cells were co-transfected with myc-KAT8, Flag-SIRT1, Flag-SIRT6 or Flag-SIRT7, and then proteins were extracted for Co-IP to detect the level of KAT8 acetylation. C HCT116 cells were co-transfected with the plasmids above, and then treated with PA at 0.2 mM for 24 h. Proteins were extracted for Co-IP to detect the level of KAT8 acetylation. D, E HCT116 cells were transfected with Flag-SIRT6 (D) or SIRT6 siRNA (E), and proteins were extracted for Co-IP to detect the level of KAT8 acetylation. F myc-KAT8 and Flag-SIRT6 were co-transfected into HCT116 cells, and then KAT8 and SIRT6 proteins were purified to perform in vitro deacetylation assay. G, H HCT116 cells were co-transfected with GFP-SIRT6 (G) or Flag-GCN5 (H) and myc-KAT8, and then treated with PA at 0.2 mM for 24 h. Cell lysates were extracted for Co-IP assay to detect the exogenous interaction between SIRT6 (G) /GCN5 (H) and KAT8. I HCT116 was transfected with GCN5 siRNA or non-specific RNA as control, protein was then extracted for Co-IP to detect the endogenous interaction between KAT8 and SIRT6. J HCT116 SIRT6-KO cells were extracted for Co-IP to detect the endogenous interaction between KAT8 and GCN5.
To clarify which is the key mediator of KAT8 acetylation, we firstly monitored the interaction between GCN5/SIRT6 and KAT8 after PA treatment. The interaction between SIRT6 and KAT8 was sharply reduced accompanied by the increased interaction between GCN5 and KAT8 after PA treatment, showing that KAT8 acetylation might be dynamically regulated by the balance of GCN5 and SIRT6 (Fig. 3G, H). Remarkably, the interaction between SIRT6 and KAT8 did not change in GCN5-specific siRNA-treated cells (Fig. 3I). On the contrary, the interaction between GCN5 and KAT8 increased significantly in SIRT6 KO cells showing that SIRT6 is the key mediator on KAT8 acetylation after PA treatment (Fig. 3J).
## KAT8 interacts with SIRT6 in vivo and in vitro
Having established that SIRT6 is a major deacetylase of KAT8, we aimed to clarify the molecular link between SIRT6 and KAT8. Exogenous, semi-exogenous and endogenous co-immunoprecipitation presented the interaction of SIRT6 with KAT8 (Fig. 4A–E). GST pull-down assay was performed to investigate whether the interaction between KAT8 and SIRT6 is direct. We found that KAT8 core domain is responsible for this interaction (Fig. 4F, G). SIRT6 CD domain was also confirmed to be responsible for the interaction (Fig. 4H, I). Together, these results indicate that KAT8 directly interacts with SIRT6 in vivo and in vitro. Fig. 4KAT8 interacts with SIRT6 in vivo and in vitro. A, B HCT116 cells were transfected with Flag-SIRT6 plasmids (A) or myc-KAT8 plasmids (B), and then proteins were extracted for Co-IP to detect the semi-exogenous interaction between KAT8 and SIRT6. C, D HCT116 cells were transfected with Flag-SIRT6 and/or myc-KAT8 plasmids, and then proteins were extracted for Co-IP to detect the exogenous interaction between KAT8 and SIRT6. E Protein of HCT116 cells was extracted for Co-IP to detect the endogenous interaction between KAT8 and SIRT6. F Schematic of plasmids encoding KAT8. G GST-KAT8 FL or fragments were incubated with His-SIRT6, and Western blotting or Coomassie staining was performed to detect the direct binding of KAT8 and SIRT6 in vitro. H Schematic of plasmids encoding SIRT6. I GST-SIRT6 FL or fragments were incubated with His-KAT8, and Western blotting or Coomassie staining was performed to detect the interaction. # indicates the specific bands.
## Lysine 168 and lysine 175 are the major acetylation sites of KAT8 and mediate lipolysis of CRC cells
Next, we identified the major acetylation sites of KAT8 by mass spectrometry (MS). The MS data indicated that KAT8 is acetylated at lysine 168 and lysine 175 residues after PA treatment (Fig. 5A). These two sites are highly conserved from EQUUS to humans (Fig. 5B). To further confirm the acetylation sites of KAT8, lysine to alanine mutant KAT8-K168R, KAT8-K175R or KAT8-2KR plasmids were generated and transfected into HCT116 cells. As shown in Fig. 5C, KAT8 acetylation was reduced significantly after PA stimulation in both mutated KAT8 transfected cells. Also, the results of co-transfected with Flag SIRT6 plasmids showed that the reduction of KAT8 acetylation in KAT8 WT transfected cells was completely recovered in KAT8 KR transfected cells (Fig. 5D). At the same time, the glycerol releases in KAT8-2KR overexpression HCT116, RKO and SW480 cells were significantly increased comparing with KAT8 WT transfected cells (Fig. 5E–G). These data demonstrate that K168 and K175 are two major functional sites of KAT8 acetylation and involve in mediating lipolysis of CRC cells. Fig. 5Lysine 168 and lysine 175 are the major acetylation sites of KAT8 and mediate lipolysis of CRC cells. A HCT116 cells were transfected with myc-KAT8 and then treated with PA at 0.2 mM for 24 h. Myc-KAT8 was purified and subsequently separated by SDS-PAGE and stained with CBB. The KAT8 band was analyzed by mass spectrometry. B Alignment of MS-characterized putative KAT8 acetylation residues among different species. C HCT116 cells were transfected with myc-KAT8 WT, K168R or K175R mutant plasmids for a Co-IP assay to detect the level of KAT8 acetylation. D HCT116 cells were transfected with the plasmids above combined with Flag-SIRT6 for a Co-IP assay to detect the level of KAT8 acetylation. E-G HCT116 (E), RKO (F) and SW480 (G) cells were transfected with myc-vector, myc-KAT8 WT or myc-KAT8 2KR mutant plasmids, and then cells were lysed to quantify the amount of glycerol inside cells by glycerol colorimetric assay kit. The bar (−) represents the means ($$n = 3$$).
## KAT8 regulates the expression of lipolysis-related genes via its acetylation
Having known that KAT8 acetylation has the ability to regulate lipolysis of CRC cells, we next explored the mechanism under this process. ATGL and HSL are two key enzymes responsible for lipolysis. We firstly detected the expression of ATGL and HSL genes and found that both genes reduced after PA stimulation (Fig. 6A). Also, the key role of SIRT6 or KAT8 was shown in regulating the expression of ATGL and HSL genes (Fig. 6B–D).Fig. 6KAT8 regulates the expression of lipolysis-related genes via its acetylation. A HCT116 cells were treated with PA at 0.2 mM for 24 h, and then the expression of ATGL and HSL were analyzed by real-time PCR. B HCT116 cells were transfected with myc-KAT8 to detect the expression of ATGL and HSL by real-time PCR. C, D HCT116 cells were transfected with Flag-SIRT6 (C) or SIRT6 siRNA (D) for 48 h, and then the expression of ATGL and HSL were analyzed by real-time qPCR. E HCT116 cells were transfected with pCMV-myc-vector or myc-KAT8 and then treated with PA to determine the changes in H4K16 acetylation patterns by Western blotting. H4 is shown as a loading control. F, G HCT116 cells were treated with or without PA at 0.2 mM for 24 h. ChIP assay was performed to detect the enrichment of H4K16ac (F) and KAT8 (G) at the gene promoter regions of ATGL and HSL. H-J HCT116 (H), RKO (I) and SW480 (J) cells were transfected with myc-KAT8 WT or myc-KAT8 2KR mutant plasmid for 48 h. The expression of ATGL and HSL was analyzed by real-time PCR. K-N HCT116 (K, L), RKO (M) and SW480 (N) cells were transfected with the Flag-vector, Flag-KAT8-WT, Flag-KAT8-2KR plasmids, ChIP assay was performed to detect the enrichment of KAT8 on the gene promoter regions of ATGL and HSL. O Protein of HCT116 cells was extracted for Co-IP to detect the endogenous interaction between KAT8 and RNA pol II. P HCT116 cells were transfected with myc-KAT8 WT or myc-KAT8 2KR mutant plasmid for Co-IP to detect the interaction between KAT8 and RNA pol II. Q HCT116 cells were transfected with the plasmids mentioned above, ChIP assay was performed to detect the enrichment of RNA pol II at the gene promoter regions of ATGL and HSL. mRNA levels of the control sample were set as 1, and relative mRNA levels of the other samples were normalized to this control. The bar (−) represents the means ($$n = 3$$).
Next, we try to elucidate the mechanism how KAT8 regulates the expression of lipolysis-related genes after PA treatment. Histone H4K16ac is a specific acetylation target of KAT8, so we firstly detected the expression of H4K16ac after PA treatment. No significant difference was detected in global H4K16ac level or in the enrichment of H4K16ac to the promoter of ATGL and HSL genes, suggesting that KAT8 may not decrease the expression of ATGL and HSL genes through H4K16ac (Fig. 6E, F). Interestingly, the binding affinity of KAT8 to the promoter of ATGL and HSL genes was found reduced sharply after PA treatment, demonstrating that KAT8 regulates the expression of both genes by changing its recruitment to the gene promoters (Fig. 6G). We then confirmed the effect of KAT8 acetylation on gene expression of CRC cells. The expression of ATGL and HSL were reduced in wild-type KAT8-transfected cells and recovered in 2KR mutant KAT8-transfected cells (Fig. 6H–J). ChIP qPCR assay revealed that the binding activity of KAT8 to the promoters of both ATGL and HSL genes were significantly increased in the 2KR mutant KAT8-transfected CRC cells comparing with control group (Fig. 6K–N). These data indicate that KAT8 acetylation mediates the binding activity of KAT8 to the promoter of lipolysis-related genes to regulate their expression in CRC cells.
To further explore the mechanism of KAT8 acetylation on regulating ATGL and HSL gene expression, we screened the possible regulatory factors, and identified that KAT8 can interact with RNA polymerase II (RNAP II) (Fig. 6O). Remarkably, the interaction between KAT8 and RNAP II was increased in the 2KR mutant KAT8-transfected cells comparing with control cells (Fig. 6P). Also, we found that the reduction of binding activity of RNAP II on the promoters of ATGL and HSL in wild-type KAT8-transfected cells was recovered in 2KR mutant KAT8-transfected cells (Fig. 6Q). Collectively, these experiments indicate that KAT8 may serve as a co-transcription factor of RNAP II to regulate the expression of lipolysis-related genes and KAT8 acetylation is response for this process through mediating KAT8 binding activity to the promoters of these genes.
## KAT8 acetylation regulates migration and invasion of CRC cells
We further demonstrated the functional significance of KAT8 acetylation in CRC progression. The results of CCK-8 viability assay and colony formation assay showed that KAT8 acetylation is not affected the proliferation of CRC cells (Fig. 7A–C). Next, a wound-healing assay was performed and we found that the rate of closure was significantly reduced after PA treatment (Fig. 7D, E). Remarkably, a significantly reduced rate of closure was found in the KAT8-WT cells compared with that in the KAT8-2KR cells after PA stimulation, indicating that KAT8 acetylation plays an important role on PA-stimulated HCT116 cell migration (Fig. 7F, G). Similar results were produced when using RKO cells (Fig. 7H, I). Also, a matrigel-coated invasion assay was established and the degree of cellular invasion was much higher in KAT8-2KR cells than KAT8-WT transfected HCT116 cells after PA stimulation (Fig. 7J, K). Similar results were produced when using SW480 cells (Fig. 7L, M). Taken together, these results suggest that KAT8 acetylation can effectively inhibit migration and invasion of CRC cells. Fig. 7KAT8 acetylation regulates migration and invasion of CRC cells. A HCT116 cells were transfected with pCMV-myc-vector, myc-KAT8-WT and myc-KAT8-2KR plasmids for CCK8 assay to generate a growth curve. B, C HCT116 cells were transfected with the plasmids mentioned above, colony formation assay was performed to identify visible colonies. D, E HCT116 cells were treated with PA at 0.2 mM for 24 h and 48 h. A wound-healing assay was performed to evaluate the migration ability of cells. F-I HCT116 (F, G) and RKO (H, I) cells were transfected with the plasmids mentioned above, and treated with or without PA at 0.2 mM for 24 h. A wound-healing assay was performed to evaluate the migration ability of cells. J-M HCT116 (J, K) and RKO cells (L, M) were transfected with the plasmids mentioned above, a transwell assay was performed to evaluate the invasive ability of cells. The bar (−) represents the means ($$n = 3$$).
## KAT8 acetylation regulates migration and invasion through Lipase HSL in CRC cells
Finally, we explored the mechanism insight into KAT8 acetylation-dependent cell migration and invasion. The expression of E-cadherin and N-cadherin, the main marker of epithelial-mesenchymal transition (EMT), were firstly examined and no significant changes were found in both KAT8-WT and KAT8-2KR transfected cells, implying that KAT8 acetylation-dependent cell migration and invasion is not driven by an EMT process (Fig. 8A). Lipase HSL was reported to regulate the invasiveness of pancreatic cancer cells [34], so we examined the effect of HSL on CRC cells migration and invasive. As shown in Fig. 8B, C, knock down HSL was found to effectively inhibit the migration of HCT116 cells. Also, HCT116 cells were transfected with myc-KAT8-2KR firstly and treated with PA at 0.1 mM for 24 h after HSL siRNA or non-specific siRNA were transfected into these cells. A wound-healing assay and matrigel-coated invasion assay were performed to evaluate the migration and invasion ability of HSL in HCT116 cells. These data showed that knocking down HSL effectively inhibited KAT8 (2KR) promoted the migration and invasion of HCT116 cells by PA stimulation (Fig. 8D–G). Also, the familiar phenomena were shown in RKO and SW480 cells (Fig. 8H–K). Together, these data demonstrate that KAT8 acetylation regulates migration and invasion through HSL in CRC cells. Fig. 8KAT8 acetylation regulates migration and invasion through lipolysis in CRC cells. A HCT116 cells were transfected with pCMV-myc-vector, myc-KAT8-WT and myc-KAT8-2KR plasmids, and Western blotting was performed to detect the expression of the EMT marker E-cadherin and N-cadherin. B, C HCT116 cells were transfected with HSL siRNA or non-specific siRNA for 48 h. A wound-healing assay was performed to evaluate the migration ability of cells. D, E HCT116 cells were transfected by myc-KAT8-2KR firstly and treated with PA at 0.1 mM for 24 h after HSL siRNA or non-specific siRNA were transfected into HCT116 cells. A wound-healing assay was performed to evaluate the migration ability of each group. F, G HCT116 cells were treated as mentioned above, a transwell assay was performed to evaluate the invasive ability of each group. H, I RKO cells were transfected by myc-KAT8-2KR firstly and then treated with PA at 0.1 mM for 24 h after HSL siRNA or non-specific siRNA were transfected into RKO cells. A wound-healing assay was performed to evaluate the migration ability of each group. J, K SW480 cells were treated as mentioned above, a transwell assay was performed to evaluate the invasive ability of each group. L A schematic showing a possible mechanism that KAT8 acetylation-controlled lipolysis affects the invasive and migratory potential of colorectal cancer cells.
## Discussion
KAT8 is associated with a wide range of cellular functions [35, 36]. Here, we found that PA treatment stimulates the release of glycerol and overexpression of KAT8 inhibits this process, showing that KAT8 involves in lipolysis of CRC cells (Fig. 1). To further mechanistically study, we found that KAT8 acetylation is response for regulating the recruitment of RNA polymerase II to the ATGL and HSL promoters to decrease the expression of these genes and further affect invasive and migratory potential in CRC cells (Fig. 8L). Recently, the effect of KAT8 on lipid metabolism has been paid on more and more attention. KAT8 was shown to acetylate fatty acid synthase (FASN) to further destabilize FASN and decrease de novo lipogenesis and tumor cell growth of human hepatocellular carcinoma [37]. Gao et al. found that signal transducer and activator of transcription 5B (STAT5B) can modulate adipocyte differentiation via KAT8 [38]. Also, KAT8 was reported to active fatty acid oxidation (FAO) to block acquisition of quiescence in ground-state ESCs [39]. Our data show the new function of KAT8 on lipid catabolism, and indicate the important role of KAT8 acetylation in human CRC progression.
In the process of identifying the acetyltransferases and deacetylase of KAT8, we found that GCN5 is main acetyltransferases of KAT8 (Fig. 2). Also, SIRT6 is confirmed as a deacetylase of KAT8 by PA stimulation (Fig. 3C–F). It is interesting to know which is the key mediator of KAT8 acetylation. We speculate that there is a dynamic balance between SIRT6 and GCN5 on KAT8 acetylation. GCN5 and SIRT6 can both interact with KAT8 in physiological condition. When SIRT6 senses the stress of lipid metabolism after PA treatment, the interaction between SIRT6 and KAT8 was significantly reduced. Meanwhile, the interaction between GCN5 and KAT8 was magically increased resulting in the increasing of KAT8 acetylation (Fig. 3G–J), indicating that SIRT6 is the key mediator on KAT8 acetylation. In our study, the acetylated sites of KAT8 are K168 and K175 residues after PA treatment. We brought up the hypothesis that auto-acetylation of KAT8 might not happened on these two residues, so SIRT6 and GCN5 are needed for the acetylation of KAT8. Here, we reported for the first time that KAT8 acetylation can be dynamically regulated by SIRT6 and GCN5 instead of its auto-acetylation after PA treatment. It adds a new layer of knowledge on KAT8 acetylation and it will be beneficial for us to further understand the molecular role of KAT8.
KAT8 was identified as a histone H4K16-specific acetyltransferase [15, 40, 41]. Knock-down KAT8 resulted in silencing of the expression of target of methylation-mediated silencing (TMS1) gene, showing that KAT8-dependent histone H4K16ac was important in the maintenance of TMS1 gene activity [42]. However, genome-wide H4K16ac distribution was analyzed and identified 25,893 DNA regions in HEK293 cells by ChIP assay indicating that centromeric regions of chromosome are largely free of H4K16ac and only a small fraction (~$10\%$) is found near promoters [43]. The results of this study imply that KAT8 mediated H4K16ac may be secondary on transcription regulation. In our study, we found that KAT8 decreased the expression of ATGL and HSL genes was not through H4K16ac (Fig. 6E, F). Remarkably, the binding activity of KAT8 on the promoter of both genes was sharply reduced after PA treatment (Fig. 6G). Importantly, KAT8 acetylation was shown to involve in this process and played an important role in regulating the expression of lipolysis-related genes (Fig. 6H–Q). These experiments provide an evidence that KAT8 may serve as a transcription co-activator to regulate the expression of lipolysis-related genes. More experiments are needed to further confirm the effect of KAT8 as a transcription co-factor on gene expression.
KAT8 depletion was shown to promote migration and invasion of tumor cells recently [44]. Here, we assessed the effect of KAT8 acetylation on CRC progression. The results showed that KAT8 acetylation effectively inhibited PA-stimulated migration and invasion without affecting the proliferation of CRC cells (Fig. 7). To explore the mechanism insight into KAT8 acetylation-dependent cell migration and invasion, we found that knowing down HSL effectively inhibited KAT8 (2KR) promoted migration and invasion of CRC cells (Fig. 8D–K). The results of q-PCR and ChIP q-PCR assay showed that KAT8 acetylation attenuates the binding activity of KAT8 to the promoter of lipolysis-related gene HSL to further lead to the decreased expression of HSL gene and inhibit the migration and invasion of CRC cells (Fig. 6G–N). The data indicate that KAT8 acetylation might be regulate migration and invasion through HSL in CRC cells.
In conclusion, our data reports that KAT8 acetylation at K$\frac{168}{175}$ residues was dynamically regulated by SIRT6 and GCN5 during PA treatment. KAT8 suppresses HSL expression to further affect the invasive and migratory potential in colorectal cancer cells. Our data reveal a novel mechanism by which KAT8 regulates lipolysis and control metastatic invasion in CRC cells. These findings expand the field of protein epigenetic regulation on lipid metabolism to mediate metastasis of CRC cells and provide a potential target for colorectal cancer therapy.
## Cell lines, cell culture and reagents
Human colon cancer HCT116 and SW480 cells, human liver cancer HepG2 cells, human pancreas cancer PANC1 cells, human gastric carcinoma MGC803 cells, human embryonic lung diploid fibroblasts 2BS were cultured by Dulbecco Modified Eagle Medium (DMEM, Macgene, China) supplemented with $10\%$ heat-inactivated fetal bovine serum (Gemini, south America). Human colon cancer RKO cells was cultured by RPMI-1640 supplemented with $5\%$ FBS. All the cells were cultured with penicillin/Streptomycin in a 37 °C incubator with a humidified, $5\%$ CO2 atmosphere. No signs of mycoplasma contamination were found for all cell lines. Short tandem repeat profiling was used for cell line authentication. Palmitic acid (PA) was purchased from SIGMA (St Louis, MO, USA) and then prepared to a stock solution and stored in room temperature.
## Plasmid construction and transfection
All the genes open reading frame were amplified from a cDNA library of HCT116 cells by PCR (R045A, TaKaRa, Japan) and cloned into 3x Flag CMV10, pCMV-myc, TF-pcold-His. The mutations of these genes were generated using a site-directed mutagenesis kit (FM111-01, transgene, China). Plasmids were transfected into cell lines by Hieff TransTM Liposomal Transfection Reagent (40802ES02, YEASEN, China) according to the manufacturer’s protocol.
## Oil red O staining
HCT116 cells were treated as required and staining with the Oil Red O working solution for 5 min. The stained lipid droplets were monitored under a microscope (Olympus, Tokyo, Japan).
## Glycerol colorimetric assay
HCT116, SW480, RKO, HepG2, MGC803 and PANC1 cells were treated as required and then lysed with the glycerol colorimetric assay kit (Applygen Technologies, Beijing, China). The lysates were then heated to 70 °C for 10 min to inactivate residue lipase activity. Glycerol in the lysates was determined by the enzyme-coupled GPO-Trinder reaction from the absorption of 550 nm.
## RNA extraction and RT-qPCR
Total RNA was extracted by Trizol agent (Applygen, China). cDNA was synthesized using Quantscript RT Kit (Promega,WI, USA) according to the instruction. RT-qPCR assay was performed by 7500 Fast RT-PCR machine with SYBR PCR mix agent (vazyme, China). The primers sequences used for RT-qPCR are available upon request.
## RNA interference (RNAi)
RNA interference was performed as described [45]. Cells were harvested after transfected with RNAi oligonucleotides and non-specific siRNA and subjected to Western blotting, RT-qPCR or a ChIP assay, respectively. The RNAi oligonucleotides sequences used are available upon request.
## Protein extraction and western blotting
Different cells were harvested after treatment and proteins were extracted to detect the expression by Western blotting as previously described with minor modifications [46]. Equal amounts of proteins were size fractionated by 9 to $15\%$ sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis. Anti-SIRT6 (2590 S, Cell Signaling Techology, Danvers, MA, USA), anti-KAT8 (ab200660, abcam, Cambridge, MA, USA), anti-Acetyl-lysine (PTM-105RM, PTMBIO, China), anti-Flag (F1804, Sigma Aldrich), anti-SIRT1 (8469 S, Cell Signaling Techology, Danvers, MA, USA), anti-SIRT7 (5360 S, Cell Signaling Techology, Danvers, MA, USA), H4K16ac (ab109463, Abcam Cambridge, MA, USA), anti-RNA Pol II (sc-47701, Santa Cruz, CA, USA), anti-E-cadherin (14472 S, Cell Signaling Techology Danvers, MA, USA), anti-N-cadherin (13116 S, Cell Signaling Techology Danvers, MA, USA), anti-Histone H4 (ab177840, Abcam, Danvers, MA, USA), anti-His (PM032, MBL), anti-GST (sc-138, Santa Cruz, CA, USA), anti-GCN5 (sc-365321, Santa Cruz, CA, USA), anti-myc (M047-3, MBL), anti-α-tubulin (BE0031, EASYBIO, Beijing, China) and anti-β-actin (4967, Cell Signaling, Danvers, MA, USA) were used and the blots were developed using an enhanced chemiluminescence kit (Amersham Corp.).
## Co-immunoprecipitation (Co-IP)
After treatment, different cells were harvested and lysed in different lysis buffer to perform Co-immunoprecipitation as described before [9]. Protein A- or G-Sepharose beads (GE Healthcare, Little Chalfont, UK) were used according to the antibody used and the proteins were analyzed by Western blotting with different antibodies.
## GST pull-down assay
GST fusion proteins or His-tagged proteins were purified as described before [9]. His-tagged proteins were incubated with GST fusion proteins in TEN buffers for 4 h at 4 °C. The beads were washed three times with TEN buffers and boiled with 2X SDS loading buffer. Proteins were analyzed by Western blotting with anti-GST or anti-His antibodies and by Coomassie brilliant blue staining.
## In vitro acetylation assay
KAT8 and His-GCN5 were purified and incubated in acetylation buffer (50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 4 mM MgCl2, 0.1 mM EDTA, 1 mM dithiothreitol (DTT), and $10\%$ glycerol) with or without acetyl-CoA (5 mM) for 1 h at 30 °C. The reactions were stopped by adding 5 x protein sample buffer and the samples were boiled at 100 °C for 5 min before SDS-PAGE and immunoblotting.
## In vitro deacetylation assay
SIRT6 and KAT8 were purified and incubated in deacetylation buffer (10 mM Tris-HCl pH 8.0, 10 mM NaCl, $10\%$ glycerol, 1 mM NAD+) for 1 h at 30 °C. The reactions were stopped by adding 5 x protein sample buffer and the samples were boiled at 100 °C for 5 min before SDS-PAGE and immunoblotting.
## Chromatin immunoprecipitation (ChIP) assay
After treatment, HCT116, SW480 or RKO cells were harvested to perform ChIP assay as described before [9]. The cross-link was reversed at 65° C overnight, and the DNA was dissolved in TE buffer and analyzed by real-time PCR. The primers for all ChIPs are available upon request.
## Cell proliferation and colony formation
For cell proliferation experiment, cells were transfected with pCMV-myc-vector, myc-KAT8 or myc-KAT8-2KR plasmids and then harvested for cell counting kit 8 (Yeasen, China) to establish a cell proliferation curve. For the colony formation assay, cells were treated as above and fixed by $4\%$ formaldehyde and stained with $0.25\%$ Coomassie brilliant blue after culturing normally for 2 weeks. All the experiments were performed in triplicate. The number of colonies was calculated using Photoshop (Adobe, American).
## Wound healing assay
HCT116 or RKO cells were seeded and gently scraped with a 10 μl sterile pipette tip when the cells achieved $90\%$ confluence. The wounded cells were continuously cultured with FBS free DMEM for 2 days. The healing width were recorded at the 0 h, 24 h and 48 h under inverted Microscope. The migration rate was measured by the ratio of the day 1 and day 2 width to original width. All the experiments were performed in triplicate.
## Transwell assay
HCT116 or SW480 cells were cultured in 24-well plate with 8-μm polyethylene terephthalate membrane filters separating the lower and upper culture chambers (Corning, American). The membrane filter was coated with Matrigel (Corning, American). Then, cells were seeded and cultured by serum-free DMEM. The bottom chamber contained DMEM with $10\%$ FBS. Cells were allowed to invade for 48 h. Before fixing, non-invasion cells on the upper side of the filter were detached using a cotton swab. Cells were fixed and stained with $0.1\%$ crystal violet. Cells were counted in three random fields. All the experiments were performed in triplicate.
## Data analysis
Measurement data analysis was conducted using PRISM and SPSS statistical analysis software (GraphPad Software, Inc., San Diego, CA). Unpaired t-tests were applied to compare data between two groups, while one-way analysis of variance (ANOVA) was used to compare data among multiple groups. The statistical information of each experiment were shown in the figures and corresponding legends.
## Supplementary information
Supplemental Data Original Data File aj-checklist The online version contains supplementary material available at 10.1038/s41419-023-05582-w.
## References
1. Lu C, Thompson CB. **Metabolic regulation of epigenetics**. *Cell Metab* (2012.0) **16** 9-17. DOI: 10.1016/j.cmet.2012.06.001
2. Kinnaird A, Zhao S, Wellen KE, Michelakis ED. **Metabolic control of epigenetics in cancer**. *Nat Rev Cancer* (2016.0) **16** 694-707. DOI: 10.1038/nrc.2016.82
3. Bensard CL, Wisidigama DR, Olson KA, Berg JA, Krah NM, Schell JC. **Regulation of Tumor Initiation by the Mitochondrial Pyruvate Carrier**. *Cell Metab* (2019.0) **31** 284-300.e7. DOI: 10.1016/j.cmet.2019.11.002
4. Riscal R, Skuli N, Simon MC. **Even Cancer Cells Watch Their Cholesterol!**. *Mol Cell* (2019.0) **76** 220-31. DOI: 10.1016/j.molcel.2019.09.008
5. You X, Tian J, Zhang H, Guo Y, Yang J, Zhu C. **Loss of mitochondrial aconitase promotes colorectal cancer progression via SCD1-mediated lipid remodeling**. *Mol Metab* (2021.0) **48** 101203. DOI: 10.1016/j.molmet.2021.101203
6. Pakiet A, Kobiela J, Stepnowski P, Sledzinski T, Mika A. **Changes in lipids composition and metabolism in colorectal cancer: a review**. *Lipids Health Dis* (2019.0) **18** 29. DOI: 10.1186/s12944-019-0977-8
7. 7.Zhu Y, Gu L, Lin X, Liu C, Lu B, Cui K, et al. Dynamic Regulation of ME1 Phosphorylation and Acetylation Affects Lipid Metabolism and Colorectal Tumorigenesis. Mol Cell. 2019;77:138–149.e5.
8. Xu YH, Song QQ, Li C, Hu YT, Song BB, Ye JM. **Bouchardatine suppresses rectal cancer in mice by disrupting its metabolic pathways via activating the SIRT1-PGC-1α-UCP2 axis**. *Eur J Pharm* (2019.0) **854** 328-37. DOI: 10.1016/j.ejphar.2019.04.029
9. Gao T, Li M, Mu G, Hou T, Zhu W-G, Yang Y. **PKCζ Phosphorylates SIRT6 to Mediate Fatty Acid β-Oxidation in Colon Cancer Cells**. *Neoplasia* (2019.0) **21** 61-73. DOI: 10.1016/j.neo.2018.11.008
10. Li M, Hou T, Gao T, Lu X, Yang Q, Zhu Q. **p53 cooperates with SIRT6 to regulate cardiolipin de novo biosynthesis**. *Cell Death Dis* (2018.0) **9** 941. DOI: 10.1038/s41419-018-0984-0
11. Lu L, Li L, Lv X, Wu X-S, Liu D-P, Liang C-C. **Modulations of hMOF autoacetylation by SIRT1 regulate hMOF recruitment and activities on the chromatin**. *Cell Res* (2011.0) **21** 1182-95. DOI: 10.1038/cr.2011.71
12. Sun B, Guo S, Tang Q, Li C, Zeng R, Xiong Z. **Regulation of the histone acetyltransferase activity of hMOF via autoacetylation of Lys274**. *Cell Res* (2011.0) **21** 1262-6. DOI: 10.1038/cr.2011.105
13. Zhong J, Li X, Cai W, Wang Y, Dong S, Yang J. **TET1 modulates H4K16 acetylation by controlling auto-acetylation of hMOF to affect gene regulation and DNA repair function**. *Nucleic acids Res* (2017.0) **45** 672-84. DOI: 10.1093/nar/gkw919
14. Dou Y, Milne TA, Tackett AJ, Smith ER, Fukuda A, Wysocka J. **Physical association and coordinate function of the H3 K4 methyltransferase MLL1 and the H4 K16 acetyltransferase MOF**. *Cell* (2005.0) **121** 873-85. DOI: 10.1016/j.cell.2005.04.031
15. Füllgrabe J, Lynch-Day MA, Heldring N, Li W, Struijk RB, Ma Q. **The histone H4 lysine 16 acetyltransferase hMOF regulates the outcome of autophagy**. *Nature* (2013.0) **500** 468-71. DOI: 10.1038/nature12313
16. Chatterjee A, Seyfferth J, Lucci J, Gilsbach R, Preissl S, Böttinger L. **MOF Acetyl Transferase Regulates Transcription and Respiration in Mitochondria**. *Cell* (2016.0) **167** 722-.e23. DOI: 10.1016/j.cell.2016.09.052
17. Luo H, Shenoy AK, Li X, Jin Y, Jin L, Cai Q. **MOF Acetylates the Histone Demethylase LSD1 to Suppress Epithelial-to-Mesenchymal Transition**. *Cell Rep.* (2016.0) **15** 2665-78. DOI: 10.1016/j.celrep.2016.05.050
18. Dong Z, Zou J, Li J, Pang Y, Liu Y, Deng C. **MYST1/KAT8 contributes to tumor progression by activating EGFR signaling in glioblastoma cells**. *Cancer Med* (2019.0) **8** 7793-808. DOI: 10.1002/cam4.2639
19. Gupta A, Hunt CR, Hegde ML, Chakraborty S, Chakraborty S, Udayakumar D. **MOF phosphorylation by ATM regulates 53BP1-mediated double-strand break repair pathway choice**. *Cell Rep.* (2014.0) **8** 177-89. DOI: 10.1016/j.celrep.2014.05.044
20. Villa R, Forné I, Müller M, Imhof A, Straub T, Becker PB. **MSL2 combines sensor and effector functions in homeostatic control of the Drosophila dosage compensation machinery**. *Mol cell* (2012.0) **48** 647-54. DOI: 10.1016/j.molcel.2012.09.012
21. Schunter S, Villa R, Flynn V, Heidelberger JB, Classen A-K, Beli P. **Ubiquitylation of the acetyltransferase MOF in Drosophila melanogaster**. *PLoS One* (2017.0) **12** e0177408-e. DOI: 10.1371/journal.pone.0177408
22. Peng L, Ling H, Yuan Z, Fang B, Bloom G, Fukasawa K. **SIRT1 negatively regulates the activities, functions, and protein levels of hMOF and TIP60**. *Mol Cell Biol* (2012.0) **32** 2823-36. DOI: 10.1128/MCB.00496-12
23. Yuan H, Rossetto D, Mellert H, Dang W, Srinivasan M, Johnson J. **MYST protein acetyltransferase activity requires active site lysine autoacetylation**. *EMBO J* (2012.0) **31** 58-70. DOI: 10.1038/emboj.2011.382
24. Singh R, Kaushik S, Wang Y, Xiang Y, Novak I, Komatsu M. **Autophagy regulates lipid metabolism**. *Nature* (2009.0) **458** 1131-5. DOI: 10.1038/nature07976
25. Onal G, Kutlu O, Gozuacik D, Dokmeci Emre S. **Lipid Droplets in Health and Disease**. *Lipids Health Dis* (2017.0) **16** 128. DOI: 10.1186/s12944-017-0521-7
26. Zechner R, Zimmermann R, Eichmann TO, Kohlwein SD, Haemmerle G, Lass A. **FAT SIGNALS-lipases and lipolysis in lipid metabolism and signaling**. *Cell Metab* (2012.0) **15** 279-91. DOI: 10.1016/j.cmet.2011.12.018
27. Vaughan M, Berger JE, Steinberg D. **Hormone-sensitive lipase and monoglyceride lipase activities in adipose tissue**. *J Biol Chem* (1964.0) **239** 401-9. DOI: 10.1016/S0021-9258(18)51692-6
28. Osuga J, Ishibashi S, Oka T, Yagyu H, Tozawa R, Fujimoto A. **Targeted disruption of hormone-sensitive lipase results in male sterility and adipocyte hypertrophy, but not in obesity**. *Proc Natl Acad Sci USA* (2000.0) **97** 787-92. DOI: 10.1073/pnas.97.2.787
29. Zimmermann R, Strauss JG, Haemmerle G, Schoiswohl G, Birner-Gruenberger R, Riederer M. **Fat mobilization in adipose tissue is promoted by adipose triglyceride lipase**. *Science* (2004.0) **306** 1383-6. DOI: 10.1126/science.1100747
30. Sakurada T, Noma A. **Subcellular localization and some properties of monoacylglycerol lipase in rat adipocytes**. *J Biochem* (1981.0) **90** 1413-9. DOI: 10.1093/oxfordjournals.jbchem.a133607
31. Kuang J, Zhang Y, Liu Q, Shen J, Pu S, Cheng S. **Fat-Specific Sirt6 Ablation Sensitizes Mice to High-Fat Diet-Induced Obesity and Insulin Resistance by Inhibiting Lipolysis**. *Diabetes* (2017.0) **66** 1159-71. DOI: 10.2337/db16-1225
32. Rea S, Xouri G, Akhtar A. **Males absent on the first (MOF): from flies to humans**. *Oncogene* (2007.0) **26** 5385-94. DOI: 10.1038/sj.onc.1210607
33. 33.Singh M, Bacolla A, Chaudhary S, Hunt CR, Pandita S, Chauhan R et al. Histone Acetyltransferase MOF Orchestrates Outcomes at the Crossroad of Oncogenesis DNA Damage Response Proliferation and Stem Cell Development ABSTRACT. Mol Cell Biol. 2020:40. 10.1128/MCB.00232-20.
34. Rozeveld CN, Johnson KM, Zhang L, Razidlo GL. **KRAS Controls Pancreatic Cancer Cell Lipid Metabolism and Invasive Potential through the Lipase HSL**. *Cancer Res* (2020.0) **80** 4932-45. DOI: 10.1158/0008-5472.CAN-20-1255
35. Mujoo K, Hunt CR, Horikoshi N, Pandita TK. **A multifaceted role for MOF histone modifying factor in genome maintenance**. *Mech Ageing Dev* (2017.0) **161** 177-80. DOI: 10.1016/j.mad.2016.03.012
36. Sheikh BN, Bechtel-Walz W, Lucci J, Karpiuk O, Hild I, Hartleben B. **MOF maintains transcriptional programs regulating cellular stress response**. *Oncogene* (2016.0) **35** 2698-710. DOI: 10.1038/onc.2015.335
37. Lin HP, Cheng ZL, He RY, Song L, Tian MX, Zhou LS. **Destabilization of Fatty Acid Synthase by Acetylation Inhibits De Novo Lipogenesis and Tumor Cell Growth**. *Cancer Res* (2016.0) **76** 6924-36. DOI: 10.1158/0008-5472.CAN-16-1597
38. Gao P, Zhang Y, Liu Y, Chen J, Zong C, Yu C. **Signal transducer and activator of transcription 5B (STAT5B) modulates adipocyte differentiation via MOF**. *Cell Signal* (2015.0) **27** 2434-43. DOI: 10.1016/j.cellsig.2015.09.010
39. Khoa LTP, Tsan YC, Mao F, Kremer DM, Sajjakulnukit P, Zhang L. **Histone Acetyltransferase MOF Blocks Acquisition of Quiescence in Ground-State ESCs through Activating Fatty Acid Oxidation**. *Cell Stem Cell* (2020.0) **27** 441-58.e10. DOI: 10.1016/j.stem.2020.06.005
40. Hilfiker A, Yang Y, Hayes DH, Beard CA, Manning JE, Lucchesi JC. **Dosage compensation in Drosophila: the X-chromosomal binding of MSL-1 and MLE is dependent on Sxl activity**. *EMBO J* (1994.0) **13** 3542-50. DOI: 10.1002/j.1460-2075.1994.tb06661.x
41. Bone JR, Kuroda MI. **Dosage compensation regulatory proteins and the evolution of sex chromosomes in Drosophila**. *Genetics* (1996.0) **144** 705-13. DOI: 10.1093/genetics/144.2.705
42. Kapoor-Vazirani P, Kagey JD, Powell DR, Vertino PM. **Role of hMOF-dependent histone H4 lysine 16 acetylation in the maintenance of TMS1/ASC gene activity**. *Cancer Res* (2008.0) **68** 6810-21. DOI: 10.1158/0008-5472.CAN-08-0141
43. Horikoshi N, Kumar P, Sharma GG, Chen M, Hunt CR, Westover K. **Genome-wide distribution of histone H4 Lysine 16 acetylation sites and their relationship to gene expression**. *Genome Integr* (2013.0) **4** 3. DOI: 10.1186/2041-9414-4-3
44. Zhu H, Wang Y, Wei T, Zhao X, Li F, Li Y. **KAT8/MOF-Mediated Anti-Cancer Mechanism of Gemcitabine in Human Bladder Cancer Cells**. *Biomol Ther (Seoul)* (2021.0) **29** 184-94. DOI: 10.4062/biomolther.2020.111
45. Yang Y, Zhao Y, Liao W, Yang J, Wu L, Zheng Z. **Acetylation of FoxO1 activates Bim expression to induce apoptosis in response to histone deacetylase inhibitor depsipeptide treatment**. *Neoplasia* (2009.0) **11** 313-24. DOI: 10.1593/neo.81358
46. Yang Y, Ikezoe T, Nishioka C, Bandobashi K, Takeuchi T, Adachi Y. **NFV, an HIV-1 protease inhibitor, induces growth arrest, reduced Akt signalling, apoptosis and docetaxel sensitisation in NSCLC cell lines**. *Br J Cancer* (2006.0) **95** 1653-62. DOI: 10.1038/sj.bjc.6603435
|
---
title: 'Acceptability of a high-protein Mediterranean-style diet and resistance exercise
protocol for cardiac rehabilitation patients: Involving service users in intervention
design using a mixed-methods participatory approach'
authors:
- Richard Kirwan
- Lisa Newson
- Deaglan McCullough
- Tom Butler
- Ian G. Davies
- Fatima Perez de Heredia
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9970995
doi: 10.3389/fnut.2023.1043391
license: CC BY 4.0
---
# Acceptability of a high-protein Mediterranean-style diet and resistance exercise protocol for cardiac rehabilitation patients: Involving service users in intervention design using a mixed-methods participatory approach
## Abstract
### Background
Current cardiac rehabilitation (CR) practices focus on aerobic-style exercise with minimal nutrition advice. This approach may not be optimal for CR patients with reduced muscle mass and elevated fat mass. Higher protein, Mediterranean-style diets combined with resistance exercise (RE) may improve muscle mass and reduce the risk of future cardiovascular events, although such an approach is yet to be trialed in a CR population.
### Objective
We explored patient perspectives on the proposed design of a feasibility study. Patients reflected on the acceptability of a proposed high-protein Mediterranean-style diet and RE protocol, emphasizing research methodology and the acceptability of the proposed recipes and exercises.
### Design
We applied quantitative and qualitative (mixed methods) approaches. The quantitative approach involved an online questionnaire ($$n = 40$$) regarding the proposed study methodology and relevance. A subset of participants ($$n = 12$$) received proposed recipe guides and were asked to prepare several dishes and complete an online questionnaire regarding their experience. Another subset ($$n = 18$$) received links to videos of the proposed RE and completed a questionnaire regarding their impressions of them. Finally, semi-structured interviews ($$n = 7$$) were carried out to explore participants’ impressions of the proposed diet and exercise intervention.
### Results
Quantitative data indicated a high level of understanding of the intervention protocol and its importance within the context of this research. There was a high degree of willingness to participate in all aspects of the proposed study (>$90\%$). The trialed recipes were enjoyed and found to be easy to make by a majority of participants (79 and $92.1\%$, respectively). For the proposed exercises $96.5\%$ of responses agreed they would be willing to perform them and, $75.8\%$ of responses agreed they would enjoy them. Qualitative analysis revealed that participants viewed the research proposal, diet, and exercise protocol in a positive light. The research materials were considered appropriate and well explained. Participants suggested practical recommendations for improving recipe guides and requested more individual-focused exercise recommendations, and more information on the specific health benefits of the diet and exercise protocols.
### Conclusion
The study methodology and the specific dietary intervention and exercise protocol were found to be generally acceptable with some suggested refinements.
## 1. Introduction
Cardiovascular disease (CVD) is the leading cause of death worldwide with almost 19.1 million deaths reported in 2020 [1]. Individuals who experience or are at high risk of, a cardiac event can be referred to cardiac rehabilitation (CR) [2], a lifestyle intervention program predominantly focusing on aerobic exercise, and may involve other components such as advice on diet quality and weight management, smoking cessation, and stress reduction [3]. The objective of CR is to reduce the risk of future cardiac events and improve quality of life, and considerable evidence points to its efficacy (4–6).
Sarcopenic obesity (SO), the combination of reduced muscle mass and function associated with aging (sarcopenia) and excessively elevated body fat (obesity) [7], has been observed to contribute to a greater risk of CVD [8, 9]. Indeed, SO may at least partially explain the phenomenon known as the obesity paradox, whereby lower body mass index (BMI) in cardiac populations is associated with increased mortality (10–12). The increased risk of CVD in those with SO may result from higher levels of pro-inflammatory cytokines produced in visceral adipose tissue (VAT), known to be elevated in individuals with SO [13]. These cytokines can contribute to detrimental changes in cardiometabolic (CM) risk factors such as insulin resistance and dyslipidaemia (14–16). These pro-inflammatory molecules may further contribute to the progression of SO through their association with reduced muscle mass and strength [17]. Reduced muscle mass and function may also contribute to reductions in physical activity levels in adults [18], which can reduce cardiorespiratory fitness and lead to an increased risk of CVD [19, 20].
Accordingly, increasing muscle mass may be an appropriate target in CR patients presenting with SO. Resistance exercise (RE) and increased protein intake are widely used strategies for increasing muscle mass and strength in older adults [21]. However, while RE may be incorporated into some CR programmes, there is a clear emphasis on aerobic exercise-based CR (22–25). Similarly, while Mediterranean diets are recommended to reduce CVD risk (26–32), there is little evidence for adherence to such dietary advice in current CR practices, particularly in non-Mediterranean populations [33, 34]. A randomized controlled trial (RCT) trialing a high-protein, Mediterranean-style diet combined with RE in a CR population would contribute evidence for the efficacy of such an approach.
Patient and public involvement (PPI) in the early stages of developing a feasibility study for such an intervention is recognized as good practice and can greatly contribute to the acceptability of such an intervention [35]. Both quantitative and qualitative methods can be employed to identify potential barriers to change, determine understanding of the relevance of specific interventions, and help to refine the proposed methodology, thereby potentially improving engagement and adherence prior to the implementation of an RCT [35, 36].
The current study was designed to assess and refine the proposed methodology for a high-protein Mediterranean-style diet and Resistance Exercise in cardiac Rehabilitation (PRiMER) [37]. Therefore, to develop a more comprehensive understanding of CVD patient impressions of the proposed intervention particularly on proposed exercises and recipe acceptability, we conducted a mixed methods study involving both qualitative and quantitative research methods [38].
## 2.1. Study design
The study consisted of (i) a cross-sectional, online questionnaire and (ii) a phone interview, conducted amongst individuals with a diagnosis, history or elevated risk of CVD, with questions focused on a proposed high-protein Mediterranean-style diet and RE intervention for CR patients [37]. This methodology was used due to COVID-19-related social distancing restrictions implemented in the UK at the time of data collection.
The principles of a “person-based” approach were used to help refine and evaluate the proposed intervention. Such a person-based approach may help those designing the intervention to better understand how potential participants, as individuals, react to the proposed methodology and identify which aspects may need to be refined for a more feasible implementation [36]. Core-elements of such an approach include (i) intervention planning, (ii) design, and (iii) evaluation of acceptability [36]. The initial planning and design of the proposed intervention were carried out in 2019 with the assistance of the Liverpool Heart and Chest Hospital (LHCH) Service Users Research Endeavour (SURE) group1, a PPI group, and CR staff from LHCH Knowsley Community Cardiovascular Service (KCCS).
## 2.2. Recruitment
Individuals registered as having a diagnosis or history of CVD or type 2 diabetes (T2D) in the Research for the Future (RftF) database were presented with a research survey link via a combination of email and announcements on the RftF website, newsletter, Facebook, and Twitter accounts in June and July of 2020. Research for the Future2 is an initiative of the National Institute for Health and Care Research Clinical Research Network (NIHR CRN)3 to facilitate recruitment to NIHR and other health research studies. Individuals with T2D were included due to the elevated risk of CVD in this population [39]. The link included a participant information sheet with information on the study as well as a consent form. Inclusion criteria were: (i) a diagnosis of a cardiovascular condition or diagnosis as high risk for a cardiac condition, and (ii) previous referral to CR. As the inclusion criteria were considered sufficient for a PPI study, no exclusion criteria were used. A study flow diagram is presented in Figure 1.
**FIGURE 1:** *Study flow diagram.*
## 2.3. Ethical approval
Ethical approval for the intervention study was granted by the National Health Service North West–Greater Manchester East Research Ethics Committee (IRAS ID: 256927, REC reference: 19/NW/0762). The study was conducted according to the ethical principles of the Declaration of Helsinki [40], and online informed consent was obtained from all participants before participation.
## 2.4. Online and telephone questionnaires
The questionnaires were administered through JISC Online Surveys (Bristol, UK4) and took approximately 30 min to complete. For the initial online questionnaire, after completing questions on demographics, participants were asked to read the proposed research plan (Supplementary Figure 1) and were told that all the following questions would relate to this plan. Briefly, the research plan included information on the relevance of the research in relation to heart disease, muscle mass, and CR. The plan also included a brief description of the proposed research intervention including: Participants then completed questions related to their understanding and thoughts regarding the proposed protocol, as well as questions regarding their previous knowledge about the Mediterranean diet and its relation to health. The two final questions asked participants if they would be willing to participate in further evaluation. The first asked participants to refer to a digital recipe book to try the healthy recipes from the study and to reflect on their use, acceptability, and feasibility. Those who agreed were provided with a PDF containing recipes for the proposed intervention (Supplementary File 1). Participants were asked to try as many recipes as they liked and reply to an online questionnaire regarding the preparation of the recipes and their impression of the finished meals. These participants were also invited to engage in a semi-structured interview concerning their thoughts, opinions and recommendations regarding the proposed intervention and recipes.
The final question asked participants to complete a follow-up questionnaire related to the proposed RE protocol (Supplementary File 2). This questionnaire explored how the proposed exercise protocol would be carried out and linked to videos of the RE to be included in the proposed intervention. Participants were asked to answer questions related to the protocol, as well as their willingness to perform, and impressions of, each of the exercises.
## 2.5. Quantitative analysis
Descriptive statistics were analyzed in R [Version 1.4.1717, R Core Team, [41]]. All quantitative data are presented as frequency of responses and percentage of total response with no further statistical analysis.
## 2.6. Qualitative analysis
For the convenience of the participants, interviews were conducted by telephone. Interviews were digitally audio recorded and later transcribed verbatim by the interviewer and first author (RK). Reflective notes were made post-interview, and interview data and analytical notes were discussed between analytical authors (RK, LN) during the analytical process. The interview transcripts were analyzed based on the interpretive-descriptive method [42] to enable the development of themes. The analysis was approached by asking, “what is important to the participants here?” and “what are we learning about the participants’ experiences?” The main themes of the collected data were developed via a constant comparative method of data analysis [43].
We acknowledge that using interpretative-descriptive method, the analysis and subsequent themes were influenced by the research team’s subjective interpretations of the data. However, throughout the analytical process, researcher reflexivity and audited discussions [44] occurred aiding researcher triangulation [45] which ensured rigor in the quality of qualitative analysis conducted [46]. The qualitative data collection and analysis was conducted by the lead author a white male, early career researcher with specific interests in nutrition in cardiac rehabilitation (RK). The qualitative analysis was led by a female Reader in Applied Health Psychology and a Registered Health Psychologist with expertise in qualitative methodology and long-term conditions (LN). The final version of the qualitative analysis was discussed further with the research team: a white female Senior Lecturer in Physiology with expertise in adipose tissue physiology (FP), a white male Reader in Nutrition with expertise in nutrition and lipidology (ID), a white male Senior Lecturer in Nutrition and Dietetics and a Registered Dietitian with expertise in cardiac rehabilitation (TB), and a white male, early career researcher with expertise in nutrition and exercise physiology (DM).
Direct quotes from a range of participants, which we felt would be transparent in context [46], acted as evidence to support commentary. The authors confirm that the raw data examples supporting this study’s findings are available within the article (see Tables 3–5). Due to the nature of this qualitative research, in line with legal and ethical processes, participants of this study did not agree for their full transcripts to be shared publicly, so supporting data beyond the sample quotation extracts is not feasible. Post-quotes, P1–7 indicates from which participant the verbatim quote has been selected. In addition, written feedback from the open-response selection of the questionnaires has also been incorporated into the analysis, to support the validation of findings for each theme; these quotes display (open response) afterwards.
## 3.1. Demographics
A total of 41 people with a history of CVD, T2D or both, participated in the quantitative questionnaire. One participant asked to be withdrawn from the study citing discontent with the proposed protocol. Demographic details of the sample are presented in Table 1. Fifty percent of participants were female, and $90\%$ were of White British ethnicity; the majority of participants ($72\%$) presented with overweight or obesity, over half of the participants ($58\%$) reported high blood pressure, and $43\%$ reported high cholesterol.
**TABLE 1**
| Characteristic | All samples (n = 40) | Male (n = 20) | Females (n = 20) |
| --- | --- | --- | --- |
| Age (years) | 64.7 ± 13.5 | 65.6 ± 11.4 | 63.9 ± 15.6 |
| Age range (years) | 21–85 | 32–85 | 21–84 |
| Body mass index (kg/m2) | 30.0 ± 6.7 | 30.5 ± 6.0 | 29.5 ± 7.5 |
| n Normal (18.5–24.9) | 11 (27.5%) | 4 (10.0%) | 7 (17.5%) |
| n Overweight (25–29.9) | 13 (32.5%) | 7 (17.5%) | 6 (15.0%) |
| n Obese (30<) | 16 (40.0%) | 9 (22.5%) | 7 (17.5%) |
| White ethnicity (%) | 36 (90%) | 17 (42.5%) | 19 (47.5%) |
| Do you have high blood pressure? (yes%) | 23 (58%) | 12 (30.0%) | 11 (27.5%) |
| Do you have high cholesterol? (yes%) | 17 (43%) | 8 (20.0%) | 9 (22.5%) |
## 3.2.1.1. Importance and willingness to participate in the proposed intervention
Data related to participants’ views on the importance of the intervention, the understanding of its relevance and their willingness to participate in such an intervention is displayed in Figure 2. All participants stated that the objectives of the study were clear, that they thought that the intervention was important and that they would be willing to participate in it. Furthermore, $85\%$ of participants replied that they thought the intervention design could to improve muscle strength and blood markers to help reduce heart disease risk.
**FIGURE 2:** *Participant responses to questions relating to intervention importance and willingness to participate. Displayed as number and (percentage) of respondents who selected each answer option (e.g., 100% would represent that all this question’s respondents chose that option).*
## 3.2.1.2. Protocol requirements
Data related to participants’ willingness to undertake the interventions dietary and exercise requirements, and their willingness to undertake the required laboratory procedures is displayed in Table 2. The majority of participants ($90\%$ and above) were willing to participate in all aspects of the proposed intervention.
**TABLE 2**
| If you were eligible for this study would you be willing to | Yes | No |
| --- | --- | --- |
| Come to 2 face-to-face appointments each lasting 1 h? | 40 (100%) | 0 |
| Come to the appointments after fasting for 12 h (only have water)? | 40 (100%) | 0 |
| Allow a blood sample be taken by trained staff? | 40 (100%) | 0 |
| Have your blood pressure measured by trained staff? | 40 (100%) | 0 |
| Have your body muscle and fat measured by trained staff? | 40 (100%) | 0 |
| Have your grip strength tested by trained staff? | 39 (97.5%) | 1 (2.5%) |
| Answer questionnaires/interviews based on your experiences? | 40 (100%) | 0 |
| Have all the above tests measured by a man or woman? | 40 (100%) | 0 |
| Be randomly put into any of the groups described in the study design for 12 weeks? | 40 (100%) | 0 |
| Travel to a local gym to exercise 3 days per week (This would be supervised by a trained person)? | 36 (90%) | 4 (10%) |
| Change your diet to a healthier diet (high-protein Mediterranean style as described)? | 39 (97.5%) | 1 (2.5%) |
| Eat 2 high protein yogurts per day? | 39 (97.5%) | 1 (2.5%) |
| Fill out a food diary for 3 days to record what you have eaten? | 40 (100%) | 0 |
| Have regular (weekly) contact by telephone or text message with a member of the research team to talk about how your exercise and diet is going? | 40 (100%) | 0 |
## 3.2.1.3. Proposed recipes
A total of 12 participants completed the recipe-related questionnaires. As each participant was instructed to try and provide feedback on as many recipes as they wished, a total of 38 responses were received. The recipes trialed, and the frequency of use of each recipe are presented in Figure 3. Due to the large number of recipes trialed by the participants it was decided to pool the results from the recipe-related questionnaires to give an overview of participants’ impressions of all the recipes trialed. A breakdown of participants’ gustatory ratings of the recipes can be seen in Figure 4. *In* general, the recipes were well received by the participants with $79\%$ stating that, overall, they “liked very much” or “extremely liked” the recipes they trialed (Figure 4). Participants’ ratings of the ease/convenience of making the recipes are displayed in Figure 5. *In* general, $73.6\%$ of respondents either agreed or strongly agreed that they would regularly make the recipe(s) they trialed.
**FIGURE 3:** *Recipes trialed by participants. Numbers presented represent the number of participants that trialed a specific recipe.* **FIGURE 4:** *Pooled participant responses to questions relating to gustatory impressions of the proposed recipes. Displayed as number and (percentage) of responses to each answer option (e.g., 100% would represent that all this question’s responses chose that option).* **FIGURE 5:** *Pooled participant responses to questions relating to ease of preparation of the proposed recipes. Displayed as number and (percentage) of responses to each answer option (e.g., 100% would represent that all this question’s responses chose that option).*
## 3.2.1.4. Proposed exercises
A total of 18 participants completed the RE-related questionnaires. Results for participants’ perceptions of RE, in general are displayed in Figure 6. The majority of participants ($61.1\%$ and above) disagreed or strongly disagreed with a number of common negative perceptions of RE such as “resistance exercise will make you look bulky or big” and “resistance exercise is bad for joints” (Figure 6). However, $94.5\%$ of participants agreed or strongly agreed that “Resistance exercise is not good for older people” (Figure 6). Participants responded positively to statements about their willingness to participate in the various aspects of the proposed RE intervention with $88.9\%$ of participants agreeing or strongly agreeing with the various participation questions (Figure 6).
**FIGURE 6:** *Pooled participant responses to questions relating to perceptions of and willingness to try resistance exercise. Displayed as number and (percentage) of responses to each answer option (e.g., 100% would represent that all this question’s responses chose that option).*
Participants also watched videos of the individual REs for the proposed intervention (leg press, Smith machine deadlift, machine chest press, machine row, machine shoulder press, lat pulldown, leg extension, hamstring curl, chest fly machine, horizontal cable row, and shoulder press machine). Due to the large number of exercise videos watched ($$n = 11$$) by the participants it was decided to pool the results from the RE-related questionnaires to give an overview of participants’ impressions of all the exercises viewed, and the results are displayed in Figure 7. *In* general, the exercises were well received by the participants with $96.5\%$ of responses agreeing or strongly agreeing that they would be willing to perform the exercises under trained supervision, and $75.8\%$ of responses agreeing or strongly agreeing that they would enjoy the exercises.
**FIGURE 7:** *Pooled participant responses to questions relating to impressions of the proposed resistance exercises. Displayed as number and (percentage) of responses to each answer option (e.g., 100% would represent that all this question’s responses chose that option).*
## 3.3.1.1. Theme 1. “Pleasantly surprised”: Support for research and practical recommendations for improvements
Overall participants welcomed the concept of this research proposal. The research materials were considered appropriate and well explained. The resources were generally seen as a supportive reminder to prioritize their health.
A few practical recommendations for information formatting and providing hard copies of resources as opposed to digital versions, were suggested. Finally, some participants requested further information on the recommendations for heart health and how it relates to this to dietary requirements (Table 3).
**TABLE 3**
| Supportive of research | Practical recommendations |
| --- | --- |
| “Anything that refreshes your memory and helps put you in the right direction doing the right stuff is always good” (P1) | “Improved somewhat by perhaps a little bit more of an index, perhaps by adding a bit more nutritional information. And if it’s from my point of view as a standalone book, these sort of references to the rest of the cardiac rehab thing” (P6) |
| “Like many overweight people I have a history of dieting and trying different things. And so, my ideal scenario based on past experience… it sounds like that combined approach sounds very attractive” (P7) | “It was alright. It was simple. There was nothing difficult, well the things that I did. There was nothing difficult. I did change one or two things put in a little bit here and there to make it more personal.” (P3) |
| “I thought the instructions were easy to follow couldn’t fault this at all” (P5) | “On the front page, it says food and exercise for a healthier heart. But there’s nothing about exercise in the recipe book. So it is part of the prime trial, that’s fine. But as a standalone, it’s, it’s a bit confusing.” (P6) |
| “If you can come up with the ideal diet and exercise programme to help people recover. Yeah.” (P1) | “The introduction, it says the idea of this recipe book is to give you an idea of blah blah blah by following this particular way of eating, but it doesn’t actually tell you what this particular way of eating is, gives you lots of recipes. So I think, I don’t know how you’re going to use it” (P6) |
| “I was pleasantly surprised” (P7) | “Some inconsistency with some of the text in the recipe book” (P4) |
| “My point of view, which is from the point of view of a non-expert cook. It was it was very good” (P6) | “And sometimes I struggled to for me being an older person perhaps it might have been handy to have (a hard copy)” (P4) |
| “People need to know and find out how exercise, how diet affects an unhealthy or a healthy heart, and to do that you need to do all the things that you’re going to do (in the study)” (P3) | |
| “Mediterranean diet is admired for improving people’s wellbeing and longevity, and I haven’t selected that type of foods most of the time. Yeah.” (P4) | |
## 3.3.1.2. Theme 2: “I definitely would eat that” evaluation of the dietary approach
This theme is explained by an overall positive response to the proposed dietary intervention. Participants typically considered they were aware of, or already engaged in similar healthy eating practices, or had aspirations to do similar. Participants described how they made tweaks to personalize recipes, but overall recipes were “not difficult,” considered easy to follow, and often used “everyday” readily available ingredients. The adapted Mediterranean dietary intervention proposed was therefore considered acceptable and appropriate to recommend to this patient population.
In addition, this theme outlines participants’ queries and recognizes that participants offered practical recommendations particularly in relation to the dietary element of the intervention. For example, there were recommendations that the recipes should use European measurements as opposed to American cup measurements for the ingredients and the recipes could include more nutritional information (e.g., the calorie and macronutrient information). It is noteworthy that participants made “tweaks” to recipes to increase perceived “healthiness,” such as reducing the amount of fat or oil used within the ingredients list. These amendments related to their (mis)understanding of the dietary approach, and it might be helpful to offer participants further information on the development and background evidence-base upon which these dietary recipes are based (Table 4).
**TABLE 4**
| Supportive of dietary approach and recipe book | Queries and amendments to dietary approach |
| --- | --- |
| “Master some of these recipes. They were well written and simple to follow…. There was nothing in it that I was frightened of, or I wouldn’t have tried” (P6) | “It listed on there, sort of how many calories there were in each dish, because I’m diabetic as well. And I’d like to keep checking the calories. So that would you know if you pick up a packet in the supermarket, it tells you how many calories you’re eating.” (P6) |
| “The Pictures were very appealing. So I like to see what I’m supposed to be making I like to see what it’s supposed to look like. So having the picture and the one page instructions were an easy thing to do” (P7) | “I find confusing it just seems to me the recipes maybe originated in the States. So that a lot of the measurements. And I find that confusing because everybody’s got different cups” (P4) |
| “Was something that we would have normally done with a little bit of a twist on it.” (P3) | “I would prefer measurements of the tomatoes to be quantified. I don’t like the American cup measurements. It would be beneficial for me to have carbohydrate value too” (open response) |
| “Super easy, super easy. I’m always on the lookout for new recipes but these were most of the stuff we already had in, you know the basics, just normal stuff you have in the cupboard. I liked the way everything was on one page so you just print out the page” (P7) | “It seemed more than necessary and tasted a bit ‘greasy’ to me. Probably too much food and too much time for a breakfast meal, at least for me! Maybe better for dinner!” (open response) |
| “Most appealing to me, but also, probably, they seem like an easy thing to do. you could do it quickly” (P1) | “We might have tried that but we cut down on one or two things because they’re on the higher fat side. So most of most of the stuff we have is chicken- based. I like Fish but my wife doesn’t so we don’t have as much as we should do.” (P2) |
| “They weren’t expensive meals. Which I think is important.” (P5) | “More oily fish dishes might be appropriate.” (P4) |
| “And I try and eat relatively healthy. So it was interesting for me to see some of the things that I thought Oh, yes, I definitely would eat that.” (P4) | “Getting used to doing these things. I mean, they’re not terribly complex. But I found with trying to eat the right food. Now I have a problem with the high cholesterol” (P1) |
| “I tried a few of the recipes, I should go on using the recipes.” (P6) | “If it (recipe) says it takes 10- 15 minutes. You know it’s gonna be double that.” (P2) |
| | “A lot of people just don’t have the time these days for food prep. You know, especially on a day like this, it’s gorgeous over here in Manchester so you don’t want to spend like 3 hours in the kitchen.” (P2) |
| | “I was a bit concerned about how many calories would be in the sausages. So I left that one out.” (P6) |
## 3.3.1.3. Theme 3. “Finding that right balance” exercise at home
Participants acknowledged that home-based exercise was feasible though most suggested that they were not motivated or did not regularly engage in physical activity at home. There was a sense that participants typically referred to other people doing or being able to exercise at home, but little acknowledgment that this was relevant to them as individuals. There was a call for more detailed information on the benefits and need of RE in this patient group.
Participants also queried the concept of home-based exercise and typically considered an external “trip to the gym” as more motivational as it offers structure and social support. There were concerns raised regarding the safety of RE at home, with thoughts that having professionals who can monitor and offer instruction, being more appropriate. References to the exercise videos were positive overall, with reference to clear instructions, although again, respondents expressed a need for confirmation that a particular exercise was suitable to their individual health status. Further consideration on the practical “how to do” and “what to do” is warranted, alongside consideration for the role of social and motivational support for a future research trial in this patient group (Table 5).
**TABLE 5**
| Sample evidence quotations for “Finding that right balance” exercise at home |
| --- |
| “More detail on the sort of exercise you’re talking about?” (P1) |
| “That’s a motivation thing. I mean you’ve got to be motivated to go to the gym anyway, its having the motivation after a hard day’s work, you got to come in and then find the motivation to do exercise at home. So unless you can find a way of motivating people, and the NHS have things like the couch potato to five k, where you can join in on social media to find other groups support to support you.” (P3) |
| “from a motivation perspective, actually going to the gym is better” (P3) |
| “I have a treadmill at home that I occasionally use when I’m too. I prefer to go out”” (P6) |
| “I go a couple of times a week to exercise in the gym only cardiovascular stuff, only static cycling, a few bits of weights. But I like to do it away from. I prefer doing it away from home. Rather than at home. I regard doing it at home as better than not doing it. but I’d rather to go out. it makes it a bit more special and gets me out of the door. But I do I do try to exercise.” (P6) |
| “If you’ve not got an instructor or, or somebody else in the gym that knows what you’re doing. So he can say, hey mate, you need to do this to improve this or that. If you’re doing it at home and doing it wrong. You could end up doing yourself more damage.” (P3) |
| “I think it’s hard for me to do that at home. But as I’m not going to the gym at the moment, I should be doing something.” (P4) |
| “I think one one commits better in a group than as an individual, you know, you think oh, I’ll just do five minutes.” (P5) |
| “Looks like a simple exercise and not overly difficult” (open response) |
| “Only concern is I could damage my back. It sounds as if this has been thought about!” (open response) |
| “Having had open heart surgery last August I would be a bit apprehensive of this exercise if the weight was too heavy as I wouldn’t want to exert too much pressure on the internal wound” (open response) |
| “I have some slight shoulder pain on my left side when extending my arm in this way. I’m happy to do the exercise but would take advice on whether it’s appropriate for me” (open response) |
## 4. Discussion
To our knowledge, this is the first mixed-methods study to determine the acceptability of using a high-protein Mediterranean-style diet and RE protocol to improve lean mass, strength and cardiometabolic risk in a UK CR population. Both our quantitative and qualitative analyses highlight the recognition of the importance of, acceptability of and willingness to participate in the research protocol presented to the participants. However, a desire for clarification on certain aspects of the protocol’s diet and exercise components and requests for more personalized guidance relating to these components were also highlighted.
The proposed research protocol presented to participants was developed in collaboration with CR practitioners and a hospital-based service-users group (LHCH SURE group) to ensure the research proposal was both understandable and applicable to the end users. This was intended not only to make the research more acceptable but to improve potential participants “health literacy” in relation to the aims and methods of the protocol. Health literacy has been described as “the cognitive and social skills which determine the motivation and ability of individuals to gain access to, understand and use information in ways which promote and maintain good health” [47]. According to the American Medical Association, “health literacy entails more than a patient being able to read written instructions; it requires the ability to comprehend and apply the information ascertained” [48]. As such, ensuring the materials provided to participants improve health literacy related to an intervention should be considered a vital aspect of intervention design. The qualitative research presented here has highlighted several areas where the proposed intervention can be improved. The formatting/presentation of study materials has been highlighted with participants requesting, for example, inclusion of an index, provision of more nutritional information for recipes (calories, carbohydrate content etc.) for recipes and inclusion of more information related to the dietary pattern recommended in the intervention. Other suggestions included changes to/standardization of the measurements used for the recipes, particularly focused on avoiding the use of cup-measures (which are not commonly used in the UK).
Of particular note were comments from participants related to the quantity of oil used in some of the recipes with some participants reluctant to use so much oil when cooking. The Mediterranean dietary pattern is characterized by its use of olive oil as the primary culinary oil, which is believed to be partially responsible for some of the noted health benefits of this way of eating [49, 50]. The inclusion of educational material explaining the potential health benefits of olive oil (and other aspects of the Mediterranean dietary pattern) may be useful to assuage any concerns participants may have regarding the use of olive oil. It should also be noted that the participants in this study only received the recipe booklet and not the full dietary guide for the proposed research intervention, which does contain such information. Of further note is the quantitative result that almost half of the participants in this study did not know or only somewhat knew what a “Mediterranean” diet was. The provision of such information in the participant guides/materials should be considered for future iterations of the intervention.
While $88.9\%$ of participants agreed or strongly agreed with the various participation questions related to the RE intervention, in contrast, $94.5\%$ of participants agreed or strongly agreed that “Resistance exercise is not good for older people.” *This is* another potential educational aspect that is worth elaborating on in future versions of the protocol. Resistance exercise has been shown to have multiple benefits for older adults including improving cardiometabolic risk markers, reducing measures of frailty and improving quality of life (51–53). It should also be noted that appropriately instructed and monitored RE is safe in older adults and even those with CVD [54]. The provision of such information in an easy-to-understand format may be useful in encouraging participation in such interventions.
Participants also commented that the act of going to a gym to perform exercise may be more beneficial and is a concept worthy of further exploration. Performing in-person/gym-based exercise may increase the likelihood of vicarious experiences (observing others be successful) and verbal persuasion (verbal cues and/or feedback that may encourage success) [55], which may lead to greater self-efficacy. Self-efficacy theory proposes that a favorable impression of one’s results can help to encourage individuals to adhere to endeavors such as exercise [56]. As such, the perception that results of in-person/gym-based exercise may be more beneficial may help individuals adhere to exercise programs such as CR [57], and accordingly, the benefits of such exercise should be elaborated on in any material/instruction provided to participants. Providing of such information along with contact with peers and CR exercise providers in in-person/gym-based settings might help encourage self-efficacy [58, 59] and exercise maintenance.
## 4.1. Strengths and limitations
This study presents a number of strengths and limitations. A particular strength of this study is the high proportion of female participants ($50\%$), which is notably higher than the proportion of female CR attendance in England, which ranges from only 15 to $38\%$ [60]. Another strength of this study is the use of the mixed methodology approach to seeking feedback and engagement for this intervention. This offers a safe forum for participants to express their experiences and not be biased by researcher expectations. As such, feedback and analysis can be considered more reflective of participants own perceptions. *The* general agreement of both quantitative and qualitative results in terms of the acceptability of the proposed intervention is also a strength of this research.
The majority of participants were of White British ethnicity ($90\%$), and this is broadly considered a representative sample of CR participants in the UK, based on a recently published report of CR demographics that reported $83.8\%$ of participants as white [60]. However, it is noteworthy that the findings may not be representative of the diverse ethnic population of the UK as a whole or globally and as such further engagement and exploration of the intervention in a more diverse patient population group is warranted. It should be noted that this study does not have data on the socio-economic status or household income of the participants. Without such information these data cannot determine if the proposed recipes and exercises would be acceptable in different socio-economic groups and as such, further research is warranted in these population groups. Furthermore, as the majority of participants had class 1 obesity research with larger numbers of participants in more diverse BMI classifications may be beneficial for tailoring the diet and exercise guidelines.
A further limitation is that participants had the freedom to choose which recipes to make, which may have biased the results of their ratings of the recipes, as participants would naturally choose recipes they expect to agree with their palate and personal tastes. Finally, participants were recruited from RftF who are likely to be a subset of people/patients very willing to participate in research and may not be representative of the wider clinical population in the UK.
## 5. Conclusion
This mixed-methods study found that the proposed high-protein, Mediterranean-style diet, and resistance exercise protocols for cardiac rehabilitation participants were generally found to be acceptable, with a high degree of willingness to participate from potentially eligible participants. Several potential areas of improvement were highlighted, particularly in regards to clarification around the benefits of the diet and exercise components and provision of more comprehensive information in participant-facing guides/documents. This information will be vital for improving future iterations of the proposed intervention protocol to help ensure acceptance and compliance in the target population, helping to increase the likelihood of positive health outcomes.
## Data availability statement
The datasets generated in this study are not publicly available due to privacy and ethical reasons but are available from the corresponding author on reasonable request. Raw qualitative data have been included as evidence via extracted quotes from verbatim transcripts as samples of evidence. Full transcript release has not received ethical approval or participant consent. For further study details, please contact the corresponding author. The authors confirm that the data supporting the findings of this study are available within the article.
## Ethics statement
The studies involving human participants were reviewed and approved by National Health Service North West–Greater Manchester East Research Ethics Committee. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
RK, DM, ID, FP, and TB conceived and designed the study. RK and DM carried out data collection. RK and LN performed data analysis and wrote the first draft of the manuscript. All authors critically revised all versions of the manuscript, read, and approved the final manuscript.
## Conflict of interest
RK received payments from Myprotein UK for the creation of educational content (unrelated to the present study) and has received protein supplements from Optimum Nutrition, for use in an intervention study related to this research. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1043391/full#supplementary-material
## References
1. Tsao C, Aday A, Almarzooq Z, Alonso A, Beaton A, Bittencourt M. **Heart disease and stroke statistics—2022 update: a report from the American heart association.**. (2022) **145** e153-639. PMID: 35078371
2. 2.National Institute for Health and Care Excellence.
Myocardial infarction: cardiac rehabilitation and prevention of further cardiovascular disease.
London: National Institute for Health and Care Excellence (2013).. (2013)
3. **BACPR standards and core components for cardiovascular disease prevention and rehabilitation**. (2017)
4. Anderson L, Thompson D, Oldridge N, Zwisler A, Rees K, Martin N. **Exercise-based cardiac rehabilitation for coronary heart disease.**. (2016) **2016**. DOI: 10.1002/14651858.CD001800.pub3
5. Sagar V, Davies E, Briscoe S, Coats A, Dalal H, Lough F. **Exercise-based rehabilitation for heart failure: systematic review and meta-analysis.**. (2015) **2**. DOI: 10.1136/openhrt-2014-000163
6. Lavie C, Milani R. **Adverse psychological and coronary risk profiles in young patients with coronary artery disease and benefits of formal cardiac rehabilitation.**. (2006) **166** 1878-83. DOI: 10.1001/archinte.166.17.1878
7. Prado C, Wells J, Smith S, Stephan B, Siervo M. **Sarcopenic obesity: a critical appraisal of the current evidence.**. (2012) **31** 583-601. DOI: 10.1016/j.clnu.2012.06.010
8. Atkins J, Whincup P, Morris R, Lennon L, Papacosta O, Wannamethee S. **Sarcopenic Obesity and risk of cardiovascular disease and mortality: a population-based cohort study of older men.**. (2014) **62** 253-60. DOI: 10.1111/jgs.12652
9. Gusmao-Sena M, Curvello-Silva K, Barreto-Medeiros J, Da-Cunha-Daltro C. **Association between sarcopenic obesity and cardiovascular risk: where are we?**. (2016) **33**. DOI: 10.20960/nh.592
10. Lavie C, De Schutter A, Patel D, Romero-Corral A, Artham S, Milani R. **Body composition and survival in stable coronary heart disease: impact of lean mass index and body fat in the “obesity paradox”.**. (2012) **60** 1374-80. DOI: 10.1016/j.jacc.2012.05.037
11. Lavie C, Osman A, Milani R, Mehra M. **Body composition and prognosis in chronic systolic heart failure: the obesity paradox.**. (2003) **91** 891-4. DOI: 10.1016/S0002-9149(03)00031-6
12. Romero-Corral A, Montori V, Somers V, Korinek J, Thomas R, Allison T. **Association of bodyweight with total mortality and with cardiovascular events in coronary artery disease: a systematic review of cohort studies.**. (2006) **368** 666-78. DOI: 10.1016/s0140-6736(06)69251-9
13. Schrager M, Metter E, Simonsick E, Ble A, Bandinelli S, Lauretani F. **Sarcopenic obesity and inflammation in the inchianti study.**. (2007) **102** 919-25. DOI: 10.1152/japplphysiol.00627.2006
14. Fox C, Massaro J, Hoffmann U, Pou K, Maurovich-Horvat P, Liu C. **Abdominal visceral and subcutaneous adipose tissue compartments: association with metabolic risk factors in the framingham heart study.**. (2007) **116** 39-48. DOI: 10.1161/circulationaha.106.675355
15. Medina-Urrutia A, Posadas-Romero C, Posadas-Sanchez R, Jorge-Galarza E, Villarreal-Molina T, Gonzalez-Salazar Mdel C. **Role of adiponectin and free fatty acids on the association between abdominal visceral fat and insulin resistance.**. (2015) **14**. DOI: 10.1186/s12933-015-0184-5
16. Ebbert J, Jensen M. **Fat depots, free fatty acids, and dyslipidemia.**. (2013) **5** 498-508. DOI: 10.3390/nu5020498
17. Rubio-Ruiz M, Guarner-Lans V, Pérez-Torres I, Soto M. **Mechanisms underlying metabolic syndrome-related sarcopenia and possible therapeutic measures.**. (2019) **20**. DOI: 10.3390/ijms20030647
18. Lee S, Tung H, Liu C, Chen L. **Physical activity and sarcopenia in the geriatric population: a systematic review.**. (2018) **19** 378-83. DOI: 10.1016/j.jamda.2018.02.003
19. Steell L, Ho F, Sillars A, Petermann-Rocha F, Li H, Lyall D. **Dose-response associations of cardiorespiratory fitness with all-cause mortality and incidence and mortality of cancer and cardiovascular and respiratory diseases: the Uk biobank cohort study.**. (2019) **53** 1371-8. DOI: 10.1136/bjsports-2018-099093
20. Sillars A, Celis-Morales C, Ho F, Petermann F, Welsh P, Iliodromiti S. **Association of fitness and grip strength with heart failure: findings from the Uk biobank population-based study.**. (2019) **94** 2230-40. DOI: 10.1016/j.mayocp.2019.04.041
21. Hou L, Lei Y, Li X, Huo C, Jia X, Yang J. **Effect of protein supplementation combined with resistance training on muscle mass, strength and function in the elderly: a systematic review and meta-analysis.**. (2019) **23** 451-8. DOI: 10.1007/s12603-019-1181-2
22. Vanhees L, Rauch B, Piepoli M, van Buuren F, Takken T, Borjesson M. **Importance of characteristics and modalities of physical activity and exercise in the management of cardiovascular health in individuals with cardiovascular disease (Part Iii).**. (2012) **19** 1333-56. DOI: 10.1177/2047487312437063
23. Piepoli M, Hoes A, Agewall S, Albus C, Brotons C, Catapano A. **2016 European guidelines on cardiovascular disease prevention in clinical practice: the sixth joint task force of the European society of cardiology and other societies on cardiovascular disease prevention in clinical practice (constituted by representatives of 10 societies and by invited experts)developed with the special contribution of the European association for cardiovascular prevention & rehabilitation (Eacpr).**. (2016) **37** 2315-81. DOI: 10.1093/eurheartj/ehw106
24. Price K, Gordon B, Bird S, Benson A. **A review of guidelines for cardiac rehabilitation exercise programmes: is there an international consensus?**. (2016) **23** 1715-33. DOI: 10.1177/2047487316657669
25. Khadanga S, Savage P, Ades P. **Resistance training for older adults in cardiac rehabilitation.**. (2019) **35** 459-68. DOI: 10.1016/j.cger.2019.07.005
26. Deanfield J, Simpson I, Bradbury K, Fox K, Boon N, Winocour P. **Joint British societies’ consensus recommendations for the prevention of cardiovascular disease (Jbs3).**. (2014) **100** 1-67. DOI: 10.1136/heartjnl-2014-305693
27. 27.National Institute for Health Care Excellence.
Myocardial infarction: cardiac rehabilitation and prevention of further cardiovascular disease.
London: National Institute for Health Care Excellence (2013).. (2013)
28. Iestra J, Knoops K, Kromhout D, de Groot L, Grobbee D, van Staveren W. **Lifestyle, mediterranean diet and survival in european post-myocardial infarction patients.**. (2006) **13** 894-900. DOI: 10.1097/01.hjr.0000201517.36214.ba
29. Panagiotakos D, Notara V, Kouvari M, Pitsavos C. **The mediterranean and other dietary patterns in secondary cardiovascular disease prevention: a review.**. (2016) **14** 442-51. DOI: 10.2174/1570161114999160719104731
30. Trichopoulou A, Bamia C, Norat T, Overvad K, Schmidt E, Tjonneland A. **Modified mediterranean diet and survival after myocardial infarction: the epic-elderly study.**. (2007) **22** 871-81. DOI: 10.1007/s10654-007-9190-6
31. Trichopoulou A, Bamia C, Trichopoulos D. **Mediterranean diet and survival among patients with coronary heart disease in Greece.**. (2005) **165** 929-35. DOI: 10.1001/archinte.165.8.929
32. de Lorgeril M, Salen P, Martin J, Monjaud I, Delaye J, Mamelle N. **Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction – final report of the lyon diet heart study.**. (1999) **99** 779-85. DOI: 10.1161/01.Cir.99.6.779
33. Linan Pinto M, Pinto R, Charneca S, Vasques J, Lemos Pires M, Borges M. **Body composition, lipid profile and mediterranean diet adherence in cardiovascular disease patients attending a long-term exercise-based cardiac rehabilitation program during covid-19 pandemic.**. (2021) **28(Suppl. 1)**. DOI: 10.1093/eurjpc/zwab061.187
34. Vanzella L, Rouse V, Ajwani F, Deilami N, Pokosh M, Oh P. **Barriers and facilitators to participant adherence of dietary recommendations within comprehensive cardiac rehabilitation programmes: a systematic review.**. (2021) **24** 4823-39. DOI: 10.1017/S1368980021002962
35. Campbell N, Murray E, Darbyshire J, Emery J, Farmer A, Griffiths F. **Designing and evaluating complex interventions to improve health care.**. (2007) **334** 455-9. DOI: 10.1136/bmj.39108.379965.BE
36. Yardley L, Ainsworth B, Arden-Close E, Muller I. **The person-based approach to enhancing the acceptability and feasibility of interventions.**. (2015) **1**. DOI: 10.1186/s40814-015-0033-z
37. McCullough D, Kirwan R, Butler T, Perez de Heredia F, Thijssen D, Lip G. **Feasibility of a high-protein mediterranean-style diet and resistance exercise in cardiac rehabilitation patients with sarcopenic obesity (Primer): study protocol for a randomised control trial.**. (2021) **45** 492-8. DOI: 10.1016/j.clnesp.2021.08.001
38. Creswell J, Clark V.. (2017)
39. Einarson T, Acs A, Ludwig C, Panton U. **Prevalence of cardiovascular disease in type 2 diabetes: a systematic literature review of scientific evidence from across the world in 2007–2017.**. (2018) **17**. DOI: 10.1186/s12933-018-0728-6
40. 40.World Medical Association.
WMA declaration of Helsinki – Ethical principles for medical researc involving human subjects.
Ferney-Voltaire: World Medical Association (2018).. (2018)
41. 41.R Core Team (2021). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available online at: https://www.R-project.org/. (2021)
42. Thorne S, Kirkham S, MacDonald-Emes J. **Interpretive description: a noncategorical qualitative alternative for developing nursing knowledge.**. (1997) **20** 169-77. DOI: 10.1002/(SICI)1098-240X(199704)20:2<169::AID-NUR9>3.0.CO;2-I
43. Corbin J, Strauss A.. (2014)
44. Noble H, Smith J. **Issues of validity and reliability in qualitative research.**. (2015) **18** 34-5. PMID: 25653237
45. Heale R, Forbes D. **Understanding triangulation in research.**. (2013) **16**
46. Reynolds J, Kizito J, Ezumah N, Mangesho P, Allen E, Chandler C. **Quality assurance of qualitative research: a review of the discourse.**. (2011) **9** 1-10. PMID: 21210997
47. Nutbeam D. **Health literacy as a public health goal: a challenge for contemporary health education and communication strategies into the 21st century.**. (2000) **15** 259-67
48. Parker R, Williams M, Weiss B, Baker D, Davis T, Doak C. **Health literacy-report of the council on scientific affairs.**. (1999) **281** 552-7
49. Trichopoulou A, Martínez-González M, Tong T, Forouhi N, Khandelwal S, Prabhakaran D. **Definitions and potential health benefits of the mediterranean diet: views from experts around the world.**. (2014) **12**. DOI: 10.1186/1741-7015-12-112
50. D’Alessandro A, De Pergola G. **The mediterranean diet: its definition and evaluation of a priori dietary indexes in primary cardiovascular prevention.**. (2018) **69** 647-59. DOI: 10.1080/09637486.2017.1417978
51. McLeod K, Jones M, Thom J, Parmenter B. **Resistance training and high-intensity interval training improve cardiometabolic health in high risk older adults: a systematic review and meta-anaylsis.**. (2022) **43** 206-18. PMID: 34320660
52. Bray N, Smart R, Jakobi J, Jones G. **Exercise prescription to reverse frailty.**. (2016) **41** 1112-6. PMID: 27649859
53. Hart P, Buck D. **The effect of resistance training on health-related quality of life in older adults: systematic review and meta-analysis.**. (2019) **9**
54. Kirkman D, Lee D, Carbone S. **Resistance exercise for cardiac rehabilitation.**. (2022) **70** 66-72. DOI: 10.1016/j.pcad.2022.01.004
55. Bandura A. **Self-Efficacy: toward a unifying theory of behavioral change.**. (1977) **84**. DOI: 10.1037//0033-295x.84.2.191
56. Desharnais R, Bouillon J, Godin G. **Self-efficacy and outcome expectations as determinants of exercise adherence.**. (1986) **59** 1155-9. DOI: 10.2466/pr0.1986.59.3.1155
57. Kwasnicka D, Dombrowski S, White M, Sniehotta F. **Theoretical explanations for maintenance of behaviour change: a systematic review of behaviour theories.**. (2016) **10** 277-96. DOI: 10.1080/17437199.2016.1151372
58. Jackson D. **How personal trainers can use self-efficacy theory to enhance exercise behavior in beginning exercisers.**. (2010) **32** 67-71
59. McAuley E, Jerome G, Marquez D, Elavsky S, Blissmer B. **Exercise self-efficacy in older adults: social, affective, and behavioral influences.**. (2003) **25**. DOI: 10.1207/S15324796ABM2501_01
60. 60.Rehabilitation NAoC.
National audit of cardiac rehabilitation (NACR) quality and outcomes report 2019.
British Heart Foundation. (2019).. (2019)
|
---
title: Analysis of the volatile organic compounds of epidural analgesia-ameliorated
metabolic disorder in pregnant women with gestational diabetes mellitus based on
untargeted metabolomics
authors:
- Si Ri Gu Leng Sana
- Yang Lv
- Guangmin Chen
- Lei Guo
- Enyou Li
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9970997
doi: 10.3389/fendo.2023.1009888
license: CC BY 4.0
---
# Analysis of the volatile organic compounds of epidural analgesia-ameliorated metabolic disorder in pregnant women with gestational diabetes mellitus based on untargeted metabolomics
## Abstract
Gestational diabetes mellitus (GDM) is a metabolic disease with an increasing annual incidence. Our previous observational study found that pregnant women with gestational diabetes had mild cognitive decline, which may be related to methylglyoxal (MGO). This study aimed to investigate whether labor pain aggravates the increase in MGO and explored the protective effect of epidural analgesia on metabolism in pregnant women with GDM based on solid-phase microextraction gas chromatography/mass spectrometry (SPME/GC-MS). Pregnant women with GDM were divided into a natural birth group (ND group, $$n = 30$$) and epidural analgesia group (PD group, $$n = 30$$). After fasting for ≥ 10 h overnight, venous blood samples were collected pre- and post-delivery to detect MGO, interleukin-6 (IL-6), and 8-epi-prostaglandin F2 alpha (8-iso-PGF2α) by ELISA. Serum samples were analyzed for volatile organic compounds (VOCs) using SPME-GC-MS. MGO, IL-6, and 8-iso-PGF2α levels in the ND group increased significantly post-delivery ($P \leq 0.05$) and were significantly higher in this group than the levels in the PD group ($P \leq 0.05$). Compared to the PD group, VOCs in the ND group increased significantly post-delivery. Further results indicated that propionic acid may be associated with metabolic disorders in pregnant women with GDM. Epidural analgesia can effectively improve the metabolism and immune function in pregnant women with GDM.
## Introduction
With the implementation of the three-child policy in China, pregnancy complications have gradually increased. Gestational diabetes mellitus (GDM) is the most common complication [1]. GDM can lead to maternal metabolic disorders and adversely affect the health of the mother and fetus [2, 3]. In pregnant women with GDM labor pain stimulation during the perinatal period further aggravates metabolic disorders. Metabolomics is the study of biological systems, more specifically on investigating the changes in low molecular metabolites in the body after stimulation [4, 5]. It mainly reflects the changes in endogenous metabolites caused by pathophysiological stimulation and disturbance. Metabolomics has great advantages in biomarker discovery, early disease diagnosis, pathogenesis research, and pharmacological and pharmacodynamic evaluations (6–12).
Childbirth is a physiological process that is accompanied by pain. labor pain is an unpleasant feeling due to the contraction of the uterus and transit of the fetus. Approximately $50\%$ of the pregnant women regard labor pain as unbearable, which can instill fear [13]. Labor pain can also be detrimental to the health of the child and the mother. Adverse effects on pregnant women and fetus include prolonged childbirth, uterine weakness, postpartum hemorrhage, fetal distress, and others [14, 15]. labor pain can also lead to neuroendocrine changes in pregnant women.
Ding et al. [ 16] showed that the incidence of postpartum depression was as high as $34.6\%$, while epidural analgesia reduced the incidence of postpartum depression to $14.0\%$. The findings highlight the need for an analgesic method during delivery that does not affect the normal childbirth process, but which relieves pain. This innovation would be valuable for the physical and mental health of the mothers. An epidural analgesia technology is needed, especially for pregnant women with GDM. We have previously reported that GDM may cause memory loss in pregnant women [17, 18]. labor pain may aggravate various GDM complications, including memory loss. This study investigated whether epidural analgesia technology in delivery can improve metabolism and serum methylglyoxal (MGO) and inflammatory, and oxidative stress factors in pregnant women with GDM.
## Subjects and protocol
Patients aged 18–35 years with American Society of Anesthesiologists physical status I-II were admitted to the study. Sixty pregnant women with GDM were selected and divided into either a natural delivery group (ND group, $$n = 30$$) or epidural analgesia group (painless delivery [PD] group, $$n = 30$$), according to whether the patient chooses labor analgesia. The study protocol was approved by the Ethics Committee of First Affiliated Hospital of Harbin Medical University and registered with the Chinese Clinical Trial Registry (registration number: ChiCTR2000038703).
Patients with pre-gestational type 1 or type 2 diabetes mellitus (T1DM and T2DM, respectively) were not included in the study. Patients with unnatural pregnancy or gestational periods <37 weeks or >41 weeks were excluded. Subjects taking medications, including corticosteroids, antidepressants, or antiepileptics, were also excluded. Additionally, subjects with chronic metabolic, endocrine, inflammatory diseases, cancer, drug or alcohol dependency, history of major brain abnormalities (e.g., tumors and hydrocephaly), epilepsy, or Parkinson’s disease were excluded. The psychological status of the pregnant women was assessed using the Hamilton Depression Scale. Those who scored > 7 and those who may have depression were excluded.
After the informed consent of the pregnant women, first of all, the pregnant women and their fetuses were monitored for oxygen inhalation and routine monitoring. At the beginning of the first stage of labor, when the uterine contraction is regular and the uterine orifice is opened to 2~3cm, let the pregnant woman take a lateral lying position, try to lower her head and hold her knees with both hands, conduct epidural puncture at the waist L2-L3 space, slowly inject the needle, place a tube 3cm to the head after the puncture is successful, and connect the syringe to draw back the bloodless and brain free spinal fluid, then inject 3 ml of $1.5\%$ lidocaine into the epidural cavity. The epidural catheter was excluded from entering the blood vessel or subarachnoid space by mistake. When the vital signs of pregnant women are stable and there is no abnormality in the anesthesia level, the loaded dose of drug is 6 ml by epidural injection, and the drug is $0.1\%$ ropivacaine+0.5 μg/mL sufentanil. Continuously observe the vital signs of pregnant women, and measure and maintain the anesthesia level at the T10 level. The continuous dose of the epidural link self-control analgesia pump is 4-6 mL/h, and the self-control dose is 2 ml. The administration will be stopped after the uterine orifice of the parturient is completely opened. During delivery, the two groups of pregnant women can drink sugar water and honey water to provide energy supply for pregnant women.
On the survey date, all the enrolled patients underwent routine medical history inquiries, physical examinations, and provided samples for laboratory measurements. The clinical research coordinators used a standard questionnaire to collect information on demographic characteristics and medical history. All pregnant women were instructed to maintain their usual physical activity and diet for at least three days before the survey. In order to evaluate the onset of analgesia, mothers’ pain was estimated using the Visual Analogue Score (VAS, 0: no pain 10: the worst pain) at analgesia request and at 20 minutes after administration of the initial bolus. After overnight fasting for ≥10 h, venous blood samples were collected to detect levels of glycosylated hemoglobin (HbA1c) and blood glucose (Glu). For each participant, blood was collected (3 mL) and centrifuged. The serum was recovered. Post-delivery, visual analog scale (VAS) scores were determined and the venous blood of each woman was collected and treated as just described. Blood samples were stored in a -80°C deep cryogenic refrigerator. MGO, interleukin-6 (IL-6), and 8-epi-prostaglandin F2 alpha (8-iso-pgf2α) were detected using ELISA. All measurements were performed within 6 months of sample collection.
## Solid-phase microextraction
Post-delivery, the venous blood of the two groups was analyzed using SPME. In this study, we selected a 75 μM extraction head. The coating material was carbon molecular sieve/polydimethylsiloxane (car/PDMS). An automatic sample injector was used for heating and extraction. The liquid sample bottle was accessed via a puncture. An extraction method for headspace extraction was adopted. The SPME fiber was inserted into evacuated 20ml glass vials and exposed to the headspace of a blood sample (blood: 2 ml, taken from the ulnar vein) for 20 min at 40°C. After the extraction and concentration of the samples, the automatic sampling device inserted the extraction head into the gas chromatography-mass spectrometry (GC-MS) injection port for analysis. The desorption of volatiles occurred in a hot GC injector at 200°C for 2 min.
## GC-MS analysis
All analyses were performed on a model QP2010 GC/MS (Shimadzu) equipped with a DB-5MS porous layer open tubular column (length: 30 m; internal diameter: 0.250 µm; film thickness, 0.25 mm: Agilent Technologies). The injections were performed in splitless mode, with a splitless time of 1 min. The injector temperature was set to 200°C and the carrier gas was helium at a flow rate of 2 mL/min. The temperature in the column was maintained at 40°C for 2 min to condense the hydrocarbons. The temperature was then increased to 200°C at 70°C/min and held for 1 min. Subsequently, the temperature was ramped to 230°C at a rate of 20°C/min and maintained for 3 min. MS analyses were performed in full-scan mode with an associated m/z range of 35–200 amu. An ionization energy of 70 eV was used for each measurement, and the ion source maintained at 200°C.
## Statistical analyses
SPSS19.0 software was used for the statistical analyses. All data were tested for normality and variance. Normally distributed data are expressed as mean ± standard deviation (mean ± SD). Continuous variables with normal distribution were compared using Student’s t-test. Variables with abnormal distribution were compared using the Mann-Whitney U test. The least significant difference method was used to make multiple comparisons among the groups. Categorical data are expressed as counts and percentages, and comparisons between groups were performed by bilateral χ 2 inspection. $P \leq 0.05$ indicated statistical significance.
SIMCA-p +11.5 software was used for multivariate data analysis and model establishment. Principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) were used for statistical analysis. At the same time, the PLS-DA sample distribution score map was used to divide different samples into different clusters. The default seven-round cross-validation method was used, and the value of the variable importance in projection (VIP) of relevant variables in the PLS-DA model was calculated. Two hundred iterative permutation tests were conducted to verify the supervision mode and to prevent overfitting of the PLS-DA model. The nonparametric Kruskal–Wallis rank sum test was used to calculate the P-value. When VIP > 1.0, and $P \leq 0.05$, the difference variable was statistically significant (i.e., the metabolite was a significantly different metabolite).
## Demographic information
There were no significant differences in body weight, blood glucose level, or HbA1c level between the two groups ($P \leq 0.05$) (Table 1). The VAS score of the PD group was significantly better than that of the ND group ($P \leq 0.05$) (Table 1).
**Table 1**
| Unnamed: 0 | PD | ND | F | P |
| --- | --- | --- | --- | --- |
| Sample | 30 | 30 | | |
| Age, years | 30.16 ± 3.27 | 30.79 ± 4.16 | 0.65 | 0.52 |
| Height, cm | 165.32 ± 2.89 | 164.91 ± 2.51 | 0.59 | 0.60 |
| Weight, kg | 79.36 ± 11.58 | 78.35 ± 13.57 | 0.31 | 0.76 |
| Glucose, mmol/LHbA1c, %VAS | 4.90 ± 1.325.31 ± 0.613.1 ± 0.7 | 5.39 ± 0.714.99 ± 0.937.8 ± 2.0 | 1.791.786.215 | 0.070.080.05 |
## Level of serum indicators
Pre-delivery, the levels of MGO, IL-6, and 8-iso-pgf2α in the ND group were not significantly higher than the levels in the PD group ($P \leq 0.05$). Compared to pre-delivery, the levels of MGO, IL-6, and 8-iso-pgf2α in the PD group post-delivery showed an upward trend, but the difference was not statistically significant. However, post-delivery, the levels of MGO, IL-6, and 8-iso-pgf2α in the ND group significantly increased ($P \leq 0.05$). These levels in the ND group post-delivery were also significantly higher than those in the PD group ($P \leq 0.05$) (Table 2).
**Table 2**
| Group | n | MGO | MGO.1 | P |
| --- | --- | --- | --- | --- |
| Group | n | PD | ND | P |
| Pernatal | 30 | 46.15 ± 8.27 | 45.34 ± 7.79 | 0.69 |
| Postpartum | 30 | 67.34 ± 18.81 | 95.35 ± 18.27 | <0.001 |
| P | | <0.001 | <0.001 | |
| Group | n | IL-6 | IL-6 | |
| Group | n | PD | ND | P |
| Pernatal | 30 | 28.31 ± 4.26 | 26.86 ± 4.01 | 0.18 |
| Postpartum | 30 | 35.51 ± 4.21 | 44.17 ± 5.51 | <0.001 |
| P | | <0.001 | <0.001 | |
| Group | n | 8-iso-PGF2α | 8-iso-PGF2α | |
| Group | n | PD | ND | P |
| Pernatal | 30 | 47.13 ± 8.26 | 48.19 ± 7.97 | 0.61 |
| Postpartum | 30 | 61.24 ± 7.22 | 81.15 ± 6.98 | <0.001 |
| P | | <0.001 | <0.001 | |
## Metabolic volatile organic compounds
Compared to the PD group, the metabolic VOCs in the ND group increased significantly post-delivery. In this study, we found 17 different metabolites (Table 3). Of these, 11 were in the ND group and six were in the PD group. In the two-dimensional PCA score diagram, data from the two groups showed a good separation trend (Figure 1A). When using a single prediction component and three orthogonal components, the PLS-DA score chart displayed the data of the two groups and also had a good separation effect (Figures 1B, C). Two hundred iterations were performed to test the supervision model. The R2 and Q2 values calculated from the converted data were lower than their original verification values, which proved the effectiveness of the supervision model (Figure 1D). Based on the heat map (Figure 1E), the expression of six types of VOCs increased in the PD group as indicated by the red color. Eleven types of VOCs were increased in the ND group.
## Discussion
Severe labor pain makes pregnant women vulnerable to fear and anxiety, and heightens their stress. These feelings can be exacerbated in pregnant women who develop GDM. These stressors have several adverse effects on the mother and fetus. During delivery, the body is in a state of stress due to stimulations that include pain and childbirth trauma [19]. The immune function of the body is disrupted, which increases the risk of maternal infection and other complications. In recent years, epidural analgesia has been used in many patients. An increasing number of pregnant women and obstetricians are accepting epidural analgesia, which has an obvious effect on reducing labor pain.
Under physiological conditions, a variety of enzymes participate in the metabolism of MGO. A small amount of MGO in the body is not sufficient to cause toxic reactions. However, metabolic disorders in patients with GDM lead to an increase in MGO production. Simultaneously, the strong stress response caused by delivery leads to an increase in levels of reactive oxygen species, which stimulate the production of MGO [20]. In a previous study, we found that pregnant women with GDM were in a state of metabolic disorder [21]. In the process of delivery, this metabolic disorder is aggravated because the delivery process itself can be considerably traumatic. In this study, the results are consistent with this phenomenon. The level of MGO in both groups increased post-delivery. However, compared with the PD group, the level of MGO in the ND group increased significantly, indicating that epidural analgesia can ameliorate metabolic disorders of MGO caused by labor pain.
Under physiological conditions, immune changes produced by delivery have protective significance for the body [22]. However, abnormal immune function disorders are closely related to perinatal diseases. Studies have shown that labor pain is the main factor that causes changes in immune function (23–25). Epidural analgesia can effectively protect the body from excessive stress and inhibit immune function. IL-6 and 8-iso-pgf2α were also detected in this study. Compared with the PD group, the levels of IL-6 and 8-iso-pgf2α in the ND group post-delivery increased significantly, indicating that epidural analgesia can reduce the inflammatory and oxidative stress response of pregnant women with GDM.
This study mainly observed the effects of different delivery methods (ND and PD) on the metabolism of pregnant women with GDM. GC-MS was used in a novel analysis of the changes in VOCs in pregnant women with GDM post-delivery. In the two groups, we found 15 kinds of VOCs, Including 10 VOCs in the ND group and 5 VOCs in the PD group. The results revealed obvious metabolic disorders in the ND group. Ten differential substances were found in the ND group, including alpha-l-galactopyranose; 6-deoxy-, cyclic 1,2:3,4-bis(butylboronate); 1-phenyl-1-(trimethylsilyloxy)ethylene; malonic acid; bis(2-trimethylsilylethyl ester); propanoic acid; trans-beta-ocimene; trans-2-dodecen-1-ol; 1-diisopropylsilyloxycyclohexane; cycloheptane; hexanal; octanal。 Among the different substances in the ND group, alcohols (trans-2-dodecen-1-ol) and aldehydes (hexanal and octanal) were identified. Active aldehydes are mainly produced during lipid and glucose metabolism (including enzymatic and non-enzymatic pathways). The enzyme pathway is usually an aldehyde intermediate or by-product produced during glucose and lipid metabolism in vivo [26]. This is also consistent with the disorder of active aldehyde metabolism observed in pregnant women with GDM. Under pathological conditions, aldehyde metabolism is disordered, resulting in abundant accumulation of aldehyde and formation of an aldehyde microenvironment [27]. Similar to saturated aldehydes, hexanal is oxidized to the corresponding caboxylic acid by aldehyde dehydrogenase mainly in the liver, but also in other tissues and cells. The acid can serve as a substrate for the Krebs cycle or is excreted as a salt. Alternatively, it can conjugate with glutathione or the sulfhydryl group of other proteins. Free radical induced lipid peroxidation may play a role in neurodegeneration and peroxidation leads to the formation of hexanal from omega-6 fatty acids. We have previously demonstrated in vitro that pyruvate dehydrogenase (PDH) catalyzes the condensation of saturated aldehydes with pyruvate to form acyloins. We have further shown in perfused rat heart that hexanal, presumably via PDH, is converted to 3-hydroxyoctan-2-one and that it in turn can be reduced to 2,3-octanediol [28].
Aldehyde metabolism disorders are involved in the occurrence and development of various diseases. α-l-*Galactopyranose is* a carbohydrate metabolized by hexose (glucose) and cyclic sugar. L-Galactose, also known as α-L-galactose or L-galactopyranose, belongs to the class of organic compounds known as hexoses. These are monosaccharides in which the sugar unit is a is a six-carbon containing moeity. L-*Galactose is* a primary metabolite. Primary metabolites are metabolically or physiologically essential metabolites. They are directly involved in an organism’s growth, development or reproduction. Based on a literature review very few articles have been published on L-Galactose. L-Galactose can be metabolized into vitamins. This may be related to the supplementary food provided to pregnant women in the ND group during delivery.
Propionic acid was identified as an important component in this study. Three different biochemical pathways for propionic acid production are succinic acid, acrylate, and propylene glycol. Propionic acid is the main end-product of succinic acid fermentation. The abundance of Bacteroides and Parabacteroides, the main producers of succinic acid, increased due to a high-fat diet and were positively correlated with body weight. Succinic acid produced by Bacteroides thetaiotaomicron supports the growth of Phascolarctobacterium and the accompanying production of propionic acid through the succinic acid pathway [29]. High concentrations of Phascolarctobacterium have been reported in severe depression, Alzheimer’s disease, autism, and other diseases, although the heterogeneity within the disease group was also high. Therefore, the increase in propionic acid was related to the tricarboxylic acid cycle, and the metabolism of propionic acid was also related to MGO (Figure 2). A previous study described a metabolic pathway in athletes after marathon competition. Veillonella metabolizes exercise-induced lactic acid into propionic acid, which improves running time, determines a natural and microbial-encoded enzyme process, and improves sports performance [30]. The delivery process is also relatively long, lasting for several hours or even longer, and the intensity is high. In pregnant women with GDM, increased glycolysis produces more pyruvate that is metabolized to lactic acid. Lactic acid is likely to be metabolized into propionic acid in the body. According to this reasoning, pyruvate can also be converted to acetone to produce MGO and aggravate nerve injury. Therefore, propionic acid may be the product of metabolic disorders in the ND group, or it may be a potential marker of cognitive dysfunction in pregnant women with GDM aggravated by labor pain. Epidural analgesia can prevent this type of injury. Malonic acid is also a kind of propionate, which may have the same metabolic pathway as propionic acid [31].
**Figure 2:** *Metabolic pathway of propionic acid.*
Six substances were found in the PD group. These included decane, heptane, and other alkanes. Alkanes were also observed in the ND group. Lipid peroxidation is a pathophysiological change in various diseases, including cancer, inflammatory diseases, atherosclerosis, and aging [32]. Alkane metabolism is unrelated to branched-chain hydrocarbons and is the product of lipid peroxidation. Unbound alkane hydrocarbons eventually appear in the blood, urine, and exhalate. Ethane and pentane are saturated hydrocarbons produced by lipid peroxidation chain reactions [33]. Therefore, these aliphatic hydrocarbons are considered biomarkers of lipid peroxidation both in vivo and in vitro [34]. This indicates that pregnant women with GDM may have an oxidative stress reaction.
This study found that both groups of pregnant women had a certain degree of stress response, but the metabolic disorder of pregnant women in the ND group was more obvious, indicating that labor pain caused an increase in oxidative stress and metabolic disorders in pregnant women with GDM. In contrast, epidural analgesia can significantly improve oxidative stress metabolites in pregnant women with GDM. This is also one of the main reasons for the changes in metabolites between the two groups.
Severe labor pain activates the hypothalamus pituitary adrenal axis and sympathetic adrenal medullary axis, causing a series of neuroendocrine disorders, which also promote the secretion of glucocorticoids and catecholamines. This stress also affects the body’s immune system, thus activating inflammatory factors and oxidative stress responses, including an increase in MGO. Neuroendocrine dysfunction is one of the pathophysiological mechanisms of GDM. In the process of delivery, epidural analgesia reduces the neuroendocrine imbalance in pregnant women with GDM, adjusts the immune response, and reduces the damage caused by stress trauma to the body.
There were several limitations in this study. The sample size was relatively small. Postpartum follow-up was not performed. Finally, the duration of labor was not considered. It is reported that decane, heptane, and other alkanes are found in patients with temporary respiratory syndrome [35]. Because GDM pregnant women are relatively obese, and sleep disorders in the late pregnancy may also be one of the reasons for these substances. However, these substances may also be part of the experimental reagent (such as the kit), so the kit can also release these substances during the research process, which needs further research.
## Conclusion
The data demonstrate for the first time that propionic acid, a volatile substance, may be a potential marker of cognitive dysfunction in pregnant women with GDM aggravated by labor pain. Epidural analgesia can improve the expression of MGO/inflammatory/oxidative stress factors induced by childbirth-related pain.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by the First Affiliated Hospital of Harbin Medical University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
SS designed the study, collected data, and wrote and revised the manuscript; YL and GC interpreted and analyzed the data; LG collected the data; EL designed the study. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Johns EC, Denison FC, Norman JE, Reynolds RM. **Gestational diabetes mellitus: Mechanisms, treatment, and complications**. *Trends Endocrinol Metab* (2018) **29**. DOI: 10.1016/j.tem.2018.09.004
2. Weinert LS. **International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy: Comment to the international association of diabetes and pregnancy study groups consensus Panel**. *Diabetes Care* (2010) **33** e97. DOI: 10.2337/dc10-0544
3. Garcia-Garcia G, Monteon-Ramos JF, Garcia-Bejarano H, Gomez-Navarro B, Reyes IH, Lomeli AM. **Renal replacement therapy among disadvantaged populations in Mexico: A report from the jalisco dialysis and transplant registry (REDTJAL)**. *Kidney Int Suppl* (2005) **68**. DOI: 10.1111/j.1523-1755.2005.09710.x
4. Junwei F, Wenyu W, Shujun S, Yang W, Qianhua L, Xiong L. **Metabolomics study of renal fibrosis and intervention effects of total aglycone extracts of scutellaria baicalensis in unilateral ureteral obstruction rats**. *J Ethnopharma- col* (2016) **192**. DOI: 10.1016/j.jep.2016.06.014
5. Priori R, Scrivo R, Brandt J, Valerio M, Casadei L, Valesini G. **Metabolomics in rheumatic diseases: The potential of an emerging methodology for improved patient diagnosis, prognosis, and treatment efficacy**. *Autoimmun Rev* (2013) **12**. DOI: 10.1016/j.autrev.2013.04.002
6. Sato Y, Ik S, Nakamura T, Bernier F, Aoshima K, Oda Y. **Identification of a new plasma biomarker of alzheimer' s disease using metabolomics technology**. *J Lipid Res* (2012) **53**. DOI: 10.1194/jlr.M022376
7. Huynh J, Xiong G, Bentley-Lewis R. **A systematic review of metabolite profiling in gestational diabetes mellitus[J]**. *Diabetologia* (2014) **57**. DOI: 10.1007/s00125-014-3371-0
8. Tan Y, Yin P, Tang L, 1Xing WB, Huang Q, Cao D. **Metabolomics study of stepwise hepatic carcino genesis from the model rats' to patients: Potential biomarkers effective for small hepatocellular carcinoma diagnosis**. *Mol Cell Proteomics* (2012) **11**. DOI: 10.1074/mcp.M111.010694
9. Legg K. **Metabolomics: Gaining insight into pain**. *Nat Rev Drug Discovery* (2012) **11**. DOI: 10.1038/nrd3685
10. Bentley-lewis R, Huynh J, Xing G, Lee H, Wenger J, Clish C. **Metabolomic profiling in the prediction of gestational diabetes mellitus**. *Diabetologia* (2015) **58**. DOI: 10.1007/s00125-015-3553-4
11. Patti GJ, Yanes O, Shriver LP, Courade JP, Tautenhahn R, Manchester M. **Metabolomics implicates altered sphingolipids in chronic pain of neuropathicorigin**. *Nat Chem Biol* (2012) **8**. DOI: 10.1038/nchembio.767
12. Brockmoller SF, Bucher E, Müller BM, Budczies J, Hilvo M, Griffin JL. **Integration of metabolomics and expression of glycerol-3-phosphate acyltransferase (GPAM) in breast cancer-link to patient survival, hormone receptor stat**. *J Proteome Res* (2012) **11**. DOI: 10.1021/pr200685r
13. Logsdon K, Smith-Morris C. **An ethnography on perceptions of pain in Dutch "Natural" childbirth[J]**. *Midwifery* (2017) **55** 67-74. DOI: 10.1016/j.midw.2017.09.004
14. O'hara MW, Wisner KL. **Perinatal mental illness: Definition, description and aetiology[J]**. *Best Pract Res Clin Obstet Gynaecol* (2014) **28** 3-12. DOI: 10.1016/j.bpobgyn.2013.09.002
15. Boudou M, Teissèdre F, Walburg V, Chabrol H. **[Association between the intensity of labor pain and the intensity of postpartum blues]**. *Encephale* (2007) **33**. DOI: 10.1016/j.encep.2006.10.002
16. Ding T, Wang DX, Qu Y, Chen Q, Zhu SN. **Epidural labor analgesia is associated with a decreased risk of postpartum depression: A prospective cohort study**. *Anesth Analg* (2014) **119**. DOI: 10.1213/ANE.0000000000000107
17. Sana S, Deng X, Guo L, Wang X, Li E. **Cognitive dysfunction of pregnant women with gestational diabetes mellitus in perinatal period**. *J Healthcare Engineering* (2021) **2021**. DOI: 10.1155/2021/2302379
18. Sana SR, Li EY, Deng XJ, Guo L. **Association between plasma dipeptidyl peptidase-4 levels and cognitive function in perinatal pregnant women with gestational diabetes mellitus**. *World J Clin cases* (2021) **9**. DOI: 10.12998/wjcc.v9.i33.10161
19. Henrique AJ, Gabrielloni MC, Rodney P, Barbieri M. **Non-pharmacological interventions during childbirth for pain relief, anxiety, and neuroendocrine stress parameters: A randomized controlled trial**. *Int J Nurs Pract* (2018) **24** e12642. DOI: 10.1111/ijn.12642
20. Rabbani N, Godfrey L, Xue M, Shaheen F, Geoffrion M, Milne R. **Glycation of LDL by methylglyoxal increases arterial atherogenicity: A possible contributor to increased risk of cardiovascular disease in diabetes**. *Diabetes* (2011) **60**. DOI: 10.2337/db11-0085
21. Pietzner M, Stewart ID, Raffler J, Khaw KT, Michelotti GA, Kastenmüller G. **Plasma metabolites to profile pathways in noncommunicable disease multimorbidity**. *Nat Med* (2021) **27**. DOI: 10.1038/s41591-021-01266-0
22. Saito S, Nakashima A, Shima T, Ito M. **Th1/Th2/Th17 and regulatory T-cell paradigm in pregnancy**. *Am J Reprod Immunol* (2010) **63**. DOI: 10.1111/j.1600-0897.2010.00852.x
23. Pomini F, Caruso A, Challis JR. **Interleukin-10 modifies the effects of interleukin-1beta and tumor necrosis factor-alpha on the activity and expression of prostaglandin hsynthase-2 and the NAD+-dependent 15-hydroxyprostaglandin dehydrogenase in cultured term human villous trophoblast and chorion trophoblast cells**. *J Clin Endcrinol Metab* (1999) **84**. DOI: 10.1210/jcem.84.12.6188
24. Akdis CA, Blaser K. **Mechanisms of interleukin-10-mediated immune suppression**. *Immunology* (2001) **103**. DOI: 10.1046/j.1365-2567.2001.01235.x
25. Asadullsh K, Sterry W, Volk HD. **Interleukin-10 therapy-review of a new approach**. *Pharmacol Rev* (2003) **55**. DOI: 10.1124/pr.55.2.4
26. Phelps RL, Metzger BE, Freinkel N. **Carbohydrate metabolism in pregnancy. XVII. diurnal profiles of plasma glucose, insulin, free fatty acids, triglycerides, cholesterol, and individual amino acids in late normal pregnancy**. *Am J Obstet Gynecol* (1981) **140**. DOI: 10.1016/0002-9378(81)90731-6
27. Goodacre R. **Metabolomics shows the way to new discoveries**. *Genome Biol* (2005) **6**. DOI: 10.1186/gb-2005-6-11-354
28. Jaar V, Ste-Marie L, Montgomery JA. **Striatal metabolism of hexanal, a lipid peroxidation product, in the Rat**. *Metab Brain Dis* **14**. DOI: 10.1023/a:1020701612639
29. Signe A, Katrin T, Heiki V, Marju P, Natalja K, Triinu V. **Degradation of fructans and production of propionic acid by bacteroides thetaiotaomicron are enhanced by the shortage of amino Acids**. *Front Nutr* (2014 D) **5**. DOI: 10.3389/fnut.2014.00021
30. Scheiman J, Luber JM, Chavkin TA, Macdonald T, Tung A, Pham LD. **Meta-omics analysis of elite athletes identifies a performance-enhancing microbe that functions**. *Nat Med* (2019) **25**. DOI: 10.1038/s41591-019-0485-4
31. Thatcher GR, Bennett BM, Dringenberg HC, Reynolds JN. **Novel nitrates as NO mimetics directed at alzheimer's disease**. *J Alzheimers Dis* (2004) **6**. DOI: 10.3233/jad-2004-6s614
32. Haick H, Broza YY, Mochalski P, Ruzsanyi V, Amann A. **Assessment, origin, and implementation of breath volatile cancer markers**. *Chem Soc Rev* (2014) **43**. DOI: 10.1039/c3cs60329f
33. Miekisch W, Schubert JK, Noeldge-Schomburg GF. **Diagnostic potential of breath analysis–focus on volatile organic compounds**. *Clin Chim Acta* (2004) **347** (1-2). DOI: 10.1016/j.cccn.2004.04.023
34. Aghdassi E, Allard JP. **Breath alkanes as a marker of oxidative stress in different clinical conditions[J]**. *Free Radic Biol Med* (2000) **28**. DOI: 10.1016/s0891-5849(00)00189-1
35. Aoki T, Nagaoka T, Kobayashi N, Kurahashi M, Tsuji C, Takiguchi H. **Editor's highlight: Prospective analyses of volatile organic compounds in obstructive sleep apnea patients**. *Toxicol Sci* (2017) **156**. DOI: 10.1093/toxsci/kfw260
|
---
title: Is Following a Cancer-Protective Lifestyle Linked to Reduced Cancer Mortality
Risk?
authors:
- Flurina Suter
- Nena Karavasiloglou
- Julia Braun
- Giulia Pestoni
- Sabine Rohrmann
journal: International Journal of Public Health
year: 2023
pmcid: PMC9970999
doi: 10.3389/ijph.2023.1605610
license: CC BY 4.0
---
# Is Following a Cancer-Protective Lifestyle Linked to Reduced Cancer Mortality Risk?
## Abstract
Objectives: This study investigates the association between a cancer protective lifestyle (defined based on the revised World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR) cancer prevention recommendations) and mortality in Switzerland.
Methods: Based on the cross-sectional, population-based National Nutrition Survey, menuCH ($$n = 2057$$), adherence to the WCRF/AICR recommendations was assessed via a score. Quasipoisson regression models were fitted to examine the association of adherence to the WCRF/AICR recommendations with mortality at the Swiss district-level. Spatial autocorrelation was tested with global Moran’s I. Integrated nested Laplace approximation models were fitted when significant spatial autocorrelation was detected.
Results: Participants with higher cancer prevention scores had a significant decrease in all-cause (relative risk 0.95; $95\%$ confidence interval 0.92, 0.99), all-cancer (0.93; 0.89, 0.97), upper aero-digestive tract cancer (0.87; 0.78, 0.97), and prostate cancer (0.81; 0.68, 0.94) mortality, compared to those with lower scores.
Conclusion: The inverse association between adherence to the WCRF/AICR recommendations and mortality points out the potential of the lifestyle recommendations to decrease mortality and especially the burden of cancer in Switzerland.
## Introduction
In 2019, cancer was the leading cause of death among men and second leading cause of death among women in Switzerland ($28.5\%$ and $22.5\%$ of all deaths, respectively) [1]. The main risk factors that contribute to cancer mortality, such as smoking, physical inactivity, and unhealthy diets, are modifiable lifestyle factors [2]. Of all cancer deaths, $30\%$–$50\%$ are assumed to be preventable through lifestyle modification [3]. Hence, there is a global interest in primary cancer prevention targeting modifiable risk factors.
In 2018, the World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR) published a revised version of their 2007 WCRF/AICR cancer prevention recommendations, providing updated guidelines on an overall healthy and cancer-protective lifestyle at the individual level [4]. The ten recommendations aimed on reducing the global burden of cancer [4]. Several studies investigated the association between adherence to the WCRF/AICR cancer prevention recommendations represented by an index and cancer risk [5, 6] or cancer mortality [7, 8] and reported the assumed inverse relationship. A higher index indicates greater concordance with the 2018 WCRF/AICR cancer prevention recommendations and therefore, is assumed to be associated with a healthier lifestyle and a decreased cancer risk.
In 2016, Lohse et al. observed an inverse association between adherence to the WCRF/AICR cancer prevention recommendations and cancer mortality for men living in Switzerland using data of the years 1977–1979 and 1983–1992 [9]. However, the study by Lohse et al. made use of rather old and crudely assessed data on diet and lifestyle [9]. To overcome this limitation, we examined the association between adherence to the WCRF/AICR recommendations and mortality in Switzerland using the first National Nutrition Survey, menuCH, which assessed dietary intake and lifestyle factors in a representative sample of the Swiss population in greater detail than previous studies [10].
## Methods
The current study was based on three data sets: the menuCH survey (2014–2015, $$n = 2057$$), the Swiss mortality data (2015–2019), and the *Swiss census* data (2015–2019) provided by the Federal Statistical Office (FSO). The data sets were linked at the district-level. The structure of this article followed the STROBE-nut guidelines [11].
## Study Design and Participants of menuCH
In $\frac{2014}{2015}$, the first National Nutrition Survey, the menuCH study ($$n = 2057$$), was conducted in ten centres across Switzerland. The survey was a cross-sectional population-based study with a target sample of 4,627,878 Swiss residents of 18–75 years of age. The study included one questionnaire [12] and two 24-hour dietary recalls (24HDR) of a representative sample of the Swiss population [10]. The target sample was representative for five age categories (18–29, 30–39, 40–49, 50–64, 65–75 years old), both sexes (male, female), the three main Swiss language regions (CH-German, CH-French, and CH-Italian), and the twelve most populous cantons of the seven major regions of Switzerland. Further details on the study recruitment have been published elsewhere [13].
## Data Collection in the menuCH Survey
Data from the menuCH participants were obtained by a questionnaire assessing lifestyle and sociodemographic factors and by two 24HDR, as described in previous studies (13–15). Briefly, in the self-administered questionnaire the following information were collected: participants’ sex (male, female), age (afterwards divided into eleven categories: 18–24, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–75 years old), language region (CH-German, CH-French, CH-Italian), nationality (Swiss, Non-Swiss, Swiss binational), civil status (single, married, divorced, other), education level (primary, secondary, tertiary), smoking status (never, former, current), physical activity (low, moderate, high; based on the short-form International Physical Activity Questionnaire (IPAQ) definitions [16]), and postal code.
Anthropometric factors were measured during the first 24HDR by trained personnel following standardized procedures [17]. Self-reported weight or height measurements were used for lactating and pregnant women and for participants, for whom a measurement was not possible. Subsequently, body mass index (BMI) was calculated and classified according to the World Health Organization as “underweight” (BMI <18.5 kg/m2), “normal weight” (18.5 kg/m2 ≤ BMI <25.0 kg/m2), “overweight” (25.0 kg/m2 ≤ BMI <30.0 kg/m2) or “obese” (BMI ≥30 kg/m2) [18].
Dietary data were collected during two non-consecutive 24HDR interviews. The interviews were performed by trained dietitians across all weekdays and seasons using the trilingual Swiss version (0.2014.02.27) of the software GloboDiet® (formerly EPIC-Soft®, International Agency for Research on Cancer (IARC), Lyon, France [19, 20]; adapted by the Federal Food Safety and Veterinary Office, Bern, Switzerland). The first interview was held on-site in one of the ten study centres, whereas the second interview was conducted two to 6 weeks later via telephone. For quantifying the consumed amounts, the participants were provided a book with about 60 actual household measures and 119 series of five to six graduated portion-sized pictures [13, 21]. After data collection, its quality was ensured by cleaning the data based on the IARC’s guidelines using an updated version of GloboDiet® (0.2015.09.28) [22]. In the end, obtained ingredients, foods, and recipes were matched to the most suited item found in the Swiss Food Composition Database [23] using the tool FoodCASE (Premotec GmbH, Winterthur, Switzerland). Assessment of dietary supplement intake was not included in the 24HDR, but was collected via self-administered questionnaire.
## Swiss Mortality and Census Data
To be consistent with the menuCH study, we used Swiss mortality and census data of the age range 18–75 years.
All-cause, all-cancer, and cancer-specific mortality based on the documented definitive cause of death were examined. Based on the existing evidence on health behaviour factors and their link to cancer prevention [2], the following cancer types encoded with the 10th revision of the International Classification of Diseases (ICD-10) [24] were investigated: all-cancer (ICD-10: C00-C97, D32-D33, and D37-D48), colorectal cancer (ICD-10: C18-C20), upper aero-digestive tract (UADT) cancer (including tissues and organs of the respiratory tract, upper part of the digestive tract, and the upper oesophagus; ICD-10: C00-C15 and C32), stomach cancer (ICD-10: C16), liver cancer (ICD-10: C22), pancreatic cancer (ICD-10: C25), breast cancer (only women; ICD-10: C50), and prostate cancer (only men; ICD-10: C61).
The mortality data provided by the FSO were linked to the dietary data by the place of residence of the participants using postal code information. Based on the indirect method, which uses the Swiss population as reference population, standardized mortality ratios (SMR) were calculated at the district-level. The SMR were standardized for sex, age, and year of death.
## 2018 WCRF/AICR Cancer Prevention Recommendations Score
In 2018, the WCRF and AICR published ten recommendations on cancer prevention [4], of which our study used the following seven to build an index: maintain a healthy weight, be physically active, eat a plant-based diet, and limit the consumption of fast-food, red and processed meat, sugar-sweetened drinks, and alcohol. In an additional sensitivity analysis, the recommendation on dietary supplement use was included. Our study excluded the recommendation on breastfeeding and the one pertinent to people with a previous cancer diagnosis, since the menuCH study did not provide enough information to determine the partial score of these two recommendations. To make studies more comparable and to consider interdependent effects of risk factors on cancer, Shams-White et al. [ 3, 25] defined an index reflecting adherence to the recommendations, to which each recommendation contributes equally. The WCRF/AICR score construction in menuCH was based on the index by Shams-White et al. [ 3, 25] and details were provided in a previous project (Karavasiloglou N et al., in review). Briefly, each of the seven included recommendations contributed equally to the final score. For each recommendation either 0, 0.5, or 1 point were assigned. Two equally weighted sub-recommendations were included for the healthy weight (based on BMI category and waist circumference) and the plant-based diet (based on fruits and vegetables intake and total fiber intake) recommendation each. Thus, for each of the latter two recommendations a partial score of 0.25 and 0.75 was possible, too [3, 25].
The physical activity recommendation was assessed using the short-form IPAQ. Dietary components were investigated as consumed average amount in grams per day (mean of the two 24HDR interviews). For the fast-food component, an adapted NOVA classification system was applied to categorize foods as ultra-processed or non-ultra-processed (Karavasiloglou N et al., in review). The NOVA classification was adapted to the cancer prevention recommendation score to ensure no double penalization for red and processed meat consumption and intake of sugar-sweetened drinks. Closer adherence to the WCRF/AICR cancer prevention recommendations lead to a higher score, indicating a healthier lifestyle. In the analyses, the score was used as a continuous variable, ranging from zero to seven points, and as a categorical variable with the following predefined three categories: low adherence (0-<3 points), moderate adherence (3-<5 points), and high adherence (5–7 points).
## Statistical Analyses
The menuCH participants’ characteristics and the SMR at the district-level were analysed descriptively.
To investigate the association between the cancer prevention score and all-cause and cancer mortality, Quasipoisson regression models, which are a generalized version of Poisson models, were fitted at the individual level. The outcome variable modeled by a Quasipoisson regression model is a count variable, that shows under- or overdispersion, i.e., the mean and the variance of the outcome data are not equal and therefore, the variance will be estimated by an additional parameter, the dispersion parameter. As the menuCH survey is not a longitudinal study and therefore lacks a mortality follow-up, an additional data set, the Swiss mortality data from 2015 until 2019, was used to determine the outcome variable. The Swiss mortality data were linked to the menuCH data geographically at the district level. The outcome variable was the observed number of deaths documented between 2015 and 2019 in the menuCH participant’s sex, age, and district group. The total number of residents in the corresponding sex, age, and district group was added as an offset term. The cancer prevention score was used as explanatory variable (either as continuous or categorical variable). The models were further adjusted for sex, age, smoking status, education level, language region, nationality, civil status, and mean daily energy intake (in kilocalories).
A sensitivity analysis excluding the physical activity recommendation due to a high percentage of missing observations was conducted. Furthermore, a second sensitivity analysis was performed, which included the seven previous recommendations and additionally the recommendation on dietary supplement use as a binary score (0 points if self-medicated intake; 1 point if prescribed by a doctor or no supplement intake).
Districts were specified as neighbouring based on a first order neighbourhood structure with rook contiguity. Furthermore, each neighbour district received a weight according to the inverse number of neighbours of the corresponding district. The global Moran’s I statistic was used to investigate the existence and degree of spatial autocorrelation [26]. A one-sided p-value based on the Z-score [27] and a one-sided p-value based on 1000 Monte Carlo (MC) simulations were calculated to check the evidence for a significant, non-random spatial pattern of the residuals aggregated at the district-level. Local Moran’s I were calculated and tested for significance based on a permutation test ($$n = 1000$$). The lower limit of the p-value is given by the number of simulations [28] and therefore, no correction for multiple testing was applied. Local indicators of spatial autocorrelation (LISA) cluster maps were used to present the local Moran’s I values. “ High-High” representing districts, which had a higher residual mean than the overall residual mean and a lagged value, which was higher than the overall mean lagged value. “ Low-Low” indicating districts, for which both values were lower than the corresponding average value. “ High-Low” representing districts, which had a higher residual mean than the overall mean and a lagged value, which was lower than the overall mean lagged value. “ Low-High” representing districts, for which the opposite was the case. Districts, which were not part of the menuCH study, and thus no data were available, are coloured in white.
An integrated nested Laplace approximation (INLA) model was fitted when evidence for spatial autocorrelation was observed. A Besag-York-Mollié model was chosen, which consists of a structured spatial component and an unstructured spatial component [29]. For both components, the default prior distribution (LogGamma with shape = 1 and rate = 0.00005) was chosen [29]. For each data set the results were pooled by taking the average of the estimates.
The menuCH participants’ data were weighted for sex, age, major living region in Switzerland, marital status, household size, and nationality. Variables on dietary factors were additionally weighted for weekday and season of the 24HDR interview day [30].
Some participants had missing values for the physical activity level ($$n = 524$$), the education level ($$n = 3$$), and smoking category ($$n = 4$$). Hence, to include all 2057 participants in the analyses, we conducted multivariate imputation by chained equations (MICE, $m = 25$). With each imputed data set the analyses were run separately. Afterwards, the results were pooled.
Analyses were conducted in GeoDa (version 1.14.0) and in the software R (version 4.1.0, R Foundation for Statistical Computing, Vienna, Austria [31]). The statistical significance level was set to 0.05 for all analyses.
## Results
In comparison to the overall study population, several differences in the menuCH participants’ characteristics were observed across the adherence categories (Table 1). Participants in the low adherence group were more likely to be 30 years of age or older, live in a CH-German language region, be of Swiss nationality only, have completed a secondary education, be current smokers, and have a higher daily energy intake. In contrast, participants with high adherence were more likely to be female, 18–29 years of age, live in the CH-French language region, be Swiss binational, be single, have completed a tertiary education, be never-smokers, and have a lower daily energy intake.
**TABLE 1**
| Variables | Overall (n = 2057) | Low adherence c (n = 227) | Moderate adherence c (n = 900) | High adherence c (n = 379) | NA (n = 551) |
| --- | --- | --- | --- | --- | --- |
| % | 100 | 11.1 | 45.5 | 17.9 | 25.5 |
| Women (%) | 50.2 | 34.1 | 43.3 | 67.5 | 57.3 |
| Age group (%) | | | | | |
| 18–29 years old | 18.8 | 13.0 | 17.7 | 23.2 | 20.2 |
| 30–44 years old | 29.9 | 31.3 | 30.9 | 26.3 | 29.9 |
| 45–59 years old | 29.8 | 32.6 | 30.7 | 28.6 | 27.8 |
| 60–75 years old | 21.6 | 23.1 | 20.7 | 21.9 | 22.1 |
| Language region d (%) | | | | | |
| German | 69.2 | 73.3 | 70.2 | 67.7 | 66.9 |
| French | 25.2 | 21.6 | 24.1 | 27.3 | 27.3 |
| Italian | 5.6 | 5.1 | 5.7 | 5.0 | 5.8 |
| Nation group (%) | | | | | |
| Swiss only | 61.4 | 67.4 | 59.4 | 60.8 | 62.9 |
| Non-Swiss | 24.8 | 21.6 | 26.5 | 24.8 | 23.1 |
| Swiss binational | 13.8 | 10.9 | 14.1 | 14.4 | 14.0 |
| Civil status (%) | | | | | |
| Single | 31.1 | 29.0 | 30.8 | 35.0 | 30.0 |
| Married | 52.2 | 53.0 | 54.0 | 48.0 | 51.7 |
| Divorced | 12.1 | 12.7 | 11.5 | 10.7 | 13.9 |
| Other | 4.4 | 5.3 | 3.7 | 6.3 | 3.9 |
| | 0.1 | 0.0 | 0.0 | 0.0 | 0.6 |
| Education level (%) | | | | | |
| Primary | 4.7 | 2.8 | 5.4 | 3.4 | 5.2 |
| Secondary | 42.6 | 48.4 | 39.4 | 36.1 | 50.3 |
| Tertiary | 52.6 | 48.8 | 55.3 | 60.5 | 44.0 |
| | 0.1 | 0.0 | 0.0 | 0.0 | 0.6 |
| Smoking (%) | | | | | |
| Never | 42.9 | 35.0 | 41.4 | 52.2 | 42.4 |
| Former | 33.6 | 34.1 | 32.9 | 34.0 | 34.5 |
| Current | 23.3 | 30.9 | 25.7 | 13.9 | 22.2 |
| | 0.2 | 0.0 | 0.0 | 0.0 | 0.8 |
| Daily energy intake [kcal] | 2130 (1711, 2600) | 2413 (1940, 2791) | 2163 (1758, 2623) | 1930 (1580, 2354) | 2042 (1637, 2597) |
From 2015 until 2019, 106,140 all-cause deaths, including 46,220 all-cancer, 4022 colorectal cancer, 2923 UADT cancer, 1525 stomach cancer, 2248 liver cancer, 3841 pancreatic cancer, 3818 breast cancer (among women), and 1913 prostate cancer deaths (among men) were reported. In Figure 1, the SMR for all-cause, all-cancer, and cancer-specific mortality are shown at the district-level. Higher all-cause and all-cancer SMR were mainly observed in the western region and lower SMR mainly in the central region of Switzerland. A clear distinction between the CH-German regions with low SMR and the CH-French and CH-Italian regions with high SMR was seen for liver cancer. For most of the individual cancer sites, no clear pattern was detectable.
**FIGURE 1:** *Standardized mortality ratios at the district-level (unweighted data number of districts = 143). Based on the indirect method using the Swiss population as reference population, mortality ratios standardized for sex, age, and year of death were computed. Breast cancer standardized mortality ratios (H) were calculated only for women. Prostate cancer standardized mortality ratios (I) were calculated only for men. For all other causes of death (A–G), the data of both sexes were included to calculate the standardized mortality ratios. Swiss mortality and Swiss census data. Switzerland. 2015–2019.*
Regarding the Quasipoisson regression models, the complete case analyses and the analyses based on imputed data revealed similar results, hence only the latter results are presented and used for subsequent analyses. When using the score as continuous variable, no statistically significant rate ratio was observed for any of the mortality outcomes (Table 2).
**TABLE 2**
| Mortality d | WCRF/AICR cancer prevention recommendations score | WCRF/AICR cancer prevention recommendations score.2 | WCRF/AICR cancer prevention recommendations score.3 | WCRF/AICR cancer prevention recommendations score.4 |
| --- | --- | --- | --- | --- |
| Mortality d | Continuous | Categorical c | Categorical c | Categorical c |
| Mortality d | Per 1-point increment RR (95% CI) | Low adherence (ref.) RR (95% CI) | Moderate adherence RR (95% CI) | High adherence RR (95% CI) |
| All-cause e , f | 1.00 (0.99, 1.01) | 1.00 - | 0.99 (0.96, 1.02) | 0.95 (0.92, 0.99) |
| All-cancer e , f | 0.99 (0.98, 1.00) | 1.00 - | 0.99 (0.95, 1.02) | 0.93 (0.89, 0.97) |
| UADT cancer e , f | 0.99 (0.96, 1.01) | 1.00 - | 0.93 (0.85, 1.01) | 0.87 (0.78, 0.97) |
| Stomach cancer e , f | 0.98 (0.94, 1.02) | 1.00 - | 0.93 (0.81, 1.05) | 0.85 (0.70, 1.00) |
| Colorectal cancer e , f | 0.99 (0.96, 1.01) | 1.00 - | 1.08 (1.00, 1.15) | 0.95 (0.87, 1.04) |
| Liver cancer e , f | 0.99 (0.96, 1.02) | 1.00 - | 1.04 (0.95, 1.13) | 0.92 (0.82, 1.03) |
| Pancreatic cancer e , f | 1.01 (0.99, 1.03) | 1.00 - | 1.03 (0.96, 1.10) | 1.00 (0.91, 1.08) |
| Breast cancer e , g | 0.98 (0.96, 1.00) | 1.00 - | 1.04 (0.96, 1.13) | 0.97 (0.88, 1.06) |
| Prostate cancer e , h | 0.98 (0.95, 1.02) | 1.00 - | 0.86 (0.77, 0.96) | 0.81 (0.68, 0.94) |
In the regression models using the score as categorical variable, there was evidence among participants to have a $5\%$ decrease in all-cause (RR = 0.95, $95\%$ CI: 0.92, 0.99), a $7\%$ decrease in all-cancer (RR = 0.93, $95\%$ CI: 0.89, 0.97), a $13\%$ decrease in UADT cancer (RR = 0.87, $95\%$ CI: 0.78, 0.97), and a $19\%$ decrease in prostate cancer (RR = 0.81, $95\%$ CI: 0.68, 0.94) mortality when comparing the high adherence with low adherence group. For prostate mortality, even participants with moderate adherence had a reduced mortality by $14\%$ compared to the low adherence group (RR = 0.86, $95\%$ CI: 0.77, 0.96). Evidence for an increase in colorectal cancer mortality by $8\%$ was observed in the moderate compared to the low adherence group (RR = 1.08, $95\%$ CI: 1.00, 1.15). Apart from the latter result, there was an overall tendency for a decrease in mortality in the moderate adherence group and an even stronger decrease in mortality in the high adherence group compared to the low adherence group.
The results of the sensitivity analyses can be found in Supplementary Tables S1, S2. Generally, sensitivity analyses showed an attenuation of the statistical significance of the results.
Supplementary Tables S3, S4 show the results of the global Moran’s I statistic based on the regression models including the score as continuous and as categorical variable, respectively. Both models revealed similar results.
Only the residuals of the regression model for liver cancer mortality revealed evidence for spatial autocorrelation at the district-level (expected global Moran’s I: −0.014; observed global Moran’s I: 0.143 (categorical score) and 0.152 (continuous score), respectively). In Figure 2, the districts with a significant local Moran’s I statistic for liver cancer mortality are visualised in a LISA cluster map, providing more detailed information on the statistically significant spatial pattern indicated by the global Moran’s I statistic. Independent of including the score as categorical or continuous variable in the regression model, four districts revealed significant evidence for spatial outliers or spatial clusters, indicating to be the core of a spatial pattern.
**FIGURE 2:** *Visualization of local Moran’s I for liver cancer mortality using a local indicators of spatial autocorrelation cluster map at the district-level (respectively. n = 143). “High-High” representing districts, which had a higher residual mean than the overall residual mean and a lagged value, which was higher than the overall mean lagged value. “Low-Low” indicating districts, for which both values were lower than the corresponding average value. “High-Low” representing districts, which had a higher residual mean than the overall mean and a lagged value, which was lower than the overall mean lagged value. “Low-High” representing districts, for which the opposite was the case. Districts, which were not part of the menuCH study, are coloured in white. A significance level of 0.05 with no correction for multiple testing was applied. Independent of including the score as categorical or continuous outcome variable in the regression model, the same four Swiss districts revealed significant evidence for spatial outliers or spatial clusters. menuCH and Swiss mortality and Swiss census data. Switzerland. 2014–2015 and 2015–2019.*
INLA models were fitted for liver cancer mortality (Table 3). The fixed effects of the INLA model were similar as the estimates of the Quasipoisson model for both the continuous and the categorical model. There was still no evidence for an association between the score and liver cancer mortality. The structured spatial component of the INLA model did not reveal evidence for any of the Swiss districts to have increased or decreased liver cancer mortality.
**TABLE 3**
| Mortality d | WCRF/AICR cancer prevention recommendations score | WCRF/AICR cancer prevention recommendations score.2 | WCRF/AICR cancer prevention recommendations score.3 | WCRF/AICR cancer prevention recommendations score.4 |
| --- | --- | --- | --- | --- |
| Mortality d | Continuous | Categorical c | Categorical c | Categorical c |
| Mortality d | Per 1-point increment RR (95% CI) | Low adherence (ref.) RR (95% CI) | Moderate adherence RR (95% CI) | High adherence RR (95% CI) |
| Liver cancer e , f | 0.98 (0.95, 1.01) | 1.00 - | 1.01 (0.92, 1.11) | 0.89 (0.80, 1.00) |
## Discussion
In this study, the association of the cancer prevention score with all-cause and cancer mortality in Switzerland was explored. Quasipoisson regression models revealed evidence for a decrease in all-cancer, UADT cancer, and prostate cancer mortality in the high compared to the low adherence group. Based on the global Moran’s I statistic, only the regression model for liver cancer mortality revealed evidence for spatial autocorrelation. However, the structured spatial component of the INLA model did not indicate evidence for any of the Swiss districts to have a significantly increased or decreased liver cancer mortality.
Our study observed lower all-cause mortality when comparing high to low adherence to the WCRF/AICR score, but did not find an inverse association on the continuous scale as compared to the systematic review by Solans et al. ( per 1-point increment: RR = 0.90; $95\%$ CI: 0.84, 0.96; $$n = 3$$) [32].
In the analyses of all-cancer mortality, we observed that the high compared to the low adherence group had a lower all-cancer mortality. The findings of the systematic review by Solans et al. on the latter association pointed towards the same direction (per 1-point increment: RR = 0.91; $95\%$ CI: 0.89, 0.92; $$n = 3$$) [32]. In addition, our results were in line with those stated in the Swiss study by Lohse et al. [ 9], who reported a significant decrease in all-cancer mortality by $7\%$ for each 1-point increment in the score (HR = 0.93; $95\%$ CI: 0.90, 0.95) [9]. However, the latter results are not directly comparable to ours, since Lohse et al. used an older version of the WCRF/AICR cancer prevention recommendations, included nine instead of seven recommendations, and used more crudely assessed *Swiss data* on diet and lifestyle [9].
The results of our study regarding colorectal cancer mortality were in contrast to the findings of Romaguera et al. [ 33], who reported a decrease in colorectal cancer mortality for the high adherence (men: 4–6 points; women: 5–7 points) compared to the low adherence group (men: 0–2 points; women: 0–3 points). Surprisingly, in our study, a borderline significant increase in colorectal cancer mortality was observed for the moderate compared to the low adherence group. The previous Swiss study by Lohse et al. [ 9] did not find significant evidence for an inverse association.
On the other hand, our study found evidence for a lower mortality from UADT and prostate cancer in the high compared to the low adherence group. These results are in line with the findings of Lohse et al. [ 9], who reported a decrease in UADT cancer mortality by $51\%$ (HR = 0.49; $95\%$ CI: 0.26, 0.92) and in prostate cancer mortality by $52\%$ (HR = 0.48; $95\%$ CI: 0.28, 0.82) when comparing high (5–9 points) with low adherence (0–3 points). In contrast, there was no association between adherence to WCRF/AICR recommendations and prostate cancer incidence in the systematic review by Solans et al. [ 32]. It might be that a healthy lifestyle is more strongly associated with aggressive disease and, hence, prostate cancer mortality than prostate cancer incidence [34].
Moreover, Lohse et al. observed in the high (5–9 points) compared to the low adherence group (0–3 points) a decrease in stomach cancer mortality by $66\%$ (HR = 0.34; $95\%$ CI: 0.14, 0.83) [9]. However, our study did not detect such an association.
As the study by Lohse et al. [ 9], our study did not find evidence for an inverse association of the score with liver, breast, and pancreatic cancer mortality. Nevertheless, an inverse association between adherence to the recommendations and liver and breast cancer risk has been reported in several previous studies [32, 35, 36]. Inverse associations with pancreatic cancer mortality have also recently been observed in an analysis of the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial [37].
The lack of significance or unexpected associations of the score with mortality in our study could be due to the lack of mortality data for menuCH participants, reverse causation inherent in the cross-sectional study design of the menuCH study, low numbers of observed deaths, and age differences across the adherence groups, leading to less observed deaths in groups with mainly younger participants.
The Swiss mortality data from the years 2015–2019 showed different geographical patterns across causes of death. The underlying causes for these various spatial patterns could be manifold, e.g., differences in diet culture [38], community social capital [39], or socioeconomic factors [39]. Based on the local Moran’s I statistic and the LISA cluster map, there was significant evidence in four districts to be spatial clusters or spatial outliers. The fixed effects estimates of the INLA model (Table 3) and the estimates of the Quasipoisson regression model for liver cancer mortality (Table 2) were alike. The structured spatial component of the INLA model did not highlight any Swiss district with a significant increase or decrease in liver cancer mortality.
In comparison to the results of the main analysis, the sensitivity analysis excluding the physical activity recommendation resulted in an attenuation of the significance of the investigated associations (Supplementary Table S1). The attenuated results indicate that the physical activity recommendation is an important component of the cancer mortality prevention index, as reported in previous studies [40, 41]. However, statistical power might have been reduced in this sensitivity analysis, since the sample size of the high adherence group was reduced by more than half after excluding the physical activity recommendation.
Furthermore, the second sensitivity analysis, including the recommendation on supplement use, lead to an attenuation of the inverse associations, such that they were no longer statistically significant (Supplementary Table S2). However, statistical power might have been reduced, since the sample size of the low adherence group was reduced by more than half in this sensitivity analysis after including the recommendation on supplement use.
Our study had several strengths. The menuCH weighting strategy allowed our final sample of 2057 participants for being representative of a target population of 4,627,878 Swiss residents. The 24HDR and the lifestyle questionnaire provided detailed information in order to operationalize seven of the ten 2018 WCRF/AICR cancer prevention recommendations for our main analyses and to follow the index by Shams-White et al. [ 3, 25] closely. For instance, we used an adapted NOVA ultra-processed food classification system to assess fast food consumption, used both measured BMI and waist circumference to calculate the healthy weight score, and distinguished the different types of meat to determine the score for red and processed meat consumption.
However, our study had some limitations. First, the menuCH survey is a cross-sectional study, which, by nature, is prone to reverse causation. Given the self-reported dietary data, assessed with two 24HDR, the possibility of recall bias leading to over- and underestimating of dietary intakes cannot be excluded. However, the assessment of consumed foods via two non-consecutive 24HDR using the software GloboDiet® has been shown to yield reliable estimates [19, 20]. Second, individual mortality data of the menuCH participants was not available. In our analyses we assumed that each menuCH participant was correctly assigned to their district and that the participants were representative for their district’s lifestyle characteristics. Last, even though our analyses were adjusted for several known confounders, residual confounding cannot be ruled out.
To conclude, using the menuCH data our study was able to overcome the main limitations of the Swiss study by Lohse et al. [ 9], and provide more up-to-date results using the latest version of the WCRF/AICR cancer prevention recommendations and more detailed dietary and lifestyle data. An inverse association of the cancer prevention score with all-cancer, UADT cancer, and prostate cancer mortality was observed, indicating the potential of the recommendations to decrease the burden of cancer in Switzerland. Significant spatial dependencies were detected only for liver cancer mortality. However, based on the structured spatial component of the INLA model, no evidence was seen for any of the Swiss districts to have a significantly increased or decreased liver cancer baseline mortality rate.
## Ethics Statement
The studies involving human participants were reviewed and approved by The Lead Ethics Committee in Lausanne (protocol $\frac{26}{13}$) approved the survey procedure, which was in accordance with the Declaration of Helsinki. The menuCH study has been registered as the international standard randomized controlled trial number (ISRCTN) 16778734. The patients/participants provided their written informed consent to participate in this study.
## Author Contributions
NK, GP, and SR designed the study and the research questions. SR received the grant for the study. FS, NK, JB, and GP prepared the data and conducted the analyses. FS, GP, and SR interpreted the results. FS wrote the manuscript. FS, NK, JB, GP, and SR critically reviewed the manuscript and approved the final version.
## Conflict of Interest
The authors declare that they do not have any conflicts of interest.
## Supplementary Material
The Supplementary Material for this article can be found online at: https://www.ssph-journal.org/articles/10.3389/ijph.2023.1605610/full#supplementary-material
## Abbreviations
WCRF, World Cancer Research Fund; AICR, American Institute for Cancer Research; 24HDR, 24-hour dietary recall; FSO, Federal Statistical Office; FSVO, Federal Food Safety and Veterinary Office; IPAQ, International Physical Activity Questionnaire; BMI, Body Mass Index; IARC, International Agency for Research on Cancer; ICD-10, 10th revision of the international classification of diseases; UADT, upper aero-digestive tract; SMR, standardized mortality ratio; MICE, multivariate imputation by chained equations; MC, Monte Carlo; LISA, local indicators of spatial autocorrelation; INLA, Integrated nested Laplace approximation; RR, rate ratio; SD, standard deviation; CI, confidence interval and credible interval, respectively; NA, missing values.
## References
1. **Specific Causes of Death**. (2021)
2. **Recommendations and Public Health and Policy Implications**. (2018)
3. Shams-White MM, Brockton NT, Mitrou P, Romaguera D, Brown S, Bender A. **Operationalizing the 2018 World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) Cancer Prevention Recommendations: a Standardized Scoring System**. *Nutrients* (2019) **11** 1572-13. DOI: 10.3390/nu11071572
4. **Cancer Prevention Recommendations - WCRF International**. (2018)
5. Kaluza J, Harris HR, Håkansson N, Wolk A. **Adherence to the WCRF/AICR 2018 Recommendations for Cancer Prevention and Risk of Cancer: Prospective Cohort Studies of Men and Women**. *Br J Cancer* (2020) **122** 1562-70. DOI: 10.1038/s41416-020-0806-x
6. Barrubés L, Babio N, Hernández-Alonso P, Toledo E, Sabio JBR, Estruch R. **Association between the 2018 WCRF/AICR and the Low-Risk Lifestyle Scores with Colorectal Cancer Risk in the Predimed Study**. *J Clin Med* (2020) **9** 1215-5. DOI: 10.3390/jcm9041215
7. Hastert TA, Beresford SAA, Sheppard L, White E. **Adherence to the WCRF/AICR Cancer Prevention Recommendations and Cancer-specific Mortality: Results from the Vitamins and Lifestyle (VITAL) Study**. *Cancer Causes and Control* (2014) **25** 541-52. DOI: 10.1007/s10552-014-0358-6
8. Vergnaud AC, Romaguera D, Peeters PH, van Gils CH, Chan DSM, Romieu I. **Adherence to the World Cancer Research Fund/American Institute for Cancer Research Guidelines and Risk of Death in Europe: Results from the European Prospective Investigation into Nutrition and Cancer Cohort Study**. *Am J Clin Nutr* (2013) **97** 1107-20. DOI: 10.3945/ajcn.112.049569
9. Lohse T, Faeh D, Bopp M, Rohrmann S. **Adherence to the Cancer Prevention Recommendations of the World Cancer Research Fund/American Institute for Cancer Research and Mortality: a Census-Linked Cohort**. *Am J Clin Nutr* (2016) **104** 678-85. DOI: 10.3945/ajcn.116.135020
10. Bochud M, Chatelan A, Blanco J-M, Beer-Borst S. **Anthropometric Characteristics and Indicators of Eating and Physical Activity Behaviors in the Swiss Adult Population - Results from menuCH 2014-2015**. (2017)
11. Lachat C, Hawwash D, Ocké MC, Berg C, Forsum E, Hörnell A. **Strengthening the Reporting of Observational Studies in Epidemiology – Nutritional Epidemiology (STROBE-Nut): An Extension of the STROBE Statement**. *Nutr Bull* (2016) **41** 240-51. DOI: 10.1111/nbu.12217
12. **menuCH - Fragebogen zum Ernährungs-und Bewegungsverhalten**. (2017)
13. Chatelan A, Beer-Borst S, Randriamiharisoa A, Pasquier J, Blanco JM, Siegenthaler S. **Major Differences in Diet across Three Linguistic Regions of Switzerland: Results from the First National Nutrition Survey menuCH**. *Nutrients* (2017) **9** 1163-17. DOI: 10.3390/nu9111163
14. Pestoni G, Krieger JP, Sych JM, Faeh D, Rohrmann S. **Cultural Differences in Diet and Determinants of Diet Quality in Switzerland: Results from the National Nutrition Survey menuCH**. *Nutrients* (2019) **11** 126-15. DOI: 10.3390/nu11010126
15. Krieger JP, Pestoni G, Cabaset S, Brombach C, Sych J, Schader C. **Dietary Patterns and Their Sociodemographic and Lifestyle Determinants in Switzerland: Results from the National Nutrition Survey menuCH**. *Nutrients* (2019) **11** 62-16. DOI: 10.3390/nu11010062
16. **Guidelines for Data Processing and Analysis of the International Physical Activity Questionnaire (IPAQ) - Short Form**. (2004)
17. **MONICA_Manual_Part III: Population Survey, Section 1: Population Survey Data Component**. (1998)
18. **WHO Technical Report Series - Obesity: Preventing and Managing the Global Epidemic**. (2000)
19. Crispim SP, de Vries JHM, Geelen A, Souverein OW, Hulshof PJM, Lafay L. **Two Non-consecutive 24h Recalls Using EPIC-Soft Software Are Sufficiently Valid for Comparing Protein and Potassium Intake between Five European Centres - Results from the European Food Consumption Validation (EFCOVAL) Study**. *Br J Nutr* (2011) **105** 447-58. DOI: 10.1017/S0007114510003648
20. Slimani N, Casagrande C, Nicolas G, Freisling H, Huybrechts I, Ocké MC. **The Standardized Computerized 24-h Dietary Recall Method EPIC-Soft Adapted for Pan-European Dietary Monitoring**. *Eur J Clin Nutr* (2011) **65** S5-15. DOI: 10.1038/ejcn.2011.83
21. Camenzind-Frey E, Zuberbühler C. **Schweizerisches Fotobuch/Livre Photo Suisse/Manuale Fotografico Svizzero**. (2014)
22. Crispim SP, Nicolas G, Casagrande C, Knaze V, Illner AK, Huybrechts I. **Quality Assurance of the International Computerised 24 H Dietary Recall Method (EPIC-Soft)**. *Br J Nutr* (2014) **111** 506-15. DOI: 10.1017/S0007114513002766
23. **Home - the Swiss Food Composition Database**. (2021)
24. World Health Organization. **ICD-10 Version: 2019 - International Statistical Classification of Diseases and Related Health Problems 10th Revision**. (2019)
25. Shams-White MM, Romaguera D, Mitrou P, Reedy J, Bender A, Brockton NT. **Further Guidance in Implementing the Standardized 2018 World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) Score**. *Cancer Epidemiol Biomark Prev* (2020) **29** 889-94. DOI: 10.1158/1055-9965.EPI-19-1444
26. Overmars KP, de Koning GHJ, Veldkamp A. **Spatial Autocorrelation in Multi-Scale Land Use Models**. *Ecol Modell* (2003) **164** 257-70. DOI: 10.1016/s0304-3800(03)00070-x
27. Al-Kindi KM, Alkharusi A, Alshukaili D, al Nasiri N, Al-Awadhi T, Charabi Y. **Spatiotemporal Assessment of COVID-19 Spread over Oman Using GIS Techniques**. *Earth Syst Environ* (2020) **4** 797-811. DOI: 10.1007/s41748-020-00194-2
28. Abrams AM, Kleinman K, Kulldorff M. **Gumbel Based P-Value Approximations for Spatial Scan Statistics**. *Int J Health Geogr* (2010) **9** 61-12. DOI: 10.1186/1476-072X-9-61
29. Blangiardo M, Cameletti M, Baio G, Rue H. **Spatial and Spatio-Temporal Models with R-INLA**. *Spat Spatiotemporal Epidemiol* (2013) **7** 33-49. DOI: 10.1016/j.sste.2012.12.001
30. Pasquier J, Chatelan A, Bochud M. **Weighting Strategy**. (2017)
31. **The R Project for Statistical Computing**. (2021)
32. Solans M, Chan DSM, Mitrou P, Norat T, Romaguera D. **A Systematic Review and Meta-Analysis of the 2007 WCRF/AICR Score in Relation to Cancer-Related Health Outcomes**. *Ann Oncol* (2020) **31** 352-68. DOI: 10.1016/j.annonc.2020.01.001
33. Romaguera D, Ward H, Wark PA, Vergnaud AC, Peeters PH, van Gils CH. **Pre-diagnostic Concordance with the WCRF/AICR Guidelines and Survival in European Colorectal Cancer Patients: a Cohort Study**. *BMC Med* (2015) **13** 107-12. DOI: 10.1186/s12916-015-0332-5
34. Romaguera D, Gracia-Lavedan E, Molinuevo A, de Batlle J, Mendez M, Moreno V. **Adherence to Nutrition-Based Cancer Prevention Guidelines and Breast, Prostate and Colorectal Cancer Risk in the MCC-Spain Case–Control Study**. *Int J Cancer* (2017) **141** 83-93. DOI: 10.1002/ijc.30722
35. Harris HR, Bergkvist L, Wolk A. **Adherence to the World Cancer Research Fund/American Institute for Cancer Research Recommendations and Breast Cancer Risk**. *Int J Cancer* (2016) **138** 2657-64. DOI: 10.1002/ijc.30015
36. Romaguera D, Vergnaud AC, Peeters PH, van Gils CH, Chan DSM, Ferrari P. **Is Concordance with World Cancer Research Fund/American Institute for Cancer Research Guidelines for Cancer Prevention Related to Subsequent Risk of Cancer? Results from the EPIC Study**. *Am J Clin Nutr* (2012) **96** 150-63. DOI: 10.3945/ajcn.111.031674
37. Zhang ZQ, Li QJ, Hao FB, Wu YQL, Liu S, Zhong GC. **Adherence to the 2018 World Cancer Research Fund/American Institute for Cancer Research Cancer Prevention Recommendations and Pancreatic Cancer Incidence and Mortality: A Prospective Cohort Study**. *Cancer Med* (2020) **9** 6843-53. DOI: 10.1002/cam4.3348
38. Pestoni G, Karavasiloglou N, Braun J, Krieger JP, Sych JM, Bopp M. **Does Diet Map with Mortality? Ecological Association of Dietary Patterns with Chronic Disease Mortality and its Spatial Dependence in Switzerland**. *Br J Nutr* (2022) **127** 1037-49. DOI: 10.1017/S0007114521001525
39. Zufferey J, Oris M. **Spatial Differentials in Mortality in Switzerland: How Do Contexts Explain the Differences between Natives and Migrants?**. *Espace Popul Soc* (2021) 1-22. DOI: 10.4000/eps.12203
40. Harnack L, Nicodemus K, Jacobs DR, Folsom AR. **An Evaluation of the Dietary Guidelines for Americans in Relation to Cancer Occurrence**. *Am J Clin Nutr* (2002) **76** 889-96. DOI: 10.1093/ajcn/76.4.889
41. Cerhan JR, Potter JD, Gilmore JME, Janney CA, Kushi LH, Lazovich D. **Adherence to the AICR Cancer Prevention Recommendations and Subsequent Morbidity and Mortality in the Iowa Women’s Health Study Cohort**. *Cancer Epidemiol Biomark Prev* (2004) **13** 1114-20. DOI: 10.1158/1055-9965.1114.13.7
|
---
title: Measurement method of tear meniscus height based on deep learning
authors:
- Cheng Wan
- Rongrong Hua
- Ping Guo
- Peijie Lin
- Jiantao Wang
- Weihua Yang
- Xiangqian Hong
journal: Frontiers in Medicine
year: 2023
pmcid: PMC9971000
doi: 10.3389/fmed.2023.1126754
license: CC BY 4.0
---
# Measurement method of tear meniscus height based on deep learning
## Abstract
Tear meniscus height (TMH) is an important reference parameter in the diagnosis of dry eye disease. However, most traditional methods of measuring TMH are manual or semi-automatic, which causes the measurement of TMH to be prone to the influence of subjective factors, time consuming, and laborious. To solve these problems, a segmentation algorithm based on deep learning and image processing was proposed to realize the automatic measurement of TMH. To accurately segment the tear meniscus region, the segmentation algorithm designed in this study is based on the DeepLabv3 architecture and combines the partial structure of the ResNet50, GoogleNet, and FCN networks for further improvements. A total of 305 ocular surface images were used in this study, which were divided into training and testing sets. The training set was used to train the network model, and the testing set was used to evaluate the model performance. In the experiment, for tear meniscus segmentation, the average intersection over union was 0.896, the dice coefficient was 0.884, and the sensitivity was 0.877. For the central ring of corneal projection ring segmentation, the average intersection over union was 0.932, the dice coefficient was 0.926, and the sensitivity was 0.947. According to the evaluation index comparison, the segmentation model used in this study was superior to the existing model. Finally, the measurement outcome of TMH of the testing set using the proposed method was compared with manual measurement results. All measurement results were directly compared via linear regression; the regression line was y0.98x−0.02, and the overall correlation coefficient was r20.94. Thus, the proposed method for measuring TMH in this paper is highly consistent with manual measurement and can realize the automatic measurement of TMH and assist clinicians in the diagnosis of dry eye disease.
## 1. Introduction
Dry eye disease (DED) is a multifactorial disease of the ocular surface that is accompanied by increased tear film osmolality and ocular surface inflammation, causing symptoms such as visual impairment and tear film instability [1, 2] and potential damage to the ocular surface, affecting the visual function of millions of people worldwide. In traditional diagnostic methods for DED, Schirmer’s test, tear break-up time measurement, and ocular surface staining score are commonly used to qualitatively and quantitatively analyze the tear film [3, 4]. Tear meniscus height (TMH) can be used to assess the tear volume and tear film status. The tear meniscus is located at the edge of the upper and lower eyelids and accounts for 75–$90\%$ of the total tear volume [5]. The lower tear meniscus is more stable, and the DED analysis mainly uses the lower tear meniscus index, which is also aimed at the lower tear meniscus. Previous studies have reported decreased tear meniscus parameters (TMH, tear meniscus volume, and tear meniscus dynamics) in DED patients (6–8). Therefore, the quantification of tear meniscus parameters is helpful in the diagnosis of DED. As a crucial parameter of the tear meniscus, the TMH has received extensive attention in recent years. In fact, in current clinical studies, although screening of the tear meniscus is performed by non-contact eye photography, the quantitative measurement of TMH is mostly manual or semi-automated. For example, physicians need to be involved in the assessment process of identifying and outlining the upper and lower edges of the tear meniscus in the image, and the measurement points of the TMH are empirically selected by the physician. These subjective assessments may lead to inconsistent results, reduced repeatability, and increased interobserver variability [9, 10]. Manual measurement of TMH is time consuming and laborious if a large number of images are involved.
As a crucial parameter of tear meniscus, TMH has received increasing attention in recent years, and screening for DED can be achieved by assessing TMH. Stegmann et al. [ 11] assessed TMH, tear meniscus area, tear meniscus depth, and tear meniscus radius using image data acquired by ultra-high resolution optical coherence tomography combined with conventional image processing algorithms. In 2019, Yang et al. [ 12] from the Human Research and Ethics Committee of Peking University Third Hospital implemented a brand-new automated tear meniscus segmentation and height measurement software ImageJ based on a multi-threshold segmentation algorithm and compared the ImageJ measurement results with the manual measurement results. Arita et al. [ 13] successfully segmented and measured the tear meniscus by interference fringes using a DR-1α tear interferometer and achieved high accuracy. However, this method is not fully automated and requires manual selection of the measurement point. In 2020, Stegmann et al. [ 14] improved the threshold-based segmentation algorithm to a convolutional neural network segmentation algorithm based on [11] and found that the use of deep learning segmentation algorithm increased the operation speed by 228 times compared with threshold segmentation algorithm. In 2021, researchers from the School of Biomedical Engineering, Department of Medicine, Shenzhen University [15] proposed a tear meniscus segmentation algorithm based on a fully convolutional neural network and combined with polynomial fitting of the upper and lower edges of the tear meniscus to measure TMH, and the measurement results of the TMH were compared with the manual measurement results. However, it was easy to deviate when the tear meniscus edge was fitted with the polynomial, and polynomial fitting was required for each measurement of a picture.
There are two main findings of this study. First, combined with the existing segmentation network structure and the characteristics of ocular surface images, a segmentation network suitable for this experiment is built to accurately segment the tear meniscus region and the central ring of the corneal projection ring (CCPR). Thereafter, combined with the image processing method, the center point of the CCPR is located, and the region that needs to be evaluated for the TMH is selected. Finally, the TMH measurement method is continuously adjusted, and the final measurement method is determined. The processes of ocular surface image acquisition, tear meniscus region segmentation, and TMH measurement are fully automatic and noninvasive [16, 17]. In addition, we compared the measured results of TMH using the method proposed in this study with those of experienced professional doctors to evaluate the feasibility of the proposed method.
## 2. Dataset
In total, 325 ocular surface images were obtained. All images were obtained from the Shenzhen Eye Hospital. All the data used in this experiment explained the purpose and possible results to the providers. All ocular surface images were acquired using a Keratograph 5 M (K5 M), during which the patients needed to place their chin on a stand in front of the K5 M, adjust the measured eye to a distance of 100 mm from the camera, face the camera, and remain still at the physician’s instructions. The images that caused unclear shooting due to unfocusing and closing eyes in 325 ocular surface images were manually eliminated. Finally, 305 clear images were obtained for the experiment. All images used for the experiment were in png format of 1,360 pixel × 1,024 pixel, as shown in Figure 1A. Furthermore, 305 ocular surface images were divided into training and testing sets, in which 270 ocular surface images were included in the training set, and 35 ocular surface images were included in the testing set. The training set is the data sample used for model fitting that performs gradient descent of training error during training and learns the trainable weight parameters. The testing set was a separate set of samples left during model training, which could be used to evaluate the performance of the model.
**FIGURE 1:** *Ocular surface image data (A) Original ocular surface image (B) Segmentation convolution neural network segmented tear meniscus region and central ring of corneal projection ring (CCPR) region (C) Schematic of selected TMH measurement region (D) Schematic of physician assessment of TMH.*
In the network training process, it is necessary to input the ocular surface image and its corresponding labeling image. All ocular surface images were labeled by a professional DED diagnostician. In the labeling process, time is not limited, and the edge of the tear meniscus region is accurately labeled to the extent possible. The labeled data are transformed into binary images, in which the labeled target region is represented by a pixel value of 1, and the background region is represented by a pixel value of 0, as shown in Figure 1B. To measure TMH, we selected several measuring points, as shown in Figure 1C, the physician selected several measuring points in the tear meniscus region near the right underneath corresponding to the center point of the CCPR, assessed TMH at these measuring points, and subsequently averaged them as the final TMH measurement, as shown in Figure 1D.
## 3. Materials and methods
In this study, the TMH was evaluated using the following steps: [1] The original ocular surface image and its corresponding tear meniscus region labeling mask in the training set were preprocessed, data augmentation was performed, the processed data were sent into the deep convolution neural network for network training, and the network parameters of the optimal model were saved. [ 2] The original ocular surface images and the CCPR in the training set were labeled with a mask for preprocessing and data augmentation. The processed data were fed into a deep convolution neural network for network training, and the network parameters of the optimal model were stored. [ 3] Load the network weights obtained in [1] and [2] to predict the tear meniscus region and the CCPR region of the ocular surface image in the testing set, respectively. [ 4] The center of CCPR was located. In this study, the center of the CCPR was located by considering the pixel coordinates and edge lengths of the upper left corner of the rectangle through an external rectangle of the predicted CCPR that requires circular fitting of the predicted CCPR to achieve a more accurate center position. [ 5] Prediction of TMH. The main flow of tear meniscus segmentation and height measurement is shown in Figure 2.
**FIGURE 2:** *Automatic measurement process of TMH.*
## 3.1. Segmentation model structure
To achieve segmentation of the tear meniscus region, we built a deep convolutional neural network based on the DeepLabv3 [18] architecture, which was initially used for semantic segmentation. To segment the tear meniscus region better, the DeepLabv3 network was adjusted and improved in this study. The segmentation of the tear meniscus region is performed through feature extraction and image size restoration to obtain the final segmentation results, which include the backbone module and ASPP [19] module. The backbone module used in this study refers to resnet50 [20] and renders certain improvements based on the characteristics of the tear meniscus image. The entire backbone consists of a 7 × 7 convolutional layer, maximum pooling layer, and four blocks. Each block consists of bottleneck1 and bottleneck2. Both bottleneck1 and bottleneck2 are residual blocks composed of several convolutional layers, linear normalization layers, rectified linear units, and shortcut branches. Bottlenboteck1 differs from bottleneck2 in that a convolutional kernel of 1 × 1 is added to the shortcut branch of bottleneck1 to reduce the dimension. The specific structures of bottleneck1 and bottleneck2 are shown in Figures 3B, C, respectively. Certain common convolutional layers in block3 and block4 are replaced by atrous convolutional layers [21], and the specific expansion coefficient setting is shown in Figure 3A. The ASPP module consists of five parallel branches, which are a convolutional layer of 1 × 1, three atrous convolutional layers of 3 × 3, and a global average pooling layer that can increase global context information (followed by a convolutional layer of 1 × 1, and subsequently, the size of the input is restored by bilinear interpolation); thereafter, the outputs of these five branches are concatenated along the channel direction, and finally the information is further fused by a convolutional layer of 1 × 1. In addition, for the three parallel atrous convolutional layers, we use the multi-grid strategy and experimentally found that in the experiments performed in this study, the best results are obtained when the multi-grid is set to [1, 1, 1]. The structure of the entire network is illustrated in Figure 3.
**FIGURE 3:** *Segmented network structure diagram (A) Segmented network structure is composed of resnet50 module, ASPP module, and upper sampling. Panels (B,C) are bottleneck1 and bottleneck2, respectively, all of which are composed of several convolutional layers, batch normalization and ReLU.*
In the process of feature extraction, the pooling and convolutional layers are generally used to increase the receptive field; however, this also reduces the size of the feature map. For segmentation, it is necessary to use upper sampling to restore the size of the feature map, and the process of feature map reduction and reamplification causes loss of accuracy. To solve this problem, the concept of atrous convolution, which can increase the receptive fields while maintaining the size of the feature maps is proposed. Atrous convolution introduces a hyperparameter called expansion rate, which defines the spacing of each value of the convolutional kernel when processing the data, as shown in Figure 4: (a) shows a common convolutional kernel with a size of 3 × 3; (b) shows an atrous convolution with a size of 5 × 5. The atrous convolutional kernel enlarges the size based on the ordinary convolutional kernel; however, the convolutional kernel unit that participates in the operation does not change; only the light blue square in the figure is the unit that participates in the operation, and the elements in the white square are filled with 0. Atrous convolution increases receptive fields by enlarging the size of the convolutional kernel, while neither increasing the computational load nor reducing the resolution of the feature map. The degree of expansion of the convolutional kernel can be controlled by the expansion factor, assuming that the expansion factor is S, the size of the common convolution kernel is K0, and the size of the convolution kernel after the expansion design is Kc, as follows:
**FIGURE 4:** *Convolutional kernel schematic: (A) Common convolution (B) atrous convolution.*
In the process of restoring the image size, bilinear interpolation is adopted in this study. To further optimize the segmentation performance of the network, this study refers to the auxiliary classifier structure in GoogleNet [22] and the FCN [23] network structure, and the output of backbone’s block3 in the model leads to an FCN head as an auxiliary output.
At present, many segmentation networks are based on the improvement of Unet [24], which has also been widely used in the field of biomedical image segmentation, and this method was proposed at the MICCAI meeting in 2015 and has now reached more than 4000 citations. Unet is characterized by an encoder–decoder structure, and through the convolutional layer and pooling layer, the input picture information is encoded into the feature information that can be recognized by the computer, and subsequently, the compressed feature map is also sampled. Compared with the previous segmentation network, Unet fuses more low-level semantic information starting from the first convolutional layer, and the output feature maps of each layer are copied and concatenated with the decoded information after the subsequent upper sampling to generate a new feature map. In the last layer of the network, Unet uses a convolutional layer of 1 × 1 instead of the fully connected layer and uses a convolutional kernel of 1 × 1 to achieve dimensionality reduction, which is a linear transformation and superposition of the combination of information between different channels. Unet plays an important role in the segmentation of medical images owing to their lightweight network structure and feature concatenation. Therefore, to further evaluate the performance of the model used in this study, the Unet series network is selected as the experimental contrast model.
## 3.2. TMH measurement method
When assessing TMH, a professional doctor mainly evaluates the height of the tear meniscus region corresponding to the vicinity directly below the center of the CCPR, that is, several measurement points are selected in the tear meniscus region directly below the center of the CCPR for TMH assessment, and the average value of the assessment is the final TMH measurement result. Because each assessment requires the physician to select the assessment point, not only is it time consuming and labor intensive, but also the assessment results are susceptible to subjective factors. After consultation with experts in the field of DED diagnosis, this study used the method of averaging multiple measurement points, which is realized as follows: the pixel coordinate of the center of the CCPR is (x,y), the pixel coordinate corresponding to the upper edge of the tear meniscus is (xi,yi), and the pixel coordinates corresponding to the lower edge of the tear meniscus is (x−i−1,y−i−1). The pixel sets of |xi−x|<=100 and |x−i−1−x|<=100 are calculated. Because the edge of the tear meniscus includes the upper and lower edges, 400 pixel coordinates can be obtained. Referring to the opinions given by professional doctors, this study selects an upper tear river coordinate value(xj,yj) and its corresponding tear river coordinate value (x−j−1,y−j−1) every 30 pixels to calculate the TMH. In addition, a previous study [15] showed that TMH in the tear meniscus region 0.5–4 mm directly below the center of the CCPR has strong robustness, and the TMH value in this region is insensitive to the selected measurement points. Therefore, a total of seven TMH measurement points were selected in the tear meniscus region 2 mm directly below the center of the CCPR, and the pixel values corresponding to these seven TMHs were averaged to obtain the final TMH, as shown in Equation 2. The height and width of all pictures used during the experiment were measured and averaged, and this step was repeated three times to finally obtain the height and width of the ocular surface image as 11.85 mm and 15.75 mm, respectively, and subsequently, the pixel value could be converted to a height value by the conversion formula of Equation 3, as shown in Figure 1C.
## 3.3. Criteria for model evaluation
The evaluation index is a key factor for measuring network performance, and tear meniscus segmentation is of practical significance only if the evaluation index meets the expectations [25]. In the field of image segmentation, many evaluation indices can describe the network segmentation accuracy. Among them, the commonly used evaluation indicators are the intersection over union (IOU), dice coefficient, intra class coefficient (ICC) and sensitivity (26–28). We recorded the target region in label A1 and the prediction of the target region as A2.
[1]The IOU describes the ratio between the intersection and merging of the real and predicted results, and the closer the ratio is to 1, the higher the coincidence degree of the two. IOU is calculated as [2] The dice coefficient describes how similar the two samples are, and the closer the two samples are, the closer the dice coefficient is to 1. The dice coefficient is calculated as follows: [3] ICC = (variance of interest)/(total variance) = (variance of interest)/(variance of interest + unwanted variance), ICC can be used to evaluate the segmentation done by the models and the observers. I would like to use ICC to evaluate my model proposed in the paper compared with doctors in measuring TMH. The ICC ranges from 0 to 1, a high ICC close to 1 indicates high reliability of the model.
[4] In addition to the IOU and dice coefficient evaluation index, the network performance can be measured using sensitivity. TP: correctly predicted as tear meniscus/CCPR; FN: incorrectly predicted background as tear meniscus/CCPR; FP: incorrectly predicted as the background of tear meniscus/CCPR; TN: correctly predicted background.
Sensitivity refers to the ratio of the predicted correct region to the predicted total region in the prediction result, i.e., the accurate measurement of the network segmentation, and its calculation formula is as follows:
## 3.4. Optimizer and learning rate updating strategy
The optimizer provides a direction for adjusting the neural network parameters in deep learning, which causes the loss function to approach the global minimum continuously and determine the global optimal solution. According to the different tasks, selecting the appropriate optimizer to optimize the parameters is necessary; otherwise, the loss function may remain in the local optimal solution, resulting in non-convergence of the network. In this study, the stochastic gradient descent algorithm [29] was used as the optimizer. The gradient is the vector pointing to the maximum value of the derivative of a function in the direction of a certain function at this point, that is, along this direction, the fastest change in the function value. Let the mean-square loss function be where θ(w1,w2,w3,….,wn) is the weight vector, and a partial derivative of each component is determined using function J(θ) to obtain the gradient g = J′(θ); accordingly, the updated θ at the next moment is whereθt is the last weight, θt+ 1is the updated weight, andα is the learning rate that determines the step size for each parameter update.
The optimizer controls the direction of the parameter optimization update, whereas the learning rate controls the speed of the parameter optimization update. Generally, the learning rate decreases with the number of iterations. At the beginning of training the network, a larger learning rate can be set to allow the network to swiftly adapt to the training samples. When training to a certain extent, reducing the learning rate is necessary, which finely adjusts the network parameters and avoids the network from shaking. In this study, we used the cosine annealing [30] strategy to update the learning rate. The learning rate decreases in the form of a cosine function. According to the characteristics of the cosine function, learning first gradually decreases, subsequently accelerates the decline, and finally decreases slowly. The learning rate decay formula is Lrt refers to the current learning rate, Lrmax and Lrmin refer to the maximum learning rate and minimum learning rate that we set in advance, and Ncur and Nmax refer to the current iteration times and total iteration times, respectively.
## 4. Results
The designed deep convolutional neural network can accurately segment the tear meniscus region and CCPR in the ocular surface image, and the TMH can be evaluated using the segmentation results. First, the center of CCPR must be determined. To locate the center of the CCPR more accurately, we performed circular fitting and cavity filling of the CCPR followed by an external rectangle (actually a square). The center of the CCPR can be located by the pixel value and edge length of the left upper vertex of the rectangle. The upper and lower edges of the tear meniscus were obtained using edge detection.
## 4.1. Segmentation of target region
Each tear meniscus ocular surface image and its corresponding mask were uniformly cropped to 480 × 480, and subsequently, the image contrast was enhanced by the HSV random enhancement method. Data augmentation was achieved using rotation, translation, and inversion [31, 32] to improve the generalization ability of the model. Experimental hardware configuration during training and testing: Intel (R) Core (TM) i7-6700 CPU @ 3.40GHz, GPUNVIDIA GeForce RTX 1080. Experimental software configuration: The operating system was Windows10 with 64 bits, PyCharm Community Edition 2021.3, Python 3.6.13. All deep convolutional segmentation neural networks were set as follows: [1] SGD (momentum = 0.9, weight decay = 0.0001) was selected by the optimizer, and the learning rate was set as 0.0001. [ 2] The batch size was set to 4, and the maximum training epoch was set to 500. [ 3] A learning rate updating strategy was applied during the experiment. This strategy allows the learning rate to update every step instead of every epoch update, such that the network can be trained more effectively. After building the experimental environment and setting the initial learning parameters, different models are trained using the training set, and different models are used to segment the tear meniscus. The change in the loss value during the training process is shown in Figure 5. After training for 500 epochs for the training set, the loss value of the network built in this study decreased the fastest and dropped below 0.1.
**FIGURE 5:** *Loss of training process for tear meniscus segmentation by different models.*
The confusion matrix of tear meniscus segmentation and CCPR segmentation used proposed network in this study showed in Figures 6, 7, respectively. To evaluate the segmentation performance of the deep convolutional segmentation neural network used in this study, we used the popular segmentation networks Unet, Resnet50-Unet, and DeepLabv3 to perform comparative experiments. In this study, the segmentation performance of different models for the tear meniscus region and CCPR were evaluated using IOU, dice coefficient, and sensitivity. The segmentation results of different models for the tear meniscus region are summarized in Table 1, with an average IOU of 0.896, dice coefficient of 0.884, and sensitivity of 0.887. The segmentation performance of our model was the best when the tear meniscus region was segmented. The segmentation results of the different models for the CCPR are summarized in Table 2, where the average IOU was 0.932, the dice coefficient was 0.926, and the sensitivity was 0.947. When the CCPR is segmented, the model segmentation effect is optimal. In summary, the proposed model can accurately segment the tear meniscus region and CCPR, which is conducive to the accurate measurement of the TMH.
**FIGURE 6:** *Confusion matrix of tear meniscus segmentation.* **FIGURE 7:** *Confusion matrix of CCPR segmentation.* TABLE_PLACEHOLDER:TABLE 1 TABLE_PLACEHOLDER:TABLE 2
## 4.2. TMH measurement
The trained segmentation network combined with the image processing method was used to measure the TMH, and three ocular surface pictures in the testing set were selected, as shown in Figure 8, to show the prediction results of TMH. We outlined the labeled tear meniscus region and the predicted tear meniscus region with red lines and green lines, respectively. Figure 8A shows that the tear meniscus region is accurately segmented, and both the true and predicted value of TMH are 0.43 mm. Figure 8B shows that the tear meniscus region is under segmented, the true value of TMH is 0.31 mm, and the predicted value is 0.27 mm, which results in the predicted value of TMH being less than the true value. Figure 8C shows that the tear meniscus region is over segmented, the true value of TMH is 0.18 mm, and the predicted value is 0.22 mm, which results in the predicted value of TMH being greater than the true value. The true and predicted values in all testing sets were compared using linear regression, as shown in Figure 9, where the true and predicted values of TMH in all images in the testing set remained consistent, with a satisfactory regression line y0.98x−0.02, and the overall correlation coefficient was r20.94. The ICC is used to evaluated the reliability of proposed model in the study, the ICC of TMH was 0.90, which showed good reliability.
**FIGURE 8:** *Example of tear meniscus segmentation; TMH-GT indicates the TMH of label, TMH-P indicates the TMH predicted in this paper: Panel (A) is accurate segmentation (B) is under segmentation (C) is over segmentation.* **FIGURE 9:** *Fitted linear regression of true value and predicted value of TMH; ordinate represents the true value of TMH, abscissa represents the predicted value of TMH, and the blue line corresponds to regression line y=0.98x−0.02,r2=0.94.*
## 5. Discussion
DED is one of the most common ocular diseases affecting visual function in 5–$30\%$ of the world’s population [33]. DED causes a series of subjective symptoms and visual damage due to tear film instability accompanied by potential ocular surface damage. As the incidence of DED increases, it affects the visual quality of patients and thus affects their daily life; therefore, the evaluation of visual quality of DED patients has gradually received considerable attention. However, there are no uniform criteria for the diagnosis of DED, and fluorescein tear break-up time [34] and Schirmer test [35] are generally used to diagnose DED; however, these traditional diagnostic methods are invasive and unrepeatable (36–38). Studies have found that TMH is an important parameter of tear meniscus, and its value can be used to distinguish normal eyes from eyes affected by DED [39, 40]; nonetheless, most of the measurements of TMH are manual or semi-automatic; for example, professionals are required to outline the upper and lower edges of the tear meniscus and select the measurement point of TMH, which is not only time consuming and laborious, but also the measured TMH is unrepeatable, which may lead to inaccurate diagnostic results. Therefore, it is highly important to design a fully automatic, noninvasive method for measuring TMH. Based on this, a method for measuring TMH is proposed in this study, in combination with deep learning and image processing methods. The acquisition of ocular surface images, detection of the tear meniscus region, and measurement of TMH are automatic and noninvasive. The prediction results of the TMH were consistent with the measurement results of professional doctors. The method proposed in this study can accurately measure TMH and can be used to assist doctors in the diagnosis of DED, which has important clinical and practical significance.
In this study, we first obtained the ocular surface image using K 5M equipment and eliminated the blurred image of the tear meniscus region caused by closing the eyes and not focusing during shooting. Subsequently, a deep convolutional neural network was built to segment the tear meniscus and CCPR regions. The segmentation network used in this study included two parts: feature extraction and image size restoration. The feature extraction part is composed of the adjusted Resnet50 and ASPP module in DeepLabv3. Segmented image size restoration was realized by bilinear interpolation sampling. In addition, an auxiliary output was elicited at the feature extraction stage by referring to the auxiliary output structure of GoogleNet and the output of the FCN. Finally, circular fitting was performed on the segmented CCPR to better locate its central point, and subsequently, the upper and lower edges of the tear meniscus were detected by edge detection to achieve the TMH, and after measuring the pixel values corresponding to the TMH, the final TMH values were obtained by 86 pixel/mm. The model used in this study exhibits an average IOU of 0.896, dice coefficient of 0.884, and sensitivity of 0.887 for the segmentation of the tear meniscus region and an average IOU of 0.932, dice coefficient of 0.926, and sensitivity of 0.947 for the segmentation of the CCPR region. The model built in this study can automatically identify and accurately segment the tear meniscus region and the CCPR region. A trained deep convolutional neural network was used to segment the ocular surface images in the testing set and predict TMH in combination with image processing methods; the regression line $y = 0.98$x−0.02 (r2=0.94) was used to fit the true and predicted values of the TMH in the testing set.
The method proposed in this study for measuring TMH is advantageous because it is noninvasive and fully automatic. The ocular surface images used to assess the TMH were obtained by a professional K5M shooting instrument without touching the patient’s eyes throughout the procedure. Segmentation of the tear meniscus region and measurement of the TMH were achieved by a computer, the measurement method of TMH was easily implementable. Furthermore, the amount of calculation was small, and the physician was not mandated to select TMH measurement points, which eliminated the problem of inconsistent results owing to subjective assessments, reduced repeatability, and increased interobserver variability. The method proposed in this study can accurately measure TMH and assist doctors in DED screening.
The shortcomings of this study are that the number of datasets is small, and the quality is uneven, and certain images with eyes closed and blurred shooting are extant, necessitating continued collection of more high-quality images. The more datasets used to train the network, the more accurate the segmentation results of the network, such that a more accurate TMH is measured. As for segmentation network, with the deepening of the convolutional layer, the obtained feature map has a larger field of view, in which the shallow network focuses on texture features and the deep network focuses on the overall information of the picture. When pooling down sampling, it inevitably loses part of the edge information of the features, and this lost information cannot be recovered by upsampling alone, whereas the Unet network achieves the retrieval of edge features through the concatenation of features, which significantly improves the segmentation fineness. In the future, we can combine different networks, such as the DeepLab series and the Unet series, to improve and further improve the accuracy of segmentation.
## 6. Conclusion
In this paper, we propose a method to automatically measure TMH using deep learning combined with image processing. The measurement results of TMH obtained using the method proposed in this paper are consistent with clinical data, and this is clinically significant. In the future, with the continuous development and optimization of algorithms and the acquisition of more high-quality datasets, the accuracy of the measurement of TMH will increase, and it can be used to screen DED.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and the institutional requirements. Written informed consent from the patients was not required to participate in this study in accordance with the national legislation and the institutional requirements.
## Author contributions
CW and RH acquired, analyzed, discussed the data and drafted the manuscript. PG and PL analyzed and discussed the data. JW, WY, and XH acquired the clinical information and revised the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Bron AJ, dePaiva CS, Chauhan SK, Bonini S, Gabison EE, Jain S. **TFOS DEWS II pathophysiology report.**. (2017.0) **15** 438-510. PMID: 28736340
2. Koh S, Ikeda C, Fujimoto H, Oie Y, Soma T, Maeda N. **Regional differences in tear film stability and meibomian glands in patients with aqueous-deficient dry eye.**. (2015.0) **42** 250-5. DOI: 10.1097/ICL.0000000000000191
3. Bron A. **Diagnosis of dry eye.**. (2001.0) **45** S221-6. DOI: 10.1016/s0039-6257(00)00201-0
4. Storås AM, Riegler MA, Grauslund J, Hammer H, Yazidi A, Halvorsen P. **Artificial intelligence in dry eye disease.**. (2022.0) **23** 74-86. PMID: 34843999
5. Holly F. **Physical chemistry of the normal and disordered tear film.**. (1985.0) **104 (Pt. 4)** 374-80. PMID: 3862270
6. Goto E, Matsumoto Y, Kamoi M, Endo K, Ishida R, Dogru M. **Tear evaporation rates in Sjögren syndrome and non-Sjögren dry eye patients.**. (2007.0) **144** 81-5. DOI: 10.1016/j.ajo.2007.03.055
7. Chen F, Shen M, Chen W, Wang J, Li M, Yuan Y. **Tear meniscus volume in dry eye after punctal occlusion.**. (2010.0) **51** 1965-9. DOI: 10.1167/iovs.09-4349
8. Tsubota K. **Tear dynamics and dry eye.**. (1998.0) **17** 565-96. DOI: 10.1016/s1350-9462(98)00004-4
9. García-Montero M, Rico-Del-Viejo L, Lorente-Velázquez A, Martínez-Alberquilla I, Hernández-Verdejo J, Madrid-Costa D. **Repeatability of noninvasive keratograph 5M measurements associated with contact lens wear.**. (2019.0) **45** 377-81. DOI: 10.1097/ICL.0000000000000596
10. Hao Y, Tian L, Cao K, Jie Y. **Repeatability and reproducibility of SMTubeMeasurement in dry eye disease patients.**. (2021.0) **2021**. DOI: 10.1155/2021/1589378
11. Stegmann H, Aranha Dos Santos V, Messner A, Unterhuber A, Schmidl D, Garhöfer G. **Automatic assessment of tear film and tear meniscus parameters in healthy subjects using ultrahigh-resolution optical coherence tomography.**. (2019.0) **10** 2744-56. DOI: 10.1364/BOE.10.002744
12. Yang J, Zhu X, Liu Y, Jiang X, Fu J, Ren X. **TMIS: a new image-based software application for the measurement of tear meniscus height.**. (2019.0) **97** e973-80. DOI: 10.1111/aos.14107
13. Arita R, Yabusaki K, Hirono T, Yamauchi T, Ichihashi T, Fukuoka S. **Automated measurement of tear meniscus height with the Kowa DR-1α tear interferometerin both healthy subjects and dry eye patients.**. (2019.0) **60** 2092-101. DOI: 10.1167/iovs.18-24850
14. Stegmann H, Werkmeister R, Pfister M, Garhofer G, Schmetterer L, dos Santos V. **Deep learn- ing segmentation for optical coherence tomography measurements of the lowertear meniscus.**. (2020.0) **11** 1539-54. DOI: 10.1364/BOE.386228
15. Deng X, Tian L, Liu Z, Zhou Y, Jie Y. **A deep learning approach for the quantification of lower tear meniscus height.**. (2021.0) **68**
16. Hong J, Sun X, Wei A, Cui X, Li Y, Qian T.. **Assessment of tear film stability in dry eye with a newly developed keratograph.**. (2013.0) **32** 716-21. PMID: 23132457
17. Arriola-Villalobos P, Fernández-Vigo JI, Díaz-Valle D, Peraza-Nieves JE, Fernández-Pérez C, Benítez-Del-Castillo JM. **Assessment of lower tear meniscus measurements obtained with Keratograph and agreement with Fourier-domain optical-coherence tomography.**. (2015.0) **99** 1120-5. PMID: 25677670
18. Chen LC, Papandreou G, Schroff F, Adam H. **Rethinking atrous convolution for semantic image segmentation.**. (2017.0). DOI: 10.48550/arXiv.1706.05587
19. Chen LC, Papandreou G, Kokkinos I, Murphy K, Yuille AL. **Semantic image segmentation with deep convolutional nets and fully connected CRFs.**. (2016.0). DOI: 10.48550/arXiv.1412.7062
20. He K, Zhang X, Ren S, Sun J. **Deep residual learning for image recognition.**. (2016.0) 770-8
21. Wang P, Chen P, Yuan Y, Liu D, Huang Z, Hou X. **Understanding convolution for semantic segmentation.**. (2018.0). DOI: 10.48550/arXiv.1702.08502
22. Szegedy C, Liu W, Jia Y, Sermanet P, Reed S, Anguelov D. **Going deeper with convolutions.**. 1-9
23. Shelhamer E, Long J, Darrell T. **Fully convolutional networks for semantic segmentation.**. (2017.0) **39** 640-51. PMID: 27244717
24. Ronneberger O, Fischer P, Brox T. **U-net: convolutional networks for biomedical image segmentation.**. (2015.0). DOI: 10.48550/arXiv.1505.04597
25. Niedernolte B, Trunk L, Wolffsohn J, Pult H, Bandlitz S. **Evaluation of tear meniscus height using different clinical methods.**. (2021.0) **104** 583-8. DOI: 10.1080/08164622.2021.1878854
26. Wan C, Shao Y, Wang C, Jing J, Yang W. **A novel system for measuring Pterygium’s progress using deep learning.**. (2022.0) **9**. DOI: 10.3389/fmed.2022.819971
27. Wan C, Wu J, Li H, Yan Z, Wang C, Jiang Q. **Optimized-Unet: novel algorithm for parapapillary atrophy segmentation.**. (2021.0) **15**. DOI: 10.3389/fnins.2021.758887
28. Wan C, Chen Y, Li H, Zheng B, Chen N, Yang W. **EAD-net: a novel lesion segmentation method in diabetic retinopathy using neural networks.**. (2021.0) **2021**. DOI: 10.1155/2021/6482665
29. Zhang Z, Hu Y, Ye Q.. (2022.0) 6-11
30. Wang Z, Hou F, Qiu Y, Ma Z, Singh S, Wang R. (2022.0) 3859-63
31. Tang Z, Zhang X, Yang G, Zhang G, Gong Y, Zhao K. **Automated segmentation of retinal nonperfusion area in fluorescein angiography in retinal vein occlusion using convolutional neural networks.**. (2020.0) **2** 648-58. DOI: 10.1002/mp.14640
32. Xu J, Zhou Y, Wei Q, Li K, Li Z, Yu T. **Three-dimensional diabetic macular edema thickness maps based on fluid segmentation and fovea detection using deep learning.**. (2022.0) **3** 495-501. DOI: 10.18240/ijo.2022.03.19
33. Smith JA, Albeitz J, Begley C, Caffery B, Nichols K, Schaumberg D. **The epidemiology of dry eye disease: report of the epidemiology subcommittee of the international dry eye workshop**. (2007.0) **5** 93-107. PMID: 17508117
34. Bron AJ, Smith JA, Calonge M. **Methodologies to diagnose and monitor dry eye disease: report of the diagnostic methodology subcommittee of the international dry eye workshop.**. (2007.0) **5pp** 108-52. DOI: 10.1016/s1542-0124(12)70083-6
35. Phadatare SP, Momin M, Nighojkar P, Askarkar S, Singh KK. **A comprehensive review on dry eye disease: diagnosis, medical management, recent developments, and future challenges.**. (2015.0) **704946**
36. Nichols KK, Mitchell GL, Zadnik K. **The repeatability of clinical measurements of dry eye.**. (2004.0) **23** 272-85. PMID: 15084861
37. Johnson ME, Murphy PJ. **The agreement and repeatability of tear meniscus height measurement methods.**. (2005.0) **82** 1030-7. PMID: 16357644
38. Koh S, Ikeda C, Watanabe S, Oie Y, Soma T, Watanabe H. **Effect of non-invasive tear stability assessment on tear meniscus height.**. (2015.0) **93** e135-9. PMID: 25308575
39. Burkat CN, Lucarelli MJ. **Tear meniscus level as an indicator of nasolacrimal obstruction.**. (2005.0) **112** 344-8. PMID: 15691573
40. Tung C, Perin A, Gumus K, Pflugfelder S. **Tear meniscus dimensions in tear dysfunction and their correlation with clinical parameters.**. (2014.0) **157** 301-10.e1. DOI: 10.1016/j.ajo.2013.09.024
|
---
title: Inhibition of diacylglycerol lipase β modulates lipid and endocannabinoid levels
in the ex vivo human placenta
authors:
- Natascha Berger
- Tom van der Wel
- Birgit Hirschmugl
- Thomas Baernthaler
- Juergen Gindlhuber
- Nermeen Fawzy
- Thomas Eichmann
- Ruth Birner-Gruenberger
- Robert Zimmermann
- Mario van der Stelt
- Christian Wadsack
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9971001
doi: 10.3389/fendo.2023.1092024
license: CC BY 4.0
---
# Inhibition of diacylglycerol lipase β modulates lipid and endocannabinoid levels in the ex vivo human placenta
## Abstract
### Introduction
Lipids and fatty acids are key components in metabolic processes of the human placenta, thereby contributing to the development of the fetus. Placental dyslipidemia and aberrant activity of lipases have been linked to diverse pregnancy associated complications, such as preeclampsia and preterm birth. The serine hydrolases, diacylglycerol lipase α and β (DAGLα, DAGLβ) catalyze the degradation of diacylglycerols, leading to the formation of monoacylglycerols (MAG), including one main endocannabinoid 2-arachidonoylglycerol (2-AG). The major role of DAGL in the biosynthesis of 2-AG is evident from various studies in mice but has not been investigated in the human placenta. Here, we report the use of the small molecule inhibitor DH376, in combination with the ex vivo placental perfusion system, activity-based protein profiling (ABPP) and lipidomics, to determine the impact of acute DAGL inhibition on placental lipid networks.
### Methods
DAGLα and DAGLβ mRNA expression was detected by RT-qPCR and in situ hybridization in term placentas. Immunohistochemistry staining for CK7, CD163 and VWF was applied to localize DAGLβ transcripts to different cell types of the placenta. DAGLβ activity was determined by in- gel and MS-based activity-based protein profiling (ABPP) and validated by addition of the enzyme inhibitors LEI-105 and DH376. Enzyme kinetics were measured by EnzChek™ lipase substrate assay. Ex vivo placental perfusion experiments were performed +/- DH376 [1 µM] and changes in tissue lipid and fatty acid profiles were measured by LC-MS. Additionally, free fatty acid levels of the maternal and fetal circulations were determined.
### Results
We demonstrate that mRNA expression of DAGLβ prevails in placental tissue, compared to DAGLα (p ≤ 0.0001) and that DAGLβ is mainly located to CK7 positive trophoblasts (p ≤ 0.0001). Although few DAGLα transcripts were identified, no active enzyme was detected applying in-gel or MS-based ABPP, which underlined that DAGLβ is the principal DAGL in the placenta. DAGLβ dependent substrate hydrolysis in placental membrane lysates was determined by the application of LEI-105 and DH376. Ex vivo pharmacological inhibition of DAGLβ by DH376 led to reduced MAG tissue levels (p ≤ 0.01), including 2-AG (p≤0.0001). We further provide an activity landscape of serine hydrolases, showing a broad spectrum of metabolically active enzymes in the human placenta.
### Discussion
Our results emphasize the role of DAGLβ activity in the human placenta by determining the biosynthesis of 2-AG. Thus, this study highlights the special importance of intra-cellular lipases in lipid network regulation. Together, the activity of these specific enzymes may contribute to the lipid signaling at the maternal-fetal interface, with implications for function of the placenta in normal and compromised pregnancies.
## Introduction
The two transmembrane enzymes diacylglycerol lipase alpha (DAGLα) and beta (DAGLβ) possess sn-1 specific hydrolytic activity for diacylglycerols (DAG), preferably hydrolyzing DAG species with mono- or polyunsaturated fatty acids (FA) at sn-2 position, leading to the formation of monoacylglycerols (MAG) [1]. Interestingly, DAGLα/β exhibit a diverse cell-type and tissue-specific abundancy and it has been shown that specific expression patterns and regulatory mechanisms converge into distinct physiological roles of these enzymes. DAGLα is predominately expressed in the central nervous system and mainly confined to neurons [2, 3]. In contrast, DAGLβ is mainly expressed in peripheral tissues where its activity is elevated in immune cells including microglia [3], macrophages [4], dendritic cells [5] and, importantly, associated with inflammatory responses. Besides the constitutive role of DAGLα/β in DAG catabolism, DAGLβ has also been proposed as polyunsaturated fatty acid-specific triacylglycerol lipase [6].
Cell membranes of the human placenta exhibit a high abundance of polyunsaturated arachidonic acid (AA) esterified phosphoglycerides [7] and quantitative analysis of placental lipid profiles revealed a high concentration of unsaturated triacylglycerol species [8]. In particular, AA is essentially involved in the development of the fetal brain during the course of pregnancy [9, 10]. DAGLα/β are renown as key components of the endocannabinoid system (ECS), regulating the biosynthesis of an AA-esterified monoglyceride, namely 2-arachidonoylglycerol (2-AG). 2-AG is one of the main endocannabinoids, acting as chemical messenger and full agonist for cannabinoid receptors 1 and 2. Both, 2-AG and AA serve as substrates for the synthesis of prostanoid-esters and prostanoids, respectively. These metabolites play an important role during parturition as they mediate processes like contractions of the myometrium and they are involved in a variety of pregnancy pathologies (11–13). The importance to tightly regulate bioactive lipid species is emphasized by several studies reporting aberrant lipase action linked to first trimester miscarriage [14, 15], endometrial cancer [16] and pregnancy associated disorders such as preeclampsia [17, 18]. Furthermore, emerging evidence demonstrates a strong association between pregnancy disorders and dyslipidemia [19, 20]. Thus, examining one of the key enzymes in lipid metabolism, contributes not only to the basic understanding of the ECS in the human placenta, but more importantly elucidates the potential role of DAGL in pregnancy pathologies.
Although many efforts have been made to describe placental hydrolases in the past decades, many functional aspects are still unknown (21–23). Furthermore, the extent to which DAGL activity may affect metabolic cellular pathways by regulating bioactive lipids in this tissue is not yet understood. Notably, animal models have major limitations to answer this question due to differences in physiology and metabolism of the human placenta compared to other species. In human cytotrophoblasts the presence of DAGLα and the main 2-AG degradative enzyme monoacylglycerol lipase (MGL) has previously been reported [24]. Epithelial-like trophoblasts build up the outermost layers of the placenta and are in direct contact with the maternal blood. The multinucleated syncytiotrophoblast derives from underlying cytotrophoblasts and facilitates the exchange of nutrients, wastes and gases between the maternal and fetal circulations. In addition, it has been demonstrated that 2-AG reduced cell viability in a choriocarcinoma cell line and showed antiproliferative effects [24]. In this study, we aimed to determine the function of DAGL enzymes in bioactive lipid metabolism in the human term placenta. Furthermore, we generated a profile of catalytically active serine hydrolases in placental tissue by using activity-based protein profiling (ABPP). Lipidomics, chemical proteomics, and ex vivo placental perfusion were applied to comprehensively study in vitro and ex vivo the effect of acute enzyme inhibition on placental lipid homeostasis.
## Experimental model and subject characteristics
Study was performed in accordance with the protocols approved by the ethical committee of the Medical University of Graz (Vote no: 29-319 ex $\frac{16}{17}$ and 24-529 ex $\frac{11}{12}$). All subjects gave written informed consent. Placentas from caesarean section and vaginal delivery were used within 20 min after delivery. Important subject characteristics of this study cohort are depicted in Table 1. In order to collect tissue samples, the placenta was divided into quadrants and a cross sectioned piece of 7-10 mm diameter was scissored from each quadrant. Tissue samples were either snap frozen in liquid nitrogen and stored at −80°C for protein isolation, or formalin fixed and embedded into paraffin for immunohistochemistry.
**Table 1**
| Term placentas, n=31 | Unnamed: 1 | Unnamed: 2 |
| --- | --- | --- |
| Mode of birth (%) | CS | 61.3 |
| | VD | 38.7 |
| Gestational age | weeks ± days | 39 ± 9 |
| Placental weight [g] | | 616 (± 93.4) |
| Fetal sex (n) | Male | 17 |
| | Female | 14 |
| Fetal [g]/[cm] | weight | 3312.4 (± 348.6) |
| | length | 50.6 (± 2.1) |
| Maternal [kg/m2] | pre-pregnancy BMI | 21.5 (± 2.4) |
| | BMI at delivery | 26.8 (± 3) |
## Quantitative real-time PCR
Frozen 20-30 mg placental tissue pieces were homogenized in 700μl *Qiazol lysis* reagent (Qiagen, Cat# 217004) for 20 seconds 6500 revolutions per minute, by MagnaLyser (Roche, Basel, Switzerland) followed by 1 minute on ice and repeated 3 times. Next, total RNA content from cells and tissue lysates were isolated using the RNeasy®Mini Kit (Qiagen, Cat# 217004). Reverse transcription was performed using 1μg of RNA and LUNA Script RT SuperMix Kit (New England Biolabs, Cat#E3010L). For RT-qPCR analysis LUNA Universal qPCR Master Mix (New England Biolabs, Cat#M3003E) and BioRad CFX384 Touch Syllabus were used. QuantiTect Primer Assays were used for gene amplification. For 18S reference gene amplification custom DNA oligos were designed (F-(5 ‘-3 ‘) CTACCACATCCAAGGAAGCA/R-(5 ‘-3 ‘) TTTTTCGTCACTACCTCCCCG). The expression of target genes DAGLα (GeneGlobe ID - QT00038164) and DAGLβ (GeneGlobe ID - QT00074319) was normalized to reference genes 18S, RPL30 (GeneGlobe ID - QT00056651) and HPRT1 (GeneGlobe ID - QT00059066). Target and reference gene ΔCT values are corrected for respective primer efficacy.
## DAGLα/β in situ hybridization
To detect and discriminate DAGLα and DAGLβ mRNA on cellular level, RNAscope® 2.5 HD Reagent Kit-RED assay (Advanced Cell Diagnostics, Cat#PN 322350) was used according to the manufacturer’s protocol. In short, 5 µm thick formaldehyde-fixed paraffin-embedded sections were de-paraffinized and pre-treated under standard pre-treatment conditions with hydrogen peroxide, target retrieval reagents and protease solution. The sections were covered with probe solution and incubated for 2 hours at 40°C using the HybEZ Hybridization System (Advanced Cell Diagnostics, Cat#PN $\frac{321710}{321720}$). The sections were treated with AMP 1 to 6 according to the manufacturer´s manual, using the HybEZ Hybridization System. The multi-step hybridization process included hybridization to alkaline phosphatase-labeled probes and resulted in the detection of signal using Fast Red as a substrate. To combine ISH with immunohistochemistry (IHC), after performing ISH IHC was started from the blocking step as described below.
## Immunohistochemistry
Placental tissue sections were blocked with $4\%$ BSA and $10\%$ secondary antibody host serum in PBS/$0.3\%$ Triton X100 and incubated overnight with primary antibody solutions. Primary antibodies for cytokeratin 7 (1:500, Abgent, Cat#AJ1229a), CD163 (1:200, Thermo Fisher Scientific, Cat#MA1-82342), and Von Willebrand Factor (1:500, Dako, Cat#A0082) were used. To detect Cytokeratin 7 (CK7) and Von Willebrand Factor (VWF) goat anti-rabbit Alexa Fluor 647 secondary antibody was used (1:500, Cell Signaling Technology, Cat#4414) and displayed in white. CD163 primary antibody incubation was followed by goat anti- mouse Alexa Fluor 488 secondary antibody application (1:500, Invitrogen, Cat#A32723) and displayed in green. Sections were counterstained with DAPI, sealed with a coverslip using VECTASHIELD® Antifade Mounting Medium with DAPI (Vector Laboratories, Cat# H-1200-10) and stored at 4°C until imaged. Representative images were captured on Nikon A1 confocal microscope (original magnification ×40) and prepared using FIJI software v.1.51h.
## Microscopy and signal quantification
For quantitative determination and localization analysis, ten z-stacks of each section were acquired using a Nikon A1 confocal with a ×40 objective at a step size of 0.5 µm. An automated image analysis was created with the software package FIJI v.1.51h. The analysis entailed a basic pre-processing, generating a maximum intensity projection and mean filter smoothing, followed by application of an algorithm-based threshold. Feature detection of channels containing CK7 or ISH information was achieved employing RenyiEntropy [25], IsoData for the CD163 channel [26] and Otsu for the VWF containing channel [27]. Generated regions of interest (ROI) of the ISH were separated by watershed and counted the ROIs of the remaining channels were used to determine the cell type specific ISH localization due to overlap.
## Gel-based activity-based protein profiling
Gel-based ABPP experiments were performed as previously described [28]. Frozen tissues were thawed on ice and homogenized in cold lysis buffer (20 mM HEPES pH 7.2, 250 mM sucrose, 1 mM MgCl2, 2.5 U/mL benzonase). After incubation on ice for 15 min, tissue debris was pelleted by centrifugation (2500 × g, 3 min, 4°C) and supernatant was transferred to a clean tube. Subsequently the supernatant was centrifuged at 30.000 × g (90-120 min, 4°C) to pellet the membrane-associated fraction and separate it from the soluble proteome. After removal of the soluble supernatant, the membrane pellet was washed with cold HEPES buffer (20 mM, pH 7.2) followed by resuspension in cold HEPES buffer by pipetting. Concentration of membrane-associated and soluble proteome was quantified (Bradford; BioRad Technologies, CA, USA) and adjusted to desired concentration (2 mg/mL) in HEPES buffer (20 mM, pH 7.2). For direct labeling, proteomes were sequentially treated with one-step activity-based probes DH379 (30 min, 1 µM, Cy3, RT) and FP-Bodipy (15 min, 500 nM, Cy2, RT) or MB064 (30 min, 250 nM, Cy3, RT) alone in a 15 µL total reaction volume. For competitive ABPP experiments, this step was preceded by incubation with DH376 in vitro or ex vivo and LEI-105 in vitro at indicated concentrations. The reactions were quenched by the addition of 5 µL 4x Laemmli-buffer (BioRad Technologies, CA, USA). After separation by SDS-PAGE ($10\%$ acrylamide) at 180V for 75 min, samples were visualized by in-gel fluorescence scanning (Cy2 $\frac{532}{28}$, Cy3 $\frac{605}{50}$, Cy5 $\frac{700}{50}$ filter settings) using a flatbed fluorescent scanner ChemiDoc™ MP Imaging System (Bio-Rad, Hercules, CA, USA). Coomassie staining was used to control the protein loading. Gel fluorescence is shown in greyscale, and optical density of the signals was determined using ImageLab 6.1 Biorad.
## DAGLβ activity assay
Membrane fractions of placental tissues were prepared as described above (see Gel-based activity-based protein profiling (ABPP)). Membrane lysates were diluted to 10 ng/µL in assay buffer (50 mM HEPES pH 7.5, $0.0025\%$ Triton X-100). Fluorescent measurements were carried out at RT in a black flat bottom 96-well plate (Thermo Fisher Scientific, MA, USA) in the presence of 0.5 µM EnzChek™ lipase substrate (Thermo Fisher Scientific, Cat#E33955) in 100 μL final volume using a Clariostar plate reader (BMG Labtech, Germany) and excitation/emission wavelengths of $\frac{477}{525.}$ For competitive experiments placental membrane lysates were pre-incubated with DH376 (100 nM) and LEI-105 (1 µM) for 30 min at RT, respectively. DMSO served as vehicle control and denatured samples ($1\%$ SDS, 5 min, 100°C) served as background controls. Background substrate hydrolysis was deducted from each measurement. Each data point is the mean of three technical replicates of $$n = 3$$ placentas, for concentration testing and $$n = 4$$ placentas for competitive experiments. The slope $t = 10$ min to $t = 60$ min was used as the enzymatic rate (RFU/min). Enzyme kinetics were plotted as curves in Graph Pad Prism 9 Software (GraphPad Software Inc., CA, USA).
## Chemical proteomics with label-free quantification
Placental tissues were homogenized and prepared as described in Gel-based activity-based protein profiling (ABPP) section. The chemical proteomics workflow is based on previously published protocol [29] and conducted with minor modifications. In short, cytosolic and membrane fractions of placental tissue lysates (250 µg protein, 1 mg/mL, $$n = 5$$) were incubated with serine hydrolase probe cocktail (10 µM MB108, 10 µM FP-Biotin, 30 min, 37 ˚C, 300 rpm). A pool of denatured vehicle control samples ($1\%$ SDS, 5 min, 100°C) was taken along as a negative control. Following steps were preformed according to protocol, including precipitation, alkylation, avidin enrichment, on-bead digestion, and sample preparation. Dried and desalted peptide samples were stored at -20°C until LC-MS analysis. Prior to measurement, samples were reconstituted in 50 µL 97:3:0.1 solution (H2O, ACN, FA) containing 10 fmol/µL yeast enolase digest (Waters, cat# 186002325) and transferred to LC-MS vials. Additionally, a quality control sample was prepared to prevent overloading the nanoLC system and the automatic gain control (AGC) of the QExactive HF mass spectrometer. LC-MS data was analyzed by MaxQuant software 2.0 applying match between runs. For further analysis, the following cut-offs were used: unique peptides ≥ 2, identified peptides ≥ 2, ratio positive over negative control ≥ 2. Additionally, targets were filtered against a putative probe-target list including human metabolic serine hydrolases.
## Ex vivo placental perfusion
Placental perfusion setup is based on the setup published by Schneider et al. [ 30] and adapted as published by Hirschmugl et al. [ 31]. In short, within 30 min after delivery of the placenta a single placental cotyledon chorionic-artery and vein pair was cannulated and flushed with perfusion buffer, containing DMEM (DMEM, phenol red free, Gibco by Life Technologies, ThermoFisher Scientific, MA, USA) mixed (3:1) with Earl's buffer (6.8 g/L NaCl, 0.4 g/L KCl, 0.14 g/L NaH2PO4, 0.2 g/L MgSO4•7H2O, 0.2 g/L CaCl2, 2.2 g/L NaHCO3, pH 7.4, all Merck, Darmstadt, Germany), amoxicillin (250 mg/L, Sigma-Aldrich, Steinheim, Germany), glucose (2 g/L Merck, Darmstadt, Germany), and essential fatty acid free bovine serum albumin (BSA) (35 g/L, Sigma-Aldrich, Seinheim, Germany). The cannulated cotyledon and surrounding tissue were placed in the pre-warmed perfusion chamber and the fetal circulation was connected to a magnetic pump (Codan, Salzburg, Austria) with a constant fetal artery inflow of 3 mL/min. The perfusion buffer is constantly fumigated by a gas exchange device (LivingSystems, St. Albans, VT, US) operated with $95\%$ N2 and $5\%$ CO2 on the fetal site during the experiment. A micro catheter pressure sensor (Millar, US) inserted into the fetal arterial cannula recorded the backflow pressure which should not exceed an average of 65 mbar. The impermeability of the perfused cotyledon was monitored within the first 30 min and each cotyledon displayed at least $95\%$ fetal flow recovery. The maternal circulation was established by inserting three rounded needles into the intervillous space of the cotyledon with a flow rate of 9 mL/min. During the experiment, the perfusion buffer was gassed with $5\%$ CO2, $20\%$ O2 and $75\%$ N2 through the gas exchanging device. DH376 (1 µM) was added to the fetal and maternal perfusion buffer reservoirs and the system changed to closed circuit in all inhibitor experiments. During the experiment maternal and fetal perfusates were collected every 30 min via a sampling port and oxygen (pO2), carbon dioxide (pCO2), pH, lactate production, and glucose consumption measurements were applied by a blood gas analyzer (Radiometer, Copenhagen, Denmark). The data sets obtained by the blood gas analyzer, magnetic pumps, and pressure sensor were registered and recorded via LabVIEWbased recording software (Beko engineering, Graz, Austria). After 4h of closed perfusion time, samples of both circuits were collected, centrifuged, and stored at −80°C. The perfused placental tissue was processed in cold PBS and snap frozen in liquid nitrogen until further analysis.
## Lipid analysis by LC-MS
Placenta (pl, ~10 mg powdered), and perfusate (pf, 140 µl) samples were extracted according to Matyash et al. [ 32]. In brief, samples were homogenized using two beads (stainless steel, 6 mm) on a Mixer Mill (Retsch, Haan, GER; 2x10sec, frequency 30/s) in 700 µl methyl-tert-butyl ether (MTBE)/methanol ($\frac{3}{1}$, v/v) containing 500 pmol butylated hydroxytoluene, $1\%$ acetic acid, and internal standards (IS; pl: 20 pmol 15:$\frac{0}{15}$:$\frac{0}{15}$:0 triacylglycerol, 13 pmol rac-17:$\frac{0}{17}$:0 diacylglycerol, rac-17:0 monoacylglycerol, 50 pmol 17:$\frac{0}{17}$:0 phosphatidylcholine, Larodan, Solna, Sweden; 133 pmol 17:$\frac{0}{17}$:0 phosphatidylethanolamine, 30 pmol 17:$\frac{0}{17}$:0 phosphatidylserine, 8 pmol 17:1 lyso-phosphatidylcholine, 30 pmol 17:1 lyso-phosphatidylethanolamine, Avanti Polar Lipids, Alabaster, AL, USA; cb and pf: 2 nmol 17:0 FA, 800 pmol C21:0 FA, Sigma-Aldrich, St. Louis, MO, USA). Total lipid extraction was performed under constant shaking for 30 min at RT. After addition of 140 µl dH2O (pl) and further incubation for 30 min at RT, samples were centrifuged at 1,000 x g for 15 min. 500 µl of the upper, organic phase were collected and dried under a stream of nitrogen. Lipids were resolved in 500 µl MTBE/methanol ($\frac{3}{1}$, v/v). Pl extracts were diluted 1:4 in 2-propanol/methanol/dH2O (SolA; $\frac{7}{2.5}$/1, v/v/v) for LC-MS analysis. To determine fatty acid levels, 200 µl (pf) were derivatized according to Bollinger et al. [ 33] using the AMP+ MS Kit (Cayman Chemical, Michigan, USA) and resolved in 500 µl SolA for LC-MS analysis. Protein precipitates of the extractions were dried, solubilized in NaOH (0.3 N) at 65°C for 4 h and the protein content was determined using Pierce™ BCA reagent (Thermo Fisher Scientific, MA, USA) and BSA as standard. Chromatographic separation was performed on a 1290 Infinity II LC system (Agilent, CA, USA) equipped with Zorbax RRHD Extend-C18 column (2.1x50 mm, 1.8 µm; Agilent, CA, USA) running a 10 min linear gradient from $60\%$ solvent A (H2O; 10 mM ammonium acetate, $0.1\%$ formic acid, 8 µM phosphoric acid) to $100\%$ solvent B (2-propanol; 10 mM ammonium acetate, $0.1\%$ formic acid, 8 µM phosphoric acid). The column compartment was kept on 50°C. A 6470 Triple Quadrupole mass spectrometer (Agilent, CA, USA) equipped with an ESI source was used for detection of lipids in positive mode. Data acquisition was done by MassHunter Data Acquisition software (B.10, Agilent, CA, USA) either in MRM (glycerol- and glycerophospholipids) or SIM (fatty acid derivatives) mode. Lipidomic data were processed using MassHunter Workstation Quantitative Analysis for QQQ (V.9, Agilent, CA, USA), normalized for recovery, extraction-, and ionization efficacy by calculating analyte/IS ratios (AU) and expressed as AU/µg protein.
## Statistical analysis
Graph Pad Prism 9.02 Software (GraphPad Software Inc., CA, USA) was used for statistical analysis and graph plotting. Data are presented as mean ± SEM. All obtained datasets were tested for normal distribution with the Shapiro-Wilk and Kolmogorov-Smirnov test. Depending on the distribution of datasets parametric or non- parametric statistical tests were applied. If two or more normal distributed groups were compared student`s t-test or one-way ANOVA, including Benjamini- Hochberg post-hoc was performed. If the dataset did not show a normal distribution Mann-Whitney U test or Kruskal-Wallis test followed by Benjamini- Hochberg post hoc was applied. Two-way ANOVA was applied comparing two or more groups including different variables using Benjamini- Hochberg post hoc for multiple comparison correction. All herein presented p-values correspond to a FDR of $1\%$ for multiple testing and p-values below 0.05 were considered statistically significant.
## Detection of DAGL mRNA and activity in placental tissue lysates
To elaborate DAGL expression in placental tissues, we first determined DAGLα/β mRNA levels by RT-qPCR. In comparison to DAGLα substantially higher DAGLβ expression was observed (Figure 1A, $p \leq 0.0001$). Serine hydrolase activities in the placenta were examined by gel-based ABPP. To obtain a broad view of placental serine hydrolase activities and DAGL-specific signals, we used a probe cocktail of non- selective FP-Bodipy [34] and the DAGLα/β directed fluorescent probe DH379 [28]. Using FP-Bodipy, we detected a broad spectrum of active enzymes in placental lysates. Application of DH379 visualized DAGLβ at the expected molecular weight of ~70 kDa (Figures 1B, C). To confirm the presence of active DAGLβ, competitive ABPP was applied using the DAGL inhibitors DH376 (IC50 3–8 nM) [28] and LEI-105 (IC50 ~32 nM) [35]. Application of DH376 and LEI-105 reduced DAGLβ activity in a dose-dependent manner. DH376 led to full inhibition at the lowest concentration of 0.1 µM (Figure 1B), while LEI-105 treatment led to a substantial reduction of DAGLβ signals at a concentration of 0.5 µM (Figure 1C). Notably, we were not able to detect DAGLα activity at the expected molecular weight of ~120 kDa, using the DAGLα tailored probe MB064 [36], likely due to low expression levels compared to DAGLβ (Supplementary Figure 1). DAGLβ activity in placenta tissue lysates was also investigated using the commercially available lipase substrate EnzChek™ (Figure 1D). We applied 100 nM of DH376, which represented the lowest inhibitor concentration leading to potent enzyme inhibition of DAGLβ in-gel (Figure 1B). In comparison to DH376, LEI-105 exhibits lower activity against DAG-lipases and acts as a reversible enzyme inhibitor. Although a substantial reduction in DAGLβ enzyme activity was observed by application of 0.5 µM (Figure 1C), we applied 1 µM LEI-105 to ensure complete enzyme inhibition over time. The application of LEI-105 and DH376 reduced hydrolase activity by $51\%$ and $70\%$, respectively (Figure 1E). The differences in inhibitor efficacy can be explained by previous observations showing that LEI-105 exhibits higher selectivity for DAGLβ than DH376. At the used inhibitor concentration, DH376 is expected to inhibit α/β hydrolase domain-containing protein 6 (ABHD6), carboxylesterase 1 and 2 (CES$\frac{1}{2}$), and hormone- sensitive lipase (HSL) [28], which can contribute to the hydrolysis of the EnzChek™ substrate. Gel-based ABPP experiments enabled us to detect specific DAGLβ activity in placental tissue and we could further decipher DAGL-dependent substrate hydrolysis by administration of DH376 and LEI-105. Overall, our observations indicate that the placenta predominantly expresses DAGLβ, which consequently can affect DAG, MAG, and fatty acid (FA) metabolism.
**Figure 1:** *Detection of DAGL mRNA and activity in placental tissue lysates. (A) Relative DAGLα/β target gene mRNA levels were detected by RT-qPCR. Results were normalized to reference genes (RefG) 18S, RLP30 and HRPT1 detected in each sample and calculated as ΔCT. For statistics student`s t-test was applied and ΔCT values are depicted as 2 -ΔCT (n=13). (B), (C) Visualization of DAGLβ activity and selectivity profile of DH376 and LEI-105 using in-gel ABPP. Placental membrane proteomes were profiled by competitive ABPP using a probe cocktail of FP-Bodipy [500 nM] (Cy2, green) and DH379 [1 µM] (Cy3, red). Samples were incubated with indicated inhibitor concentrations or DMSO as a vehicle control. Concentration-dependent inhibition of DAGLβ by DH376 (B) and LEI-105 (C). Coomassie staining served as a protein loading control. (D) Hydrolase activity was determined using EnzChek™ lipase substrate [0.5µM] and displayed as relative fluorescence units (RFU) per time. DAGLβ activity determined by applying LEI-105 [1 µM] and DH376 [100 nM]. (E) The slope of the linear interval t=10 to t=60 min was used to calculate the enzymatic rate (RFU/min). One-way ANOVA for multiple comparisons followed by Benjamini- Hochberg post hoc was applied to quantify the differences of enzymatic activities (n=4). Data are depicted in mean ± SEM; ****p ≤ 0.0001.*
## Activity profiling of placental metabolic serine hydrolases
ABPP using FP-Bodipy already suggested that the placenta expresses a broad spectrum of serine hydrolases (Figures 1B, C). To get a profile of these enzymes, we performed mass spectrometry-based chemical proteomics utilizing the biotinylated non- selective probes MB108 and FP-Biotin for target identification. While gel-based ABPP experiments strongly rely on specific inhibitors for target identification, MS- based ABPP enables target enrichment and provides high sensitivity. To increase the resolution of proteins, tissue lysates were separated into membrane and cytosolic fraction. This approach resulted in the identification of 38 and 33 different serine hydrolases in membrane and cytosolic fractions, respectively (Figure 2). Activities of several α/β hydrolase domain-containing protein family members (ABHD) and phospholipases such as DDHD2, patatin-like phospholipase domain-containing proteins (PNPLA) and members of the phospholipase A2 family (PLA2) were identified. Furthermore, lipases involved in DAG, MAG and FA metabolism, including HSL, CES$\frac{1}{2}$ and acyl-coenzyme A thioesterase (ACOT1) were detected. Within the 2-AG biosynthetic active enzymes, we again exclusively detected DAGLβ activity. Taken together, performing chemical proteomics allowed us to generate an overview of the lipolytic proteome of human placental tissue and demonstrated a broad spectrum of metabolic hydrolase activities.
**Figure 2:** *Activity profiling of placental metabolic serine hydrolases. Membrane and cytosolic tissue protein fractions were labeled with hydrolase probe cocktail (MB108, FP-Biotin [10 µM]) and analyzed by chemical proteomics. Absolute abundance refers to the mean of LFQ intensities of vehicle perfused placentas and is depicted in alphabetical order as heat map (blue scale, log2), not detected proteins are depicted in grey (n=5).*
## DAGLβ expression is mainly confined to trophoblasts
As the human placenta is composed of different highly specialized cell types, we aimed to determine the localization of DAGLβ in situ, by using specific RNA probes. To localize the transcripts to distinct cell types of the placenta, immunofluorescence (IF) was applied. The visualization of DAGLβ transcripts was combined with cytokeratin 7 (CK7) staining, representing trophoblasts, which are building up the first structural barrier between maternal and fetal compartment and CD163 as a pan macrophage marker (Figure 3A). To identify feto-placental endothelial cells, which are lining placental vessels and are in direct contact with fetal blood, Von- Willebrand Factor (VWF) was applied (Figure 3B). We localized DAGLβ transcripts mainly to CK7 positive trophoblasts (Figure 3A). Quantitative analysis of the signals revealed that $54\%$ of DAGLβ mRNA was localized to trophoblasts (T), while negligible signals of $2\%$ and $3\%$ were detected in endothelial cells (E) and macrophages (M), respectively (Figure 3C). In contrast to DAGLβ, we could not detect clear signals for DAGLα, confirming our observation that this enzyme is poorly expressed in the human placenta (Supplementary Figure 2).
**Figure 3:** *DAGLβ expression is mainly confined to trophoblasts. DAGLβ transcripts were detected in placental tissues using RNAscope® 2.5 HD RED assay. (A) DAGLβ transcripts detected in CK7 positive trophoblasts and CD163 positive placental macrophages. (B) DAGLβ mRNA was localized to VWF stained endothelial cells. Nuclei were counterstained with DAPI (blue). Arrowheads in merged micrographs indicate probe co-localization to different cell types. Magnification x40, Scale bar 100 µm. For quantitative determinations, ten images of four individual placentas were captured on Nikon A1 confocal microscope. Probe co-localization was quantified by Fiji software. (C) Relative distribution of DAGLβ transcripts to trophoblasts (T), endothelial cells (E) and placental macrophages (M) based on ISH signals (n=4). Statistical analysis was performed using one- way ANOVA, followed by Benjamini- Hochberg post hoc test (n=4). Representative stainings are shown in (A) and (B). Data are depicted in mean ± SEM; ****p < 0.0001.*
## Inhibition of DAGLβ activity leads to reduced DAG, MAG and FA levels in perfused placental tissue
To better understand the specific function of DAGLβ in the intact organ, the lipid profile was examined after tissue perfusion with/out inhibitor. First, we examined the extent of DAGLβ inhibition obtained after perfusion by in-gel ABPP. Using DH379, we again identified the DAGLβ band (~70 kDa) in the vehicle perfused proteome (Figure 4A; Supplementary Figure 3). By co-application of DH376 [1µM], signal intensity was strongly reduced by $87\%$, confirming target engagement (Figure 4B). Thus, our results suggest that the application of DH376 at a concentration of 1 µM in an ex vivo setting leads to substantial inhibition of DAGL, which may also be accompanied by changes in the lipid profile. To examine the consequences of DAGLβ inhibition under the applied conditions, we analyzed changes in lipid species in perfused placental tissue. In accordance with the proposed cellular function of DAGL, total MAG levels were decreased by $60\%$ (Figure 4C), resembled by a reduction of all detected MAG species (Figure 4D). Notably, the obtained data further demonstrated significant decreased 2-AG levels (MAG 20:4) upon inhibition of DAGLβ (Figure 4D). Interestingly, decreased MAG levels were not accompanied by augmented DAG concentrations, but a trend to decreased levels could be detected (Figure 4E). In fact, we observed significant reductions in specific DAG species, including DAG 32:0-16:0, DAG 34:2-18:2 and DAG 36:2-18:2 (Figure 4F). Furthermore, FA concentrations showed a downward trend in inhibitor perfused tissues (Figure 4G), of which eicosenoic acid (FA 20:1) was significantly reduced, indicating an effect on the hydrolysis of sn-1 FA of DAGs (Figure 4H). In contrast, FA levels in the maternal and fetal circuit remained unchanged (Supplementary Figures 4A, B).
**Figure 4:** *Inhibition of DAGLβ activity leads to reduced DAG, MAG and FA levels in perfused placental tissue. (A)
In vitro labeling of enzymes in DH376 [1 µM] and vehicle perfused placental membrane proteomes by direct ABPP using DH379. (B) Densiometric quantification of in gel ABPP demonstrated significantly decreased DAGLβ activity in DH376 perfused placentas compared to vehicle controls. Student´s t-test was performed for statistical testing (n=5). (C) Total monoacylglycerol (MAG) tissue levels in DH376 [1 µM] and vehicle perfused tissues. (D) Depiction of all measured MAG species by LC-MS. (E) LC-MS analysis of total diacylglycerol tissue levels (DAG) and diacylglycerol species in vehicle and DH376 perfused placental tissues (F). (G) Total tissue FA levels and corresponding FA species (H). Lipid levels are expressed as arbitrary units (AU) and were normalized to total tissue protein (µg). Student´s t-test and multiple t-test followed by Benjamini- Hochberg post hoc was performed, respectively (n=3 lipid levels, n=5 FA levels). Data are depicted as mean ± SEM; *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p< 0.0001.*
## Discussion
DAG lipases occupy a central role in multiple lipid signaling pathways by regulating DAG [37], endocannabinoid levels and downstream inflammatory mediators [4, 28]. Studies in mice demonstrated that acute blockade of DAGL by DH376 in vivo led to significant reductions in endocannabinoid-, AA-, and prostaglandin levels in the central nervous system [28]. Moreover, it has been shown that pharmacological inhibition and genetic disruption of DAGLα/β suppress induction of 2-AG and prostaglandin levels upon lipopolysaccharide treatment, which was accompanied by decreased pro-inflammatory cytokine secretion [4, 5, 28]. Recently, Shin et al. identified DAGLβ as polyunsaturated fatty acid- specific triacylglycerol lipase by demonstrating robust hydrolysis activity for triarachidonin (C20:4 FA) or tridocosahexaenoin (C22:6 FA) in vitro [6]. Hence, understanding the role of these versatile enzymes in human physiology and disease gained considerable interest. Here, we aimed to study the role of DAGL in placental lipid homeostasis.
ABPP assays were used to screen for serine hydrolase activities in the human placenta and particularly to examine the functional state of DAGL enzymes. The use of a fluorescent DAGL-tailored probe enabled us to detect DAGLβ activity at the corresponding molecular weight of ~70 kDa. Competitive ABPP with selective inhibitors confirmed that the signal was DAGLβ. DH376 has been described as potent, central active and covalent DAGL inhibitor (IC50 of 3-8 nM) [28]. Since it has been reported that this compound cross-reacts with several other lipases such as CES$\frac{1}{2}$, HSL and ABHD6, we decided to include LEI-105 to verify our findings. LEI-105 is described as highly selective, but reversible DAGL inhibitor (IC50 ~32 nM) [35]. Importantly, it has been shown that this compound did not affect the activity of other endocannabinoid-related hydrolases such as ABHD6 [35]. Complete enzyme inhibition of placental DAGLβ could be achieved by applying relatively high inhibitor concentrations, as LEI-105 represents a non-covalent inhibitor and shows ~4-fold lower activity against DAGL enzymes compared to DH376. In conclusion, administration of LEI-105 validated our observations in-gel and contributed to the determination of DAGL- dependent substrate hydrolysis. Conversely, DAGLα activity was not detectable by ABPP, suggesting that DAGLβ is the principal active DAG-lipase in the human term placenta. The predominance of DAGLβ over α in placental tissue was further corroborated on transcriptional level in vitro as well as in situ. These findings are in accordance with previous studies showing that DAGLβ is mainly found in peripheral metabolically active tissues such as the liver, where DAGLβ-/- mice showed $90\%$ reductions in 2-AG levels [2]. As placental tissue is composed of different cell types, we specifically looked at the spatial expression of DAGLβ. DAGLβ transcripts were mainly located to CK7-postive trophoblasts, lining the first cellular barrier between the maternal and fetal circulation. Co-localization of DAGLβ to trophoblasts, which reflect the main site of action upon maternally derived signals, is in concordance with the expression sites of other lipid related enzymes in this organ [38].
In order to assess the importance of DAGL in the spectrum of the lipolytic enzymes in placental tissue we further generated an activity- based profile of serine hydrolases. The chemoproteomic analysis revealed a broad spectrum of hydrolases, which determine lipid metabolism and signaling. Within the 2-AG biosynthetic enzymes we again exclusively detected DAGLβ. Further, specific activity of enzymes involved in degradation of 2-AG, such as MGL, fatty acid amide hydrolase (FAAH), ABHD6 and ABHD12 was detected. It has been shown that beside MGL, which is the main enzyme for 2-AG hydrolysis, ABHD$\frac{6}{12}$ and FAAH also possess 2-AG hydrolytic activity [39, 40]. Blankman et al. suggested that the simultaneous occurrence of different 2-AG hydrolytic enzymes could be explained by the regulation of distinct subcellular 2-AG pools. In accordance with previous observations, ABHD$\frac{6}{12}$ were exclusively identified in placental membrane preparations, whereas MGL activity was found in the cytosolic and membrane fraction [40]. Interestingly, ABHD12 showed the highest activity in our dataset, compared to other 2-AG metabolizing enzymes. The major function of ABHD12 is the hydrolysis of lysophosphatidylserine as shown by genetic depletion in mice [41]. The specific role of this enzyme in the placenta remains to be investigated, since only descriptive data is available yet [22]. Besides 2-AG, anandamide (AEA) was one of the first discovered endocannabinoids and FAAH is the main catabolic enzyme in this pathway [42]. Moreover, ABHD4 activity was detected, which contributes to AEA biosynthesis. In addition, several hydrolases determining lipid and FA metabolism, including (lyso)phospholipases, HSL and CES$\frac{1}{2}$ were identified.
This study set out with the aim of assessing the importance of DAGLβ activity in the lipid homeostasis of the human term placenta. Therefore, we looked at the functional consequences of pharmacological enzyme inhibition ex vivo, by applying DH376 as an inhibitor targeting DAGL activity. Lipidomic analysis of perfused tissue samples showed that acute inhibition of DAGLβ led to significantly reduced total MAG tissue levels, confirming the well-described role of DAGL in DAG catabolism. In fact, we could observe a significant decrease in 2-AG levels and a trend towards reduced saturated as well as mono- and polyunsaturated MAG species. In contrast to previously published data, the decrease in MAG levels was not followed by an increase in respective DAGs, suggesting that DAGs are efficiently metabolized in the absence of DAGL. Interestingly, specific DAG species showed a significant reduction upon DAGLβ inhibition, indicating that the enzyme may possess hydrolase activity against TAGs. In this context, DAGLβ has been previously described as polyunsaturated- specific TAG lipase in mice using genetic and pharmacological approaches [6]. The decline in eicosenoic acid (FA 20:1), in inhibitor perfused tissues, suggests that DAGLβ prefers 20:1 species at sn-1 position of DAGs. Interestingly, we could not detect considerable differences of FA levels neither in the maternal nor in the fetal circulation. In contrast, in vivo pharmacological inhibition of DAGLα/β by DH376 led to significant reductions of AA levels in murine central nervous tissues [28]. Unchanged AA levels could be explained by compensatory or bypass activities ensuring a constant supply of polyunsaturated fatty acids to the fetus. Furthermore, Hirschmugl et al. demonstrated that only a small proportion of free FA are directly transferred across the placenta and emphasized the tightly regulated release of FA out of metabolic pools of the placenta [31]. It is also important to note that DAGLβ is predominantly expressed in trophoblasts and lipid extracts were obtained from whole tissue. Nonetheless, we observed a substantial decrease in MAGs indicating that DAGLβ activity strongly affects lipid homeostasis distinctively in this specific placental cell type.
This study describes for the first time that 2-AG is dramatically reduced after DAGL inactivation in the human placenta. Since 2-AG represents only one of the main endocannabinoids, this study is limited by the lack of information on the potential simultaneous regulation of AEA. Although, both endocannabinoids exhibit distinct synthesis, transport and degradation processes, one may speculate that alterations in lipid levels affect concomitant lipid signaling pathways, and compensatory or bypass mechanisms could be activated to restore lipid homeostasis. In particular, a potential mechanism for synaptic crosstalk and feed-back regulatory mechanisms between the two pathways has already been described in tissues of the central nervous system [43, 44]. Furthermore, only one inhibitor concentration and experimental timepoint was used for ex vivo experiments. Since DAGLα exhibits a short half- life (< 4 h) and cumulative evidence supports the on-demand model of endocannabinoid biosynthesis [28, 45], it would be of great interest to study the consequences of enzyme blockage on the dynamic composition of placental lipids over time.
In summary, our study demonstrates that the application of small molecule inhibitors in perfusion experiments provides a very useful tool to investigate enzyme function as close as possible to the in vivo situation. In addition, ABPP is a powerful technique to visualize the active pool of enzymes and confirm target engagement in such ex vivo tissue experiments. The integration of these two approaches provided evidence that inhibition of DAGLβ affects tissue lipid homeostasis with no direct effect on the FA profile in the maternal or fetal compartment. We expect that further experiments by utilizing serine hydrolase inhibitors will strongly improve our understanding of the role of these enzymes in lipid signaling and metabolism at the maternal-fetal interface and reveal important insights related to placental function in normal and compromised pregnancies.
## Data availability statement
The mass spectrometry proteomics data have been deposited to the ProteomeXchange *Consortium via* the PRIDE [46] partner repository with the dataset identifier PXD039930.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethical Committee of the Medical University of Graz, Graz, Austria. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
Conceptualization, CW and NB. Methodology, NB, BH, TB, TW, TE and NF. Software, JG. Investigation, NB, TE and NF. Writing – Original Draft, NB, CW, BH and RZ. Funding Acquisition, CW. Resources, MS, RB-G and RZ. Supervision, CW, BH, TW, MS and RZ. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1092024/full#supplementary-material
## References
1. Bisogno T, Howell F, Williams G, Minassi A, Cascio MG, Ligresti A. **Cloning of the first sn1-DAG lipases points to the spatial and temporal regulation of endocannabinoid signaling in the brain**. *J Cell Biol* (2003) **163**. DOI: 10.1083/jcb.200305129
2. Gao Y, Vasilyev DV, Goncalves MB, Howell FV, Hobbs C, Reisenberg M. **Loss of retrograde endocannabinoid signaling and reduced adult neurogenesis in diacylglycerol lipase knock-out mice**. *J Neurosci* (2010) **30**. DOI: 10.1523/JNEUROSCI.5693-09.2010
3. Viader A, Ogasawara D, Joslyn CM, Sanchez-Alavez M, Mori S, Nguyen W. **A chemical proteomic atlas of brain serine hydrolases identifies cell type-specific pathways regulating neuroinflammation**. *Elife* (2016) **5**. DOI: 10.7554/eLife.12345
4. Hsu KL, Tsuboi K, Adibekian A, Pugh H, Masuda K, Cravatt BF. **DAGLβ inhibition perturbs a lipid network involved in macrophage inflammatory responses**. *Nat Chem Biol* (2012) **8** 999-1007. DOI: 10.1038/nchembio.1105
5. Shin M, Buckner A, Prince J, Bullock TNJ, Hsu KL. **Diacylglycerol lipase-β is required for TNF-α response but not CD8+ T cell priming capacity of dendritic cells**. *Cell Chem Biol* (2019) **26** 1036-1041.e3. DOI: 10.1016/j.chembiol.2019.04.002
6. Shin M, Ware TB, Hsu KL. **DAGL-beta functions as a PUFA-specific triacylglycerol lipase in macrophages**. *Cell Chem Biol* (2020) **27** 314-321.e5. DOI: 10.1016/j.chembiol.2020.01.005
7. Bitsanis D, Crawford MA, Moodley T, Holmsen H, Ghebremeskel K, Djahanbakhch O. **Arachidonic acid predominates in the membrane phosphoglycerides of the early and term human placenta**. *J Nutr* (2005) **135**. DOI: 10.1093/jn/135.11.2566
8. Brown SHJ, Eather SR, Freeman DJ, Meyer BJ, Mitchell TW. **A lipidomic analysis of placenta in preeclampsia: Evidence for lipid storage**. *PloS One* (2016) **11**. DOI: 10.1371/journal.pone.0163972
9. Basak S, Mallick R, Banerjee A, Pathak S, Duttaroy AK. **Maternal supply of both arachidonic and docosahexaenoic acids is required for optimal neurodevelopment**. *Nutrients* (2021) **13**. DOI: 10.3390/nu13062061
10. Brittis PA Walsh FS, Doherty PSJ. **Fibroblast growth factor receptor function is required for the orderly projection of ganglion cell axons in the developing mammalian retina**. *Mol Cell Neurosci* (1996) **8**. DOI: 10.1006/mcne.1996.0051
11. Mitchell MD. **Prostaglandins during pregnancy and the perinatal period**. *Reproduction* (1981) **62**. DOI: 10.1530/jrf.0.0620305
12. Szczuko M, Kikut J, Komorniak N, Bilicki J, Celewicz Z, Ziętek M. **The role of arachidonic and linoleic acid derivatives in pathological pregnancies and the human reproduction process**. *Int J Mol Sci* (2020) **21**. DOI: 10.3390/ijms21249628
13. Phillips RJ, Fortier MA, López Bernal A. **Prostaglandin pathway gene expression in human placenta, amnion and choriodecidua is differentially affected by preterm and term labour and by uterine inflammation**. *BMC Pregnancy Childbirth* (2014) **14**. DOI: 10.1186/1471-2393-14-241
14. Maccarrone M. **Low fatty acid amide hydrolase and high anandamide levels are associated with failure to achieve an ongoing pregnancy after IVF and embryo transfer**. *Mol Hum Reprod* (2002) **8**. DOI: 10.1093/molehr/8.2.188
15. Habayeb OMH, Taylor AH, Evans MD, Cooke MS, Taylor DJ, Bell SC. **Plasma levels of the endocannabinoid anandamide in women–a potential role in pregnancy maintenance and labor**. *J Clin Endocrinol Metab* (2004) **89**. DOI: 10.1210/jc.2004-0681
16. Guida M, Ligresti A, De Filippis D, D’Amico A, Petrosino S, Cipriano M. **The levels of the endocannabinoid receptor CB2 and its ligand 2-arachidonoylglycerol are elevated in endometrial carcinoma**. *Endocrinology* (2010) **151**. DOI: 10.1210/en.2009-0883
17. Fügedi G, Molnár M, Rigó J, Schönléber J, Kovalszky I, Molvarec A. **Increased placental expression of cannabinoid receptor 1 in preeclampsia: An observational study**. *BMC Pregnancy Childbirth* (2014) **14**. DOI: 10.1186/s12884-014-0395-x
18. Abán C, Leguizamón GF, Cella M, Damiano A, Franchi AM, Farina MG. **Differential expression of endocannabinoid system in normal and preeclamptic placentas: effects on nitric oxide synthesis**. *Placenta* (2013) **34** 67-74. DOI: 10.1016/j.placenta.2012.10.009
19. Smith CJ, Baer RJ, Oltman SP, Breheny PJ, Bao W, Robinson JG. **Maternal dyslipidemia and risk for preterm birth**. *PloS One* (2018) **13** e0209579. DOI: 10.1371/JOURNAL.PONE.0209579
20. Wojcik-Baszko D, Charkiewicz K, Laudanski P. **Role of dyslipidemia in preeclampsia–a review of lipidomic analysis of blood, placenta, syncytiotrophoblast microvesicles and umbilical cord artery from women with preeclampsia**. *Prostaglandins Other Lipid Mediat* (2018) **139** 19-23. DOI: 10.1016/J.PROSTAGLANDINS.2018.09.006
21. Lindegaard MLS, Olivecrona G, Christoffersen C, Kratky D, Hannibal J, Petersen BL. **Endothelial and lipoprotein lipases in human and mouse placenta**. *J Lipid Res* (2005) **46**. DOI: 10.1194/jlr.M500277-JLR200
22. Maia J, Fonseca BM, Cunha SC, Braga J, Gonçalves D, Teixeira N. **Impact of tetrahydrocannabinol on the endocannabinoid 2-arachidonoylglycerol metabolism: ABHD6 and ABHD12 as novel players in human placenta**. *Biochim Biophys Acta Mol Cell Biol Lipids* (2020) **1865**. DOI: 10.1016/j.bbalip.2020.158807
23. Besenboeck C, Cvitic S, Lang U, Desoye G, Wadsack C. **Going into labor and beyond: phospholipase A2 in pregnancy**. *Reproduction* (2016) **151** R91-102. DOI: 10.1530/REP-15-0519
24. Costa MA, Fonseca BM, Keating E, Teixeira NA, Correia-Da-Silva G. **2-arachidonoylglycerol effects in cytotrophoblasts: Metabolic enzymes expression and apoptosis in BeWo cells**. *Reproduction* (2014) **147**. DOI: 10.1530/REP-13-0563
25. Shanbhag AG. **Utilization of information measure as a means of image thresholding**. *CVGIP Graph Model Image Process* (1994) **56**. DOI: 10.1006/cgip.1994.1037
26. Ridler TW, Calvard S. **Picture Thresholding Using an Iterative Selection Method**. *IEE Transactions on Systems, Man, and Cybernetics* (1978) **8**. DOI: 10.1109/TSMC.1978.4310039
27. Otsu N. **A threshold selection method from Gray-level histograms**. *IEEE Trans Systems Man Cybernetics* (1979) **9**. DOI: 10.1109/TSMC.1979.4310076
28. Ogasawara D, Deng H, Viader A, Baggelaar MP, Breman A, den Dulk H. **Rapid and profound rewiring of brain lipid signaling networks by acute diacylglycerol lipase inhibition**. *Proc Natl Acad Sci* (2016) **113** 26-33. DOI: 10.1073/pnas.1522364112
29. van Rooden EJ, Florea BI, Deng H, Baggelaar MP, van Esbroeck ACM, Zhou J. **Mapping**. *Nat Protoc* (2018) **13**. DOI: 10.1038/nprot.2017.159
30. Schneider H, Huch A. **Dual**. *Contrib Gynecol Obstet* (1985) **13**. DOI: 10.1159/000410668
31. Hirschmugl B, Perazzolo S, Sengers BG, Lewis RM, Gruber M, Desoye G. **Placental mobilization of free fatty acids contributes to altered materno-fetal transfer in obesity**. *Int J Obes* (2021) **45**. DOI: 10.1038/s41366-021-00781-x
32. Matyash V, Liebisch G, Kurzchalia TV, Shevchenko A, Schwudke D. **Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics**. *J Lipid Res* (2008) **49**. DOI: 10.1194/jlr.D700041-JLR200
33. Bollinger JG, Thompson W, Lai Y, Oslund RC, Hallstrand TS, Sadilek M. **Improved sensitivity mass spectrometric detection of eicosanoids by charge reversal derivatization**. *Anal Chem* (2010) **82**. DOI: 10.1021/ac100720p
34. Janssen APAA, van der Vliet D, Bakker AT, Jiang M, Grimm SH, Campiani G. **Development of a multiplexed activity-based protein profiling assay to evaluate activity of endocannabinoid hydrolase inhibitors**. *ACS Chem Biol* (2018) **13**. DOI: 10.1021/acschembio.8b00534
35. Baggelaar MP, Chameau PJPP, Kantae V, Hummel J, Hsu KL, Janssen F. **Highly selective, reversible inhibitor identified by comparative chemoproteomics modulates diacylglycerol lipase activity in neurons**. *J Am Chem Soc* (2015) **137**. DOI: 10.1021/jacs.5b04883
36. Baggelaar MP, Janssen FJ, van Esbroeck ACM, den Dulk H, Allarà M, Hoogendoorn S. **Development of an activity-based probe and in silico design reveal highly selective inhibitors for diacylglycerol lipase-α in brain**. *Angew Chemie Int Ed* (2013) **52**. DOI: 10.1002/anie.201306295
37. Eichmann TO, Lass A. **DAG tales: The multiple faces of diacylglycerol - stereochemistry, metabolism, and signaling**. *Cell Mol Life Sci* (2015) **72**. DOI: 10.1007/s00018-015-1982-3
38. Gauster M, Hiden U, Blaschitz A, Frank S, Lang U, Alvino G. **Dysregulation of placental endothelial lipase and lipoprotein lipase in intrauterine growth-restricted pregnancies**. *J Clin Endocrinol Metab* (2007) **92**. DOI: 10.1210/jc.2006-2403
39. Savinainen JR, Saario SM, Laitinen JT. **The serine hydrolases MAGL, ABHD6 and ABHD12 as guardians of 2-arachidonoylglycerol signalling through cannabinoid receptors**. *Acta Physiol* (2012) **204**. DOI: 10.1111/j.1748-1716.2011.02280.x
40. Blankman JL, Simon GM, Cravatt BF. **A comprehensive profile of brain enzymes that hydrolyze the endocannabinoid 2-arachidonoylglycerol**. *Chem Biol* (2007) **14**. DOI: 10.1016/j.chembiol.2007.11.006
41. Blankman JL, Long JZ, Trauger SA, Siuzdak G, Cravatt BF. **ABHD12 controls brain lysophosphatidylserine pathways that are deregulated in a murine model of the neurodegenerative disease PHARC**. *Proc Natl Acad Sci USA.* (2013) **110**. DOI: 10.1073/PNAS.1217121110/SUPPL_FILE/PNAS.201217121SI.PDF
42. Giang DK, Cravatt BF. **Molecular characterization of human and mouse fatty acid amide hydrolases**. *Proc Natl Acad Sci USA.* (1997) **94**. DOI: 10.1073/pnas.94.6.2238
43. Maccarrone M, Rossi S, Bari M, De Chiara V, Fezza F, Musella A. **Anandamide inhibits metabolism and physiological actions of 2-arachidonoylglycerol in the striatum**. *Nat Neurosci* (2008) **11**. DOI: 10.1038/nn2042
44. Musella A, Fresegna D, Rizzo FR, Gentile A, Bullitta S, De Vito F. **A novel crosstalk within the endocannabinoid system controls GABA transmission in the striatum**. *Sci Rep* (2017) **7** 7363. DOI: 10.1038/s41598-017-07519-8
45. Hashimotodani Y, Ohno-Shosaku T, Tanimura A, Kita Y, Sano Y, Shimizu T. **Acute inhibition of diacylglycerol lipase blocks endocannabinoid-mediated retrograde signalling: Evidence for on-demand biosynthesis of 2-arachidonoylglycerol**. *J Physiol* (2013) **591**. DOI: 10.1113/jphysiol.2013.254474
46. Perez-Riverol Y, Bai J, Bandla C, Hewapathirana S, García-Seisdedos D, Kamatchinathan S. **The PRIDE database resources in 2022: A Hub for mass spectrometry-based proteomics evidences**. *Nucleic Acids Res* (2022) **50**
|
---
title: 'Self-Reported Sleep Disturbance is an Independent Predictor of All-Cause Mortality
and Respiratory Disease Mortality in US Adults: A Population-Based Prospective Cohort
Study'
authors:
- Xinran Hou
- Jiajia Hu
- E Wang
- Jian Wang
- Zongbin Song
- Jie Hu
- Jian Shi
- Chengliang Zhang
journal: International Journal of Public Health
year: 2023
pmcid: PMC9971003
doi: 10.3389/ijph.2023.1605538
license: CC BY 4.0
---
# Self-Reported Sleep Disturbance is an Independent Predictor of All-Cause Mortality and Respiratory Disease Mortality in US Adults: A Population-Based Prospective Cohort Study
## Abstract
Objective: Self-reported sleep disturbance is common but its association with mortality has rarely been investigated.
Methods: This prospective cohort analysis included 41,257 participants enrolled in the National Health and Nutrition Examination Survey from 2005 to 2018. Self-reported sleep disturbance in the present study refers to the patients who have ever consulted doctors or other professionals for trouble sleeping. Univariate and multivariate survey-weighted Cox proportional hazards models were used to evaluate the association of self-reported sleep disturbance with all-cause and disease-specific mortality.
Results: Approximately $27.0\%$ of US adults were estimated to have self-reported sleep disturbance. After adjusting for all sociodemographic variables, health behavioral factors, and common comorbidities, participants with self-reported sleep disturbance tend to have higher all-cause mortality risk with a hazard ratio (HR) of 1.17 ($95\%$ CI, 1.04–1.32) and chronic lower respiratory disease mortality risk (HR, 1.88; $95\%$ CI, 1.26–2.80), but not cardiovascular disease mortality risk (HR, 1.19; $95\%$ CI, 0.96–1.46) and cancer mortality risk (HR, 1.10; $95\%$ CI, 0.90–1.35).
Conclusion: Self-reported sleep disturbance could be associated with higher mortality in adults, and may need to be paid more attention in public health management.
## Introduction
Sleep is a fundamental biological process integral to human health, and sleep problems are linked with cognitive, metabolic, cardiovascular, and immunological impairment [1]. Sleep duration is the most common indicator to evaluate sleep and numerous epidemical studies have demonstrated that both short and long sleep duration could be a causal risk factor for mortality (2–4) or major cardiovascular events (5–7). However, the measurement bias of sleep duration affects the robustness of its association with mortality risk [8], and the optimal sleep duration varies from person to person, since some individuals may complain of trouble sleeping, sleepiness, or anxiety even if they spend enough time in bed [9, 10]. Moreover, to confirm the diagnosis of a certain type of sleep disorder, a series of questionnaires and examinations (e.g., polysomnography) [11, 12] are necessary, which limits the evaluation of sleep health in the general population. On the other hand, subjective sleepiness is valid to indicate the drive for sleep [13] and is more effective in predicting voluntary decreases in social activity than sleep duration [10]. Excessive daytime sleepiness is associated with high risks of cardiovascular mortality [14]. Thus, self-evaluation of sleep is also useful and noteworthy in public health management.
Some researchers evaluated sleep disorders by asking participants whether they had ever told a doctor they had trouble sleeping (15–17), in which the medical help seeking behavior reflected the severity of the sleep disorder. Therefore, in our study, self-reported sleep disturbance referred to the condition of ever consulting a doctor or other professionals because of trouble sleeping, which was reported to be prevalent in $19.2\%$–$33.2\%$ of US adults [18]. It is no doubt that self-reported sleep disturbance seriously affects daily life, but whether the presence of such a sleep disorder affects long-term outcomes has not been investigated.
In the present study, we aimed to depict the prevalence of self-reported sleep disturbance and evaluate the association of self-reported sleep disturbance with all-cause and disease-specific mortality based on the National Health and Nutrition Examination Survey (NHANES) 2005 to 2018 dataset with linkage to the National Death Index (NDI) mortality files.
## Study Population
The NHANES is a continuous program to assess the health and nutritional status of the civilian population in the United States.; it is conducted by the National Center for Health Statistics (NCHS), which belongs to the U.S. Centers for Disease Control and Prevention (CDC). Details about the NHANES study design, study protocol, and data collection have been described elsewhere [19]. NHANES is approved by NCHS Ethics Review Board and written informed consent was obtained from all participants. The data in the present study were publicly released and were used in compliance with its data usage guidelines.
The present study population was limited to seven cycles of continuous NHANES (2005–2018). After excluding individuals less than 18 years old, during pregnancy, and without definite responses about sleep disorder, the remaining participants were included in the final analysis.
## Assessment of Self-Reported Sleep Disturbance
The questionnaire about sleep disorder (SLQ) has been added to NHANES since 2005, so we extracted sleep disorder data from 2005. In all the seven cycles of NHANES (2005–2018), the specific question, SLQ050, “Have you ever told a doctor or other health professionals that you have trouble sleeping?” was asked, and those participants responding “Yes” were considered to have self-reported sleep disturbance; participants responding “No” were considered not to have self-reported sleep disturbance; participants who chose “Refused” or “Don’t know” or did not answer this question were excluded in our analysis.
## Ascertainment of All-Cause Mortality and Disease-Specific Mortality
All-cause and disease-specific mortality were determined by linking to the NDI [20], in which the public-use Linked Mortality Files (LMF) are available for 1999–2018 NHANES. Death from cardiovascular disease (CVD) was defined as codes I00–I09, I11, I13, I20–I51, and I60–I69; death from cancer was defined as codes C00-C97; death from chronic lower respiratory disease (LRD) was defined as codes J40-J47, according to the 10th revision of the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD-10) guidelines.
The follow-up time has been calculated using months from the date of the interview to the date of death or the end of the mortality period (31 December 2019) for each participant of the NHANES.
## Assessment of Other Covariates
Demographics, basic physical examination (namely body mass index, BMI, calculated as weight in kilograms divided by the square of height in meters), behavioral factors, and chronic conditions collected when enrolled into the NHANES (baseline), along with an interview regarding sleep problems, were analyzed as covariates based on previous literature to reduce the effect of potential confounding (21–23).
Demographic covariates included age (years), sex (male, female), race (Mexican American, non-Hispanic white, non-Hispanic black, and others), education (less than 9th grade, 9–11th grade, high school, college or AA degree, college graduate or above), marital status (married, widowed, divorced, separated, never married, and living with partner), and family income to poverty ratio (PIR, a ratio of family income to poverty threshold, continuous from 0 to 4.99, values are 5 if the ratio is 5 or over).
Behavioral covariates included cigarette smoking (defined as never, former, and now smoker, based on responses to the following two questions: “Have you smoked at least 100 cigarettes in your entire life?” and “Do you now still smoke cigarettes?”); alcohol use, defined as never (had <12 drinks in lifetime), former (had ≥12 drinks in lifetime but did not drink last year), mild (≤1 drink per day for women or ≤2 drinks per day for men), moderate (≤2 drinks per day for women or ≤3 drinks per day for men), and heavy (>2 drinks per day for women or >3 drinks per day for men); caffeine consumption (evaluated as the mean consumption from 2 typical days and recorded in mg/day) [24]; and overall diet quality (assessed by the Healthy Eating Index, HEI-2015) [25].
Chronic comorbidity conditions, namely hypertension, diabetes, and coronary heart disease (CHD) were obtained through questionnaires, based on the response of participants to the question of whether or not they were ever told by a doctor or other health professionals that they had a specific disease.
## Statistical Analyses
Statistical analyses were performed according to NHANES recommended guidelines [26]. Since NHANES used a complex, multistage probability sampling design to select participants, and generated a representative sample of the US civilian non-institutionalized resident population, this survey design was considered and survey sample weights were used in most of our analyses unless otherwise noted. Data are presented as survey-weighted mean ($95\%$ confidence intervals, CIs) for continuous variables, or number (survey-weighted percentage, %) for categorical variables, respectively. Survey-weighted linear regression (svyglm) and survey-weighted Chi-square test (svytable) were used to detect statistical differences between the means and proportions between the two groups.
Univariate and multivariate Cox regression models were applied to evaluate the association between sleep disorders and mortality. For multivariate regression models, collinearity diagnostics were performed to ensure variance inflation factors (VIF) of all independent variables less than 5; then multiple stepwise regression was performed to determine the inclusion of each covariate. The crude and adjusted hazard ratios (HRs) and $95\%$ CIs were calculated for the risk of sleep disturbance associated with all-cause or disease-specific death.
According to the recommendation of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [27], we simultaneously showed the results unadjusted and with different covariate adjustment strategies in three models. Minimally adjusted models (Model 1) were adjusted for age, sex, and race. Partially adjusted models (Model 2) were additionally adjusted for education, marital status, PIR, BMI, smoking, alcohol use, caffeine consumption, and HEI-2015. Fully adjusted models (Model 3) were additionally adjusted for common chronic comorbidities, namely hypertension, diabetes, and CHD.
The stratified analyses were performed to demonstrate the effect of self-reported sleep disturbance in different subgroups using the fully adjusted model except for the specific stratification variable and interactions of self-reported sleep disturbance with stratification variables, which were inspected by the likelihood ratio test.
Data were analyzed with the use of the statistical packages R (The R Foundation; http://www.r-project.org; version 4.2.0) and EmpowerStats (www.empowerstats.net, X&Y solutions, Inc. Boston, Massachusetts; version 4.1). A two-sided $p \leq 0.05$ was considered significantly different.
## Selection of Participants
The total number of participants in 7 cycles of NHANES (2005–2018) was 70,190. After excluding 28,047 individuals with ages less than 18 years old, another 737 pregnant individuals, another 121 individuals ineligible for mortality follow-up (insufficient identifying data), and another 28 responding to the question, “Have you ever told a doctor or other health professional that you have trouble sleeping?”, with “Refused” or “Don’t know” or no response, 41,257 participants were finally included in our study (Figure 1).
**FIGURE 1:** *Flowchart of the present study indicating the included and excluded participants (National Health and Nutrition Examination Survey, the United States, 2005–2018).*
## Baseline Characteristics of Participants and Sleep Characteristics Based on the Presence of Self-Reported Sleep Disturbance
The enrolled 41,257 participants represented 228.6 million US adults. The survey-weighted mean age was 46.6 years old ($95\%$ CI: 46.2–47.1). The cohort comprised 20,897 women (survey-weighted percentage, $51.1\%$); 16,879 non-Hispanic White ($66.6\%$), 9,063 non-Hispanic Black ($11.4\%$), 6,595 Mexican American ($8.6\%$), and 8,702 participants of other ethnicities ($13.4\%$) (Table 1).
**TABLE 1**
| Variables | Self-reported sleep disturbance | Self-reported sleep disturbance.1 | Self-reported sleep disturbance.2 | Self-reported sleep disturbance.3 |
| --- | --- | --- | --- | --- |
| Variables | Total (n = 41,257) | No (n = 31,188) | Yes (n = 10,069) | P-value |
| Sex | | | | <0.001 |
| Female | 20,862 (50.6) | 14,981 (48.4) | 5,881 (58.6) | |
| Male | 20,395 (49.4) | 16,207 (51.6) | 4,188 (41.4) | |
| Age (years) | 46.6 (46.2, 47.1) | 45.2 (44.7, 45.7) | 50.5 (49.9, 51.0) | <0.001 |
| Race | | | | <0.001 |
| White | 16,897 (41) | 11,803 (63.7) | 5,094 (74.4) | |
| Black | 9,063 (22) | 6,937 (11.9) | 2,126 (10.1) | |
| Mexican | 6,595 (16) | 5,506 (9.9) | 1,089 (5.1) | |
| Other | 8,702 (21.1) | 6,942 (14.5) | 1760 (10.4) | |
| Education | | | | <0.001 |
| Less than 9th grade | 4,309 (11.1) | 3,401 (6.1) | 908 (4.6) | |
| 9–11th grade | 5,570 (14.3) | 4,225 (11.0) | 1,345 (9.7) | |
| High school graduate/GED | 8,932 (23) | 6,634 (23.1) | 2,298 (24.2) | |
| Some college or AA degree | 6,776 (17.4) | 4,790 (17.9) | 1986 (21.5) | |
| College graduate or above | 13,255 (34.1) | 10,001 (41.9) | 3,254 (40.0) | |
| Marital status | | | | <0.001 |
| Never married | 7,583 (19.2) | 5,984 (19.6) | 1,599 (15.5) | |
| Living with partner | 3,121 (7.9) | 2,401 (8.2) | 720 (7.7) | |
| Married | 19,768 (50.1) | 15,180 (55.8) | 4,588 (51.9) | |
| Separated | 1,325 (3.4) | 909 (2.2) | 416 (3.1) | |
| Divorced | 4,266 (10.8) | 2,781 (8.8) | 1,485 (14.1) | |
| Widowed | 3,361 (8.5) | 2,320 (5.4) | 1,041 (7.7) | |
| PIR | 3.0 (2.9, 3.0) | 3.0 (2.9, 3.0) | 3.0 (2.9, 3.1) | 0.951 |
| BMI (kg/m2) | 28.9 (28.7, 29.1) | 28.4 (28.3, 28.6) | 30.1 (29.9, 30.4) | <0.001 |
| Smoker | | | | <0.001 |
| Never | 22,263 (56) | 17,611 (58.8) | 4,652 (46.5) | |
| Former | 9,381 (23.6) | 6,586 (22.3) | 2,795 (29.4) | |
| Now | 8,126 (20.4) | 5,651 (18.9) | 2,475 (24.0) | |
| Alcohol user | | | | <0.001 |
| Never | 5,183 (15.3) | 4,177 (12.6) | 1,006 (8.7) | |
| Former | 5,478 (16.2) | 3,778 (12.1) | 1700 (16.2) | |
| Mild | 11,173 (33) | 8,272 (35.8) | 2,901 (37.1) | |
| Moderate | 5,147 (15.2) | 3,767 (17.0) | 1,380 (18.4) | |
| Heavy | 6,861 (20.3) | 5,285 (22.5) | 1,576 (19.6) | |
| Caffeine consumption (mg/day) | 167.3 (162.8, 171.7) | 161.4 (156.4, 166.3) | 182.8 (177.1, 188.6) | <0.001 |
| HEI-2015 | 50.4 (50.1, 50.8) | 50.6 (50.2, 51.0) | 50.1 (49.7, 50.5) | 0.03 |
| Hypertension | | | | <0.001 |
| No | 26,947 (65.4) | 21,882 (73.6) | 5,065 (55.9) | |
| Yes | 14,248 (34.6) | 9,257 (26.4) | 4,991 (44.1) | |
| Diabetes | | | | <0.001 |
| No | 35,196 (85.4) | 27,317 (90.5) | 7,879 (83.2) | |
| Borderline | 899 (2.2) | 584 (1.7) | 315 (2.9) | |
| Yes | 5,131 (12.4) | 3,265 (7.7) | 1866 (13.9) | |
| CHD | | | | <0.001 |
| No | 37,099 (95.8) | 27,994 (97.2) | 9,105 (94.79) | |
| Yes | 1,637 (4.2) | 998 (2.8) | 639 (5.21) | |
The self-reported sleep disturbance was estimated to affect $27.0\%$ ($95\%$ CI: $26.1\%$–$27.8\%$) of the whole population. Table 1 also shows the demographics, behavioral variables, and chronic comorbidities in all participants stratified by those with or without the self-reported sleep disturbance. Subjects reporting sleep disturbance were likely to be older (50.5 vs. 45.2), female ($58.6\%$ vs. $48.4\%$), white ($74.4\%$ vs. $63.7\%$), smoker ($22.6\%$ vs. $18.1\%$ for current smoker and $28.1\%$ vs. $21.5\%$ for former smoker), with higher BMI (30.1 vs. 28.4), diabetes ($13.0\%$ vs. $7.3\%$), hypertension ($42.5\%$ vs. $25.4\%$), and CHD ($5.2\%$ vs. $2.8\%$) at the time of their participation in NHANES, than those not complaining about sleep.
To depict the detailed manifestations of self-reported sleep disturbance, general sleep symptoms were compared according to whether participants had self-reported sleep disturbance. As shown in Table 2, it was more common for individuals with self-reported sleep disturbance to have shorter sleep duration, to have trouble falling asleep, to wake up during the night, to wake up too early in the morning, to snore or stop breathing during sleep, to take pills to help sleep, to feel like they are not getting enough sleep, to feel unrested and overly sleepy during the day.
**TABLE 2**
| Unnamed: 0 | Self-reported sleep disturbance | Self-reported sleep disturbance.1 | Self-reported sleep disturbance.2 |
| --- | --- | --- | --- |
| | No, mean (95% CI) or n (%) | Yes, mean (95% CI) or n (%) | P-value |
| How much sleep do you get (hours)? | 7.0 (7.0, 7.1) | 6.5 (6.4, 6.6) | <0.001 |
| How long to fall asleep (minutes)? | 18.9 (18.2, 19.6) | 29.8 (28.6, 31.0) | <0.001 |
| How often have trouble falling asleep? | | | <0.001 |
| Never | 4,425 (45.9) | 457 (16.9) | |
| Rarely (1 time a month) | 1830 (23.7) | 356 (16.7) | |
| Sometimes (2–4 times a month) | 1720 (20.6) | 653 (25.8) | |
| Often (5–15 times a month) | 528 (6.5) | 511 (20.5) | |
| Almost always (16–30 times a month) | 334 (3.4) | 526 (20.1) | |
| How often wake up during night? | | | <0.001 |
| Never | 4,020 (41.8) | 440 (17.0) | |
| Rarely (1 time a month) | 1809 (22.4) | 312 (14.0) | |
| Sometimes (2–4 times a month) | 1911 (22.8) | 648 (26.0) | |
| Often (5–15 times a month) | 737 (9.5) | 600 (23.9) | |
| Almost always (16–30 times a month) | 358 (3.6) | 500 (19.2) | |
| How often wake up too early in morning? | | | <0.001 |
| Never | 4,546 (49.4) | 640 (25.6) | |
| Rarely (1 time a month) | 1,650 (20.5) | 366 (17.0) | |
| Sometimes (2–4 times a month) | 1,607 (18.6) | 574 (24.0) | |
| Often (5–15 times a month) | 669 (8.0) | 517 (18.7) | |
| Almost always (16–30 times a month) | 359 (3.5) | 404 (14.6) | |
| How often do you snort or stop breathing? | | | <0.001 |
| Never | 13,224 (81.2) | 3,372 (67.4) | |
| Rarely (1–2 nights/week) | 1,577 (10.2) | 697 (13.5) | |
| Occasionally (3–4 nights/week) | 835 (5.0) | 501 (8.9) | |
| Frequently—5 or more nights a week | 588 (3.6) | 542 (10.2) | |
| How often take pills to help you sleep? | | | <0.001 |
| Never | 7,952 (89.0) | 1,413 (54.7) | |
| Rarely (1 time a month) | 309 (4.0) | 148 (6.4) | |
| Sometimes (2–4 times a month) | 311 (3.8) | 260 (11.4) | |
| Often (5–15 times a month) | 101 (1.3) | 205 (8.6) | |
| Almost always (16–30 times a month) | 163 (1.9) | 478 (18.8) | |
| How often did you not get enough sleep? | | | <0.001 |
| Never | 3,320 (30.3) | 374 (12.5) | |
| Rarely (1 time a month) | 1,634 (20.6) | 325 (13.2) | |
| Sometimes (2–4 times a month) | 2,258 (29.1) | 682 (28.8) | |
| Often (5–15 times a month) | 1,055 (13.6) | 542 (23.6) | |
| Almost always (16–30 times a month) | 547 (6.4) | 570 (22.0) | |
| How often feel unrested during the day? | | | <0.001 |
| Never | 3,436 (32.0) | 415 (13.2) | |
| Rarely (1 time a month) | 1,486 (18.7) | 285 (11.7) | |
| Sometimes (2–4 times a month) | 2,313 (29.1) | 666 (27.2) | |
| Often (5–15 times a month) | 1,047 (14.0) | 574 (25.2) | |
| Almost always (16–30 times a month) | 541 (6.2) | 562 (22.8) | |
| How often feel overly sleepy during day? | | | <0.001 |
| Never | 5,478 (26.0) | 782 (11.6) | |
| Rarely (1 time a month) | 4,081 (25.8) | 929 (16.6) | |
| Sometimes (2–4 times a month) | 4,965 (31.1) | 1768 (32.2) | |
| Often (5–15 times a month) | 1958 (12.7) | 1,306 (25.2) | |
| Almost always (16–30 times a month) | 851 (4.4) | 847 (14.4) | |
| Have you ever been told by doctors to have sleep disorder? | | | <0.001 |
| No | 22,234 (97.9) | 5,003 (73.8) | |
| Yes | 432 (2.1) | 1901 (26.2) | |
During a median follow-up of 88 months (interquartile: 49–129), 4,521 deaths were documented until 31 December 2019. Among them, 1,363 deaths were due to cardiovascular disease (CVD); 1,026 deaths were due to malignant neoplasms; 228 deaths were due to chronic lower respiratory diseases (LRD), and more details about the less common causes of death were shown in Supplementary Figure S1.
## Univariate Analysis of the Association of Self-Reported Sleep Disturbance With All-Cause and Disease-Specific Mortality
The unadjusted weighted Kaplan-Meier curves for all-cause mortality and disease-specific mortality by self-reported sleep disturbance were shown in Figure 2, followed by weighted log-rank tests. Then, univariate Cox regression models were used to quantify the risk of self-reported sleep disturbance with HR of 1.47 ($95\%$ CI: 1.36–1.59) for all-cause mortality, HR of 1.32 (1.13–1.54) for CVD mortality HR of 1.30 (1.10–1.53) for cancer mortality, and HR of 2.49 (1.85–3.35) for chronic LRD mortality (Supplementary Table S1). Univariate analysis of other covariates was also summarized in Supplementary Table S1.
**FIGURE 2:** *Survey-weighted Kaplan-Meier survival curves and log-rank tests comparing mortality due to all-cause (A), cardiovascular disease (B), cancer (C), and chronic lower respiratory disease (D) in participants with or without self-reported sleep disturbance. CVD, cardiovascular disease; LRD, lower respiratory diseases (National Health and Nutrition Examination Survey, the United States, 2005–2018).*
## Multivariate Analysis of the Association Between Self-Reported Sleep Disturbance and All-Cause and Disease-Specific Mortality
After adjustment for age, sex, and race, self-reported sleep disturbance was associated with elevated risks of all-cause mortality (adjusted HR: 1.26, $95\%$ CI:1.16–1.38), CVD mortality (1.17, 1.01–1.35), and chronic LRD mortality (2.08, 1.51–2.87), but no cancer mortality (1.13, 0.95–1.36) (Table 3).
**TABLE 3**
| Outcome | Unadjusted | Unadjusted.1 | Model 1 | Model 1.1 | Model 2 | Model 2.1 | Model 3 | Model 3.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Outcome | HR (95% CI) | P-value | HR (95% CI) | P-value | HR (95% CI) | P-value | HR (95% CI) | P-value |
| All-cause mortality | 1.47 (1.36, 1.59) | <0.001 | 1.26 (1.16, 1.38) | <0.001 | 1.23 (1.10, 1.38) | <0.001 | 1.17 (1.04, 1.32) | 0.011 |
| CVD mortality | 1.32 (1.13, 1.54) | <0.001 | 1.17 (1.01, 1.35) | 0.038 | 1.26 (1.03, 1.54) | 0.023 | 1.19 (0.96, 1.46) | 0.114 |
| Cancer mortality | 1.30 (1.10, 1.53) | 0.002 | 1.13 (0.95, 1.36) | 0.173 | 1.11 (0.91, 1.35) | 0.312 | 1.10 (0.90, 1.35) | 0.370 |
| LRD mortality | 2.49 (1.85, 3.35) | <0.001 | 2.08 (1.51, 2.87) | <0.001 | 2.04 (1.39, 3.00) | <0.001 | 1.88 (1.26, 2.80) | 0.002 |
After further adjustment of other demographics and behavioral factors, namely education, marital status, PIR, BMI, smoking, alcohol use, caffeine consumption, and HEI-2015, self-reported sleep disturbance was associated with elevated risks of all-cause mortality (1.23, 1.10–1.38), CVD mortality (1.26, 1.03–1.54), chronic LRD mortality (2.04, 1.39–3.00), but no cancer mortality (1.11, 0.91–1.35, Table 3).
Finally, in the fully adjusted model of all covariates, with chronic comorbidities (diabetes, hypertension, CHD) included, self-reported sleep disturbance was still associated with elevated risks of all-cause mortality (1.17, 1.04–1.32) and chronic LRD mortality (1.88, 1.26–2.80), but not CVD mortality (1.19, 0.96–1.46) and cancer mortality (1.10, 0.90–1.35). As shown in Table 3, the associations in fully-adjusted models were largely attenuated, compared with that in non-adjusted models.
## The Stratified Analyses and Interaction Between Self-Reported Sleep Disturbance and Stratification Variables
Stratified analyses were performed to demonstrate the effect of self-reported sleep disturbance in different subgroups using the fully adjusted model, which was adjusted for sex, age, race, education, marital status, PIR, BMI, smoke, alcohol use, caffeine consumption, HEI-2015 and comorbidities of hypertension, diabetes, and CHD, except for the specific stratification variable of that analysis. The self-reported sleep disturbance was a significant risk factor for all-cause mortality in populations of males (adjusted HR: 1.28, $95\%$ CI: 1.09–1.50), less than 60 years old (1.47, 1.19–1.81), married (1.24, 1.03–1.49), with an education level of high school or below (1.20, 1.01–1.23), with low PIR (1.28, 1.09–1.51), with low BMI (1.23, 1.02–1.48), smoker (1.23, 1.07–1.41), with more caffeine consumption (1.23, 1.04–1.45), and without diabetes (1.20, 1.06–1.36) (Figure 3).
**FIGURE 3:** *Stratified analyses of the associations of self-reported sleep disturbance and all-cause mortality and the interaction test between self-reported sleep disturbance and stratification variables. HR, hazard ratio; PIR, family income-to-poverty ratio; BMI, body mass index; HEI, Healthy Eating Index; CHD, coronary heart disease, P50, 50th percentile (National Health and Nutrition Examination Survey, the United States, 2005–2018).*
The interaction analyses were performed to test the possible interaction of self-reported sleep disturbance with the stratification variable after adjusting all other variables. The interaction with the dichotomous age of 60 years old is remarkable ($p \leq 0.001$), indicating that self-reported sleep disturbance would be a more significant risk factor for all-cause mortality in the young adults than in the old population. The interactions with other variables are not statistically significant (Figure 3).
## Discussion
In this nationally population-based prospective cohort study, we demonstrated that self-reported sleep disturbance affected about $27\%$ of US adults (approximately 61.7 million), more prevalent in populations who are older, female, white, and smokers, and those who have higher BMI, more caffeine consumption, diabetes, hypertension, and CHD. Individuals with self-reported sleep disturbance tended to have a higher risk of mortality from all-cause, CVD, cancer, and chronic LRD. After adjusting all covariates, the HR of self-reported sleep disturbance was largely attenuated, but it still was associated with a higher risk of mortality from all-cause and chronic LRD, suggesting the potential beneficial role to prevent and treat respiratory disease in individuals with self-reported sleep disturbance. There was potential interaction between self-reported sleep disturbance and age; the self-reported sleep disturbance tended to be associated with a higher risk of all-cause mortality in young adults than in old adults.
According to the International Classification of Sleep Disorders, third edition (ICSD-3), sleep disorders can be classified into insomnia, sleep-disordered breathing, central disorders of hypersomnolence, circadian rhythm sleep-wake disorders, parasomnias, and sleep-related movement disorders [28, 29]. Insomnia is the most common type and affects nearly one-third of the adult population; and even if only taking into account insomnia with symptoms severe enough to cause daytime consequences, the prevalence is still more than $10\%$ in adults and higher in women than men ($17.6\%$ vs. $10.1\%$) [30]. The prevalence of other types of sleep disorders varied greatly based on the specific condition [30, 31]. However, the conventional diagnosis of sleep disorder may underestimate the prevalence of sleep disorder, because of the delayed seeking of medical advice, complicated diagnostic procedures, and some uncategorized subtypes of sleep disorder [32, 33]. If all sleep disorders that require consultation with doctors were taken as a whole, whether it could affect long-term outcomes and what the effect size would be, have not been explored.
In the present study, we demonstrated that self-reported sleep disturbance is prevalent in $27\%$ of US adults, consistent with previous reports [18], and is associated with higher risks of mortality from all-cause by $47\%$, from CVD by $32\%$, from cancer by $30\%$, and from chronic LRD by $149\%$, regardless of other factors. To explore the independent role of self-reported sleep disturbance, we adjusted demographic, socioeconomic, lifestyle, and comorbidities conditions step by step, and we found the HRs of self-reported sleep disturbance were gradually attenuated. This was probably because sleep disturbance could be influenced by these behavioral factors and it could also be an accompanying symptom of some chronic comorbidities, such as diabetes [31, 34, 35], hypertension (17, 36–38), and CHD [39, 40]. Actually, sleep intervention is already one of the recommended preventive and therapeutic strategies for such chronic diseases (41–44).
Many researchers have demonstrated that sleep disorders could promote the incidence of CVD and its mortality [7, 40, 45], possibly by increasing adverse cardiometabolic risk [39, 46], and healthy sleep behavior is recommended to promote ideal cardiac health, along with efforts to address other established risk factors including blood pressure, cholesterol, diet, blood glucose, physical activity, weight, and smoking cessation [46]. However, in our fully adjusted model, the HR of self-reported sleep disturbance to CVD mortality did not reach the statistical significance level. This was probably because we have balanced major CVD mortality-related risk factors, such as BMI, hypertension, and CHD, which might be the main mechanism of sleep disorders affecting CVD mortality.
In the present study, we found that self-reported sleep disturbance was significantly associated with higher mortality from chronic LRD. It has been shown that the prevalence of sleep disturbance is higher in individuals with chronic respiratory disease, such as chronic obstructive pulmonary disease (COPD) [47] and asthma [48], which was probably because sleep disturbance was a premonitory or accompanying symptom of respiratory disease. Another possible explanation would be that sleep disturbances usually overlapped with some respiratory diseases (49–51), which was associated with a worse prognosis (52–54). Our results indicated that effective prevention and treatment of respiratory diseases may be worth advocating for in individuals with sleep disturbances.
In our stratified analysis and interaction analysis, we found that after all other covariates were adjusted, the association of self-reported sleep disturbance with all-cause mortality was more pronounced in a population of less than 60 years old and the interaction with age is statistically significant. This is consistent with previous research [45, 55], and the possible reasons are that bad habits, such as more sedentary time, more screen use time, and more sleep deprivation appear more in young adults and were not adjusted in the model; then similarly the higher prevalence of chronic diseases in senior adults weaken the effect of sleep disturbance since they were adjusted in the model. Recent research indicated that poor sleep behavior is independently associated with an increased risk of subclinical multi-territory atherosclerosis in middle-aged participants free of known CVD history [56], suggesting that sleep disturbance in the seemingly healthy population should be noticeable to public health practitioners.
Several possible mechanisms contributed to the mortality increase due to sleep disturbances, such as dysfunction of the autonomic nervous system [57, 58], endothelial function [59], metabolic regulation [60], inflammatory factors secretion (61–63), and arrhythmogenic effects on the heart [64, 65]. However, due to the complexity and heterogeneity of sleep disturbance, more well-designed experimental research is needed to verify these mechanisms.
The strengths of our study included the nationally representative sampling of participants, the corresponding survey-weighted analyses, the relatively large sample size (n > 40,000), and the adjustment of many confounding factors, which made the findings applicable to the US adult population. In addition, this was the first study to treat self-reported sleep disturbance as an exposure factor and to evaluate its association with mortality as far as we know.
However, some limitations of the present study must be acknowledged. Firstly, the participants were sampled from US non-institutional civilians, so we should be cautious when attempting to apply the results to other populations. Secondly, the effect of sleep disturbance on health is chronic and continuous, so distributing those participants into the self-reported sleep disturbance group, whose sleep function was just disturbed by some transient factors, may lead to the underestimation of the correlation between sleep disturbance and mortality. Thirdly, the definition of self-reported sleep disturbance in the present study is dichotomous and somewhat simple, although facilitating its application in clinical management, and evaluation with a period of follow-up or objective measurement with more detailed classification is needed from the perspective of academic rigor. Finally, residual or unmeasured confounding cannot be completely ruled out despite every effort made to adjust the major confounding factors.
## Conclusion
Self-reported sleep disturbance is associated with a higher risk of all-cause mortality. Among the major leading causes of death, a higher risk of chronic LRD mortality was associated with the self-reported sleep disturbance. Future large-scale longitudinal studies with a more elaborate classification of sleep disturbances are still needed to clarify the cause-and-effect relationship. Finally, complaints about sleep disturbance should be paid attention to in public health evaluation; moreover, effective prevention and treatment of respiratory disease in individuals with sleep disturbances are recommended.
## Ethics Statement
The studies involving human participants were reviewed and approved by Research Ethics Review Board of the National Health and Nutrition Examination Survey. The patients/participants provided their written informed consent to participate in this study.
## Author Contributions
Conceptualization: XH and CZ; methodology: XH, CZ, and JW; software: XH, CZ, and JW; validation: CZ, JiaH, and JieH; formal analysis: XH, CZ, JS, and JW; investigation: JieH, JiaH, and XH; resources: EW, ZS, and CZ; data curation: XH, CZ, and JW; writing-original draft preparation: CZ; writing-review and editing: XH and CZ; visualization: XH, ZS, and JW; supervision: EW, ZS, and CZ; project administration: XH, CZ, and JW; funding acquisition: EW, ZS, and CZ; All authors have read and agreed to the published version of the manuscript.
## Conflict of Interest
The authors declare that they do not have any conflicts of interest.
## Supplementary Material
The Supplementary Material for this article can be found online at: https://www.ssph-journal.org/articles/10.3389/ijph.2023.1605538/full#supplementary-material
## References
1. Lane JM, Qian J, Mignot E, Redline S, Scheer F, Saxena R. **Genetics of Circadian Rhythms and Sleep in Human Health and Disease**. *Nat Rev Genet* (2022) **24** 4-20. DOI: 10.1038/s41576-022-00519-z
2. Wang C, Bangdiwala SI, Rangarajan S, Lear SA, AlHabib KF, Mohan V. **Association of Estimated Sleep Duration and Naps with Mortality and Cardiovascular Events: a Study of 116 632 People from 21 Countries**. *Eur Heart J* (2019) **40** 1620-9. DOI: 10.1093/eurheartj/ehy695
3. Svensson T, Saito E, Svensson AK, Melander O, Orho-Melander M, Mimura M. **Association of Sleep Duration with All- and Major-Cause Mortality Among Adults in Japan, China, Singapore, and Korea**. *JAMA Netw open* (2021) **4** e2122837. DOI: 10.1001/jamanetworkopen.2021.22837
4. García-Perdomo HA, Zapata-Copete J, Rojas-Cerón CA. **Sleep Duration and Risk of All-Cause Mortality: a Systematic Review and Meta-Analysis**. *Epidemiol Psychiatr Sci* (2019) **28** 578-88. DOI: 10.1017/S2045796018000379
5. Daghlas I, Dashti HS, Lane J, Aragam KG, Rutter MK, Saxena R. **Sleep Duration and Myocardial Infarction**. *J Am Coll Cardiol* (2019) **74** 1304-14. DOI: 10.1016/j.jacc.2019.07.022
6. Ai S, Zhang J, Zhao G, Wang N, Li G, So HC. **Causal Associations of Short and Long Sleep Durations with 12 Cardiovascular Diseases: Linear and Nonlinear Mendelian Randomization Analyses in UK Biobank**. *Eur Heart J* (2021) **42** 3349-57. DOI: 10.1093/eurheartj/ehab170
7. Fan M, Sun D, Zhou T, Heianza Y, Lv J, Li L. **Sleep Patterns, Genetic Susceptibility, and Incident Cardiovascular Disease: a Prospective Study of 385 292 UK Biobank Participants**. *Eur Heart J* (2020) **41** 1182-9. DOI: 10.1093/eurheartj/ehz849
8. Kurina LM, McClintock MK, Chen JH, Waite LJ, Thisted RA, Lauderdale DS. **Sleep Duration and All-Cause Mortality: a Critical Review of Measurement and Associations**. *Ann Epidemiol* (2013) **23** 361-70. DOI: 10.1016/j.annepidem.2013.03.015
9. Kocevska D, Lysen TS, Dotinga A, Koopman-Verhoeff ME, Luijk M, Antypa N. **Sleep Characteristics across the Lifespan in 1.1 Million People from the Netherlands, United Kingdom and United States: a Systematic Review and Meta-Analysis**. *Nat Hum Behav* (2021) **5** 113-22. DOI: 10.1038/s41562-020-00965-x
10. Holding BC, Sundelin T, Schiller H, Åkerstedt T, Kecklund G, Axelsson J. **Sleepiness, Sleep Duration, and Human Social Activity: An Investigation into Bidirectionality Using Longitudinal Time-Use Data**. *Proc Natl Acad Sci United States America* (2020) **117** 21209-17. DOI: 10.1073/pnas.2004535117
11. Kapur VK, Auckley DH, Chowdhuri S, Kuhlmann DC, Mehra R, Ramar K. **Clinical Practice Guideline for Diagnostic Testing for Adult Obstructive Sleep Apnea: An American Academy of Sleep Medicine Clinical Practice Guideline**. *J Clin Sleep Med : JCSM : official Publ Am Acad Sleep Med* (2017) **13** 479-504. DOI: 10.5664/jcsm.6506
12. Morin CM, Belleville G, Bélanger L, Ivers H. **The Insomnia Severity Index: Psychometric Indicators to Detect Insomnia Cases and Evaluate Treatment Response**. *Sleep* (2011) **34** 601-8. DOI: 10.1093/sleep/34.5.601
13. Shochat T, Santhi N, Herer P, Dijk DJ, Skeldon AC. **Sleepiness Is a Signal to Go to Bed: Data and Model Simulations**. *Sleep* (2021) **44** zsab123. DOI: 10.1093/sleep/zsab123
14. Li J, Covassin N, Bock JM, Mohamed EA, Pappoppula LP, Shafi C. **Excessive Daytime Sleepiness and Cardiovascular Mortality in US Adults: A NHANES 2005-2008 Follow-Up Study**. *Nat Sci Sleep* (2021) **13** 1049-59. DOI: 10.2147/NSS.S319675
15. Rahman HH, Niemann D, Yusuf KK. **Association of Urinary Arsenic and Sleep Disorder in the US Population: NHANES 2015-2016**. *Environ Sci Pollut Res Int* (2022) **29** 5496-504. DOI: 10.1007/s11356-021-16085-6
16. Beydoun HA, Beydoun MA, Jeng HA, Zonderman AB, Eid SM. **Bisphenol-A and Sleep Adequacy Among Adults in the National Health and Nutrition Examination Surveys**. *Sleep* (2016) **39** 467-76. DOI: 10.5665/sleep.5466
17. Li C, Shang S. **Relationship between Sleep and Hypertension: Findings from the NHANES (2007-2014)**. *Int J Environ Res Public Health* (2021) **18** 7867. DOI: 10.3390/ijerph18157867
18. **QuickStats: Percentage* of Adults Aged ≥20 Years Who Ever Told A Doctor that They Had Trouble Sleeping,† by Age Group and Sex - National Health and Nutrition Examination Survey, 2013-2014**. *MMWR Morbidity mortality weekly Rep* (2016) **65** 1323. DOI: 10.15585/mmwr.mm6546a11
19. 19.
Centers for Disease Control and Prevention NCfHS. NHANES Questionnaires, Datasets, and Related Documentation (2022). Available from: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx
(Accessed on September 3, 2022).. *NHANES Questionnaires, Datasets, and Related Documentation* (2022)
20. 20.
Statistics NCFH (2022). National Death Index. Available from: https://www.cdc.gov/nchs/ndi/index.htm (Accessed on September 3, 2022).. *National Death Index* (2022)
21. Huang BH, Duncan MJ, Cistulli PA, Nassar N, Hamer M, Stamatakis E. **Sleep and Physical Activity in Relation to All-Cause, Cardiovascular Disease and Cancer Mortality Risk**. *Br J Sports Med* (2022) **56** 718-24. DOI: 10.1136/bjsports-2021-104046
22. Jin Q, Yang N, Dai J, Zhao Y, Zhang X, Yin J. **Association of Sleep Duration with All-Cause and Cardiovascular Mortality: A Prospective Cohort Study**. *Front Public Health* (2022) **10** 880276. DOI: 10.3389/fpubh.2022.880276
23. Wang YH, Wang J, Chen SH, Li JQ, Lu QD, Vitiello MV. **Association of Longitudinal Patterns of Habitual Sleep Duration with Risk of Cardiovascular Events and All-Cause Mortality**. *JAMA Netw open* (2020) **3** e205246. DOI: 10.1001/jamanetworkopen.2020.5246
24. Chaudhary NS, Grandner MA, Jackson NJ, Chakravorty S. **Caffeine Consumption, Insomnia, and Sleep Duration: Results from a Nationally Representative Sample**. *Nutrition (Burbank, Los Angeles County, Calif)* (2016) **32** 1193-9. DOI: 10.1016/j.nut.2016.04.005
25. Shan Z, Rehm CD, Rogers G, Ruan M, Wang DD, Hu FB. **Trends in Dietary Carbohydrate, Protein, and Fat Intake and Diet Quality Among US Adults, 1999-2016**. *JAMA* (2019) **322** 1178-87. DOI: 10.1001/jama.2019.13771
26. Johnson CL, Paulose-Ram R, Ogden CL, Carroll MD, Kruszon-Moran D, Dohrmann SM. **National Health and Nutrition Examination Survey: Analytic Guidelines, 1999-2010. Vital and Health Statistics Series 2**. *Data Eval Methods Res* (2013) **2013** 1-24
27. Vandenbroucke JP, von Elm E, Altman DG, Gøtzsche PC, Mulrow CD, Pocock SJ. **Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): Explanation and Elaboration**. *Ann Intern Med* (2007) **147** W163-94. DOI: 10.7326/0003-4819-147-8-200710160-00010-w1
28. Sateia MJ. **International Classification of Sleep Disorders-Third Edition: Highlights and Modifications**. *Chest* (2014) **146** 1387-94. DOI: 10.1378/chest.14-0970
29. Karna B, Sankari A, Tatikonda G. *Sleep Disorder. StatPearls* (2022)
30. Morin CM, Jarrin DC, Ivers H, Mérette C, LeBlanc M, Savard J. **Incidence, Persistence, and Remission Rates of Insomnia over 5 Years**. *JAMA Netw open* (2020) **3** e2018782. DOI: 10.1001/jamanetworkopen.2020.18782
31. Schipper SBJ, Van Veen MM, Elders PJM, van Straten A, Van Der Werf YD, Knutson KL. **Sleep Disorders in People with Type 2 Diabetes and Associated Health Outcomes: a Review of the Literature**. *Diabetologia* (2021) **64** 2367-77. DOI: 10.1007/s00125-021-05541-0
32. Holder S, Narula NS. **Common Sleep Disorders in Adults: Diagnosis and Management**. *Am Fam Physician* (2022) **105** 397-405. PMID: 35426627
33. Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. **The Occurrence of Sleep-Disordered Breathing Among Middle-Aged Adults**. *New Engl J Med* (1993) **328** 1230-5. DOI: 10.1056/NEJM199304293281704
34. Khalil M, Power N, Graham E, Deschênes SS, Schmitz N. **The Association between Sleep and Diabetes Outcomes - A Systematic Review**. *Diabetes Res Clin Pract* (2020) **161** 108035. DOI: 10.1016/j.diabres.2020.108035
35. Reutrakul S, Mokhlesi B. **Obstructive Sleep Apnea and Diabetes: A State of the Art Review**. *Chest* (2017) **152** 1070-86. DOI: 10.1016/j.chest.2017.05.009
36. Jarrin DC, Alvaro PK, Bouchard MA, Jarrin SD, Drake CL, Morin CM. **Insomnia and Hypertension: A Systematic Review**. *Sleep Med Rev* (2018) **41** 3-38. DOI: 10.1016/j.smrv.2018.02.003
37. Li L, Gan Y, Zhou X, Jiang H, Zhao Y, Tian Q. **Insomnia and the Risk of Hypertension: A Meta-Analysis of Prospective Cohort Studies**. *Sleep Med Rev* (2021) **56** 101403. DOI: 10.1016/j.smrv.2020.101403
38. Bathgate CJ, Fernandez-Mendoza J. **Insomnia, Short Sleep Duration, and High Blood Pressure: Recent Evidence and Future Directions for the Prevention and Management of Hypertension**. *Curr Hypertens Rep* (2018) **20** 52. DOI: 10.1007/s11906-018-0850-6
39. Tobaldini E, Fiorelli EM, Solbiati M, Costantino G, Nobili L, Montano N. **Short Sleep Duration and Cardiometabolic Risk: from Pathophysiology to Clinical Evidence**. *Nat Rev Cardiol* (2019) **16** 213-24. DOI: 10.1038/s41569-018-0109-6
40. Kwok CS, Kontopantelis E, Kuligowski G, Gray M, Muhyaldeen A, Gale CP. **Self-Reported Sleep Duration and Quality and Cardiovascular Disease and Mortality: A Dose-Response Meta-Analysis**. *J Am Heart Assoc* (2018) **7** e008552. DOI: 10.1161/JAHA.118.008552
41. Arnett DK, Blumenthal RS, Albert MA, Buroker AB, Goldberger ZD, Hahn EJ. **2019 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Circulation* (2019) **140** e596-e646. DOI: 10.1161/CIR.0000000000000678
42. Cai H, Wang XP, Yang GY. **Sleep Disorders in Stroke: An Update on Management**. *Aging Dis* (2021) **12** 570-85. DOI: 10.14336/AD.2020.0707
43. Jaarsma T, Hill L, Bayes-Genis A, La Rocca HB, Castiello T, Čelutkienė J. **Self-care of Heart Failure Patients: Practical Management Recommendations from the Heart Failure Association of the European Society of Cardiology**. *Eur J Heart Fail* (2021) **23** 157-74. DOI: 10.1002/ejhf.2008
44. Valenzuela PL, Carrera-Bastos P, Gálvez BG, Ruiz-Hurtado G, Ordovas JM, Ruilope LM. **Lifestyle Interventions for the Prevention and Treatment of Hypertension**. *Nat Rev Cardiol* (2021) **18** 251-75. DOI: 10.1038/s41569-020-00437-9
45. Kadier K, Qin L, Ainiwaer A, Rehemuding R, Dilixiati D, Du YY. **Association of Sleep-Related Disorders with Cardiovascular Disease Among Adults in the United States: A Cross-Sectional Study Based on National Health and Nutrition Examination Survey 2005-2008**. *Front Cardiovasc Med* (2022) **9** 954238. DOI: 10.3389/fcvm.2022.954238
46. St-Onge MP, Grandner MA, Brown D, Conroy MB, Jean-Louis G, Coons M. **Sleep Duration and Quality: Impact on Lifestyle Behaviors and Cardiometabolic Health: A Scientific Statement from the American Heart Association**. *Circulation* (2016) **134** e367-e386. DOI: 10.1161/CIR.0000000000000444
47. Vaidya S, Gothi D, Patro M. **Prevalence of Sleep Disorders in Chronic Obstructive Pulmonary Disease and Utility of Global Sleep Assessment Questionnaire: An Observational Case-Control Study**. *Ann Thorac Med* (2020) **15** 230-7. DOI: 10.4103/atm.ATM_85_20
48. Khan FS, Marshall ZW. *Assessment of Sleep Disturbance in Patients with Atopic Conditions. Annals of Allergy, Asthma & Immunology* (2022)
49. Prasad B, Nyenhuis SM, Imayama I, Siddiqi A, Teodorescu M. **Asthma and Obstructive Sleep Apnea Overlap: What Has the Evidence Taught Us?**. *Am J Respir Crit Care Med* (2020) **201** 1345-57. DOI: 10.1164/rccm.201810-1838TR
50. Adir Y, Humbert M, Chaouat A. **Sleep-related Breathing Disorders and Pulmonary Hypertension**. *Eur Respir J* (2021) **57** 2002258. DOI: 10.1183/13993003.02258-2020
51. McNicholas WT, Hansson D, Schiza S, Grote L. **Sleep in Chronic Respiratory Disease: COPD and Hypoventilation Disorders**. *Eur Respir Rev : official J Eur Respir Soc* (2019) **28** 190064. DOI: 10.1183/16000617.0064-2019
52. Li SQ, Sun XW, Zhang L, Ding YJ, Li HP, Yan YR. **Impact of Insomnia and Obstructive Sleep Apnea on the Risk of Acute Exacerbation of Chronic Obstructive Pulmonary Disease**. *Sleep Med Rev* (2021) **58** 101444. DOI: 10.1016/j.smrv.2021.101444
53. Shah AJ, Quek E, Alqahtani JS, Hurst JR, Mandal S. **Cardiovascular Outcomes in Patients with COPD-OSA Overlap Syndrome: A Systematic Review and Meta-Analysis**. *Sleep Med Rev* (2022) **63** 101627. DOI: 10.1016/j.smrv.2022.101627
54. Lechat B, Appleton S, Melaku YA, Hansen K, McEvoy RD, Adams R. **Comorbid Insomnia and Sleep Apnoea Is Associated with All-Cause Mortality**. *Eur Respir J* (2022) **60** 2101958. DOI: 10.1183/13993003.01958-2021
55. Kim J, Jo I. **Age-dependent Association between Sleep Duration and Hypertension in the Adult Korean Population**. *Am J Hypertens* (2010) **23** 1286-91. DOI: 10.1038/ajh.2010.166
56. Domínguez F, Fuster V, Fernández-Alvira JM, Fernández-Friera L, López-Melgar B, Blanco-Rojo R. **Association of Sleep Duration and Quality with Subclinical Atherosclerosis**. *J Am Coll Cardiol* (2019) **73** 134-44. DOI: 10.1016/j.jacc.2018.10.060
57. Kim H, Jung HR, Kim JB, Kim DJ. **Autonomic Dysfunction in Sleep Disorders: From Neurobiological Basis to Potential Therapeutic Approaches**. *J Clin Neurol* (2022) **18** 140-51. DOI: 10.3988/jcn.2022.18.2.140
58. Grimaldi D, Carter JR, Van Cauter E, Leproult R. **Adverse Impact of Sleep Restriction and Circadian Misalignment on Autonomic Function in Healthy Young Adults**. *Hypertension* (2016) **68** 243-50. DOI: 10.1161/HYPERTENSIONAHA.115.06847
59. Routledge FS, Dunbar SB, Higgins M, Rogers AE, Feeley C, Ioachimescu O. **Insomnia Symptoms Are Associated with Abnormal Endothelial Function**. *J Cardiovasc Nurs* (2017) **32** 78-85. DOI: 10.1097/JCN.0000000000000295
60. Tomoda A, Kawatani J, Joudoi T, Hamada A, Miike T. **Metabolic Dysfunction and Circadian Rhythm Abnormalities in Adolescents with Sleep Disturbance**. *Neuroimage* (2009) **47** T21-6. DOI: 10.1016/j.neuroimage.2009.02.038
61. Grandner MA, Sands-Lincoln MR, Pak VM, Garland SN. **Sleep Duration, Cardiovascular Disease, and Proinflammatory Biomarkers**. *Nat Sci Sleep* (2013) **5** 93-107. DOI: 10.2147/NSS.S31063
62. Mullington JM, Simpson NS, Meier-Ewert HK, Haack M. **Sleep Loss and Inflammation**. *Best Pract Res Clin Endocrinol Metab* (2010) **24** 775-84. DOI: 10.1016/j.beem.2010.08.014
63. Mills PJ, von Känel R, Norman D, Natarajan L, Ziegler MG, Dimsdale JE. **Inflammation and Sleep in Healthy Individuals**. *Sleep* (2007) **30** 729-35. DOI: 10.1093/sleep/30.6.729
64. Genuardi MV, Ogilvie RP, Saand AR, DeSensi RS, Saul MI, Magnani JW. **Association of Short Sleep Duration and Atrial Fibrillation**. *Chest* (2019) **156** 544-52. DOI: 10.1016/j.chest.2019.01.033
65. Chokesuwattanaskul R, Thongprayoon C, Sharma K, Congrete S, Tanawuttiwat T, Cheungpasitporn W. **Associations of Sleep Quality with Incident Atrial Fibrillation: a Meta-Analysis**. *Intern Med J* (2018) **48** 964-72. DOI: 10.1111/imj.13764
|
---
title: BMI growth trajectory from birth to 5 years and its sex-specific association
with prepregnant BMI and gestational weight gain
authors:
- Jinting Xie
- Yan Han
- Lei Peng
- Jingjing Zhang
- Xiangjun Gong
- Yan Du
- Xiangmei Ren
- Li Zhou
- Yuanhong Li
- Ping Zeng
- Jihong Shao
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9971005
doi: 10.3389/fnut.2023.1101158
license: CC BY 4.0
---
# BMI growth trajectory from birth to 5 years and its sex-specific association with prepregnant BMI and gestational weight gain
## Abstract
### Objective
The purpose of the study was to identify the latent body mass index (BMI) z-score trajectories of children from birth to 5 years of age and evaluate their sex-specific association with prepregnant BMI and gestational weight gain (GWG).
### Methods
This was a retrospective longitudinal cohort study performed in China. In total, three distinct BMI-z trajectories from birth to 5 years of age were determined for both genders using the latent class growth modeling. The logistic regression model was used to assess the associations of maternal prepregnant BMI and GWG with childhood BMI-z growth trajectories.
### Results
Excessive GWG increased the risks of children falling into high-BMI-z trajectory relative to adequate GWG (OR = 2.04, $95\%$ CI: 1.29, 3.20) in boys; girls born to mothers with prepregnancy underweight had a higher risk of low-BMI-z trajectory than girls born to mothers with prepregnancy adequate weight (OR = 1.85, $95\%$ CI: 1.22, 2.79).
### Conclusion
BMI-z growth trajectories of children from 0 to 5 years of age have population heterogeneity. Prepregnant BMI and GWG are associated with child BMI-z trajectories. It is necessary to monitor weight status before and during pregnancy to promote maternal and child health.
## 1. Introduction
The high prevalence of childhood overweight or obesity is a significant public health issue [1]. According to the World Health Organization reports [2], an estimated $5.7\%$ or 38.9 million children under the age of 5 around the world were affected by overweight in 2020. The Report on the Status of Nutrition and Chronic Diseases of Chinese Residents [2020] [3] shows that the prevalence of childhood overweight or obesity was $19\%$ among children aged 6–17 and $10.4\%$ among children under the age of 6. Early childhood overweight or obesity is critical for lifelong health [4, 5]. Most prior studies examining the associated gene and environmental determinants of childhood overweight or obesity have focused on childhood BMI at just one point in time (6–8). Compared with the developmental assessments at a single time point, longitudinal child growth trajectories comprehensively evaluate the growth and development level of children from a dynamic perspective and detect abnormal growth in a timely manner [9].
Most studies showed that childhood growth trajectories have population heterogeneity (10–12). The early childhood growth trajectories were proved to be predictive of obesity risk in later life [13], cardiometabolic risk [14, 15], and adult diabetes [16]. A recent birth cohort study evaluated childhood BMI z-score trajectories from age of 2 to 18 and showed that preschool age is a critical window that could predict growth patterns during puberty [17]. Therefore, it is necessary to closely monitor the early childhood growth trajectories, focusing on those at higher risk of later overweight or obesity status, and helping to target specific groups for early intervention.
Accumulating evidence has supported that prepregnant BMI and gestational weight gain (GWG) may influence childhood overweight or obesity (18–21). Most studies showed that excessive GWG might increase the risk of childhood OWOB (22–24). However, the association between inadequate GWG with childhood BMI status remains unclear [25]. Furthermore, there are still gaps in our knowledge regarding the associations of prepregnant BMI and GWG with BMI growth trajectories in early childhood. Therefore, the primary aim of this study was to identify the latent BMI-z growth trajectories of children from birth to 5 years of age in different genders and evaluate their independent association with prepregnant BMI and GWG.
## 2.1. Study subjects
The present study was a retrospective longitudinal cohort study. The study was approved by the Ethics Committee of Xuzhou Maternity and Child Health Care Hospital (No.201901).
Participants were singleton offspring born at term in Xuzhou Maternity and Child Health Care Hospital between 1 January 2016 and 31 December 2016. The inclusion criteria included [1] singleton offspring born at 28–42 completed weeks of gestation; and [2] mother–child pairs with recorded information, such as maternal gestational age, education level, prepregnant BMI, GWG, delivery type, child sex, birth weight, feeding mode in 6 months, and children physical check with at least 4 height/length and weight measurement recorded at 1 year (±2 months), 2 years (±2 months), 3 years (±3 months), 4 years (±3 months), and 5 years (±3 months). Exclusion criteria included [1] offspring born with congenital disabilities or postnatal diseases that could interfere with body composition development and [2] offspring with missing covariate data. Figure 1 depicts the study cohort derivation. A total of 2,190 mother–child pairs were enrolled in this study, and written informed consent was obtained from the parents of the subjects at recruitment.
**Figure 1:** *Study cohort derivation.*
Data of mothers and children were collected retrospectively from the Jiangsu Maternal and Child Health Management Information System, including the maternal and child health information on prenatal, antenatal, and child healthcare electronic records with the standard quality control measures.
## 2.2. Prepregnant BMI and GWG
Height (m) and weight (kg) before pregnancy were collected at enrollment and calculated prepregnant BMI (kg/m2). BMI was calculated as weight (kg) divided by square of the length/height (m2). According to the BMI standards for Chinese adults [26], mothers were categorized as underweight with a BMI of <18.5 kg/m2, adequate weight with 18.5 kg/m2 ≤ BMI < 24.0 kg/m2, overweight with 24.0 kg/m2 ≤ BMI < 28.0 kg/m2, and obese with BMI ≥ 28.0 kg/m2. Based on their prepregnant BMI, GWG (kg) was calculated by subtracting prepregnancy weight (kg) from maternal weight at delivery (kg) and was categorized based on the standard of recommendation for weight gain during the pregnancy period (WS/T801-2022): underweight mothers (BMI < 18.5 kg/m2) who gained 11.0–16.0 kg, normal-weight mothers (18.5 kg/m2 ≤ BMI < 24.0 kg/m2) gained 8.0–14.0 kg, overweight mothers (24.0 kg/m2 ≤ BMI < 28.0 kg/m2) gained 7.0–11.0 kg, and obese mothers (BMI ≥ 28.0 kg/m2) gained 5.0–9.0 kg were categorized as adequate GWG; mothers who gained weight above or below this criterion were categorized as excessive or inadequate GWG, respectively. The BMI-z values that were more than ±5 were set to missing.
## 2.3. Child BMI-z
Children’s weight and length/height were measured at each annual healthcare visit by trained staff. Body weight was measured using a digital scale (measuring range: 5–150 kg, measurement resolution: 0.1 kg, and measurement accuracy: ±$0.3\%$). Recumbent length was obtained at the first-year and second-year visits, and standing height was obtained for those 3 years or older, all to the nearest 0.1 cm. BMI z-scores (BMI-z) were generated based on the age and sex-specific BMI reference from the WHO Child Growth Standards [2006].
## 2.4. Confounding factors
The potential confounding factors included maternal and children information. Maternal information included household income (￥), mother’s education, prepregnancy BMI (kg/m2), age at pregnancy (years), type of delivery, and parity. In terms of children’s information, gestational age of delivery (years), infant feeding mode from birth to 6 months, and time of complementary foods introduction (months) were considered. Child birth weight was categorized as <2.5 kg, 2.5–4 kg, and ≥ 4 kg. Infant feeding mode from birth to 6 months was classified as exclusive breastfeeding [27], mixed feeding, and formula feeding. The time of complementary food introduction was classified as ≤5 months, 6 months, and ≥ 7 months.
## 2.5. Statistical analysis
The latent class growth modeling (LCGM) approach was used to identify the subgroups shared a similar underlying trajectory based on the children’s BMI z-scores with the Mplus 8.0. Model fit indices include Akaike’s Information Criteria (AIC), Bayesian Information Criteria (BIC) and sample size adjusted BIC (aBIC), entropy, and a value of p for bootstrapped likelihood ratio test (BLRT) and Vuong-Lo–Mendell–Rubin likelihood ratio test (VLRT). The smaller the first three indices, the better the model fitting effect. The significant value of p for BLRT and VLRT indicates that a model with k−1 class should be rejected in favor of a model with k classes. The value of entropy >0.70 and the number of subjects in each trajectory group ≥$5\%$ indicate a good model fit. The maximum likelihood robust estimator was used to account for missing data when fitting the trajectories. Three distinct BMI-z trajectories were determined for both genders using LCGM.
Prepregnant BMI and GWG were compared among latent BMI trajectory groups using the chi-square test for proportions and the ANOVA F test for means. Logistic regression models were used to examine the association of prepregnant BMI and GWG with child BMI-z growth trajectory in different genders with the adjustment for the potential confounding factors, including household income, mother’s education, age at pregnancy, type of delivery, parity, birth weight, and time of complementary food introduction. Covariate selection was based on a compulsory entry procedure and other potential confounders identified in the literature [28]. Crude and adjusted odd ratios (ORs), along with $95\%$ confidence intervals (CIs), were calculated. Data were analyzed using Statistic Product Service Solutions 23.0. All statistical tests were two-sided, and a value of p of <0.05 was considered statistically significant.
## 3. Results
The study population consisted of 2,190 mother–child pairs, of which 1,165 were boys and 1,025 were girls. Table 1 summarizes the maternal and child characteristics of the participants. A higher rate of exclusive breastfeeding in the first 6 months was found in girls than in boys (53.2 vs. $51.1\%$), and a higher proportion of girls with the time of complementary food introduction at 6 months was found (56.0 vs. $51.9\%$).
**Table 1**
| Unnamed: 0 | All children n = 2,190 | Boy n = 1,165 | Girl n = 1,025 | p-value |
| --- | --- | --- | --- | --- |
| | n (%) | n (%) | n (%) | |
| Maternal characteristics | | | | |
| Household income per month, ￥ | | | | 0.41 |
| <2,500 | 110 (5.0) | 57 (4.9) | 53 (5.2) | |
| 2,500–4,000 | 140 (6.4) | 82 (7.0) | 58 (5.7) | |
| ≥4,000 | 1940 (88.6) | 1,026 (88.1) | 914 (89.2) | |
| Mother’s level of education | | | | 0.52 |
| Junior high or below | 296 (13.5) | 158 (13.6) | 138 (13.5) | |
| High school/technical secondary school | 425 (19.4) | 225 (19.3) | 200 (19.5) | |
| Junior college/vocational college | 508 (23.2) | 284 (24.4) | 224 (21.9) | |
| College degree or above | 961 (43.9) | 498 (42.7) | 463 (45.2) | |
| Age at pregnancy, years | | | | 0.04* |
| 18~ | 417 (19.0) | 209 (17.9) | 208 (20.3) | |
| 25~ | 1,096 (50.0) | 569 (48.8) | 527 (51.4) | |
| ≥30 | 677 (30.9) | 387 (33.2) | 290 (28.3) | |
| Type of delivery | | | | 0.24 |
| Vaginal | 1,019 (46.6) | 529 (45.4) | 491 (47.9) | |
| Cesarean | 1,170 (53.4) | 636 (54.6) | 534 (52.1) | |
| Parity | | | | 0.06 |
| 0 | 1,375 (62.8) | 710 (60.9) | 665 (64.9) | |
| 1+ | 815 (37.2) | 455 (39.1) | 360 (35.1) | |
| Child characteristics | | | | |
| Gestational age of delivery, weeks | | | | 0.26 |
| <37 | 106 (4.8) | 62 (5.3) | 44 (4.3) | |
| ≥37 | 2084 (95.2) | 1,103 (94.7) | 981 (95.7) | |
| Child birth weight, kg | | | | <0.001* |
| <2.5 | 35 (1.6) | 16 (1.4) | 19 (1.9) | |
| 2.5–4 | 1857 (84.8) | 951 (81.6) | 906 (88.4) | |
| ≥4 | 298 (13.6) | 198 (17.0) | 100 (9.8) | |
| Infant feeding mode from birth to 6 months | | | | 0.49 |
| Exclusive breastfeeding | 1,140 (52.1) | 595 (51.1) | 545 (53.2) | |
| Mixed feeding | 924 (42.2) | 498 (42.7) | 426 (41.6) | |
| Formula feeding | 126 (5.8) | 72 (6.2) | 54 (5.3) | |
| Time of complementary foods introduction, months | | | | 0.05 |
| ≤5 | 516 (23.6) | 298 (25.6) | 218 (21.3) | |
| 6 | 1,179 (53.8) | 605 (51.9) | 574 (56.0) | |
| ≥7 | 495 (34.6) | 262 (22.5) | 233 (22.7) | |
The number of classes and the shape of the BMI z-scores trajectories pattern were estimated according to the data of 1,165 boys and 1,025 girls. Complete BMI-z data were available for 1,043 ($89.5\%$) boys and 917 ($89.5\%$) girls, and the remaining children had one missing BMI-z data. The proportions of missing BMI-z data at each time point are shown in Supplementary Table 1. We tested from one to four possible trajectory classes. Considering the model fit indices, the three-class model was identified as the optimal model for both boys and girls. Supplementary Table 2 shows the fit statistics for the trajectory classes estimated.
Boys and girls shared similar patterns of growth trajectories but differed in their proportions. For the three latent trajectories, boys were more likely to have stable and moderate growth trajectories than girls. According to the relative position of the estimated three trajectories and combining professional significance, Class 1 was named as “moderate-BMI-z” ($69.5\%$ for boys and $63.9\%$ for girls), Class 2 as “high-BMI-z” ($11.6\%$ for boys and $14.5\%$ for girls), and Class 3 as “low-BMI-z” ($18.9\%$ for boys and $21.6\%$ for girls). Moderate-BMI-z trajectory group represented children who had relatively stable BMI-z scores around 0 with a low increasing trend. The high-BMI-z trajectory group exhibited a relatively high initial BMI-z, a rapid increase until the age of 2 years, and a slight decrease after that, with the overall BMI-z ranging from 1 to 2.5. The low-BMI-z trajectory group was characterized by a relatively low initial BMI-z value, which tends to decrease rapidly until the age of 2 years and then increases slightly into the normal range, with the overall BMI-z ranging from −1.5 to −0.5. Figure 2 shows the BMI-z growth trajectory for trends and sizes of the three kinds of trajectories. Supplementary Table 3 shows the parameter estimation results of the LCGA model for children’s BMI-z growth trajectory.
**Figure 2:** *Child BMI-z growth trajectories from birth to 5 years. Class 1: moderate-BMI-z trajectory, Class 2: high-BMI-z trajectory, Class 3: low-BMI-z trajectory. Line of circle symbol represents sample means; line of triangle symbol represents estimated means.*
Table 2 presents the distribution of prepregnant BMI and GWG overall and according to child BMI-z growth trajectory classes stratified for sex. The average prepregnant BMI was 20.83 ± 2.58 kg/m2, and the maternal mean GWG was 15.53 ± 5.85 kg. The prevalence of maternal prepregnancy overweight and obesity was 8.4 and $1.6\%$, respectively, and GWG was classified as excessive for $60.8\%$ of mothers and inadequate for $8.6\%$. Mothers with excessive GWG were more likely to have children within the high-BMI-z trajectory group among boys ($$p \leq 0.018$$). Mothers with prepregnancy underweight were more likely to have children within the low-BMI-z trajectory group among girls ($$p \leq 0.001$$).
**Table 2**
| Unnamed: 0 | Mean ± SD or n (%) | Mean ± SD or n (%).1 | Mean ± SD or n (%).2 | Mean ± SD or n (%).3 | Unnamed: 5 |
| --- | --- | --- | --- | --- | --- |
| | All | Class 1 | Class 2 | Class 3 | p-value |
| Boys | | | | | |
| Prepregnant BMI, kg/m2 | 20.86 ± 2.57 | 20.92 ± 2.62 | 21.06 ± 2.53 | 20.53 ± 2.38 | 0.085 |
| Prepregnant BMI category | | | | | |
| Underweight | 169 (14.5) | 119 (14.7) | 17 (12.6) | 33 (15.0) | 0.564 |
| Adequate weight | 876 (75.2) | 607 (74.9) | 99 (73.3) | 170 (77.3) | |
| Overweight | 101 (8.7) | 69 (8.5) | 17 (12.6) | 15 (6.8) | |
| Obesity | 19 (1.6) | 15 (1.9) | 2 (1.5) | 2 (0.9) | |
| GWG, kg | 15.32 ± 5.82 | 15.25 ± 5.88 | 16.18 ± 5.92 | 15.07 ± 5.44 | 0.172 |
| GWG | | | | | |
| Adequate | 357 (30.6) | 265 (32.7) | 31 (23.0) | 61 (27.7) | 0.018 |
| Inadequate | 105 (9.0) | 72 (8.9) | 7 (5.2) | 26 (11.8) | |
| Excessive | 703 (60.3) | 473 (58.4) | 97 (71.9) | 133 (60.5) | |
| Girls | | | | | |
| Prepregnant BMI, kg/m2 | 20.80 ± 2.58 | 20.81 ± 2.55 | 21.28 ± 2.69 | 20.44 ± 2.55 | 0.009 |
| Prepregnant BMI category | | | | | |
| Underweight | 156 (15.2) | 92 (14.0) | 12 (8.1) | 52 (23.5) | 0.001 |
| adequate weight | 768 (74.9) | 501 (76.5) | 116 (77.9) | 151 (68.3) | |
| Overweight | 84 (8.2) | 50 (7.6) | 19 (12.8) | 15 (6.8) | |
| Obesity | 17 (1.7) | 12 (1.8) | 2 (1.3) | 3 (1.4) | |
| GWG, kg | 15.77 ± 5.88 | 15.70 ± 5.59 | 15.08 ± 5.36 | 16.41 ± 6.91 | 0.095 |
| GWG | | | | | |
| Adequate | 312 (30.4) | 201 (30.7) | 42 (28.2) | 69 (31.2) | 0.705 |
| Inadequate | 84 (8.2) | 52 (7.9) | 10 (6.7) | 22 (10.0) | |
| Excessive | 629 (61.4) | 402 (61.4) | 97 (65.1) | 130 (58.8) | |
Table 3 shows the independent association of prepregnant BMI and GWG with child BMI-z growth trajectory. After adjusting for household income, mother’s education, age at pregnancy, type of delivery, parity, birth weight, time of complementary food introduction, and defining the moderate-BMI-z trajectory group as the reference category, boys of maternal excessive GWG were more likely to have high-BMI-z trajectory than their adequate GWG counterparts (OR = 2.04, $95\%$ CI:1.29, 3.20); girls born to the mothers with prepregnancy underweight were more likely to have low-BMI-z trajectory than girls born to the mothers with prepregnancy adequate weight (OR = 1.85, $95\%$ CI: 1.22, 2.79).
**Table 3**
| Unnamed: 0 | Boys | Boys.1 | Girls | Girls.1 |
| --- | --- | --- | --- | --- |
| | High-BMI-z trajectory | Low-BMI-z trajectory | High-BMI-z trajectory | Low-BMI-z trajectory |
| | OR (95%CI) | OR (95%CI) | OR (95%CI) | OR (95%CI) |
| Model 1a | | | | |
| Prepregnant BMI category | | | | |
| Adequate weight | 1.00 | 1.00 | 1.00 | 1 |
| Underweight | 1.02 (0.58, 1.80) | 0.97 (0.63, 1.50) | 0.58 (0.30, 1.11) | 1.88 (1.26, 2.81)b |
| Overweight | 1.52 (0.86, 2.70) | 0.78 (0.43, 1.39) | 1.64 (0.93, 2.89) | 0.99 (0.54, 1.82) |
| Obesity | 0.75 (0.17, 3.34) | 0.48 (0.11, 2.11) | 0.71 (0.16, 3.22) | 0.83 (0.23, 2.98) |
| Gestational weight gain | | | | |
| Adequate | 1 | 1 | 1 | 1 |
| Inadequate | 0.83 (0.35, 1.98) | 1.56 (0.92, 2.66) | 0.95 (0.44, 2.02) | 1.16 (0.65, 2.06) |
| Excessive | 1.76 (1.14, 2.72) | 0.48 (0.11, 2.11) | 1.07 (0.71, 1.61) | 1.06 (0.75, 1.51) |
| Model 2 | | | | |
| Prepregnant BMI category | | | | |
| Adequate weight | 1.00 | 1.00 | 1.00 | 1 |
| Underweight | 1.10 (0.60, 2.00) | 0.96 (0.61, 1.51) | 0.63 (0.33, 1.23) | 1.85 (1.22, 2.79) |
| Overweight | 1.41 (0.76, 2.59) | 0.84 (0.46, 1.54) | 1.46 (0.80, 2.65) | 0.94 (0.50, 1.76) |
| Obesity | 0.39 (0.08, 1.84) | 0.37 (0.08, 1.72) | 0.57 (0.12, 2.72) | 0.84 (0.23, 3.08) |
| Gestational weight gain | | | | |
| Adequate | 1.00 | 1.00 | 1.00 | 1 |
| Inadequate | 0.84 (0.35, 2.03) | 1.54 (0.89, 2.69) | 0.72 (0.33, 1.61) | 1.28 (0.71, 2.32) |
| Excessive | 2.04 (1.29, 3.20) | 1.41 (0.99, 2.01) | 1.18 (0.77, 1.80) | 1.07 (0.75, 1.52) |
## 4. Discussion
This study was conducted to identify the childhood BMI-z trajectories from birth to 5 years among children born at term in Xuzhou Maternity and Child Health Care Hospital between 1 January 2016 and 31 December 2016 in different genders and to assess the association between BMI-z trajectories with prepregnant BMI and GWG. In total, three main findings are worthy of further attention and discussion. Data from this retrospective longitudinal cohort study showed that childhood BMI-z trajectories from birth to 5 years could be classified into three latent groups for both boys and girls, characterized as moderate-BMI-z trajectory group, high-BMI-z trajectory group, and low-BMI-z trajectory group. Prepregnant BMI and GWG were significantly associated with childhood BMI-z trajectories. Excessive GWG predicted the increased risk for the high-BMI-z trajectory group for boys, and prepregnancy underweight predicted the increased risk for the low-BMI-z trajectory group for girls.
With the application of longitudinal data analysis methods in childhood growth trajectories, accumulating studies have documented the potential heterogeneity of childhood growth trajectories [29, 30]. The growth trajectory of early childhood is particularly important. Children at risk for overweight and obesity may have unique developmental trajectories during early childhood [31], which may influence the subsequent development of overweight or obesity and other health issues [14, 32]. The classification and description of early childhood BMI-z growth trajectory in previous studies (33–36) can be roughly summarized into three types, including stable-moderate BMI-z growth trajectory, stable-low BMI-z growth trajectory, and stable-high BMI-z growth trajectory, which are consistent with the trajectories observed in our study. Furthermore, similar to these studies, it was concluded that children with stable-moderate BMI-z growth trajectory were in the majority, approximately $60\%$ account, with the average BMI-z score range of around 0. However, Zhang et al. [ 37] reported four latent BMI-z growth trajectory patterns from birth to the age of 60 months. In addition to the three categories mentioned above, a catch-up BMI-z growth trajectory was also identified. The reason this result differs from our study may be due to the different fitting methods used [38]. In future research, it would be beneficial to examine the application of different longitudinal data analysis methods to childhood growth trajectories.
In our analysis, after adjusting for the potential confounders, we found that boys of mothers with excessive GWG were significantly associated with an increased risk of high-BMI-z trajectory from birth to 5 years of age. Similarly, a retrospective longitudinal cohort study of 71,892 children suggested that children’s high and increasing BMI trajectories were modestly associated with excessive GWG [28]. However, the index used in this study to fit the growth trajectory of children is the raw BMI value, and the timeframe for the trajectory is 2–6 years old. Compared with the BMI trajectory, the BMI-z trajectory can better reflect the change in BMI values relative to their peers. Additionally, Montazeri et al. [ 39] found that excessive GWG was positively associated with the BMI trajectory of higher birth size and subsequent accelerated BMI gain, and inadequate GWG was associated with the BMI trajectory of lower birth size and slower BMI gain. However, our study did not observe a significant association between inadequate GWG and the low-BMI-z trajectory after adjustment. This may be due to the differences in the timeframe and classification of childhood growth trajectories. More in-depth studies are required to determine how inadequate GWG affects childhood growth trajectory.
Previous studies have reported positive associations between maternal prepregnant obesity and child high BMI growth trajectory [24, 40], which were consistent with the results of our study, though not statistically significant. Our study found that maternal prepregnant underweight predicted the increased risk for the low-BMI-z trajectory group for girls, which may be explained mainly by long-term changes in fetal endocrine and metabolic disorders [41].
There are several strengths in our study. First, multiple assessment points of BMI-z score were collected to identify the childhood growth trajectories with the LCGM approach, revealing the potential heterogeneity of growth trajectories in early childhood. Second, the study was conducted based on the Jiangsu Maternal and Child Health Management Information System to collect maternal and children’s health data, which was electronically recorded with the standard quality control measures. Third, GWG was classified according to the standard of recommendation for weight gain during the pregnancy period (WS/T801-2022), which was based on the BMI of Chinese adults as the tangent point and more suitable for evaluating maternal weight status of Chinese women than the Institute of Medicine guidelines in 2009. Finally, a relatively large sample size was used in the study. It took comprehensive covariates into inclusion in evaluating the association between prepregnant BMI and GWG with childhood growth trajectories.
Nevertheless, several limitations should be mentioned as well. First, the retrospective cohort study was used in this study, and the credibility of the evidence is insufficient. Hence, more prospective studies are needed to corroborate the results of this study in the future. Second, the information on energy balance-related behavior was not collected and adjusted for in our study, which needs to be considered in the future.
## 5. Conclusion
The BMI-z trajectory from birth to 5 years of age was identified as three latent groups both for boys and girls with the approach of LCGM. Excessive GWG is associated with the increased risk for the high-BMI-z trajectory group for boys, prepregnancy underweight predicted the increased risk for the low-BMI-z trajectory group for girls. Pre-school age is the key window for the formation of trajectory patterns. Maternal weight should be managed precisely, and physical surveillance and intervention should be carried out for children at high risk of obesity, which can help to move the threshold of prevention and control of childhood overweight or obesity forward.
## Data availbility statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Author contributions
JS, YH, and LP designed and conceptualized the study. XG, YH, and PZ were responsible for the methodology. JX, YH, and JZ conducted the formal analysis. YD, JZ, and JX were the investigators. JX and JZ prepared the original draft of the manuscript. JS, XR, LZ, and YL reviewed and edited the manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported by the National Natural Science Foundation of China (no. 82204056), the Jiangsu Provincial Maternal and Child Health Research Project in 2018 (F201805), the Key Lab of Human Genetics and Environmental Medicine, School of Public Health, Xuzhou Medical University, and the Key Lab of Environment and Health, School of Public Health, Xuzhou Medical University.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2023.1101158/full#supplementary-material
## References
1. **Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128.9 million children, adolescents, and adults**. *Lancet* (2017) **390** 2627-42. DOI: 10.1016/S0140-6736(17)32129-3
2. 2.World Health Organization. (2021). Levels and Trends in Child Malnutrition. The UNICEF/World Health Organization/World Bank Group.. *Levels and Trends in Child Malnutrition* (2021)
3. 3.The Report on the Status of Nutrition and Chronic Diseases of Chinese Residents. (2020). Chinese Center for Disease Control and Prevention/National Cancer Center/National Center for Cardiovascular Diseases.. *Chinese Center for Disease Control and Prevention/National Cancer Center/National Center for Cardiovascular Diseases* (2020)
4. Dong SS, Zhang K, Guo Y, Ding JM, Rong Y, Feng JC. **Phenome-wide investigation of the causal associations between childhood bmi and adult trait outcomes: a two-sample mendelian randomization study**. *Genome Med* (2021) **13** 48. DOI: 10.1186/s13073-021-00865-3
5. Drozdz D, Alvarez-Pitti J, Wojcik M, Borghi C, Gabbianelli R, Mazur A. **Obesity and cardiometabolic risk factors: from childhood to adulthood**. *Nutrients* (2021) **13** 13. DOI: 10.3390/nu13114176
6. Ontai LL, Sutter C, Sitnick S, Shilts MK, Townsend MS. **Parent food-related behaviors and family-based dietary and activity environments: associations with bmi z-scores in low-income preschoolers**. *Child Obes* (2020) **16** S-55-63. DOI: 10.1089/chi.2019.0105
7. Tang X, Liu Y, Hu J, Zhai L, Jia L, Ding N. **Association of waist circumference with blood pressure and familial dietary habits in preschool children: a cross-sectional study in northeastern China**. *Ital J Pediatr* (2022) **48** 53. DOI: 10.1186/s13052-022-01236-3
8. Medina M, Barreto P, Natero V, Moratorio X, Severi C. **Prevalence of malnutrition among children and women of reproductive age in Uruguay by socio-economic status and educational level**. *Public Health Nutr* (2020) **23** s101-7. DOI: 10.1017/S1368980020000804
9. Massara P, Keown-Stoneman CD, Erdman L, Ohuma EO, Bourdon C, Maguire JL. **Identifying longitudinal-growth patterns from infancy to childhood: a study comparing multiple clustering techniques**. *Int J Epidemiol* (2021) **50** 1000-10. DOI: 10.1093/ije/dyab021
10. Mattsson M, Maher GM, Boland F, Fitzgerald AP, Murray DM, Biesma R. **Group-based trajectory modelling for bmi trajectories in childhood: a systematic review**. *Obes Rev* (2019) **20** 998-1015. DOI: 10.1111/obr.12842
11. Widen EM, Burns N, Daniels M, Backlund G, Rickman R, Foster S. **Gestational weight change and childhood body composition trajectories from pregnancy to early adolescence**. *Obesity (Silver Spring)* (2022) **30** 707-17. DOI: 10.1002/oby.23367
12. Min J, Wen X, Xue H, Wang Y. **Ethnic disparities in childhood bmi trajectories and obesity and potential causes among 29, 250 us children: findings from the early childhood longitudinal study-birth and kindergarten cohorts**. *Int J Obes* (2018) **42** 1661-70. DOI: 10.1038/s41366-018-0091-4
13. Buscot MJ, Thomson RJ, Juonala M, Sabin MA, Burgner DP, Lehtimaki T. **Bmi trajectories associated with resolution of elevated youth bmi and incident adult obesity**. *Pediatrics* (2018) **141** 141. DOI: 10.1542/peds.2017-2003
14. Aris IM, Rifas-Shiman SL, Li LJ, Kleinman KP, Coull BA, Gold DR. **Patterns of body mass index milestones in early life and cardiometabolic risk in early adolescence**. *Int J Epidemiol* (2019) **48** 157-67. DOI: 10.1093/ije/dyy286
15. Wibaek R, Vistisen D, Girma T, Admassu B, Abera M, Abdissa A. **Body mass index trajectories in early childhood in relation to cardiometabolic risk profile and body composition at 5 years of age**. *Am J Clin Nutr* (2019) **110** 1175-85. DOI: 10.1093/ajcn/nqz170
16. Zhang T, Xu J, Li S, Bazzano LA, He J, Whelton PK. **Trajectories of childhood bmi and adult diabetes: the Bogalusa heart study**. *Diabetologia* (2019) **62** 70-7. DOI: 10.1007/s00125-018-4753-5
17. Liang J, Zheng S, Li X, Xiao D, Wang P. **Associations of community, famliy and early individual factors with body mass index z-scores trajectories among chinese children and adolescents**. *Sci Rep* (2021) **11** 14535. DOI: 10.1038/s41598-021-93949-4
18. Wang Q, Yang M, Deng X, Wang S, Zhou B, Li X. **Explorations on risk profiles for overweight and obesity in 9501 preschool-aged children**. *Obes Res Clin Pract* (2022) **16** 106-14. DOI: 10.1016/j.orcp.2022.02.007
19. Meyer DM, Stecher L, Brei C, Hauner H. **Mid-pregnancy weight gain is associated with offspring adiposity outcomes in early childhood**. *Pediatr Res* (2021) **90** 390-6. DOI: 10.1038/s41390-020-01202-x
20. Rito AI, Buoncristiano M, Spinelli A, Salanave B, Kunesova M, Hejgaard T. **Association between characteristics at birth, breastfeeding and obesity in 22 countries: the who european childhood obesity surveillance initiative - cosi 2015/2017**. *Obes Facts* (2019) **12** 226-43. DOI: 10.1159/000500425
21. Ziauddeen N, Huang JY, Taylor E, Roderick PJ, Godfrey KM, Alwan NA. **Interpregnancy weight gain and childhood obesity: analysis of a Uk population-based cohort**. *Int J Obes* (2022) **46** 211-9. DOI: 10.1038/s41366-021-00979-z
22. Baran J, Weres A, Czenczek-Lewandowska E, Leszczak J, Kalandyk-Osinko K, Luszczki E. **Excessive gestational weight gain: long-term consequences for the child**. *J Clin Med* (2020) **9** 9. DOI: 10.3390/jcm9123795
23. Voerman E, Santos S, Patro GB, Amiano P, Ballester F, Barros H. **Maternal body mass index, gestational weight gain, and the risk of overweight and obesity across childhood: an individual participant data meta-analysis**. *PLoS Med* (2019) **16** e1002744. DOI: 10.1371/journal.pmed.1002744
24. Chang R, Mei H, Zhang Y, Xu K, Yang S, Zhang J. **Early childhood body mass index trajectory and overweight/obesity risk differed by maternal weight status**. *Eur J Clin Nutr* (2022) **76** 450-5. DOI: 10.1038/s41430-021-00975-6
25. Wan N, Cai L, Tan W, Zhang T, Yang J, Chen Y. **Associations of gestational weight gain with offspring thinness and obesity: by prepregnancy body mass index**. *Reprod Health* (2018) **15** 149. DOI: 10.1186/s12978-018-0585-5
26. **Appropriate body-mass index for asian populations and its implications for policy and intervention strategies**. *Lancet* (2004) **363** 157-63. DOI: 10.1016/S0140-6736(03)15268-3
27. 27.World Health Organization. Report of expert consultation on the optimal duration of exclusive breastfeeding. Gevena: WHO Press (2001). 10 p.. *Report of expert consultation on the optimal duration of exclusive breastfeeding* (2001) 10
28. Wang X, Martinez MP, Chow T, Xiang AH. **Bmi growth trajectory from ages 2 to 6 years and its association with maternal obesity, diabetes during pregnancy, gestational weight gain, and breastfeeding**. *Pediatr Obes* (2020) **15** e12579. DOI: 10.1111/ijpo.12579
29. Bichteler A, Gershoff ET. **Identification of children's bmi trajectories and prediction from weight gain in infancy**. *Obesity (Silver Spring)* (2018) **26** 1050-6. DOI: 10.1002/oby.22177
30. Zhou Y, Yu P, Zhang Y, Wang T, Wang A. **Distinct patterns of urban-rural and sex disparities in children's bmi trajectories from 2013 to 2018**. *Front Public Health* (2021) **9** 689021. DOI: 10.3389/fpubh.2021.689021
31. Braun JM, Kalkwarf HJ, Papandonatos GD, Chen A, Lanphear BP. **Patterns of early life body mass index and childhood overweight and obesity status at eight years of age**. *BMC Pediatr* (2018) **18** 161. DOI: 10.1186/s12887-018-1124-9
32. Barraclough JY, Garden FL, Toelle BG, Marks GB, Baur LA, Ayer JG. **Weight gain trajectories from birth to adolescence and cardiometabolic status in adolescence**. *J Pediatr* (2019) **208** 89-95.e4. DOI: 10.1016/j.jpeds.2018.12.034
33. Zhao Q, Hu Z, Kocak M, Liu J, Fowke JH, Han JC. **Associations of prenatal metabolomics profiles with early childhood growth trajectories and obesity risk in african americans: the candle study**. *Int J Obes* (2021) **45** 1439-47. DOI: 10.1038/s41366-021-00808-3
34. Nedelec R, Miettunen J, Mannikko M, Jarvelin MR, Sebert S. **Maternal and infant prediction of the child bmi trajectories; studies across two generations of northern Finland birth cohorts**. *Int J Obes* (2021) **45** 404-14. DOI: 10.1038/s41366-020-00695-0
35. Shao ZY, Wang P, Li P, Sun Y, Li PP, Zhu P. **Socioeconomic disparities and infancy growth trajectory: a population-based and longitudinal study**. *BMC Pediatr* (2021) **21** 549. DOI: 10.1186/s12887-021-02995-4
36. Hu Z, Tylavsky FA, Han JC, Kocak M, Fowke JH, Davis RL. **Maternal metabolic factors during pregnancy predict early childhood growth trajectories and obesity risk: the candle study**. *Int J Obes* (2019) **43** 1914-22. DOI: 10.1038/s41366-019-0326-z
37. Zheng M, Campbell KJ, Baur L, Rissel C, Wen LM. **Infant feeding and growth trajectories in early childhood: the application and comparison of two longitudinal modelling approaches**. *Int J Obes* (2021) **45** 2230-7. DOI: 10.1038/s41366-021-00892-5
38. Becnel JN, Williams AL. **Using latent class growth modeling to examine longitudinal patterns of body mass index change from adolescence to adulthood**. *J Acad Nutr Diet* (2019) **119** 1875-81. DOI: 10.1016/j.jand.2019.04.025
39. Montazeri P, Vrijheid M, Martinez D, Basterrechea M, Fernandez-Somoano A, Guxens M. **Maternal metabolic health parameters during pregnancy in relation to early childhood bmi trajectories**. *Obesity (Silver Spring)* (2018) **26** 588-96. DOI: 10.1002/oby.22095
40. Sha T, Gao X, Chen C, Li L, He Q, Wu X. **Associations of prepregnancy bmi, gestational weight gain and maternal parity with the trajectory of weight in early childhood: a prospective cohort study**. *Int J Environ Res Public Health* (2019) **16** 1110. DOI: 10.3390/ijerph16071110
41. Skrypnik D, Bogdanski P, Zawiejska A, Wender-Ozegowska E. **Role of gestational weight gain, gestational diabetes, breastfeeding, and hypertension in mother-to-child obesity transmission**. *Pol Arch Intern Med* (2019) **129** 267-75. DOI: 10.20452/pamw.4426
|
---
title: Long-term intensive endurance exercise training is associated to reduced markers
of cellular senescence in the colon mucosa of older adults
authors:
- Marco Demaria
- Beatrice Bertozzi
- Nicola Veronese
- Francesco Spelta
- Edda Cava
- Valeria Tosti
- Laura Piccio
- Dayna S. Early
- Luigi Fontana
journal: NPJ Aging
year: 2023
pmcid: PMC9971019
doi: 10.1038/s41514-023-00100-w
license: CC BY 4.0
---
# Long-term intensive endurance exercise training is associated to reduced markers of cellular senescence in the colon mucosa of older adults
## Body
Regular physical exercise is one of the key pillars for health promotion since ancient times, although our ancestors did not know the biological processes responsible for its beneficial effects. Data from animal and human randomized trials indicate that aerobic exercise training improves glucose tolerance, insulin sensitivity and lipid metabolism through multiple mechanisms, including mitochondrial biogenesis, increased expression of the insulin responsive glucose transporter type 4 (GLUT4) and lipoprotein lipase in the skeletal muscle1,2. Regular exercise training also promotes visceral fat loss, reduces inflammation and oxidative stress, and improves left ventricular diastolic function in overweight men and women1,3–6.
Accumulating data show that physical activity evokes profound metabolic and molecular responses not only in key metabolic organs (skeletal muscle, adipose tissues, and liver), but also in tissues at high risk of neoplastic transformation. Epidemiological studies suggest an inverse association between physical activity and risk for 13 different types of cancer, in particular for colon and breast cancer7–9. Regular exercise training can also improve prognosis among breast and colorectal cancer survivors10,11 by long-term regulation of various metabolic, inflammatory and aging pathways that promote DNA and cellular repair, proteostasis, replicative stress resistance, and apoptosis of permanently damaged cells12.
One of the fundamental cellular mechanisms regulating aging and tumor development is cellular senescence13. Cellular senescence is a state of irreversible proliferative arrest triggered by diverse DNA or mitochondrial damages to prevent propagation of damaged cells14. Senescent cells are characterized by the engagement of the Cyclin-dependent kinase inhibitors p16Ink4a (p16) and p21CIP1 (p21)15, enhanced lysosomal activity, and a hypersecretory phenotype known as Senescence-Associated Secretory Phenotype (SASP). The SASP remain highly heterogeneous and dependent on various intrinsic and extrinsic factors16,17. However, persistent senescent cells can cause chronic low-level inflammation and aberrant tissue growth and remodeling via SASP factors18–23.
Lifestyle factors can have consequences on induction and accumulation of cellular senescence. For example, caloric restriction (CR), a well-known and highly conserved anti-aging and anti-cancer intervention, is associated to reduced accumulation of senescent cells in both mice and humans24,25. Interestingly, we have also recently shown that, at least in mice, high-protein and high-fat diets lead to premature hepatic accumulation of hyper-inflammatory senescent cells26. Besides dietary approaches, it has been shown that a 12-week exercise program reduces circulating senescence biomarkers in older adults27 and a recent human study suggests that the number of senescent cells of the adipose tissue is inversely correlated to physical function in older women28. However, whether regular vigorous aerobic exercise can prevent accumulation of age-associated senescence, especially in highly proliferating cancer prone tissues, remains controversial29. Here, we studied the effects of chronic intensive endurance exercise training on cardiometabolic health and candidate biomarkers of cell senescence in colon mucosa biopsies of master athletes who ran an average 48 miles/week (range 30 to 90 miles/week) for an average of 21 years (range 5–35 years).
Participants in this study were endurance runners (mean age 57 ± 10 years) consuming usual American diets (EX); age- and sex-matched sedentary (regular exercise < 1 h per week) controls eating Western diets (WD-o); and very young (mean age 24.3 ± 2 years) sedentary controls (WD-y) who should have negligible numbers of senescent cells. Average calorie intake in the EX group was 2806 ± 618 kcal/day, 13 and $7\%$ higher than in the WD-o (2443 ± 407 kcal/day) and WD-y (2618 ± 712 kcal/day) groups, respectively ($p \leq 0.05$ for EX vs. WD-o). The percentages of total energy intake derived from protein, carbohydrate, and fat were similar among the groups: $15.7\%$, $51.8\%$, and $32.5\%$, respectively, in EX; $15.1\%$, $50.4\%$, and $32.8\%$, in WD-o; $17.2\%$, $48.2\%$, and $33.4\%$ in WD-y.
In Table 1, we reported the study sample’s summary statistics, including the distribution of age, sex, body mass index, DXA body fat percentage and lean mass, and a range of fitness and cardiometabolic parameters. BMI, body fat, resting heart rate, LDLc, total cholesterol HDL ratio, triglycerides, triglycerides to HDL ratio, fasting glucose, fasting insulin, HOMA-IR, and total white blood cell count were significantly lower in the EX group than in the WD-o group ($p \leq 0.05$). As expected, EX volunteers had significantly higher VO2max and HDLc than WD-o participants ($p \leq 0.05$).Table 1Characteristics of the study subjects. EX groupWD-o groupWD-y groupAmong group P($$n = 44$$)($$n = 44$$)($$n = 6$$)Age (years)57 ± 1057 ± 924.3 ± 2a,c<0.001Sex (M:F)37:737:74:2–Height (m)1.75 ± 0.11.76 ± 0.1a1.79 ± 0.1NSWeight (Kg)70.0 ± 1078.8 ± 14a82.6 ± 13b<0.001BMI (Kg/m2)22.7 ± 425.3 ± 2.7a25.7 ± 1b<0.001Body fat (% body weight)14.8 ± 6.525.2 ± 6.6a17.9 ± 8.2d<0.001Lean mass (kg)56.1 ± 854.7 ± 1163.3 ± 14NSResting heart rate (b/min)52 ± 863 ± 10a66 ± 12b<0.001VO2max (ml/Kg/min)51 ± 1033 ± 7a–<0.001SBP (mm Hg)126 ± 19129 ± 14129 ± 11NSDBP (mm Hg)73 ± 1079 ± 9b78 ± 110.055LDL-c (mg/dl)92 ± 22115 ± 28a94 ± 240.004HDL-c (mg/dl)68 ± 1755 ± 15a64 ± 18<0.001Triglycerides (mg/dl)64 ± 22120 ± 71a61 ± 33d<0.001TChol/HDL ratio2.6 ± 0.53.8 ± 1.0a2.9 ± 0.9d<0.001TG/HDL ratio1 ± 0.42.5 ± 2.0a0.9 ± 0.6d<0.001Fasting glucose (mg/dl)90 ± 894 ± 9b82 ± 5c0.001Fasting insulin (mg/dl)3.0 ± 2.37.6 ± 5.4a6.6 ± 3<0.001HOMA-IR0.7 ± 0.61.8 ± 1.3a1.3 ± 0.6<0.001WBC (K/cumm)4.5 ± 1.25.8 ± 1.6a4.9 ± 0.5<0.001hsCRP (mg/L)0.7 ± 0.61.8 ± 1.30.8 ± 0.3NSAll values are means ± SD.Significantly different from EX group, aP ≤ 0.003, bP ≤ 0.05.Significantly different from WD-o group, cP ≤ 0.003, dP ≤ 0.05.
To investigate the effects of long-term EX on biomarkers of cell senescence, we collected colon mucosa biopsies in a subset of 11 middle-aged (58.6 ± 8.3 years), weight-stable and lean (BMI, 24.5 ± 2.8 kg/m2) master athletes, 10 age- and sex-matched nonobese (BMI, 27.1 ± 2.3 kg/m2) and 6 sedentary young (24.3 ± 2 years) and lean (BMI, 25.7 ± 1 kg/m2) control subjects (Supplementary table 1). Because p16 is still considered one of the most relevant cell senescence markers in human specimens, and because p16 measurements were included in most of the previous studies on the effect of physical exercise on senescence markers, we measured its mRNA abundance. As expected, p16 levels were markedly higher in older than younger sedentary individuals consuming Western diets (Fig. 1A). Strikingly, this upregulation was significantly blunted in endurance runners (Fig. 1A). p21 is another important regulator of cell cycle arrest often dysregulated during senescence. Similar to what observed for p16, p21 levels were upregulated in older sedentary individuals compared to young sedentary or endurance runners (Fig. 1B). However, the difference in expression between groups did not reach statistical significance. The detrimental functions of cellular senescence are, at least partly, mediated by pro-inflammatory secreted factors. The composition of pro-inflammatory SASP is variable, but IL-6 remains one of the most consistent SASP factors13. In accordance, IL-6 mRNA levels of colon mucosa were significantly higher in old sedentary individuals consuming Western diet, whereas IL-6 levels in master athletes where low and similar to those of very young sedentary people consuming Western diets (Fig. 1C). We then measured levels of two additional SASP factors, IL8 and MMP3. We observed a trend for the upregulation of MMP3 and IL8 in older sedentary individuals compared to young sedentary and endurance runners, but there was no statistical significance (Fig. 1D, E). p16 mRNA levels correlated linearly with p21 ($r = 0.758$; $p \leq 0.001$) and IL-6 ($r = 0.798$; $p \leq 0.001$) mRNA levels (not shown). To evaluate potential links between the senescence burden and metabolic alterations, we then studied the association of p16 levels with various metabolic parameters. Strikingly, p16 mRNA levels were linearly correlated with the triglycerides to HDL ratio, a well-accepted marker of metabolic syndrome, coronary heart disease and colon adenoma risk (Fig. 1F)30–32. Interestingly, in patients with early-stage colorectal cancer, the combination of obesity and low HDL-cholesterol and high triglycerides levels predicts worst cancer survival33.Fig. 1Expression of senescence-associated genes in the colon mucosa of master athletes and sedentary controls. RNA was extracted from the sigmoid portion of the colon of human volunteers. The groups were: EX, exercised volunteers of average age 57 ± 10 years; age-matched sedentary controls (SED); young, volunteers of average age 24.3 ± 2 years. mRNA encoding p16 (A), p21 (B), IL6 (C), IL8 (D), and MMP3 (E) were quantified by qRT-PCR. mRNA encoding tubulin was used as internal control ($$n = 5$$–11 with each sample indicated by an individual dot). Panel F shows the relationship between p16 mRNA levels and the triglycerides to HDL ratio. All values are represented together with means and SEM. One-way Anova, *$p \leq 0.05.$
Our results are preliminary and limited, in particular in light of characterizing senescence-associated phenotypes and the type of cells more affected by these changes. Nevertheless, the findings shown here suggest that chronic high-volume high-intensity, unlike low-volume34, endurance exercise can play a major role in preventing the accumulation of senescent cells in cancer prone tissues like colon mucosa with age. This is important because data from transgenic mouse models, including our p16-3MR mouse19, have shown that ablation of senescent cells is sufficient to systemically reduce inflammation, rejuvenate tissue functions, alleviate various age-related conditions, improve health and extend longevity13. As senescent cells are likely contributor to dysregulated inflammatory responses, our data are in line with a previous report showing that the level of stress-induced (acute exercise) inflammatory markers is reduced in muscle and blood of lifelong aerobic exercising older men compared to old healthy nonexercisers35. In addition, prevention of cell senescence could partly explain the anti-cancer effect of lifelong aerobic exercise36. Future studies should be focusing on understanding which tissues are most affected, and what are the molecular and cellular mechanisms that mediate the senopreventative effect of endurance exercise training. Moreover, it will be key to analyze individuals following different physical exercise regimens, including resistance and high-intensity interval training.
## Abstract
Regular endurance exercise training is an effective intervention for the maintenance of metabolic health and the prevention of many age-associated chronic diseases. Several metabolic and inflammatory factors are involved in the health-promoting effects of exercise training, but regulatory mechanisms remain poorly understood. Cellular senescence—a state of irreversible growth arrest—is considered a basic mechanism of aging. Senescent cells accumulate over time and promote a variety of age-related pathologies from neurodegenerative disorders to cancer. Whether long-term intensive exercise training affect the accumulation of age-associated cellular senescence is still unclear. Here, we show that the classical senescence markers p16 and IL-6 were markedly higher in the colon mucosa of middle-aged and older overweight adults than in young sedentary individuals, but this upregulation was significantly blunted in age-matched endurance runners. Interestingly, we observe a linear correlation between the level of p16 and the triglycerides to HDL ratio, a marker of colon adenoma risk and cardiometabolic dysfunction. Our data suggest that chronic high-volume high-intensity endurance exercise can play a role in preventing the accumulation of senescent cells in cancer-prone tissues like colon mucosa with age. Future studies are warranted to elucidate if other tissues are also affected, and what are the molecular and cellular mechanisms that mediate the senopreventative effects of different forms of exercise training.
## Patients and tissue collection
This study sample includes three groups of volunteers, named from hereafter EX, WD-o, and WD-y. The EX group consisted of 44 master athletes who ran at least 30 miles/week (range 30–90 miles/week) or expended similar amount of energy by cycling or swimming, for at least the previous three years (range 3–35 years). The control (WD-o) group comprised 44 age- and sex-matched individuals reporting less than 1 hour of physical activity per week, recruited from the St. Louis metropolitan area. A third control group with negligible numbers of senescent cells consisted of 6 very young (mean age 24.3 ± 2 years) sedentary controls (WD-y) consuming Western diets. All the participants reported weight stability, defined as less than a 2-kg change in body weight in the preceding 6 months. Participants recorded all food and beverage intake for 7 consecutive days. Food records were analyzed by our dietitian by using the NDS-R pro-gram (v.4.03_31) and used to define the western diet consumers. None of the participants had evidence of chronic disease, smoked cigarettes, or took medications that could affect the outcome variables. The present study (HRPO #: 01-0804) was approved by the Human Studies Committee of Washington University School of Medicine, and all subjects gave written informed consent before their participation. Height and body weight were obtained in the morning after an overnight fast, with the participants wearing only underwear and a hospital gown. Total body fat mass and fat-free mass were determined by dual-energy X-ray absorptiometry (DXA; QDR 1000/w; Hologic). VO2max was determined by indirect calorimetry during an incremental exercise test to exhaustion37. Participants walked on a level treadmill at a pace that elicited 60–$70\%$ of age-predicted maximal heart rate for a 5-minute warm-up. The speed was then set at the fastest comfortable pace, and the grade was increased 1–$2\%$ every 1–2 minutes until volitional exhaustion, electrocardiographic changes, or other abnormalities that rendered it unsafe to continue. Blood pressure was measured with an oscillometric blood pressure monitor (Dinamap Procare 200; GE Healthcare, Waukesha, WI) in the morning after a 12-h fast. In the EX group, blood pressure was measured at least 48 h after the last exercise session. A venous blood sample was taken to determine lipid and hormone concentrations after subjects had fasted overnight. In the EX group, blood samples were obtained ≥48 h after the last exercise session. Measurement of serum lipid and lipoprotein-cholesterol concentrations, glucose, insulin, C-reactive protein was performed in the Barnes-Jewish Hospital Laboratory by automated enzymatic, radioimmunoassay and ELISA commercial kits. Insulin resistance was calculated using homeostasis model assessment of insulin resistance 9HOMA-IR = [fasting glucose (mmol/l) × fasting insulin 59]/22.5).
## RNA isolation and cDNA synthesis
Biopsy specimens of normal-appearing sigmoidal colon mucosa were collected from a subset of 11 EX, 11 WD-o, and 5 WD-y volunteers in the morning after an overnight fast and a preparation with an enema containing water. Colonic mucosal specimens were immediately washed in PBS and then flash-frozen in liquid nitrogen and stored at −80 °C until processed. Tissues were homogenized in liquid nitrogen. For each sample, 20 mg of tissue powder was used to isolate total RNA using the Isolate II RNA Mini Kit (Bioline). In all, 250–500 ng of RNA was reverse transcribed into cDNA using a kit (Applied Biosystems).
## Real time-qPCR
qRT-PCR reactions were performed with the LightCycler 480 Instrument II (Roche) using UPL system (Roche) with a SensiFast Probe kit (Bioline). The reactions were carried out in a total volume of 10 μl using a TaqMan assay. Tubulin was used for normalization of the CT values. List of primers/probe combination: Tubulin: FW- cttcgtctccgccatcag; RV-cgtgttccaggcagtagagc; Probe #40 Cdkn2a (p16): FW-gagcagcatggagcctc; RV-cgtaactattcggtgcgttg; Probe #67 Cdkn1a (p21): FW-tcactgtcttgtacccttgtgc; RV-ggcgtttggagtggtagaaa; Probe #32
IL6: FW-caggagcccagctatgaact; RV-gaaggcagcaggcaacac; Probe #45 IL8: FW-gagcactccataaggcacaaa; RV-atggttccttccggtggt; Probe #72 MMP3: FW-caaaacatatttctttgtagaggacaa; RV-ttcagctatttgcttgggaaa; Probe #36
## Statistical analysis
One-way analysis of variance (ANOVA) was used to compare group variables, followed by Tukey post-hoc testing when indicated. One-way ANOVA with Games-Howell was performed for distributions where equal variances could not be assumed. Pearson correlation was used to assess associations between continuous variables. Statistical significance was set at $P \leq 0.05$ for all tests. All data were analyzed by using SPSS software, version 28.0 (SPSS Inc, Chicago). Data are expressed as mean ± SEM or SD (indicated). A difference with P-values < 0.05 were considered statistically significant.
## Supplementary information
Supplementary table 1 The online version contains supplementary material available at 10.1038/s41514-023-00100-w.
## References
1. Weiss EP. **Improvements in glucose tolerance and insulin action induced by increasing energy expenditure or decreasing energy intake: a randomized controlled trial**. *Am. J. Clin. Nutr.* (2006.0) **84** 1033-1042. DOI: 10.1093/ajcn/84.5.1033
2. Fontana L. **Calorie restriction or exercise: effects on coronary heart disease risk factors. A randomized, controlled trial**. *Am. J. Physiol.-Endocrinol. Metab.* (2007.0) **293** E197-E202. DOI: 10.1152/ajpendo.00102.2007
3. Racette SB. **One year of caloric restriction in humans: feasibility and effects on body composition and abdominal adipose tissue**. *J. Gerontol. Seri. A: Biol. Sci. Med. Sci.* (2006.0) **61** 943-950. DOI: 10.1093/gerona/61.9.943
4. Hofer T. **Long-term effects of caloric restriction or exercise on DNA and RNA oxidation levels in white blood cells and urine in humans**. *Rejuvenation Res.* (2008.0) **11** 793-799. DOI: 10.1089/rej.2008.0712
5. Fontana L, Klein S, Holloszy JO. **Effects of long-term calorie restriction and endurance exercise on glucose tolerance, insulin action, and adipokine production**. *Age* (2010.0) **32** 97-108. DOI: 10.1007/s11357-009-9118-z
6. Riordan MM. **The effects of caloric restriction- and exercise-induced weight loss on left ventricular diastolic function**. *Am. J. Physiol. Heart Circ. Physiol.* (2008.0) **294** H1174-H1182. DOI: 10.1152/ajpheart.01236.2007
7. Matthews CE. **Amount and intensity of leisure-time physical activity and lower cancer risk**. *J. Clin. Oncol.* (2020.0) **38** 686-697. DOI: 10.1200/JCO.19.02407
8. Moore SC. **Association of leisure-time physical activity with risk of 26 types of cancer in 1.44 million adults**. *JAMA Int. Med.* (2016.0) **176** 816-825. DOI: 10.1001/jamainternmed.2016.1548
9. Giovannucci E. **Physical activity, obesity, and risk for colon cancer and adenoma in men**. *Ann. Int. Med.* (1995.0) **122** 327-334. DOI: 10.7326/0003-4819-122-5-199503010-00002
10. Meyerhardt JA. **Physical activity and survival after colorectal cancer diagnosis**. *J. Clin. Oncol.* (2006.0) **24** 3527-3534. DOI: 10.1200/JCO.2006.06.0855
11. Zhong S. **Association between physical activity and mortality in breast cancer: a meta-analysis of cohort studies**. *Eur. J. Epidemiol.* (2014.0) **29** 391-404. DOI: 10.1007/s10654-014-9916-1
12. McTiernan A. **Mechanisms linking physical activity with cancer**. *Nat. Rev. Cancer* (2008.0) **8** 205-211. DOI: 10.1038/nrc2325
13. Borghesan M, Hoogaars WMH, Varela-Eirin M, Talma N, Demaria M. **A senescence-centric view of aging: implications for longevity and disease**. *Trends Cell Biol.* (2020.0) **30** 777-791. DOI: 10.1016/j.tcb.2020.07.002
14. Sharpless NE, Sherr CJ. **Forging a signature of in vivo senescence**. *Nat. Rev. Cancer* (2015.0) **15** 397-408. DOI: 10.1038/nrc3960
15. Kuilman T, Michaloglou C, Mooi WJ, Peeper DS. **The essence of senescence**. *Genes Dev.* (2010.0) **24** 2463-2479. DOI: 10.1101/gad.1971610
16. Hernandez-Segura A. **Unmasking transcriptional heterogeneity in senescent cells**. *Curr. Biol.* (2017.0) **27** 2652-2660. e2654. DOI: 10.1016/j.cub.2017.07.033
17. Casella G. **Transcriptome signature of cellular senescence**. *Nucleic Acids Res.* (2019.0) **47** 7294-7305. DOI: 10.1093/nar/gkz555
18. van Deursen JM. **The role of senescent cells in ageing**. *Nature* (2014.0) **509** 439-446. DOI: 10.1038/nature13193
19. Demaria M. **An essential role for senescent cells in optimal wound healing through secretion of PDGF-AA**. *Dev. Cell* (2014.0) **31** 722-733. DOI: 10.1016/j.devcel.2014.11.012
20. Baker DJ. **Naturally occurring p16(Ink4a)-positive cells shorten healthy lifespan**. *Nature* (2016.0) **530** 184-189. DOI: 10.1038/nature16932
21. Burd CE. **Monitoring tumorigenesis and senescence in vivo with a p16(INK4a)-luciferase model**. *Cell* (2013.0) **152** 340-351. DOI: 10.1016/j.cell.2012.12.010
22. Yamakoshi K. **Real-time in vivo imaging of p16Ink4a reveals cross talk with p53**. *J. Cell Biol.* (2009.0) **186** 393-407. DOI: 10.1083/jcb.200904105
23. Xu M. **Senolytics improve physical function and increase lifespan in old age**. *Nat. Med.* (2018.0) **24** 1246-1256. DOI: 10.1038/s41591-018-0092-9
24. Fontana L. **The effects of graded caloric restriction: XII. Comparison of mouse to human impact on cellular senescence in the colon**. *Aging Cell* (2018.0) **17** e12746. DOI: 10.1111/acel.12746
25. Green CL, Lamming DW, Fontana L. **Molecular mechanisms of dietary restriction promoting health and longevity**. *Nat. Rev. Mol. Cell Biol.* (2022.0) **23** 56-73. DOI: 10.1038/s41580-021-00411-4
26. 26.Nehme, J. et al. High dietary protein and fat contents exacerbate hepatic senescence and SASP in mice. FEBS J. 10.1111/febs.16292 (2021).
27. Englund DA. **Exercise reduces circulating biomarkers of cellular senescence in humans**. *Aging Cell* (2021.0) **20** e13415. DOI: 10.1111/acel.13415
28. Justice JN. **Cellular senescence biomarker p16INK4a+ cell burden in thigh adipose is associated with poor physical function in older women**. *J. Gerontol. A Biol. Sci. Med. Sci.* (2018.0) **73** 939-945. DOI: 10.1093/gerona/glx134
29. Chen XK. **Is exercise a senolytic medicine? A systematic review**. *Aging Cell* (2021.0) **20** e13294. DOI: 10.1111/acel.13294
30. Manninen V. **Joint effects of serum triglyceride and ldl cholesterol and Hdl cholesterol concentrations on coronary heart-disease risk in the helsinki heart-study - implications for treatment**. *Circulation* (1992.0) **85** 37-45. DOI: 10.1161/01.CIR.85.1.37
31. McLaughlin T. **Use of metabolic markers to identify overweight individuals who are insulin resistant**. *Ann. Int. Med.* (2003.0) **139** 802-809. DOI: 10.7326/0003-4819-139-10-200311180-00007
32. Bayerdorffer E. **Decreased high-density-lipoprotein cholesterol and increased low-density cholesterol levels in patients with colorectal adenomas**. *Ann. Int. Med.* (1993.0) **118** 481-487. DOI: 10.7326/0003-4819-118-7-199304010-00001
33. Cespedes Feliciano EM. **Metabolic dysfunction, obesity, and survival among patients with early-stage colorectal cancer**. *J. Clin. Oncol.* (2016.0) **34** 3664. DOI: 10.1200/JCO.2016.67.4473
34. Kammire MS. **Does walking during chemotherapy impact p16(INK4a) levels in women with early breast cancer**. *J. Clin. Lab. Anal.* (2022.0) **36** e24753. DOI: 10.1002/jcla.24753
35. Lavin KM. **Effects of aging and lifelong aerobic exercise on basal and exercise-induced inflammation in women**. *J. Appl. Physiol. (1985)* (2020.0) **129** 1493-1504. DOI: 10.1152/japplphysiol.00655.2020
36. Nilsson MI. **Lifelong aerobic exercise protects against inflammaging and cancer**. *PLoS ONE* (2019.0) **14** e0210863. DOI: 10.1371/journal.pone.0210863
37. Kohrt WM. **Effects of gender, age, and fitness level on response of VO2max to training in 60–71 yr olds**. *J. Appl. Physiol. (1985)* (1991.0) **71** 2004-2011. DOI: 10.1152/jappl.1991.71.5.2004
|
---
title: Nucleated red blood cells explain most of the association between DNA methylation
and gestational age
authors:
- Kristine L. Haftorn
- William R. P. Denault
- Yunsung Lee
- Christian M. Page
- Julia Romanowska
- Robert Lyle
- Øyvind E. Næss
- Dana Kristjansson
- Per M. Magnus
- Siri E. Håberg
- Jon Bohlin
- Astanand Jugessur
journal: Communications Biology
year: 2023
pmcid: PMC9971030
doi: 10.1038/s42003-023-04584-w
license: CC BY 4.0
---
# Nucleated red blood cells explain most of the association between DNA methylation and gestational age
## Abstract
Determining if specific cell type(s) are responsible for an association between DNA methylation (DNAm) and a given phenotype is important for understanding the biological mechanisms underlying the association. Our EWAS of gestational age (GA) in 953 newborns from the Norwegian MoBa study identified 13,660 CpGs significantly associated with GA (pBonferroni<0.05) after adjustment for cell type composition. When the CellDMC algorithm was applied to explore cell-type specific effects, 2,330 CpGs were significantly associated with GA, mostly in nucleated red blood cells [nRBCs; $$n = 2$$,030 ($87\%$)]. Similar patterns were found in another dataset based on a different array and when applying an alternative algorithm to CellDMC called Tensor Composition Analysis (TCA). Our findings point to nRBCs as the main cell type driving the DNAm–GA association, implicating an epigenetic signature of erythropoiesis as a likely mechanism. They also explain the poor correlation observed between epigenetic age clocks for newborns and those for adults.
CpG sites associated with gestational age are predominantly found in nucleated red blood cells, which point to an epigenetic signature of erythropoiesis as being partly responsible for this association.
## Introduction
Gestational age (GA) is intimately linked to fetal development. Even slight variations in GA at birth are associated with a wide variety of perinatal health outcomes, some of which have important clinical consequences1–5. Epigenetic modifications, such as DNA methylation (DNAm), play a critical role in fetal development6–8. DNAm has also been shown to be robustly associated with GA at thousands of CpG sites5,9–12. The strong association between DNAm and GA probably reflects biological processes related to fetal development, but the specific mechanisms underlying this association are still unknown. Thus, elucidating these mechanisms may provide a deeper understanding of the molecular processes involved in normal as well as aberrant fetal growth and development.
Most of the previous epigenome-wide association studies (EWASs) of GA were based on DNAm data generated on the Illumina Infinium HumanMethylation450 array (450k) or its predecessor, the Illumina Infinium HumanMethylation27 array (27k)5,9,12. These arrays were designed to cover mainly gene promoters and protein-coding regions13,14. In December 2015, 450k was replaced by the more comprehensive Illumina Infinium MethylationEPIC array (EPIC), which employs the same technology as 450k for measuring DNAm but contains almost twice the number of CpG sites (~850,000) and has a higher coverage of CpGs in regulatory regions13. Despite the substantial improvement in genome-wide coverage of regulatory regions and the higher reproducibility and reliability of EPIC13, studies investigating the association between GA and DNAm data generated on EPIC are lacking. It is also uncertain whether the extra regulatory CpGs on EPIC are useful in explaining the association between GA and DNAm.
Most studies exploring the link between DNAm and GA are based on samples from cord blood, which comprises a mixture of cell types15. As cell-type proportions vary substantially across individuals and DNAm is highly cell-type specific16, it is customary to adjust for cell-type proportions in statistical models in order to avoid bias17. Several cellular deconvolution algorithms and cord-blood reference panels are available to infer cell-type proportions from heterogeneous samples and adjust for cord blood cell-type composition in newborn DNAm data18–20. However, including cell-type proportions as covariates in the statistical model will not necessarily provide insight as to how cell types influence the association between the explanatory variable and DNAm. One solution is to perform an EWAS in isolated cell types. However, cell sorting of whole-blood samples is costly, especially in large cohort studies with hundreds of thousands of participants.
To counter this, statistical algorithms have been developed to allow the detection of cell-type specific differential DNAm within a heterogeneous mixture of cells without the need for cell sorting or single-cell methods21–24. One example is CellDMC, by Zheng et al.24, which incorporates interaction terms between the phenotype of interest and the estimated cell-type fractions in a linear modeling framework. Another example is Tensor Composition Analysis (TCA), by Rahmani et al.23. which employs matrix factorization to infer cell-type specific DNAm signals that are subsequently used to search for associations in each cell type separately. Exploring cell-type specific associations can be essential to decipher the biological underpinnings of an association between DNAm and a specific phenotype of interest25. Whilst changes in cord blood cell-type proportions have been reported for GA26,27, studies on cell-type specific epigenetic associations with GA are lacking.
To bridge these knowledge gaps, we investigate the association between cord blood DNAm and GA using an EPIC-derived DNAm dataset comprising 953 newborns and a 450k-derived dataset comprising 1062 newborns. Both datasets are from the Norwegian Mother, Father, and Child Cohort Study (MoBa)28. We apply CellDMC to these datasets to determine the relationship between cell-type specific DNAm and GA. We also apply TCA as an alternative method for cell-type-specific analysis. The results show many CpGs associated with GA, predominantly in nucleated red blood cells (nRBCs). This association reflects an epigenetic signature of erythropoiesis in fetal development and provides a biologically plausible rationale for the consistently observed strong association between DNAm and GA. It also helps explain the observed incompatibility between epigenetic age clocks for newborns and those for adults.
## Study sample characteristics
We analysed cord blood DNAm in newborns from two substudies in MoBa. The main study sample consisted of 953 naturally conceived newborns from the Study of Assisted Reproductive Technology (START), in which DNAm was measured using the EPIC array29,30. We also used another dataset consisting of 1062 newborns (referred to as MoBa1 hereafter) with DNAm measured using the 450k array10. GA ranged from 216–305 days (mean 280.1 days, SD ± 10.7 days) in START and 209–301 days (mean 279.8 days, SD ± 10.8 days) in MoBa1. Table 1 summarizes the key demographic and clinical characteristics of these two datasets. More MoBa1 mothers continued to smoke during pregnancy compared to START mothers ($$p \leq 0.033$$, Table 1). There were also more boys in MoBa1 than in START ($$p \leq 0.007$$, Table 1).Table 1Characteristics of the mothers and newborns in START and MoBa1.CharacteristicsSTART $$n = 956$$MoBa1 $$n = 1062$$p valueaMothersAge (years), mean (SD)29.9 (4.7)29.9 (4.3)0.800Smoking, n (%)0.033No smoking before or during pregnancy478 ($50\%$)522 ($49\%$)Smoked, but quit before pregnancy245 ($26\%$)233 ($22\%$)Smoked, but quit early in pregnancy131 ($14\%$)154 ($15\%$)Continued smoking during pregnancy102 ($11\%$)153 ($14\%$)NewbornsGA in days, mean (SD)280.1 (10.7)279.8 (10.8)0.400GA in days, min216209GA in days, max305301Birth weight in grams, mean (SD)3657 [521]3643 [539]0.500Sex (male), n (%)455 ($47\%$)569 ($54\%$)0.007SD standard deviation, GA gestational age.aWilcoxon rank-sum test; Pearson’s Chi-squared test.
## Analyses of cell-type composition
We estimated the proportion of each of the seven main cell types in cord blood (B-cells, CD4 + T-cells, CD8 + T-cells, granulocytes, monocytes, natural killer cells, and nRBCs) separately in START and MoBa1, using a combined reference dataset consisting of cell-type specific DNAm profiles in cord blood19 (Fig. 1 and Supplementary Data 1). As expected from the reference data, granulocytes and nRBCs were the two most abundant cell types in both datasets. The results of a principal component analysis (PCA) of cell-type proportions in START further confirmed that granulocytes and nRBCs explained most of the variance in cell-type composition (Supplementary Fig. 1 and Supplementary Table 1).Fig. 1Estimated proportions of seven main cell types in cord blood.a Estimated proportions of cell types in the START dataset ($$n = 953$$, EPIC-based). b Estimated proportions of cell types in the MoBa1 dataset ($$n = 1062$$, 450k-based). The upper and lower box limits correspond to the interquartile range (25 to $75\%$ of the values for each cell type) and the horizontal line in the box represents the median value. The whiskers outstretch 1.5 times the box height from the top and bottom of the box. The dots outside the whiskers represent outliers beyond the interquartile range. The percentage below each cell type denotes the median proportion of that cell type. Bcell B-cell, CD4T CD4 + T-cell, CD8T CD8 + T-cell, Gran granulocyte, Mono monocyte, NK natural killer cell, nRBC nucleated red blood cell.
We examined the proportion of each cell type in START and found significant correlations with GA in B-cells (Pearson correlation r = –0.21, $$p \leq 6.30$$ × 10−11), CD4 + T-cells (r = −0.10, $$p \leq 0.002$$), granulocytes ($r = 0.20$, $$p \leq 5.77$$ × 10−10), and nRBCs (r = −0.08, $$p \leq 0.010$$; see Supplementary Fig. 2 for more details).
## Conventional EWAS of GA
First, we applied a linear mixed effects regression model to the EPIC-derived START dataset where the outcome was DNAm level at each CpG, the exposure was GA, and the following were included as covariates: cell-type proportions, newborn sex, maternal age, maternal smoking, and batch (see Methods for details). This model is referred to as the conventional EWAS model throughout this paper, since this framework is routinely adopted in the majority of published EWASs. We identified 13,660 CpGs significantly associated with GA after applying a Bonferroni correction for multiple testing (Bonferroni-corrected p value (pB) <0.05, Fig. 2a and Supplementary Data 2). About 7639 ($56\%$) of the GA-associated CpGs were only present on the EPIC array and were distributed across the genome (Supplementary Fig. 3). Most of the GA-associated CpGs in the conventional EWAS were hypermethylated [$$n = 9503$$ ($70\%$), Fig. 3a].Fig. 2Manhattan plots of the epigenome-wide DNAm associated with GA in START ($$n = 953$$).a Results from the conventional EWAS where we adjusted for the estimated cell-type proportions (see Methods for details of the statistical model). b–h Results for each of the seven cell types from the cell-type specific analysis using CellDMC. CpG loci are aligned on the x-axis according to their genomic coordinates. The y-axis represents the −log10 p values. The dashed black line denotes the Bonferroni-corrected genome-wide significance threshold (pB < 0.05).Fig. 3Volcano plots of the epigenome-wide DNAm associated with GA in START ($$n = 953$$).a Results from the conventional EWAS in which we adjusted for estimated cell-type proportions (see Methods for details of the statistical model). b–h Results for each of the seven cell types from the cell-type specific analysis using CellDMC. Gray dots indicate nonsignificant associations and colored dots indicate those that are Bonferroni-significant (pB < 0.05). Blue-colored dots show CpGs with a negative effect size and orange dots show CpGs with a positive effect size. The x-axis represents coefficient estimates (β-values) for the DNAm–GA association, and the y-axis the corresponding -log10 p values. The horizontal dashed line denotes the Bonferroni-corrected genome-wide significance threshold (pB < 0.05).
## Cell-type specific analyses of the association between DNAm and GA
We applied CellDMC to investigate cell-type specific DNAm in the START dataset and identified 2,330 CpGs significantly associated with GA (pB <0.05, Fig. 2b–h). Most of these CpGs ($$n = 2030$$, $87\%$) were specific for nRBCs (Fig. 2h), and only a few of the CpGs ($$n = 31$$–157 and 1.3–$6.7\%$) were identified in the other cell types. Moreover, 522 of the 2330 cell-type-specific CpGs associated with GA were also identified in the conventional EWAS. Detailed results of the CellDMC analyses are provided in Supplementary Data 3.
CpGs that were associated with GA in CD4 + T-cells and monocytes were predominantly hypermethylated [CD4 + T-cells: $$n = 67$$ ($65\%$), Fig. 3c; monocytes: $$n = 29$$ ($78\%$), Fig. 3f]. We found an almost equal number of hyper- and hypomethylated CpGs associated with GA in B-cells [hypermethylated $$n = 29$$ ($55\%$); hypomethylated $$n = 24$$ ($45\%$); Fig. 3b] and CD8 + T-cells [hypermethylated $$n = 13$$ ($42\%$); hypomethylated $$n = 18$$ ($58\%$); Fig. 3d]. In contrast, GA-associated CpGs specific for granulocytes, natural killer cells, and nRBCs were predominantly hypomethylated [granulocytes: $$n = 97$$ ($71\%$), Fig. 3e; natural killer cells: $$n = 97$$ ($62\%$), Fig. 3g; nRBCs: $$n = 1888$$ ($93\%$), Fig. 3h].
## Impact of the type of DNAm array: 450k versus EPIC
To determine whether the type of DNAm array had an impact on the cell-type specific results, given the lower coverage of regulatory CpGs on 450k compared to EPIC, we repeated the CellDMC analysis on MoBa1 ($$n = 1062$$ newborns) in which DNAm was measured using 450k. The results showed a similar pattern of cell-type specific DNAm associated with GA, despite fewer significant CpGs overall ($$n = 373$$, pB < 0.05, Supplementary Data 4 and Supplementary Fig. 4). Specifically, $62\%$ ($$n = 231$$) of the Bonferroni-significant CpGs mapped to nRBCs.
To further assess the robustness of our findings, we used the r value approach of ref. 31 to compare the results from START and MoBa1. This approach tests if a CpG is significantly associated in two separate studies and then computes the corresponding false discovery rate (FDR) value of this test, which is referred to as the r value (see Methods for details). If the r value was <0.05, we deemed a GA–CpG association detected in START as successfully replicated in MoBa1. Among 1129 nRBC-specific CpGs detected in START that were also available on the 450k array, 174 CpGs were significantly replicated in MoBa1 (r < 0.05, Fig. 4 and Supplementary Data 5). The results were also consistent in terms of the direction of effect, except for one CpG (cg13746414). Importantly, there was no overlap in CpGs between START and MoBa1 for the remaining six cell types (r < 0.05, Supplementary Fig. 5).Fig. 4Comparison of nRBC-specific CpGs associated with GA in the EPIC-based START dataset ($$n = 953$$) and the 450k-based MoBa1 dataset ($$n = 1062$$).Gray dots indicate nonsignificant CpGs, blue dots CpGs significantly associated only in MoBa1 (pB < 0.05), green dots CpGs significantly associated only in START (pB < 0.05), and orange dots CpGs significantly associated in both datasets (r < 0.05). Black isolines indicate the density of the points, increasing towards the crossing point of the axes. The x and y axes represent z-scores (i.e., the coefficient estimate divided by the standard error) for START and MoBa1, respectively.
## Validation with a different cell-type specific method
To further validate the cell-type specific associations between DNAm and GA, we applied TCA to the START dataset using two different approaches. First, we applied a one-stage implementation of TCA which runs marginal conditional tests for each cell type, analogous to CellDMC. We then applied a two-stage implementation of TCA, by first extracting the cell-type tensors additionally adjusted for the above-mentioned covariates and then performing separate EWAS regressions on each tensor with respect to GA. With the one-stage approach, we identified 979 GA-associated CpGs (pB <0.05), whereas with the two-stage approach, we identified 4714 GA-associated CpGs (pB <0.05). Both approaches map most of the cell-type specific significant CpGs to nRBCs [$$n = 836$$ ($85\%$) in the one-stage approach (Supplementary Fig. 6) and $$n = 3130$$ ($66\%$) in the two-stage approach (Supplementary Fig. 7)]. For all cell types, more CpGs were statistically significant using the two-stage approach compared to the one-stage approach. In granulocytes specifically, 1668 CpGs were identified as significantly associated with GA, of which 829 were also mapped to nRBCs. The results from the one-stage and two-stage TCA analyses can be found in Supplementary Data 6 and 7, respectively.
Among the 2030 nRBC-specific CpGs detected by CellDMC, 623 CpGs were also detected when applying the one-stage TCA (Supplementary Fig. 8). Overall, 260 nRBC-specific CpGs were detected by both CellDMC and the two-stage TCA approach (Supplementary Fig. 9). The results from the one-stage TCA analysis were also generally consistent with those of the CellDMC analysis for the other six cell types (Supplementary Fig. 8), while the two-stage TCA results showed more divergent associations for the other cell types (Supplementary Fig. 9).
## Location of GA-associated CpGs
We scrutinized the GA-associated CpGs identified by the conventional EWAS and CellDMC analyses according to their location in the genome (Fig. 5 and Supplementary Data 2 and 3). The 2030 nRBC-specific CpGs that were significantly associated with GA in START were predominantly localized to gene bodies ($48\%$ of the nRBC-specific CpGs versus $30\%$ of all CpGs on EPIC, $$p \leq 2.5$$ × 10−67, Fig. 5a), open sea ($75\%$ versus $56\%$, $$p \leq 2.2$$ × 10−69, Fig. 5b), and CpG island shelves ($8.2\%$ versus $7.1\%$, $$p \leq 0.023$$, Fig. 5b). Markedly fewer nRBC-specific CpGs were in promoter regions (22 versus $38\%$, $$p \leq 2.8$$ ×10−55, Fig. 5a), shores (12 versus $18\%$, $$p \leq 1.0$$ × 10−12, Fig. 5b), and CpG islands ($4.7\%$ versus $19\%$, $$p \leq 5.3$$ × 10−77, Fig. 5b). We discovered a similar pattern of CpG localization in the nRBC-specific MoBa1 results. The corresponding patterns for the other cell types showed more variation between the two datasets (Supplementary Fig. 10), which may be due to a substantially lower number of CpGs in each category. Fig. 5Position enrichment results of CpGs associated with GA compared to all CpGs on the EPIC array. Position enrichment results of all the CpGs on the EPIC array ($$n = 770$$,586; denoted as EPIC on the x-axis), those specifically associated with GA in the conventional EWAS ($$n = 13$$,660; EWAS), and each cell type in the CellDMC analyses in START (Bcell, $$n = 53$$; CD4T, $$n = 103$$; CD8T, $$n = 31$$; Gran, $$n = 136$$; Mono, $$n = 37$$; NK, $$n = 157$$; nRBC, $$n = 2030$$). a The proportion of CpGs in the promoter (orange), gene body (yellow), and intergenic (blue) regions. b The proportion of CpGs in CpG islands (orange), shores (green), shelves (yellow), and open sea (blue). Bcell B-cell, CD4T CD4 + T-cell, CD8T CD8 + T-cell, Gran granulocyte, Mono monocyte, NK natural killer cell, nRBC nucleated red blood cell.
## Gene annotation and enrichment analysis of nRBC-specific CpGs associated with GA
We used the online Genomic Regions Enrichment of Annotations Tool (GREAT)32 to examine whether the 2030 GA-associated CpGs for nRBC were located near or within any gene of known pathway annotation. 2836 genes were identified using this approach (Supplementary Data 8), 198 of which were associated with more than three differentially methylated CpGs. A foreground/background hypergeometric test was performed on the 2030 GA-associated nRBC-specific CpGs. The results of this test revealed four clusters of Gene Ontology (GO) biological processes significantly enriched in our data (Supplementary Data 9). These processes were related to (i) response to corticosteroid (75 CpGs/55 genes, pB = 0.0001), (ii) response to purine-containing compound (65 CpGs/45 genes, pB = 0.002), (iii) granulocyte migration (34 CpGs/23 genes, pB = 0.006), and (iv) stress-activated protein kinase signaling cascade (58 CpGs/32 genes, pB = 0.01). When the analyses were restricted to only those CpGs that are present on both 450k and EPIC, we did not find any significantly enriched biological pathways.
## Discussion
Although epigenome-wide associations between GA and DNAm in cord blood are now well established, little is known about the contribution of different cell types and the biological mechanisms underlying these associations. In this study, we explored the association between GA and DNAm using data from two types of DNAm arrays (EPIC and 450k) and conducted both a conventional EWAS as well an investigation of cell-type specific associations. We found that most of the cell-type-specific associations between DNAm and GA were restricted to nRBCs. These results were robust across different datasets, DNAm arrays, and analysis methods. Our results point to a strong link between red blood cell development (erythropoiesis) in fetal life and fetal growth as measured by GA, providing critical insights and implications for further studies on the relationship between DNAm and GA.
In the conventional EWAS, we identified 13,660 CpGs linked to 8669 genes as being differentially methylated with GA. Slightly more of the significant CpGs were specific for the EPIC array ($56\%$), despite only $48\%$ of the CpGs being EPIC-specific. Bohlin et al.10 previously applied a similar model to the MoBa1 dataset and identified 5474 CpGs associated with GA. 2556 of the CpGs and 1741 of the genes identified in that study overlap with our results in the START EPIC-based dataset. We also compared our results to the “all births model” from a recent meta-analysis by Merid et al.5 where the authors investigated GA and DNAm measured on 450k in cord-blood DNA from 6885 newborns in 20 different cohorts. The authors identified 17,095 CpGs significantly associated with GA, of which 4688 CpGs and 4437 genes overlap with our results. Of note, MoBa1 and yet another MoBa-based dataset (MoBa2) were also included in the meta-analysis by Merid et al. Nevertheless, these comparisons show that the results from our conventional EWAS model are concordant with those of previous studies on DNAm and GA.
As a primary step to explore cell-type specific changes in DNAm with GA, we used the interaction-based algorithm CellDMC that has been validated in several EWAS datasets and data in which the actual cell-type composition is known24,33,34. We identified 2330 differentially methylated CpGs associated with GA, with an overwhelming number of the significant CpGs confined to nRBCs (2030 CpGs linked to 2836 genes). This is particularly striking given that nRBCs are not the dominant cell type in terms of variation and abundance. Taken together, these findings strongly suggest that DNAm changes in nRBCs are responsible for the observed DNAm–GA association.
It is nevertheless important to account for the limited sensitivity of CellDMC when including seven different cell types in the analysis33. To assess this limited sensitivity and verify that the nRBC-specific results were not an array-based artifact, we repeated the CellDMC analyses in MoBa1, which is a 450k-based dataset stemming from the same source population as the START dataset (MoBa). We observed a similar pattern of cell-type-specific association with GA as with the START dataset, although there were fewer significant CpGs in the MoBa1 dataset. Moreover, 174 nRBC-specific CpGs were significantly associated with GA in both datasets, as opposed to no such overlap in CpGs across the other six cell types. One option to further increase the power of the CellDMC analysis would have been to merge the two datasets over the common set of 450k CpGs. Even though this would have increased the sample size substantially, such an approach has several major drawbacks. First, one would lose the much greater coverage of the EPIC array and possibly miss important associations between GA and CpGs that are only detectable using EPIC-derived DNAm data. Second, merging the datasets would introduce a new batch variable that would need to be accounted for in the model. We thus opted to keep the analyses of the two datasets separate.
To further validate our results, we applied another method for cell-type specific analysis, TCA, to the START data. TCA utilizes a statistical framework based on matrix factorization23. The results from both the one-stage and two-stage applications of TCA showed a similar pattern of cell-type specific association with GA as observed with CellDMC. Our findings are also consistent with a previous study on nRBCs pointing to extensive DNAm changes in nRBCs between preterm and term newborns35. In that study, the authors identified 9258 differentially methylated sites when comparing nRBCs from preterm and term newborns. These sites were predominantly hypomethylated and enriched in gene body and intergenic regions35. Taken together, these results strengthen the interpretation that nRBCs are the primary cell type driving the association between DNAm and GA in cord blood.
nRBCs are an integral part of erythropoiesis, the process by which mature red blood cells (erythrocytes) are produced in adult and fetal bone marrow, fetal liver, and the embryonic yolk sac. Erythropoiesis is crucial for embryonic and fetal growth. During the third trimester of pregnancy, the production of erythrocytes is approximately three to five times that of the adult steady-state levels36. Although nRBCs circulate in the fetal bloodstream throughout pregnancy, they stay in circulation for only a few days after birth37. *Several* genes annotated to the nRBC-specific CpGs that we found to be associated with GA are implicated in a wide array of biological processes involved in erythropoiesis. A subset of the genes related to these processes are described in more detail in Supplementary Data 10. Briefly, these processes include cell-cycle progression and cytokinesis38,39, chromatin condensation39,40, hemoglobin synthesis38, mitochondrial function and iron metabolism38,41,42, degradation of proteins and organelles34,43, erythroblastic island formation44, and enucleation39,40. Moreover, several of the genes are essential for the switch from fetal to adult hemoglobin, which occurs shortly after birth45. Taken together, our findings provide strong support for fetal erythropoiesis representing an important biological mechanism underlying the association between DNAm and GA.
To learn more about the mechanisms contributing to the nRBC-specific association between DNAm and GA, we searched for the enrichment of specific biological pathways in the set of nRBC-specific CpGs. One of the main clusters of biological pathways was the response to corticosteroids, and more specifically, the response to glucocorticoids. Glucocorticoids are a class of corticosteroids that are essential for a wide variety of biological processes, including proliferation, differentiation, and apoptosis of many cell types in response to stress. They also play a pivotal role in pregnancy and normal fetal development46, even though prenatal overexposure to glucocorticoids has also been reported to be detrimental to fetal growth and postnatal physiology47,48. Glucocorticoids are known regulators of erythroid progenitors49,50, and the glucocorticoid receptor encoded by NR3C1 controls several processes involved in erythropoiesis51–53. In particular, the glucocorticoid receptor controls erythroid response to stress54–56. Stress, such as hypoxia, leads to the glucocorticoid receptor-dependent activation of the BMP4-dependent stress erythropoiesis pathway, in which many new erythrocytes are generated to maintain homeostasis57. Interestingly, stress erythropoiesis shares several similarities with fetal erythropoiesis58.
The link between erythropoiesis and GA is not unprecedented. Several of the genes found to be relevant for erythropoiesis in our data have previously been identified in other studies of GA. A few examples include NCOR25,10,59,60, HDAC45,10,60, CASP85,10,60,61, and RAPGEF25,60,62. The nuclear receptor co-repressor encoded by NCOR2 interacts with the transcription factor BCL11A in regulating the expression of fetal hemoglobin63. NCOR2 also promotes chromatin condensation, which is a crucial step during terminal erythropoiesis. Histone deacetylase 4 (HDAC4) also plays a key role in chromatin condensation and associates directly with the key erythroid transcription factor GATA164. CASP8 encodes the protease Caspase 8, which is a key activator of effector caspases required for terminal erythroid differentiation65. Finally, RAPGEF2 encodes a guanine nucleotide exchange factor known to play an important role in embryonic hematopoiesis66.
The results of our study, as well as those of others described above, strongly suggest that DNAm patterns related to erythropoiesis are at least partly responsible for the observed association between DNAm and GA. Our findings of predominantly hypomethylated nRBC-specific CpGs are in line with previous studies showing progressive global DNA hypomethylation involved in erythroid lineage commitment and differentiation as well as chromatin condensation and enucleation of nRBCs during erythropoiesis67,68. Other studies have consistently shown a higher proportion of hypomethylated CpGs amongst those associated with GA5,10,59,61.
Further, the findings that nRBCs are the primary drivers behind the association between DNAm and GA may help explain the poor correlation observed between epigenetic clocks for newborn GA and those for chronological age in adults10,11. Indeed, GA-related changes in cord blood DNAm do not persist through childhood and adolescence, as shown in a longitudinal analysis of DNAm associated with GA59 and a meta-analysis of several EWASs of GA5. This could be due to the rapid loss of nRBCs with increasing GA and its subsequent disappearance from the bloodstream of healthy newborns within the first few days after birth. In other words, the disappearance of nRBCs shortly after birth implies that the main driver behind the GA-related changes in cord blood DNAm also disappears. Moreover, the association between GA and specific DNAm changes in nRBCs, as demonstrated by our study, may also help explain why GA acceleration (GAA, defined as the discrepancy between GA predicted from DNAm data and GA determined by clinical measurements) has been linked to several adverse outcomes11,69,70. In this regard, it is interesting to note that increased nRBC counts at birth are associated with a higher risk of mortality and adverse neonatal outcomes and have been suggested as a predictive marker for perinatal hypoxia, intrauterine growth restriction, and preeclampsia71–75. Further studies are needed to determine if GAA is indeed related to these or other adverse outcomes, and if differences in nRBCs may be driving these associations.
The results of our study may have important clinical implications. For instance, fetal nRBCs are routinely isolated from the mother’s peripheral blood during pregnancy for prenatal diagnostics, and several experimental approaches are available for the rapid isolation of nRBCs76,77. Our findings may help pave the way for the development of DNAm-based GA prediction during pregnancy based on nRBC-specific assays, which may provide a more targeted assessment of fetal growth and prenatal development.
One important limitation of our study is the use of in silico estimations of cell-type proportions. Although we have used a reference-based method with validated cord blood-specific reference data, it is important to bear in mind that the proportions we have used here are only estimates. In addition, since the cell-type proportions are essentially fractions that sum up to one, they are not independent of each other, and the correlation between them may impact our analyses. However, since our results were robust despite the use of different DNAm arrays, datasets, and methods, our findings are unlikely to be severely affected by these limitations.
In conclusion, the results of our study strongly indicate that nRBCs are the primary drivers behind the observed DNAm–GA association. Importantly, an epigenetic signature of erythropoiesis seems to be partly responsible for this association, providing a biologically compelling mechanism that links GA, DNAm, and nRBCs. Furthermore, our findings provide an explanation for the poor correlation observed between epigenetic clocks for newborn GA and those for chronological age in adults, contributing important mechanistic insights into the epigenetic regulation of fetal growth and development.
## Study population
MoBa is a population-based pregnancy cohort study in which ~114,500 newborns, 95,200 mothers, and 75,200 fathers were recruited from all over Norway from 1999 to 200828. The mothers consented to participation in $41\%$ of the pregnancies. The study participants have been followed at different time points via self-administered questionnaires and linkage to the Medical Birth Registry of Norway (MBRN). Further details on MoBa have been provided in our previous publications28,78.
For this study specifically, we used two non-overlapping subsamples: (i) the Study of Assisted Reproductive Technology (START; $$n = 953$$ newborns) and (ii) MoBa1 ($$n = 1062$$ newborns). Both datasets are based on cord blood samples from the same source population (MoBa). However, they differ in the methylation array used to generate the DNAm data: START used EPIC whereas MoBa1 used 450k (see below for details). An overview of the sample selection and analysis flow is provided in Supplementary Fig. 11. Detailed characteristics and eligibility criteria for the START and MoBa1 datasets have been provided in our previous work29,79.
## Sample processing, DNAm measurement, and quality control
The sample processing, DNAm measurement, and quality control pipeline used for data cleaning have been extensively detailed in our previous works29,79. Briefly, cord blood samples taken by a midwife immediately after birth were frozen. For the START dataset, DNAm was measured at 885,000 CpG sites using the Illumina Infinium MethylationEPIC BeadChip (Illumina, San Diego, USA). The raw iDAT files were processed in four batches using the R package RnBeads80. Cross-hybridizing probes81 and probes that had a detection p value above 0.01 were removed using the greedycut algorithm in RnBeads. We also excluded probes in which the last three bases overlapped with a single-nucleotide polymorphism (SNP). The remaining DNAm signal was processed using BMIQ82 to normalize the type I and type II probe chemistries83. The RnBeads output of control probes were visually inspected for all samples, and those with low overall signals were removed. The greedycut algorithm was used to remove outliers with markedly different DNAm signals than the rest of the samples, resulting in the removal of 58 samples in total. For consistency, CpG sites excluded from one batch due to poor quality and low detection p value were also removed from all subsequent batches.
For the MoBa1 samples, DNAm was measured at 485,577 CpG sites using the Illumina Infinium HumanMethylation450 BeadChip (Illumina, San Diego, USA). Arrays not fulfilling the $5\%$ detection p value were removed together with all duplicates. Within-array normalization was carried out using BMIQ from the wateRmelon R package84.
## Variables
Information on GA, newborn sex and birth weight, maternal age, parity, and whether the birth was induced was extracted from MBRN. GA at birth was estimated by ultrasound measurements around week 18 of pregnancy. Since newborn sex may occasionally be incorrectly recorded in MBRN, we inferred sex from the DNAm data. As a result, one female was reclassified as male, and five males were reclassified as females. Information on maternal smoking was derived from the MoBa questionnaires and was included as a four-level categorical variable: (i) no smoking before or during pregnancy; (ii) smoked, but quit before pregnancy; (iii) smoked, but quit early in pregnancy; and (iv) continued smoking during pregnancy.
## Estimation of cell-type proportions
To estimate cell-type proportions in our samples, we used the filtered and combined reference dataset “FlowSorted. CordBloodCombined.450 k” from ref. 19, which specifies seven main cell types in cord blood (B-cells, CD4 + T-cells, CD8 + T-cells, granulocytes, monocytes, natural killer cells, and nRBCs). We used the estimateCellCounts2 function in the FlowSorted. Blood. EPIC R package85 and the IDentifying Optimal Libraries (IDOL) probe selection to perform cellular deconvolution and noob preprocessing.
## Statistics and reproducibility
After quality control, the sample available for the current analyses in the START dataset consisted of 770,586 autosomal CpGs and 953 newborns conceived naturally and for whom we had information on ultrasound-based GA (Supplementary Fig. 9). For the MoBa1 dataset, the sample available for the current analyses comprised 473,731 autosomal CpGs and 1062 newborns with information on ultrasound-based GA (Supplementary Fig. 9).
Principal component analysis (PCA) of estimated cell-type proportions was conducted using the prcomp R function. The R package robustbase86 for MM-type robust regression was used to assess the relationship between cell-type composition and GA. Bonferroni correction was applied to the results from the conventional EWAS and cell-type-specific models to control for multiple testing. A Bonferroni p value (pB) <0.05 was declared statistically significant.
## Analyses in START
In the conventional EWAS model, we screened for associations between DNAm in cord blood and GA at birth by applying a linear mixed-effect model to each of the 770,586 CpG sites remaining after quality control. The β-values of the individual CpGs were used as the response (dependent) variables and GA was used as the explanatory (independent) variable, with adjustments made for newborn sex, maternal age, maternal smoking, cell-type proportions, and array plate in the regression model.
To assess interactions between cell-type specific DNAm and GA, we performed epigenome-wide analyses using the CellDMC framework as outlined in ref. 24 and the corresponding CellDMC function in the EpiDISH R package. Briefly, CellDMC runs a linear model similar to that used in our conventional EWAS, but it also includes an interaction term to inform the model whether there is a significant interaction between the exposure and the corresponding fraction of each specific cell type. Estimates of the regression coefficients and p values are calculated for each cell type using least squares. As with the conventional EWAS, newborn sex, maternal age, maternal smoking, and plate were also included as covariates in the CellDMC model. Bonferroni correction was applied to all the results from the conventional EWAS and CellDMC models to control for multiple testing. As before, a Bonferroni p value (pB) <0.05 was declared statistically significant.
Besides CellDMC, we also applied the TCA framework developed by ref. 23 to detect cell-type specific DNAm–GA associations. In contrast to CellDMC, TCA is based on the concept of matrix factorization. Specifically, TCA uses the DNAm measurements from the mixed samples along with information on cell-type proportions (in our case, the ones that are estimated) for each individual and calculates a three-dimensional tensor of DNAm values for each cell type in each individual. The TCA framework further allows a search for statistical associations between cell-type specific signals and an outcome or exposure of interest. We used two different approaches for TCA based on the available functions in the TCA package23. First, we applied a one-stage approach using the tca function, which fits a model for all cell types jointly and tests the effect of each cell type separately for statistical significance. We included the same covariates in the TCA model as in the CellDMC and conventional EWAS models (newborn sex, maternal age, maternal smoking, and array plate). Additionally, we applied a two-stage approach, where a tensor for each cell type is first inferred and then an EWAS of GA is conducted for each tensor. This was carried out by first using the tca function to fit a model including all covariates mentioned above except GA. The model resulting from the tca function was subsequently added as input for the tensor function, obtaining new DNAm tensors for each cell type. An EWAS of GA was then performed for each cell-type-specific tensor.
## Analyses in MoBa1
To test whether array type had an impact on the findings obtained from the analysis of the EPIC-based START dataset, we re-ran the CellDMC analysis on the 450k-based MoBa1 dataset, testing all the 473,731 CpGs available in this dataset.
To compare the CellDMC results from MoBa1 with those from START, we applied the r value approach suggested by ref. 31, which allows a rigorous assessment of the replication of findings. In short, we tested each CpG for association with GA in both datasets (MoBa1 and START) and computed an r value (the lowest FDR level at which the finding was replicated). We chose the r value approach over other approaches, such as those used in a standard meta-analysis or a two-step replication study, for the following reasons. First, a meta-analysis tests whether there is any signal across the two studies; however, it does not test whether the two studies show appropriate significance. Second, assessing replicability in a two-step replication study is not straightforward, as this requires adequate control of the type I error in both studies. This may involve a different number of tests, especially as we use two types of DNAm arrays (EPIC and 450k). Thus, the approach of ref. 31 provides a simpler solution for assessing replicability and for controlling the type I error.
## Location of CpGs
Information on CpG location and regulatory regions was extracted from the respective Illumina Manifest Files (Infinium MethylationEPIC v1.0 B4 for START and HumanMethylation450 v1.2 for MoBa1). One-tailed hypergeometric tests were conducted to assess the relative enrichment of CpGs in specific regions of interest.
## Gene annotation and enrichment analysis
CpGs were annotated using the online Genomic Regions Enrichment of Annotations Tool (GREAT32) using the human genome build hg19 (GRCh37). GREAT was selected amongst other competing methods because it considers both proximal (5.0 kb upstream and 1.0 kb downstream) and distal (up to 1000 kb) regulatory regions. This is an advantage over other methods that only take proximal regions into account, because taking distal regulatory regions into account enables an assessment of the extra information gained from detecting DNAm on distal regulatory CpGs on the EPIC array. *For* gene enrichment analysis, GREAT performed a foreground/background hypergeometric test over genomic regions using the total number of CpGs surviving quality control as background (770,586 CpGs for the EPIC analyses and 473,731 CpGs for the 450k analyses). Finally, GREAT extracts information from Gene Ontology (GO) and other ontologies covering human and mouse phenotypes32.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Supplementary Data 7 Supplementary Data 8 Supplementary Data 9 Supplementary Data 10 Reporting Summary The online version contains supplementary material available at 10.1038/s42003-023-04584-w.
## Peer review information
Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Manuel Breuer.
## References
1. Ghartey K. **Neonatal respiratory morbidity in the early term delivery**. *Am. J. Obstet. Gynecol.* (2012.0) **207** 292.e291-294. DOI: 10.1016/j.ajog.2012.07.022
2. 2.Knight, A. K., Conneely, K. N. & Smith, A. K. Gestational age predicted by DNA methylation: potential clinical and research utility. Epigenomics10.2217/epi-2016-0157 (2017).
3. Raby BA. **Low-normal gestational age as a predictor of asthma at 6 years of age**. *Pediatrics* (2004.0) **114** e327-e332. DOI: 10.1542/peds.2003-0838-L
4. Yang S, Bergvall N, Cnattingius S, Kramer MS. **Gestational age differences in health and development among young Swedish men born at term**. *Int. J. Epidemiol.* (2010.0) **39** 1240-1249. DOI: 10.1093/ije/dyq070
5. Merid SK. **Epigenome-wide meta-analysis of blood DNA methylation in newborns and children identifies numerous loci related to gestational age**. *Genome Med.* (2020.0) **12** 25. DOI: 10.1186/s13073-020-0716-9
6. Guo H. **The DNA methylation landscape of human early embryos**. *Nature* (2014.0) **511** 606-610. DOI: 10.1038/nature13544
7. Lim YC. **A complex association between DNA methylation and gene expression in human placenta at first and third trimesters**. *PLoS ONE* (2017.0) **12** e0181155. DOI: 10.1371/journal.pone.0181155
8. Slieker RC. **DNA methylation landscapes of human fetal development**. *PLoS Genet.* (2015.0) **11** e1005583. DOI: 10.1371/journal.pgen.1005583
9. Akhabir L. **DNA methylation changes in cord blood and the developmental origins of health and disease - a systematic review and replication study**. *BMC Genomics* (2022.0) **23** 221. DOI: 10.1186/s12864-022-08451-6
10. Bohlin J. **Prediction of gestational age based on genome-wide differentially methylated regions**. *Genome Biol.* (2016.0) **17** 207. DOI: 10.1186/s13059-016-1063-4
11. Knight AK. **An epigenetic clock for gestational age at birth based on blood methylation data**. *Genome Biol.* (2016.0) **17** 206. DOI: 10.1186/s13059-016-1068-z
12. York TP. **Replicated umbilical cord blood DNA methylation loci associated with gestational age at birth**. *Epigenetics* (2020.0) **15** 1243-1258. DOI: 10.1080/15592294.2020.1767277
13. Pidsley R. **Critical evaluation of the illumina methylationEPIC BeadChip microarray for whole-genome DNA methylation profiling**. *Genome Biol.* (2016.0) **17** 208. DOI: 10.1186/s13059-016-1066-1
14. Sandoval J. **Validation of a DNA methylation microarray for 450,000 CpG sites in the human genome**. *Epigenetics* (2011.0) **6** 692-702. DOI: 10.4161/epi.6.6.16196
15. Reinius LE. **Differential DNA methylation in purified human blood cells: implications for cell lineage and studies on disease susceptibility**. *PLoS ONE* (2012.0) **7** e41361. DOI: 10.1371/journal.pone.0041361
16. Teschendorff AE, Zheng SC. **Cell-type deconvolution in epigenome-wide association studies: a review and recommendations**. *Epigenomics* (2017.0) **9** 757-768. DOI: 10.2217/epi-2016-0153
17. Titus AJ, Gallimore RM, Salas LA, Christensen BC. **Cell-type deconvolution from DNA methylation: a review of recent applications**. *Hum. Mol. Genet.* (2017.0) **26** R216-r224. DOI: 10.1093/hmg/ddx275
18. Houseman EA. **DNA methylation arrays as surrogate measures of cell mixture distribution**. *BMC Bioinforma.* (2012.0) **13** 86. DOI: 10.1186/1471-2105-13-86
19. Gervin K. **Systematic evaluation and validation of reference and library selection methods for deconvolution of cord blood DNA methylation data**. *Clin. Epigenetics* (2019.0) **11** 125. DOI: 10.1186/s13148-019-0717-y
20. Teschendorff AE, Zhu T, Breeze CE, Beck S. **EPISCORE: cell type deconvolution of bulk tissue DNA methylomes from single-cell RNA-Seq data**. *Genome Biol.* (2020.0) **21** 221. DOI: 10.1186/s13059-020-02126-9
21. Li Z, Wu Z, Jin P, Wu H. **Dissecting differential signals in high-throughput data from complex tissues**. *Bioinforma.* (2019.0) **35** 3898-3905. DOI: 10.1093/bioinformatics/btz196
22. Luo X, Yang C, Wei Y. **Detection of cell-type-specific risk-CpG sites in epigenome-wide association studies**. *Nat. Commun.* (2019.0) **10** 3113. DOI: 10.1038/s41467-019-10864-z
23. Rahmani E. **Cell-type-specific resolution epigenetics without the need for cell sorting or single-cell biology**. *Nat. Commun.* (2019.0) **10** 3417. DOI: 10.1038/s41467-019-11052-9
24. Zheng SC, Breeze CE, Beck S, Teschendorff AE. **Identification of differentially methylated cell types in epigenome-wide association studies**. *Nat. Methods* (2018.0) **15** 1059-1066. DOI: 10.1038/s41592-018-0213-x
25. Bauer M. **Cell-type-specific disturbance of DNA methylation pattern: a chance to get more benefit from and to minimize cohorts for epigenome-wide association studies**. *Int. J. Epidemiol.* (2018.0) **47** 917-927. DOI: 10.1093/ije/dyy029
26. Glasser L, Sutton N, Schmeling M, Machan JT. **A comprehensive study of umbilical cord blood cell developmental changes and reference ranges by gestation, gender and mode of delivery**. *J. Perinatol.* (2015.0) **35** 469-475. DOI: 10.1038/jp.2014.241
27. Pérez A, Gurbindo MD, Resino S, Aguarón A, Muñoz-Fernández MA. **NK cell increase in neonates from the preterm to the full-term period of gestation**. *Neonatology* (2007.0) **92** 158-163. DOI: 10.1159/000101567
28. 28.Magnus, P. et al. Cohort profile update: the Norwegian Mother and Child Cohort Study (MoBa). Int. J. Epidemiol. 45, 382–388 (2016).
29. Håberg SE. **DNA methylation in newborns conceived by assisted reproductive technology**. *Nat. Commun.* (2022.0) **13** 1896. DOI: 10.1038/s41467-022-29540-w
30. Haftorn KL. **An EPIC predictor of gestational age and its application to newborns conceived by assisted reproductive technologies**. *Clin. Epigenetics* (2021.0) **13** 82. DOI: 10.1186/s13148-021-01055-z
31. Heller R, Bogomolov M, Benjamini Y. **Deciding whether follow-up studies have replicated findings in a preliminary large-scale omics study**. *Proc. Natl Acad. Sci. USA* (2014.0) **111** 16262-16267. DOI: 10.1073/pnas.1314814111
32. McLean CY. **GREAT improves functional interpretation of cis-regulatory regions**. *Nat. Biotechnol.* (2010.0) **28** 495-501. DOI: 10.1038/nbt.1630
33. You C. **A cell-type deconvolution meta-analysis of whole blood EWAS reveals lineage-specific smoking-associated DNA methylation changes**. *Nat. Commun.* (2020.0) **11** 4779. DOI: 10.1038/s41467-020-18618-y
34. Lu T. **Detecting cord blood cell type-specific epigenetic associations with gestational diabetes mellitus and early childhood growth**. *Clin. Epigenetics* (2021.0) **13** 131. DOI: 10.1186/s13148-021-01114-5
35. de Goede OM, Lavoie PM, Robinson WP. **Cord blood hematopoietic cells from preterm infants display altered DNA methylation patterns**. *Clin. Epigenetics* (2017.0) **9** 39. DOI: 10.1186/s13148-017-0339-1
36. Palis J, Segel GB. **Developmental biology of erythropoiesis**. *Blood Rev.* (1998.0) **12** 106-114. DOI: 10.1016/S0268-960X(98)90022-4
37. Hermansen MC. **Nucleated red blood cells in the fetus and newborn**. *Arch. Dis. Child Fetal Neonatal Ed.* (2001.0) **84** F211-F215. DOI: 10.1136/fn.84.3.F211
38. Sen T, Chen J, Singbrant S. **Decreased PGC1β expression results in disrupted human erythroid differentiation, impaired hemoglobinization and cell cycle exit**. *Sci. Rep.* (2021.0) **11** 17129. DOI: 10.1038/s41598-021-96585-0
39. Mei Y, Liu Y, Ji P. **Understanding terminal erythropoiesis: An update on chromatin condensation, enucleation, and reticulocyte maturation**. *Blood Rev.* (2021.0) **46** 100740. DOI: 10.1016/j.blre.2020.100740
40. Menon V, Ghaffari S. **Erythroid enucleation: a gateway into a “bloody” world**. *Exp. Hematol.* (2021.0) **95** 13-22. DOI: 10.1016/j.exphem.2021.01.001
41. Chen K. **Resolving the distinct stages in erythroid differentiation based on dynamic changes in membrane protein expression during erythropoiesis**. *Proc. Natl Acad. Sci. USA* (2009.0) **106** 17413-17418. DOI: 10.1073/pnas.0909296106
42. Liang R. **A systems approach identifies essential FOXO3 functions at key steps of terminal erythropoiesis**. *PLoS Genet.* (2015.0) **11** e1005526. DOI: 10.1371/journal.pgen.1005526
43. Mathangasinghe Y, Fauvet B, Jane SM, Goloubinoff P, Nillegoda NB. **The Hsp70 chaperone system: distinct roles in erythrocyte formation and maintenance**. *Haematologica* (2021.0) **106** 1519-1534. DOI: 10.3324/haematol.2019.233056
44. Chasis JA, Mohandas N. **Erythroblastic islands: niches for erythropoiesis**. *Blood* (2008.0) **112** 470-478. DOI: 10.1182/blood-2008-03-077883
45. Sankaran VG, Orkin SH. **The switch from fetal to adult hemoglobin**. *Cold Spring Harb. Perspect. Med.* (2013.0) **3** a011643. DOI: 10.1101/cshperspect.a011643
46. Solano ME, Arck PC. **Steroids, pregnancy and fetal development**. *Front. Immunol.* (2019.0) **10** 3017. DOI: 10.3389/fimmu.2019.03017
47. Moisiadis VG, Matthews SG. **Glucocorticoids and fetal programming part 1: outcomes**. *Nat. Rev. Endocrinol.* (2014.0) **10** 391-402. DOI: 10.1038/nrendo.2014.73
48. Tang JI, Seckl JR, Nyirenda MJ. **Prenatal glucocorticoid overexposure causes permanent increases in renal erythropoietin expression and red blood cell mass in the rat offspring**. *Endocrinology* (2011.0) **152** 2716-2721. DOI: 10.1210/en.2010-1443
49. Flygare J, Rayon Estrada V, Shin C, Gupta S, Lodish HF. **HIF1alpha synergizes with glucocorticoids to promote BFU-E progenitor self-renewal**. *Blood* (2011.0) **117** 3435-3444. DOI: 10.1182/blood-2010-07-295550
50. von Lindern M. **The glucocorticoid receptor cooperates with the erythropoietin receptor and c-Kit to enhance and sustain proliferation of erythroid progenitors in vitro**. *Blood* (1999.0) **94** 550-559. DOI: 10.1182/blood.V94.2.550
51. Lee HY. **PPAR-α and glucocorticoid receptor synergize to promote erythroid progenitor self-renewal**. *Nature* (2015.0) **522** 474-477. DOI: 10.1038/nature14326
52. Nicolaides NC, Galata Z, Kino T, Chrousos GP, Charmandari E. **The human glucocorticoid receptor: molecular basis of biologic function**. *Steroids* (2010.0) **75** 1-12. DOI: 10.1016/j.steroids.2009.09.002
53. Stellacci E. **Interaction between the glucocorticoid and erythropoietin receptors in human erythroid cells**. *Exp. Hematol.* (2009.0) **37** 559-572. DOI: 10.1016/j.exphem.2009.02.005
54. Bauer A. **The glucocorticoid receptor is required for stress erythropoiesis**. *Genes Dev.* (1999.0) **13** 2996-3002. DOI: 10.1101/gad.13.22.2996
55. Dolznig H. **Erythroid progenitor renewal versus differentiation: genetic evidence for cell autonomous, essential functions of EpoR, Stat5 and the GR**. *Oncogene* (2006.0) **25** 2890-2900. DOI: 10.1038/sj.onc.1209308
56. Leonard MO, Godson C, Brady HR, Taylor CT. **Potentiation of glucocorticoid activity in hypoxia through induction of the glucocorticoid receptor**. *J. Immunol.* (2005.0) **174** 2250-2257. DOI: 10.4049/jimmunol.174.4.2250
57. Paulson RF, Hariharan S, Little JA. **Stress erythropoiesis: definitions and models for its study**. *Exp. Hematol.* (2020.0) **89** 43-54.e42. DOI: 10.1016/j.exphem.2020.07.011
58. Porayette P, Paulson RF. **BMP4/Smad5 dependent stress erythropoiesis is required for the expansion of erythroid progenitors during fetal development**. *Dev. Biol.* (2008.0) **317** 24-35. DOI: 10.1016/j.ydbio.2008.01.047
59. Simpkin AJ. **Longitudinal analysis of DNA methylation associated with birth weight and gestational age**. *Hum. Mol. Genet.* (2015.0) **24** 3752-3763. DOI: 10.1093/hmg/ddv119
60. Parets SE. **Fetal DNA methylation associates with early spontaneous preterm birth and gestational age**. *PLoS ONE* (2013.0) **8** e67489. DOI: 10.1371/journal.pone.0067489
61. Schroeder JW. **Neonatal DNA methylation patterns associate with gestational age**. *Epigenetics* (2011.0) **6** 1498-1504. DOI: 10.4161/epi.6.12.18296
62. Lee H. **DNA methylation shows genome-wide association of NFIX, RAPGEF2 and MSRB3 with gestational age at birth**. *Int. J. Epidemiol.* (2012.0) **41** 188-199. DOI: 10.1093/ije/dyr237
63. Xu J. **Corepressor-dependent silencing of fetal hemoglobin expression by BCL11A**. *Proc. Natl Acad. Sci. USA* (2013.0) **110** 6518-6523. DOI: 10.1073/pnas.1303976110
64. Varricchio L. **Identification of NuRSERY, a new functional HDAC complex composed by HDAC5, GATA1, EKLF and pERK present in human erythroid cells**. *Int J. Biochem. Cell Biol.* (2014.0) **50** 112-122. DOI: 10.1016/j.biocel.2014.02.019
65. Zermati Y. **Caspase activation is required for terminal erythroid differentiation**. *J. Exp. Med.* (2001.0) **193** 247-254. DOI: 10.1084/jem.193.2.247
66. Satyanarayana A. **RapGEF2 is essential for embryonic hematopoiesis but dispensable for adult hematopoiesis**. *Blood* (2010.0) **116** 2921-2931. DOI: 10.1182/blood-2010-01-262964
67. Schulz VP. **A unique epigenomic landscape defines human erythropoiesis**. *Cell Rep.* (2019.0) **28** 2996-3009.e2997. DOI: 10.1016/j.celrep.2019.08.020
68. Yu Y. **High resolution methylome analysis reveals widespread functional hypomethylation during adult human erythropoiesis**. *J. Biol. Chem.* (2013.0) **288** 8805-8814. DOI: 10.1074/jbc.M112.423756
69. Khouja JN. **Epigenetic gestational age acceleration: a prospective cohort study investigating associations with familial, sociodemographic and birth characteristics**. *Clin. Epigenetics* (2018.0) **10** 86. DOI: 10.1186/s13148-018-0520-1
70. Girchenko P. **Associations between maternal risk factors of adverse pregnancy and birth outcomes and the offspring epigenetic clock of gestational age at birth**. *Clin. Epigenetics* (2017.0) **9** 49. DOI: 10.1186/s13148-017-0349-z
71. Davari-Tanha F, Kaveh M, Nemati S, Javadian P, Salmanian B. **Nucleated red blood cells count in pregnancies with idiopathic intra-uterine growth restriction**. *J. Fam. Reprod. Health* (2014.0) **8** 77-81
72. Gasparović VE, Ahmetasević SG, Colić A. **Nucleated red blood cells count as first prognostic marker for adverse neonatal outcome in severe preeclamptic pregnancies**. *Coll. Antropol.* (2012.0) **36** 853-857. PMID: 23213944
73. Hebbar S, Misha M, Rai L. **Significance of maternal and cord blood nucleated red blood cell count in pregnancies complicated by preeclampsia**. *J. Pregnancy* (2014.0) **2014** 496416. DOI: 10.1155/2014/496416
74. Morton SU, Brettin K, Feldman HA, Leeman KT. **Association of nucleated red blood cell count with mortality among neonatal intensive care unit patients**. *Pediatr. Neonatol.* (2020.0) **61** 592-597. DOI: 10.1016/j.pedneo.2020.07.009
75. 75.Sokou, R. et al. Nucleated red blood cells: could they be indicator markers of illness severity for neonatal intensive care unit patients? Children10.3390/children7110197 (2020).
76. Byeon Y, Ki CS, Han KH. **Isolation of nucleated red blood cells in maternal blood for Non-invasive prenatal diagnosis**. *Biomed. Microdevices* (2015.0) **17** 118. DOI: 10.1007/s10544-015-0021-3
77. Singh R. **Fetal cells in maternal blood for prenatal diagnosis: a love story rekindled**. *Biomark. Med.* (2017.0) **11** 705-710. DOI: 10.2217/bmm-2017-0055
78. 78.Paltiel, L. et al. The biobank of the Norwegian Mother and Child Cohort Study – present status. Norsk Epidemiologi10.5324/nje.v24i1-2.1755 (2014).
79. Engel SM. **Neonatal genome-wide methylation patterns in relation to birth weight in the Norwegian Mother and Child Cohort**. *Am. J. Epidemiol.* (2014.0) **179** 834-842. DOI: 10.1093/aje/kwt433
80. Müller F. **RnBeads 2.0: comprehensive analysis of DNA methylation data**. *Genome Biol.* (2019.0) **20** 55. DOI: 10.1186/s13059-019-1664-9
81. McCartney DL. **Identification of polymorphic and off-target probe binding sites on the illumina infinium methylationEPIC beadChip**. *Genomics Data* (2016.0) **9** 22-24. DOI: 10.1016/j.gdata.2016.05.012
82. Teschendorff AE. **A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data**. *Bioinformatics* (2013.0) **29** 189-196. DOI: 10.1093/bioinformatics/bts680
83. Zhou W, Laird PW, Shen H. **Comprehensive characterization, annotation and innovative use of infinium DNA methylation BeadChip probes**. *Nucleic Acids Res.* (2017.0) **45** e22. PMID: 27924034
84. Pidsley R. **A data-driven approach to preprocessing Illumina 450K methylation array data**. *BMC Genomics* (2013.0) **14** 293. DOI: 10.1186/1471-2164-14-293
85. Salas LA. **An optimized library for reference-based deconvolution of whole-blood biospecimens assayed using the Illumina HumanMethylationEPIC BeadArray**. *Genome Biol.* (2018.0) **19** 64. DOI: 10.1186/s13059-018-1448-7
86. 86.Maechler, M. et al. robustbase: Basic Robust Statistics v. R package 0.93-6. http://robustbase.r-forge.r-project.org/ (2020).
87. 87.R: a language and environment for statistical computing (R Foundation for Statistical Computing, 2021).
|
---
title: The Pathogenic Role of Oxidative Stress, Cytokine Expression, and Impaired
Hematological Indices in Diabetic Cardiovascular Diseases
authors:
- Howaida Saad
- Hanan A. Soliman
- Basant Mahmoud
- Adel Abdel Moneim
- Mohamed Y. Zaky
journal: Inflammation
year: 2022
pmcid: PMC9971070
doi: 10.1007/s10753-022-01718-w
license: CC BY 4.0
---
# The Pathogenic Role of Oxidative Stress, Cytokine Expression, and Impaired Hematological Indices in Diabetic Cardiovascular Diseases
## Abstract
A simultaneous increase in the prevalence of diabetes mellitus (DM), a risk factor for cardiovascular diseases (CVDs), has contributed to the escalation of CVD related mortalities. To date, oxidative stress and inflammation are increasingly recognized as significant drivers of cardiovascular complications in patients with diabetes. Therefore, this study aims to explore the correlation between oxidative stress, inflammation, and hematological indices in diabetic patients with CVDs. Patients were allocated into five groups: healthy controls; nondiabetic patients with myocardial infarction; diabetic patients with myocardial infarction; nondiabetic patients with heart failure; and diabetic patients with heart failure. The results revealed that the malondialdehyde levels were increased; whereas superoxide dismutase enzyme activities were markedly reduced in all CVD groups compared with those of healthy controls. Although the mRNA expression levels of interleukin (IL)-6, IL-18, and IL-38 were significantly increased, those of the anti-inflammatory cytokine, IL-35, have been reduced in all CVD groups compared with healthy controls. Regarding hematological indices, hematocrit, red blood cell distribution width, mean platelet (PLT) volume, plateletcrit, PLT distribution width, leukocyte count, and PLT-to-lymphocyte and neutrophil-to-lymphocyte ratios were markedly increased in the diabetic and nondiabetic CVD groups compared with those of the healthy controls. Oxidative stress and cytokine biomarkers may play a significant role in the complications of diabetic cardiomyopathy. Moreover, hematological indices are particularly sensitive to systemic inflammatory changes and are novel markers for the early detection of diabetic cardiomyopathy.
## INTRODUCTION
Diabetes mellitus (DM) is a global health issue and is concomitant with cellular, metabolic, and blood cell abnormalities [1]. The American Heart Association regards diabetes as one of the key risk factors for cardiovascular diseases (CVDs) [2]. DM is often accompanied by cardiovascular disturbances, including coronary artery disease (CAD) and peripheral artery disease. Diabetes-induced pathological disorders may cause tissue damage in one-third to one-half of people with diabetes [3].
Oxidative stress is an imbalance in reactive oxygen species (ROS) and antioxidant defense mechanisms [4]. Cardiovascular insulin resistance, diabetic cardiomyopathy, and heart failure are accelerated by oxidative stress [5]. Recently, hypertriglyceridemia was demonstrated to increase CVD-related deaths in patients with diabetes [6]. According to human and animal studies, hyperglycemia may trigger an inflammatory response through oxidative pathways, which results in cardiac fibrosis and cardiomyocyte death, followed by cardiac dysfunction [7]. Indeed, inflammation, which is common in diabetes, can impact cardiomyocyte contractility and survival through various mechanisms, such as peroxynitrite formation and alterations in the extracellular matrix composition; dynamics are hypothesized to contribute to cytokine-induced cardiac contractile failure [8]. Interleukin (IL)-6 plays a vital role in the pathophysiology of CVD and its complications. Different immune responses and susceptibility to CAD are produced by genetic differences in IL-6 [9]. IL-38, a recently revealed member of the IL-1 cytokine family with anti-inflammatory activity, is generally recognized as a key regulator of inflammation. IL-38 is commonly expressed in the heart, and IL-38 polymorphisms are associated with coronary artery syndrome [10]. Moreover, IL-35, an anti-inflammatory cytokine, is involved in various disorders, including CVD and diabetes [11]. Hematological alterations in diabetes may occur due to ROS production as a consequence of long-term hyperglycemia [12]. Excessive ROS production causes oxidative stress, which leads to tissue damage and hematological alterations, including endothelial dysfunction, platelet (PLT) hyperactivity, and red blood cell (RBC) dysfunction [13]. Hematological alterations have been encountered in patients with DM, including alterations in the structure, function, and metabolism of PLT, RBC, white blood cell (WBC), and the coagulation system [12].
The role of inflammation in diabetic CVD has been recently reported; however, its mechanism requires more exploration. In particular, the interplay between DM-induced inflammation and hematological disorders in the risk of CVD requires further investigation. Therefore, this study aims to explore the relationship between oxidative stress, inflammation, and alteration in hematological indices and the risk of developing diabetic cardiomyopathy.
## Study Population
A total of 120 patients with both myocardial infarction (diabetic type 2 and nondiabetic) and heart failure (diabetic type 2 and nondiabetic) who visited the Cardiac and Chest Intensive Care Unit of Beni-Suef University Hospital (Beni-Suef, Egypt) were enrolled in this cross-sectional study. The data was collected using a paper questionnaire and clinical laboratory reports. Eligible patients were allocated into four groups based on the clinical and biochemical investigations. Moreover, thirty normal healthy participants were selected as healthy controls. Moreover, all patients provided a written consent and blood samples during the period from March 2019 to November 2019 following protocol approval by the Beni-Suef University Hospital Ethics Committee (BNS/$\frac{2019}{2}$).
## Inclusion and Exclusion Criteria and Experimental Design
The adult patients (males, 72; females, 78; aged 40–70 years), including healthy controls, diabetic and nondiabetic patients with heart failure, and diabetic and nondiabetic patients with myocardial infarction, were enrolled in this study (Fig. 1). The healthy subjects consisted of participants who had no significant health-related issues. All participants (healthy controls and patients) were free of asthma symptoms, alcohol abuse, infectious diseases, allergies, kidney failure, eczema, thyroid diseases, autoimmune disorders, and liver dysfunction. Fig. 1Schematic diagram presenting inclusion and exclusion criteria and experimental design.
## Biochemical Investigations
After overnight fasting, blood samples were collected in sodium fluoride, ethylenediaminetetraacetic acid (EDTA), and plain tubes (5 mL each). The EDTA blood samples were used to estimate complete blood count and glycated hemoglobin (HbA1c). Samples were stored at − 40 °C until use. Sodium fluoride tubes were used to collect blood samples to determine fasting blood sugar (FBS) level. Plasma FBS and serum cholesterol, high-density lipoprotein (HDL), and triglycerides levels were assessed using a reagent kit (Spinreact Co., Spain). Subsequently, low-density lipoprotein (LDL) and very low-density lipoprotein (vLDL) levels were calculated based on the formula of Friedewald et al. [ 14]. Cardiovascular risks (1 and 2) were calculated following the formula of Ross [15]. The HbA1c level was determined using a reagent kit (Stanbio Laboratory, TX, USA). Furthermore, serum insulin was assessed using radioimmunoassay kits (Diagnostic Products Corporation, LA, USA). Homeostatic model assessment for insulin resistance (HOMA-IR) values was calculated using the following equation: HOMA-IR = fasting insulin (U/L) × fasting glucose (mg/dL)/405 [16]. The C-reactive protein (CRP) was estimated using a reagent kit (Spinreact Co., Spain). Serum CK-MB and troponin I levels were determined using test kits (HUMAN, Germany). Moreover, malondialdehyde (MDA) levels and superoxide dismutase (SOD) activities were determined using a reagent kit (BioVision, Milpitas, CA, USA). Hematology parameters, including hemoglobin, RBCs, hematocrit, mean corpuscular hemoglobin, mean corpuscular hemoglobin content, mean corpuscular volume, RBC distribution width (RDW), PLT count, plateletcrit (PCT), PLT distribution width (PDW), mean PLT volume (MPV), WBC count, and differential leukocyte count, were determined using a MICROS ABX autoanalyzer. All procedures were performed following the kit manufacturers’ instructions.
## Real-Time Polymerase Chain Reaction
RNA was isolated from the blood samples using a GeneJET™ RNA purification kit (Thermo Fisher Scientific Inc., Rochester, NY, USA). RNA was purified and spectrophotometrically quantified. Consequently, the target DNA cDNA was amplified using GoTaq Green Master Mix (Promega, WI, USA) using the following sets of primers: 5ˋ-GGTACATCCTCGACGGCATCT-3ˋ (forward primer) and 5ˋ-GT GCCTCTTTGCTGCTTTCAC-3ˋ (reverse primer) for IL-6, 5ˋ -GCTTCCTCTCGC AAC AAA C-3ˋ(forward primer) and 5ˋ -CACTTCACAGAGATAGTTACAGCC-3ˋ(reverse primer) for IL-18, 5ˋ-TGTTCTCCATGGCTCCCTA-3ˋ(forward primer) and 5ˋ-TTATGAAAGGCACGAAGCTG-3ˋ(reverse primer) for IL-35, 5ˋ- AAGAAGGACCTCCGGCTCT -3ˋ(forward primer) and 5ˋTGACTCAGAATCTGGC5GTATTTC-3ˋ (reverse primer) for IL-38, and 5ˋ- TCACCCTGAAGTACCCCATGGAG-3ˋ(forward primer) and 5ˋ TTGGCCTTGGGGTTCAGGGGG -3ˋ (reverse primer) for β-actin. Green Master Mix (Promega, WI, USA) and T100TM thermal cycler (Bio-Rad Laboratories, CA, USA) were used for PCR under the following conditions: initial denaturation at 95 °C for 5 min, 35 cycles set at 95 °C (1 min) for denaturation, 60 °C (1 min) for annealing, and 72 °C (1 min) for extension, and ultimately at 72 °C (5 min) to finish the extension reaction. Values were normalized to the quantity of β-actin. All molecular assays were conducted at the Molecular Biology Laboratory of CliniLab (Cairo, Egypt).
## Statistical Analysis
Results were presented as mean values and SEM. Shapiro–Wilk normality test showed that the data were not normally distributed, and thus a non-parametric Kruskal–Wallis test followed by the post hoc test using a pairwise multiple-comparative analysis was used to determine the statistical differences between groups using a computer software package (SPSS version 20, IBM Corp., 2011). A simple linear correlation study was conducted using Spearman’s correlation analysis to estimate the degree of the relationship between the variables. A P value < 0.05 was considered statistically significant.
## RESULTS
The results revealed that the body mass index values were significantly higher ($p \leq 0.001$) in all CVD patient groups compared with the healthy control group. Meanwhile, all patient groups showed a significant ($p \leq 0.001$) increase in systolic and diastolic blood pressures compared with the healthy control group. Moreover, the diabetic myocardial infarction and diabetic heart failure groups showed a significant ($p \leq 0.001$) increase in the FBS level compared with the nondiabetic myocardial infarction, nondiabetic heart failure, and healthy control groups. However, the diabetic CVD groups exhibited a significant ($p \leq 0.001$) decline in the insulin level compared with the nondiabetic CVD and healthy control groups. The diabetic myocardial infarction and diabetic heart failure groups revealed a significant ($p \leq 0.001$) increase in HbA1c% and HOMA-IR values compared with the nondiabetic CVD and healthy control groups. All CVD patient groups showed a highly significant ($p \leq 0.001$) increase in serum CK-MB and troponin I levels compared with the healthy control group, with a marked increase in both myocardial infarction groups (Table 1).Table 1Demographic and Biochemical Characteristics of Healthy Controls and Diabetic and Nondiabetic CVD Dysfunction GroupsAge48.7±4.061.2±10.8***59.2±10.5***59.0±11.2***65.3±10.0*** GenderM 22 ($73.3\%$)M 20 ($66.7\%$) M 22 ($73.3\%$)M 20 ($66.7\%$)M 22 ($73.3\%$)F 8 ($26.7\%$) F 10 ($33.3\%$)F 8 ($26.7\%$)F10 ($33.3\%$)F 8 ($26.7\%$)BMI23.8±0.627.9±2.2*27.6±2.8*28.9±2.2*28.9±3.0* Smoking 2 ($6.7\%$)18 ($60\%$)15 ($50.0\%$)14 ($46.7\%$)14 ($46.7\%$)Duration of CVD --3.3±1.44.5±1.7 N.S.8.2±2.58.9 ±2.5 N.S.Duration of DM -- --11.3±3.5 --12.1±4.5 Systolic BP 105.3±6.3154.7±12.8***153.3±12.7***158.0±14.9***156.0±14.8***Diastolic BP 75.3±5.198.7±8.2***96.7±8.8***98.7±9.0***98.7±9.7***FBS (mg/dl)72.5±0.872.9±0.7 N.S.197.3±7.9***###72.9±0.7 N.S.179.1±5.7***### Insulin (mIU/ml)10.1±0.110.0±0.1 N.S.7.4±0.2***##10.1±0.1 N.S. 8.6±0.2***##HbAIC%5.0±0.14.6±0.1 N.S.9.3±0.4***###4.7±0.1 N.S.9.9±0.3*** ###HOMA-IR1.8±0.01.9±0.1N.S.3.6±0.1***###2.0±0.0 N.S.3.7±0.0***###Troponin I (ng/ml)0.2±0.02.9±0.1***3.1±0.1***1.7±0.0***1.8±0.1***CK-MB (U/L)11.3±0.366.8±2.6***68.8±4.2***53.2±2.7***50.0±1.8***Mean values and SEM are represented. A non-parametric Kruskal–Wallis test followed by the post hoc test using a pairwise multiple-comparative analysis was used to determine the statistical difference between groups, number of individuals = 150. ** $p \leq 0.01$ and ***$p \leq 0.001$ compared with the healthy control group and with ##$p \leq 0.01$ and ###$p \leq 0.001$ compared with the respective nondiabetic groupNS not significant, BMI body mass index, BP blood pressure, FBS fasting blood sugar, HbAIc glycosylated hemoglobin, and CK-MB creatine kinase myocardial band activity Regarding lipid profile results, all CVD groups exhibited a significant ($p \leq 0.001$) increase in serum triglycerides, cholesterol, LDL-c, and vLDL-c levels, as well as cardiovascular risks 1 and 2 values, than healthy controls. However, HDL and anti-atherogenic index values showed a notable ($p \leq 0.001$) reduction in the diabetic CVD groups compared with those in the nondiabetic CVD and healthy control groups (Fig. 2).Fig. 2The changes in the values of A cholesterol, B HDL, C LDL, D triglycerides, E vLDL F risk factor 1, G risk factor 2, and H anti-atherogenic index among healthy controls, myocardial infarction (non-diabetic and diabetic), and heart failure (non-diabetic and diabetic) groups. Mean values and SEM are represented. A non-parametric Kruskal–Wallis test followed by the post hoc test using a pairwise multiple-comparative analysis was used to determine the statistical difference between groups, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ as compared to healthy control and with #$p \leq 0.05$, ##$p \leq 0.01$ as compared with non-diabetic groups. N.S. was not significant. LDL, low-density lipoprotein cholesterol; HDL, high-density lipoprotein cholesterol; vLDL, very low-density lipoprotein cholesterol.
The expression of pro-inflammatory markers, IL-6, IL-18, and IL-38 showed a significant ($p \leq 0.001$) upregulation in CVD groups compared with those in the healthy control group and exhibited a significant increase in diabetic patients with CVD compared with that in nondiabetic CVD patients. In contrast, the diabetic and nondiabetic CVD groups showed a highly significant ($p \leq 0.001$) downregulation in the IL-35 mRNA expression compared with the healthy control group, with lower levels in diabetic patients than that in nondiabetic CVD patients. All CVD patient groups exhibited a highly significant ($p \leq 0.001$) increase in serum CRP levels compared with the healthy control group. Regarding oxidative stress markers, the CVD patient groups showed a highly significant ($p \leq 0.001$) increase in MDA levels and a remarkable ($p \leq 0.001$) decline in SOD levels compared with the healthy control group. However, the MDA levels exhibited a significant ($p \leq 0.01$) increase in the diabetic CVD patient groups compared with those in the healthy control group. Furthermore, SOD levels were significantly decreased ($p \leq 0.05$) in CVD diabetic groups than in CVD non-diabetic groups (Fig. 3).Fig. 3The changes in the values of A IL-6, B IL-18, C IL-35, D IL-38, E CRP, F MDA, and G SOD among healthy controls, myocardial infarction (non-diabetic and diabetic), and heart failure (non-diabetic and diabetic) groups. Mean values and SEM are represented. A non-parametric Kruskal–Wallis test followed by the post hoc test using a pairwise multiple-comparative analysis was used to determine the statistical difference between groups, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ when compared to healthy control and with #$p \leq 0.05$, ##$p \leq 0.01$, ###$p \leq 0.001$ when compared with non-diabetic groups. N.S. was not significant. IL, interleukin; MDA, malondialdehyde; SOD, superoxide dismutase; C-RP, C-reactive protein.
In the present study, both heart failure patient groups exhibited a significant decrease in the hemoglobin content, RBC count, and hematocrit values compared with the healthy control group (Table 2). Moreover, the heart failure patient groups revealed a significant ($p \leq 0.001$) increase in RDW% compared with the healthy control group. Additionally, all CVD patient groups displayed a significant ($p \leq 0.001$) increase in PDW% compared with the healthy control group (Fig. 3). Regarding PLT indices, the diabetic heart failure group displayed a significant ($p \leq 0.05$) increase in the PLT count compared with the healthy control group (Table 2), whereas all CVD patient groups revealed a marked increase in MPV values, except for the diabetic myocardial infarction group, compared with the healthy control group. Meanwhile, the diabetic CVD patient groups showed a marked ($p \leq 0.05$) increase in PCT% compared with the healthy control and nondiabetic groups (Fig. 4).Table 2Hematological Profile of Healthy Controls and Diabetic and Nondiabetic CVD GroupsGroupHealthy ControlsNon Diabetic myocardial infractionDiabetic myocardial infractionNon Diabetic Heart failureDiabetic Heart failureParameter($$n = 30$$)($$n = 30$$)($$n = 30$$)($$n = 30$$)($$n = 30$$) Hb (g/dl) 12.7±0.9 12.8±0.3 N.S. 13.5±0.3N.S. 11.0±0.3*** 11.6±0.3* WBCS (10³/µl) 5.9 ±0.3 8.7±0.6* 12.4±1.2***# 9.1±1.0* 7.2±0.6N.S. RBCS (M/µl) 4.7±0.1 4.6±0.1N.S. 5.0±0.1 N.S. 4.0±0.1*** 4.2±0.1* Platelets (10³/µl) 231.5±9.0 249.5±13.3 N.S. 247.7±9.9 N.S. 266.8±15.1N.S. 273.7±18.3* HCT% 38.5±0.7 39.5±0.9 N.S. 42.4±1.0 34.0±0.9** 35.5±1.1* MCV(fl) 82.2±1.0 86.9±1.4* 84.9±0.7 N.S. 85.2±1.4 N.S. 85.8±0. 9 * MCH (pg)27.3±0.3 28.2±0.6 N.S. 27.1±0.4 N.S. 27.4±0.5 N.S. 28.0±1.7 N.S. MCHC(g/dl) 33.2±0.3 32.5±0.4 N.S. 31.9±0.3N.S. 32.3±0.3 N.S. 32.7±0.4N.S. N% 57.9±2.4 70.4±1.9*** 73.9±2.0*** 65.2±2.4* 65.6±2.8* L% 37.8±2.5 20.2±1.7*** 20.2±1.7*** 28.7±2.5* 29.1±2.8* M% 3.1±0.2 3.5±0.4 N.S. 4.0±0.4 N.S. 4.1±0.6N.S. 3.7±0.5 N.S. Mean values and SEM are represented. A non-parametric Kruskal–Wallis test followed by the post hoc test using a pairwise multiple-comparative analysis was used to determine the statistical difference between groups, number of individuals = 150. * $p \leq 0.05$, **$p \leq 0.01$, and ***$p \leq 0.001$ compared with the healthy control group and with #$p \leq 0.05$ compared with the respective nondiabetic groupsNS not significant, Hb hemoglobin, WBCs white blood cells, RBCs red blood cells, HCT hematocrit, MCV mean corpuscular volume, MCH mean corpuscular hemoglobin, MCHC mean corpuscular hemoglobin concentration, N neutrophil, L lymphocytes, and M monocytesFig. 4The changes in the values of A RDW%, B MPV, C PCT%, D PDW%, E N/L ratio, F L/M ratio, and G P/L ratio among healthy controls, myocardial infarction (non-diabetic and diabetic), and heart failure (non-diabetic and diabetic) groups. Mean values and SEM are represented. A non-parametric Kruskal–Wallis test followed by the post hoc test using a pairwise multiple-comparative analysis was used to determine the statistical difference between groups, *$p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ when compared to healthy control and with #$p \leq 0.05$ when compared with non-diabetic groups. N.S. was not significant. PDW, platelets distribution width; RDW, red distribution width; PCT, plateletcrit; MPV, mean platelets volume; N/L, neutrophil/lymphocyte, N/L, lymphocyte/monocyte; P/L, platelets/lymphocyte.
The WBC count increased significantly ($p \leq 0.001$) in all CVD patient groups, except for the diabetic heart failure group, compared with that in the healthy control group (Table 2). However, the neutrophil count exhibited a significant increase in all CVD patient groups compared with that in the healthy control group. Conversely, the lymphocyte count showed a remarkable decline in all CVD patient groups compared with that in the healthy control group (Table 2). Concerning PLT-to-lymphocyte (P/L) and neutrophil-to-lymphocyte (N/L) ratios, the results showed a noticeable increase in all CVD patient groups compared with those in the healthy control group. Additionally, the P/L and N/L ratios exhibited remarkable increases in the diabetic CVD patient groups compared with those in the nondiabetic CVD patient groups. However, the lymphocyte-to-monocyte (L/M) ratio revealed a significant decrease in both myocardial infarction patient groups compared with that in the healthy control group, with a remarkably high decrease in the diabetic myocardial infarction patient group (Fig. 4).
In the diabetic CVD groups, HbA1c% displayed a significant positive correlation with IL-6, IL-18, IL-38, MDA, and PDW%, whereas a negative correlation was detected with IL-35 and SOD. In addition, CK-MB exhibited a significant positive correlation with IL-6, IL-18, IL-38, MDA, PDW%, and P/L and N/L ratios, whereas a significant negative correlation was observed with IL-35 and SOD. Moreover, the anti-atherogenic index demonstrated a significant positive correlation with IL-36 and SOD, whereas a significant negative correlation was observed with IL-6, IL-18, IL-38, MDA, PDW%, PCT%, and P/L and N/L% ratios (Table 3).Table 3Correlation of HbAIc%, CK-MB, and Anti-Atherogenic with IL-38, IL-6, IL-38, IL-36, MDA, SOD, PDW, PCT, P/L Ratio, and N/L Ratio in Diabetic CVD GroupsP valuer valueCorrelated parameters < 0.010.226HbA1c %/IL-6 < 0.010.225HbA1c %/IL-18 < 0.01 − 0.346HbA1c %/IL-35 < 0.010.258HbA1c %/IL-38 < 0.010.204HbA1c %/MDA < 0.001 − 0.428HbA1c %/SOD < 0.010.278HbA1c %/PDW% > 0.050.084HbA1c %/PCT > 0.050.067HbA1c %/P/L > 0.050.021HbA1c %/N/L < 0.010.492CK-MB/IL-6 < 0.0010.593CK-MB/IL-18 < 0.01 − 0.346CK-MB/IL-35 < 0.0010.539CK-MB/IL-38 < 0.010.580CK-MB/MDA < 0.001 − 0.368CK-MB/SOD < 0.010.496CK-MB/PDW% > 0.050.190CK-MB/PCT% < 0.0010.484CK-MB/P/L < 0.010.447CK-MB/N/L < 0.01 − 0.572Anti-atherogenic/IL-6 < 0.001 − 0.571Anti-atherogenic/IL-18 < 0.010.481Anti-atherogenic/IL-35 < 0.01 − 0.565Anti-atherogenic/IL-38 < 0.001 − 0.620Anti-atherogenic/MDA < 0.010.560Anti-atherogenic/SOD < 0.01 − 0.547Anti-atherogenic/PDW% < 0.01 − 0.220Anti-atherogenic/PCT% < 0.01 − 0.333Anti-atherogenic/P/L < 0.01 − 0.256Anti-atherogenic/N/LThe correlations between HbAIc%, CK-MB, and anti-atherogenic with IL-38, IL-6, IL-38, IL-36, MDA, SOD, PDW, PCT, P/L ratio, and N/L ratio were established by Spearman’s correlation analysis. r, Spearman’s correlation coefficient
## DISCUSSION
After controlling for inflammatory markers and classical risk factors, patients with diabetes had almost twice the risk of cardiovascular mortality compared with those without diabetes [1]. Furthermore, the increase in the number of patients with T2DM highlights the necessity of early CVD detection in patients with diabetes [6]. Thus, this study was anticipated to explore the interplay of DM-induced oxidative stress and inflammation with hematological abnormalities in CVD risk.
In our study, the myocardial infarction and heart failure groups exhibited a significant increase in serum cholesterol, triglyceride, LDL, vLDL-c levels, and risks factors 1 and 2, compared with the healthy control group. However, HDL levels and the anti-atherogenic index showed a notable reduction in the diabetic CVD groups compared with those in the nondiabetic CVD and healthy control groups. High glucose levels and dyslipidemia are complicated in atherosclerosis pathogenesis that eventually leads to CVD deaths. Chronic inflammation is one of the leading causes of atherosclerosis development [17]. The increase in serum CK-MB, troponin I, and lipid profile markers, including risks factors 1 and 2, confirmed the cardiac dysfunction of patients with diabetes, which leads to exploring diabetes cardiomyopathy complications.
In this study, MDA displayed a significant increase in the CVD patient groups with a marked elevation in the diabetic groups compared with the healthy control and nondiabetic groups. At the same time, SOD levels exhibited a remarkable decline in all CVD groups with a significant decrease in diabetic patients with CVD compared with those in the healthy control nondiabetic groups. Our results revealed a significant correlation among both MDA and SOD and HbA1c, CK-MB, and the anti-atherogenic index, indicating that dyslipidemia- and hyperglycemia-associated oxidative stress may induce the development of diabetic cardiomyopathy complications. These results are consistent with the study of Fathelbab et al. [ 18] in which it was observed that MDA levels were significantly increased in diabetic patients with CVD, and SOD levels in the diabetic CVD group were significantly reduced compared with those in the control group, indicating the role of oxidative stress-mediated tissue injury in diabetic patients with CVD. Moreover, a marked reduction in all antioxidant enzymes was observed, including SOD, in patients with metabolic syndrome (patients with dyslipidemia, hypertension, and T2DM) compared with the healthy controls [19]. Oxidative stress in heart failure occurs as a consequence of the excessive ROS production that can enhance lipid peroxidation and oxidize proteins to inactive states and cause DNA damage [20]. In addition, high ROS production may trigger maladaptive signaling pathways, which results in cell death and promotes abnormal cardiac remodeling, ultimately leading to diabetic cardiomyopathy-related functional abnormalities [21]. SOD is a member of the metalloproteinase family, and its overproduction is related to hypertension, diabetes, and CVD [22]. Additionally, SOD decline was associated with endothelial dysfunctions and hypertension [19]. When acute myocardial infarction occurs, it almost reduces SOD’s ability to scavenge free radicals. Increased MDA levels, followed by ROS accumulation, resulted in acute myocardial infarction [23]. Moreover, low antioxidant enzyme levels led to excessive insulin resistance that enhanced stress-related pathways, leading to diabetic cardiovascular events [24].
CVD is an inflammatory condition, whereas CRP is an acute-phase protein. Notably, clinical studies have demonstrated that CRP is a predictor of CVD [25] and associated complications [26]. The current study showed that all CVD groups (myocardial infarction and heart failure) exhibited higher CRP levels compared with the healthy controls. Moreover, CVD groups with diabetes had higher CRP levels than the corresponding normoglycemic groups. Furthermore, our results suggest a strong correlation between inflammatory cytokines and oxidative biomarkers in diabetic CVD groups. Our results are consistent with the results of previous studies, which declared that CRP is a novel marker for predicting CVDs in patients with diabetes mellitus [27]. Thus, our findings indicate that diabetes may worsen inflammatory responses and oxidative stress and could contribute to adverse outcomes in CVD.
Notably, the current data found a strong correlation between pro-inflammatory cytokines (IL-6, IL-18, and IL-38) and HbA1c, CK-MB, and anti-atherogenic index biomarkers. Additionally, the results showed that IL-6, IL-18, and IL-38 mRNA expressions exhibited a significant upregulation in the diabetic and nondiabetic CVD groups compared with those in the healthy control group, with a highly significant increase in diabetic patients with CVD. Cytokines, TNF-α, IL-1β, and IL-6, were found to be elevated in patients with heart failure, and their higher levels appear to be directly related to LVEF dysfunction [28]. Myocardial damage, heart failure, and mortality are linked to high levels of circulating IL-6 during and immediately after an acute myocardial infarction [29]. IL-6 levels in patients with left ventricular diastolic dysfunction (LVDD) were found to be substantially linked with the levels of fibrotic parameters, suggesting that IL-6 may play a role in increasing myocardial fibrosis and LV remodeling, finally leading to LVDD [30].
IL-18 is a pro-inflammatory cytokine that is predominantly produced by macrophages and binds to its receptor on the membranes of endothelial cells, lymphocytes, and smooth muscle cells to cause the production of interferon gamma, endothelial dysfunction, and plaque instability [31]. Our results are consistent with the results of Xiao et al. [ 32] in which it was revealed that increased IL-18 levels have been related to an increased risk of CVD. Several studies found that IL-18 was considerably higher in the plasma of patients who had coronary events [33], which agrees with our results. Notably, continued IL-18 suppression by IL-18-binding protein leads to cardiac fibrosis reduction and NF-κB phosphorylation, diastolic function improvement, electrical remodeling normalization, and IL-18-mediated ventricular tachycardia reduction in mice [34]. The IL-38 mRNA expression was upregulated in the mouse heart following myocardial infarction and exhibited a decrease in dendritic cell–mediated immune response [35]. Consequently, targeting IL-38 may hold new therapeutic approaches in the treatment of patients with myocardial infarction.
Moreover, our data demonstrated a correlation among IL-35 and HbA1c, CK-MB, and anti-atherogenic index biomarkers; it may be a novel therapeutic target in patients at risk for diabetic CVD. IL-35 has been linked to several cardiovascular disorders, including atherosclerosis and viral myocarditis [36]. IL-35 reduces doxorubicin-induced heart damage by increasing STAT3 signaling, lowering oxidative stress, and blocking mitochondrial-related apoptotic pathways [37]. The abovementioned studies are consistent with our study, showing a significant decrease in IL-35 levels in the CVD patient groups compared with those in the healthy control group. IL-35 can reduce myocardial ischemia/reperfusion (I/R) dysfunction by reducing mtROS. Furthermore, it was believed that IL-35 protects cardiomyocytes by inhibiting apoptosis, thereby minimizing cardiac I/R injury [38].
Furthermore, anemia is related to microvascular complications, diabetic nephropathy, neuropathy, and CVD [39]. HB content, RBCs count, and HCT% exhibited a significant decrease in CVD groups compared with healthy controls. Moreover, RBCs of a patient with diabetes combined more willingly, thereby augmenting whole blood viscosity and harming the microcirculation, finally leading to microangiopathy [40]. Moreover, all CVD patient groups revealed a significant increase in PDW% compared with the healthy control group. A high RDW is a disease severity metric associated with a variety of adverse outcomes, including CV and non-CV death, in patients with T2DM who have recently had an acute coronary syndrome [41]. Despite the completely unclear mechanisms between the RDW and adverse health consequences, it was recommended that it might be associated with the increase in ROS and pro-inflammatory cytokine levels [42].
PLTs are essential for maintaining normal homeostasis, and MPV is the indicator of their function. PLTs play a substantial role in atherosclerosis development and acute thrombotic cardiovascular event progression [43]. Our findings of higher MPV among patients with T2DM are consistent with those of several studies [44] that reported that the diabetic CVD group showed a significant increase in MPV compared with the healthy group. Elevated MPV has been seen in diabetic patients with retinopathy, nephropathy, and coronary heart disease; it represents alterations in PLT stimulation or PLT production rate [45]. PDW is a specific PLT reactivity biomarker that can help predict CVD [46], which is consistent with our study that showed that the CVD group showed a significant increase in PDW compared with the healthy control group. In our study, the CVD groups showed a significantly increased PCT% compared with the healthy control group. PLTs release various mediators, including thromboxanes, which can cause inflammation; increased PCT levels at the time of admission have been linked to poor long-term outcomes in patients with myocardial infarction [47].
WBCs are a biological systemic inflammation indicator, and an increased WBC count is linked to an increased risk of CAD, mortality rate, and stroke [48]. In this study, the CVD groups demonstrated a significant elevation in the neutrophil count compared with the healthy control group. The current results revealed that CK-MB and the anti-atherogenic index exhibited a positive correlation with the P/L and N/L ratios. Neutrophils are leukocytes that serve as the first line of defense against pathogens and damages caused by inflammation. The increase in neutrophil blood count has been linked to the severity of coronary damage and the prevalence of heart failure [49]. Lymphocytes reflect a calm and controlled immune response that causes less cardiac damage; lymphocyte levels decrease as apoptosis increases. In individuals with chronic heart failure, a low blood lymphocyte count has been linked to worse cardiovascular outcomes [50].
Recently, PDW, MPV, and N/L and P/L ratios were identified as microvascular diabetic issue indicators and announced as novel inflammatory biomarkers in cardiac diseases. The results of the previous studies [51, 52] are consistent with the results of our study, revealing that the N/L ratio had a significant increase in the CVD groups compared with that in the healthy control group. The N/L ratio, a simple, inexpensive, and innovative inflammatory biomarker, was shown to be higher in patients with diabetes and was related to poor glycemic control and may be prognostic predictors in cardiovascular events [53]. In patients with heart failure, neutrophilia, lymphopenia, and a higher N/L ratio have been linked to heart failure severity [54].
Additionally, our results showed that the N/L, L/M, and P/L ratios beside MPV, PDW, and PCT% were considered novel inflammatory biomarkers, confirming that the impaired hematological indices are one of the diagnostic markers in the development of diabetic cardiomyopathy. Also, the P/L and N/L ratios exhibited remarkable elevations in the diabetic CVD patient groups compared with those in the nondiabetic CVD patient groups. The P/L ratio is significantly and independently linked with the occurrence of in-hospital severe adverse cardiovascular events as a novel inflammatory marker. In patients with acute myocardial infarction, there was additional evidence of a link between the P/L ratio and long-term prognoses [55]. Additionally, the combination of the N/L and P/L ratios had both short- and long-term predictive significance [56]. These previous studies [55, 56] agree with our study, which reported a significant increase in the P/L ratio in the CVD groups compared with that in the healthy control group. Moreover, our findings revealed that the CVD groups exhibited a significant decrease in the L/M ratio compared with the healthy control group. These findings were consistent with those of [57] which reported that the L/M ratio was also linked to the prognosis of individuals with myocardial infarction, heart failure, and stable CAD. Notably, a lower L/M ratio is linked to a higher frequency of unstable angina pectoris or myocardial infarction in patients with CVD. Furthermore, the L/M ratio is an independent indicator of heart failure re-hospitalization in patients with CVD [58]. The study had some limitations, however, including the sample size and separation of the results based on age, sex, and disease duration. Moreover, this study lacks data on physical activity, drug use, and the serum protein levels of the tested cytokines.
In conclusion, our findings suggest a possible association of an oxidative stress represented by excess of MDA production and reduced SOD activity, and inflammatory state represented by increased mRNA expressions of the pro-inflammatory cytokines, IL-6, IL-18, and IL-38 and reduced expression of anti-inflammatory IL-35, and the progression of the diabetic cardiovascular diseases. Moreover, impaired hematological indices, including RDW, PLT, PDW%, PDW, PCT%, and N/L, L/M, and P/L ratios, were associated with the development of diabetic cardiomyopathy. Hematological indices are particularly sensitive to systemic inflammatory changes and may be novel markers for predict the progression of diabetic CVD.
## References
1. Daryabor G, Atashzar MR, Kabelitz D, Meri S, Kalantar K. **The effects of type 2 diabetes mellitus on organ metabolism and the immune system**. *Frontiers in immunology* (2020.0) **11** 1582. DOI: 10.3389/fimmu.2020.01582
2. Arnett DK, Blumenthal RS, Albert MA, Buroker AB, Goldberger ZD, Hahn EJ. **ACC/AHA guideline on the primary prevention of cardiovascular disease: A report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Journal of the American College of Cardiology* (2019.0) **74** 1376-1414. DOI: 10.1016/j.jacc.2019.03.009
3. Bharath LP, Rockhold JD, Conway R. **Selective autophagy in hyperglycemia-induced microvascular and macrovascular diseases**. *Cells* (2021.0) **10** 2114. DOI: 10.3390/cells10082114
4. Zhou Y, Murugan DD, Khan H, Huang Y, Cheang WS. **Roles and therapeutic implications of endoplasmic reticulum stress and oxidative stress in cardiovascular diseases**. *Antioxidants* (2021.0) **10** 1167. DOI: 10.3390/antiox10081167
5. Salvatore T, Pafundi PC, Galiero R, Albanese G, Di Martino A, Caturano A. **The diabetic cardiomyopathy: The contributing pathophysiological mechanisms**. *Frontiers in Medicine* (2021.0) **8** 695792. DOI: 10.3389/fmed.2021.695792
6. Tahir A, Martinez PJ, Ahmad F, Fisher-Hoch SP, McCormick J, Gay JL. **An evaluation of lipid profile and pro-inflammatory cytokines as determinants of cardiovascular disease in those with diabetes: A study on a Mexican American cohort**. *Scientific Reports* (2021.0) **11** 1-12. PMID: 33414495
7. Kaur N, Guan Y, Raja R, Ruiz-Velasco A, Liu W. **Mechanisms and therapeutic prospects of diabetic cardiomyopathy through the inflammatory response**. *Frontiers in Physiology* (2021.0) **12** 694864. DOI: 10.3389/fphys.2021.694864
8. Byrne NJ, Rajasekaran NS, Abel ED, Bugger H. **Therapeutic potential of targeting oxidative stress in diabetic cardiomyopathy**. *Free Radical Biology and Medicine* (2021.0) **169** 317-342. DOI: 10.1016/j.freeradbiomed.2021.03.046
9. Su JH, Luo MY, Liang N, Gong SX, Chen W, Huang WQ. **Interleukin-6: A novel target for cardio-cerebrovascular diseases**. *Frontiers in Pharmacology* (2021.0) **12** 745061. DOI: 10.3389/fphar.2021.745061
10. Huard A, Do HN, Frank AC, Sirait-Fischer E, Fuhrmann D, Hofmann MC. **IL-38 ablation reduces local inflammation and disease severity in experimental autoimmune encephalomyelitis**. *The Journal of Immunology* (2021.0) **206** 1058-1066. DOI: 10.4049/jimmunol.2000923
11. Ye C, Yano H, Workman CJ, Vignali DA. **Interleukin-35: Structure, function and its impact on immune-related diseases**. *Journal of Interferon & Cytokine Research* (2021.0) **41** 391-406. DOI: 10.1089/jir.2021.0147
12. Arkew M, Yemane T, Mengistu Y, Gemechu K, Tesfaye G. **Hematological parameters of type 2 diabetic adult patients at Debre Berhan Referral Hospital, Northeast Ethiopia: A comparative cross-sectional study**. *PLoS ONE* (2021.0) **16** e0253286. DOI: 10.1371/journal.pone.0253286
13. Mahdi A, Cortese-Krott MM, Kelm M, Li N, Pernow J. **Novel perspectives on redox signaling in red blood cells and platelets in cardiovascular disease**. *Free Radical Biology and Medicine* (2021.0) **168** 95-109. DOI: 10.1016/j.freeradbiomed.2021.03.020
14. Friedewald WT, Levy RI, Fredrickson DS. **Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge**. *Clinical chemistry* (1972.0) **18** 499-502. DOI: 10.1093/clinchem/18.6.499
15. Ross R. **The pathogenesis of atherosclerosis: A perspective for the 1990s**. *Nature* (1993.0) **362** 801-809. DOI: 10.1038/362801a0
16. Wallace TM, Levy JC, Matthews DR. **Use and abuse of HOMA modeling**. *Diabetes Care* (2004.0) **27** 1487-1495. DOI: 10.2337/diacare.27.6.1487
17. Poznyak A, Grechko AV, Poggio P, Myasoedova VA, Alfieri V, Orekhov AN. **The diabetes mellitus–atherosclerosis connection: The role of lipid and glucose metabolism and chronic inflammation**. *International journal of molecular sciences* (2020.0) **21** 1835. DOI: 10.3390/ijms21051835
18. Fathelbab M, Fahmy EM, Elshormilisy AA, Gaafar AE, Waly NE. **A putative role for oxidative stress in pathophysiology of diabetic cardiomyopathy**. *Egyptian Journal of Obesity, Diabetes and Endocrinology* (2017.0) **3** 95-99
19. Abdelazeem AH, Abuelsaad AS, Abdel-Moniem A, Abdel-Gabbar M. **Association of metabolic syndrome components with alterations in oxidative stress and cytokines expression**. *Journal of Taibah University for Science* (2021.0) **15** 928-940. DOI: 10.1080/16583655.2021.2009680
20. Neri M, Fineschi V, Di Paolo M, Pomara C, Riezzo I, Turillazzi E. **Cardiac oxidative stress and inflammatory cytokines response after myocardial infarction**. *Current vascular pharmacology* (2015.0) **13** 26-36. DOI: 10.2174/15701611113119990003
21. Lee WS, Kim J. **Diabetic cardiomyopathy: Where we are and where we are going**. *The Korean journal of internal medicine* (2017.0) **32** 404-421. DOI: 10.3904/kjim.2016.208
22. Gómez-Marcos MA, Blázquez-Medela AM, Gamella-Pozuelo L, Recio-Rodriguez JI, García-Ortiz L, Martínez-Salgado C. **Serum superoxide dismutase is associated with vascular structure and function in hypertensive and diabetic patients**. *Oxidative Medicine and Cellular Longevity* (2016.0) **2016** 9124676. DOI: 10.1155/2016/9124676
23. Aladağ N, Asoğlu R, Ozdemir M, Asoğlu E, Derin AR, Demir C. **Oxidants and antioxidants in myocardial infarction (MI): Investigation of ischemia modified albumin, malondialdehyde, superoxide dismutase and catalase in individuals diagnosed with ST elevated myocardial infarction (STEMI) and non-STEMI (NSTEMI)**. *Journal of Medical Biochemistry* (2021.0) **40** 286-294. DOI: 10.5937/jomb0-28879
24. Darogha SN. **Serum levels of TNF-a and IFN-g gene polymorphism in type 2 diabetes mellitus in Kurdish patients**. *Cellular and Molecular Biology* (2021.0) **67** 171-177. DOI: 10.14715/cmb/2021.67.2.27
25. Fu Y, Wu Y, Liu E. **C-reactive protein and cardiovascular disease: From animal studies to the clinic**. *Experimental and Therapeutic Medicine* (2020.0) **20** 1211-1219. DOI: 10.3892/etm.2020.8840
26. Zhu L, Zou Y, Wang Y, Luo X, Sun X. **Prognostic significance of plasma high-sensitivity C-reactive protein in patients with hypertrophic cardiomyopathy**. *Journal of the American Heart Association* (2017.0) **6** e004529. DOI: 10.1161/JAHA.116.004529
27. Koosha P, Roohafza H, Sarrafzadegan N, Vakhshoori M, Talaei M, Sheikhbahaei E. **High sensitivity C-reactive protein predictive value for cardiovascular disease: A nested case control from Isfahan cohort study (ICS)**. *Global Heart* (2020.0) **15** 3. DOI: 10.5334/gh.367
28. 28.Hage, C., E. Michaëlsson, C. Linde, E. Donal, J.C. Daubert, L.M. Gan, et al. 2017. Inflammatory biomarkers predict heart failure severity and prognosis in patients with heart failure with preserved ejection fraction: a holistic proteomic approach. Circulation: Cardiovascular Genetics 10: e001633.
29. Fanola CL, Morrow DA, Cannon CP, Jarolim P, Lukas MA, Bode C. **Interleukin-6 and the risk of adverse outcomes in patients after an acute coronary syndrome: Observations from the SOLID-TIMI 52 (stabilization of plaque using darapladib—thrombolysis in myocardial infarction 52) trial**. *Journal of the American Heart Association* (2017.0) **6** e005637. DOI: 10.1161/JAHA.117.005637
30. 30.Franssen, C., S. Chen, A. Unger, H.I. Korkmaz, G.W. De Keulenaer, C. Tschöpe, et al. 2016. Myocardial microvascular inflammatory endothelial activation in heart failure with preserved ejection fraction. Journal of the American College of Cardiology: Heart Failure 4: 312–324.
31. Yasuda K, Nakanishi K, Tsutsui H. **Interleukin-18 in health and disease**. *International journal of molecular sciences* (2019.0) **20** 649. DOI: 10.3390/ijms20030649
32. Xiao H, Li H, Wang JJ, Zhang JS, Shen J, An XB. **IL-18 cleavage triggers cardiac inflammation and fibrosis upon β-adrenergic insult**. *European heart journal* (2018.0) **39** 60-69. DOI: 10.1093/eurheartj/ehx261
33. Bahrami A, Sathyapalan T, Sahebkar A. **The role of interleukin-18 in the development and progression of atherosclerosis**. *Current Medicinal Chemistry* (2021.0) **28** 1757-1774. DOI: 10.2174/0929867327666200427095830
34. Wei Y, Lan Y, Zhong Y, Yu K, Xu W, Zhu R. **Interleukin-38 alleviates cardiac remodeling after myocardial infarction**. *Journal of Cellular and Molecular Medicine* (2020.0) **24** 371-384. DOI: 10.1111/jcmm.14741
35. Gupta A, Fei YD, Kim TY, Xie A, Batai K, Greener I. **IL-18 mediates sickle cell cardiomyopathy and ventricular arrhythmias**. *Blood, The Journal of the American Society of Hematology* (2021.0) **137** 1208-1218
36. Sha X, Meng S, Li X, Xi H, Maddaloni M, Pascual DW. **Interleukin-35 inhibits endothelial cell activation by suppressing MAPK-AP-1 pathway**. *Journal of Biological Chemistry* (2015.0) **290** 19307-19318. DOI: 10.1074/jbc.M115.663286
37. Ye J, Huang Y, Que B, Chang C, Liu W, Hu H. **Interleukin-12p35 knock out aggravates doxorubicin-induced cardiac injury and dysfunction by aggravating the inflammatory response, oxidative stress, apoptosis and autophagy in mice**. *eBioMedicine* (2018.0) **35** 29-39. DOI: 10.1016/j.ebiom.2018.06.009
38. Zhou F, Feng T, Lu X, Wang H, Chen Y, Zhang Q. **Interleukin 35 protects cardiomyocytes following ischemia/reperfusion-induced apoptosis via activation of mitochondrial STAT3**. *Acta Biochimica et Biophysica Sinica* (2021.0) **53** 410-418. DOI: 10.1093/abbs/gmab007
39. Hosseini MS, Rostami Z, Saadat A, Saadatm SM, Naeimi E. **Anemia and microvascular complications in patients with type 2 diabetes mellitus**. *Nephrourology monthly* (2014.0) **6** e19976
40. Cho YI, Mooney MP, Cho DJ. **Hemorheological disorders in diabetes mellitus**. *Journal of diabetes science and technology* (2008.0) **2** 1130-1138. DOI: 10.1177/193229680800200622
41. Nada AM. **Red cell distribution width in type 2 diabetic patients**. *Diabetes, Metabolic Syndrome and Obesity: Targets and Therapy* (2015.0) **8** 525-333. DOI: 10.2147/DMSO.S85318
42. Lippi G, Targher G, Montagnana M, Salvagno GL, Zoppini G, Guidi GC. **Relation between red blood cell distribution width and inflammatory biomarkers in a large cohort of unselected outpatients**. *Archives of pathology & laboratory medicine* (2009.0) **133** 628-632. DOI: 10.5858/133.4.628
43. Coenen DM, Heinzmann A, Karel M, Cosemans JM, Koenen RR. **The multifaceted contribution of platelets in the emergence and aftermath of acute cardiovascular events**. *Atherosclerosis* (2021.0) **319** 132-141. DOI: 10.1016/j.atherosclerosis.2020.12.017
44. Lebas H, Yahiaoui K, Martos R, Boulaftali Y. **Platelets are at the nexus of vascular diseases**. *Frontiers in cardiovascular medicine* (2019.0) **6** 132. DOI: 10.3389/fcvm.2019.00132
45. Milosevic D, Panin VL. **Relationship between hematological parameters and glycemic control in type 2 diabetes mellitus patients**. *Journal of Medical Biochemistry* (2019.0) **38** 164-171. DOI: 10.2478/jomb-2018-0021
46. Kern A, Gil RJ, Bojko K, Sienkiewicz E, Januszko-Giergielewicz B, Górny J. **Platelet distribution width as the prognostic marker in coronary bifurcation treatment**. *European Journal of Clinical Investigation* (2017.0) **47** 524-530. DOI: 10.1111/eci.12773
47. Feldman C, Anderson R. **Platelets and their role in the pathogenesis of cardiovascular events in patients with community-acquired pneumonia**. *Frontiers in Immunology* (2020.0) **11** 577303. DOI: 10.3389/fimmu.2020.577303
48. Okyere AD, Tilley DG. **Leukocyte-dependent regulation of cardiac fibrosis**. *Frontiers in Physiology* (2020.0) **11** 301. DOI: 10.3389/fphys.2020.00301
49. Woitas RP, Schwab S. **Neutrophil counts and their potential prognostic value in cardiovascular diseases**. *Journal of Laboratory and Precision Medicine* (2017.0) **2** 30. DOI: 10.21037/jlpm.2017.06.03
50. Ibrahim MS, Khalil AA, Ali AS. **Neutrophils-to-lymphocytes ratio in children with acute heart failure**. *The Egyptian Journal of Hospital Medicine* (2020.0) **81** 1878-1883. DOI: 10.21608/ejhm.2020.121015
51. Demirtas L, Degirmenci H, Akbas EM, Ozcicek A, Timuroglu A, Gurel A. **Association of hematological indices with diabetes, impaired glucose regulation and microvascular complications of diabetes**. *International journal of clinical and experimental medicine* (2015.0) **8** 11420-11427. PMID: 26379958
52. Abdel-Moneim A, Semmler M, Abdel-Reheim ES, Zanaty MI, Addaleel W. **Association of glycemic status and interferon-γ production with leukocytes and platelet indices alterations in type2 diabetes**. *Diabetes & Metabolic Syndrome: Clinical Research & Reviews* (2019.0) **13** 1963-1969. DOI: 10.1016/j.dsx.2019.04.046
53. Duman TT, Aktas G, Atak BM, Kocak MZ, Erkus E, Savli H. **Neutrophil to lymphocyte ratio as an indicative of diabetic control level in type 2 diabetes mellitus**. *African health sciences* (2019.0) **19** 1602-1606. DOI: 10.4314/ahs.v19i1.35
54. Hussain M, Babar M, Akhtar L, Hussain MS. **Neutrophil lymphocyte ratio (NLR): A well assessment tool of glycemic control in type 2 diabetic patients**. *Pakistan journal of medical sciences* (2017.0) **33** 1366-1370. DOI: 10.12669/pjms.336.12900
55. Li XT, Fang H, Li D, Xu FQ, Yang B, Zhang R. **Association of platelet to lymphocyte ratio with in-hospital major adverse cardiovascular events and the severity of coronary artery disease assessed by the Gensini score in patients with acute myocardial infarction**. *Chinese Medical Journal* (2020.0) **133** 415-423. DOI: 10.1097/CM9.0000000000000650
56. 56.Toprak, C., M.M. Tabakci, Z. Simsek, U. Arslantas, H.I. Durmus, L. Ocal, et al. 2015. Platelet/lymphocyte ratio was associated with impaired myocardial perfusion and both in-hospital and long-term adverse outcome in patients with ST-segment elevation acute myocardial infarction undergoing primary coronary intervention. Postępy w Kardiologii Interwencyjnej= Advances in Interventional Cardiology 11: 288–297.
57. Zhou Z, Liang M, Wu H, Huang S, Weng R, Hou J. **Preoperative lymphocyte-to-monocyte ratio as a prognostic predictor of long-term mortality in cardiac surgery patients: A propensity score matching analysis**. *Frontiers in Cardiovascular Medicine* (2021.0) **8** 639890. DOI: 10.3389/fcvm.2021.639890
58. Fan W. **Epidemiology in diabetes mellitus and cardiovascular disease**. *Cardiovascular endocrinology* (2017.0) **6** 8-16. DOI: 10.1097/XCE.0000000000000116
|
---
title: Onset and mortality of Parkinson’s disease in relation to type II diabetes
authors:
- Gianni Pezzoli
- Emanuele Cereda
- Paolo Amami
- Santo Colosimo
- Michela Barichella
- Giorgio Sacilotto
- Anna Zecchinelli
- Michela Zini
- Valentina Ferri
- Carlotta Bolliri
- Daniela Calandrella
- Maria Grazia Bonelli
- Viviana Cereda
- Elisa Reali
- Serena Caronni
- Erica Cassani
- Margherita Canesi
- Francesca del Sorbo
- Paola Soliveri
- Luigi Zecca
- Catherine Klersy
- Roberto Cilia
- Ioannis U. Isaias
journal: Journal of Neurology
year: 2022
pmcid: PMC9971073
doi: 10.1007/s00415-022-11496-y
license: CC BY 4.0
---
# Onset and mortality of Parkinson’s disease in relation to type II diabetes
## Abstract
### Objectives
There is growing evidence that Parkinson’s disease and diabetes are partially related diseases; however, the association between the two, and the impact of specific treatments, are still unclear. We evaluated the effect of T2D and antidiabetic treatment on age at PD onset and on all-cause mortality.
### Research design and methods
The standardized rate of T2D was calculated for PD patients using the direct method and compared with subjects with essential tremor (ET) and the general Italian population. Age at onset and survival were also compared between patients without T2D (PD-noT2D), patients who developed T2D before PD onset (PD-preT2D) and patients who developed T2D after PD onset (PD-postT2D).
### Results
We designed a retrospective and prospective study. The T2D standardized ratio of PD ($$n = 8380$$) and ET ($$n = 1032$$) patients was $3.8\%$ and $6.1\%$, respectively, while in the *Italian* general population, the overall prevalence was $5.3\%$. In PD-preT2D patients, on antidiabetic treatment, the onset of PD was associated with a + 6.2 year delay ($p \leq 0.001$) while no difference was observed in PD-postT2D. Occurrence of T2D before PD onset negatively affected prognosis (adjusted hazard ratio = 1.64 [$95\%$ CI 1.33–2.02]; $p \leq 0.001$), while no effect on survival was found in PD-postT2D subjects (hazard ratio = 0.86, [$95\%$ CI 0.53–1.39]; $$p \leq 0.54$$).
### Conclusions
T2D, treated with any antidiabetic therapy before PD, is associated with a delay in its onset. Duration of diabetes increases mortality in PD-preT2D, but not in PD-postT2D. These findings prompt further studies on antidiabetic drugs as a potential disease-modifying therapy for PD.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00415-022-11496-y.
## Introduction
Parkinson’s disease (PD) and related syndromes constitute the second most common group of neurodegenerative conditions. It has been noted that from 1990 to 2015 the number of people with PD doubled to over 6 million and this figure is predicted to double again to over 12 million by 2040, mainly because of the increasing average age of the population [1, 2]. The increasing prevalence of PD is of concern due to the huge burden this disease places on patients, caregivers and public healthcare budgets.
PD is mainly managed symptomatically using dopaminergic drugs, with an excellent response in the early years of the disease. Disease progression demonstrates that motor symptoms re-emerge after 5–7 years of relative well-being [3]. Efforts to find disease-modifying agents have not met with success, and none of the randomized trials have shown convincing effects of putative agents on PD progression [4].
PD and type II diabetes (T2D) have been established to share underlying mechanisms, including mitochondrial dysfunction, under-expression of the transcriptional regulator PPARγ coactivator 1α, oxidative stress, and inflammation [5]. Interestingly, a significant decrease in 11C-donepezil (a high-affinity ligand for acetylcholinesterase) signal was demonstrated in the pancreas of PD patients [6], as in patients with type I diabetes [7], indicating parasympathetic denervation in PD. Furthermore, cytoplasmic phosphorylated alpha-synuclein deposits were found in the pancreatic beta-cells of subjects with PD and T2D [8, 9]. Insulin resistance was found in $60\%$ of PD patients, $30\%$ of whom developed glucose intolerance [10], and increased levels of alpha-synuclein negatively affect glucose-stimulated insulin secretion in pancreatic beta-cells of the Rip/Snca transgenic mice model [11].
There are several epidemiologic studies connecting PD and T2D; however, there is no agreement on the risk of diabetic patients to develop PD: reviews and meta-analyses have reached opposite conclusions [12–16]. Concordant are instead the clinical studies, with cross-sectional observations suggesting that T2D is associated with a more aggressive PD phenotype and an enhanced progression of PD symptoms [17–19]. A recent prospective study has reported a comparable, accelerated progression associated with good or poor glycemic control, respectively [20]. It seems that T2D, when treated, has no detrimental effect on PD clinical phenotype. Further evidence suggests that some antidiabetic drugs may protect against neurodegeneration. A randomized controlled trial of exenatide—a glucagon-like peptide-1 (GLP-1) receptor agonist used to treat T2D—found that the drug had positive effects on practically all off-medication motor features, which persisted even beyond the period of drug exposure [21]. It has also been observed that the incidence of PD in patients with T2D was significantly lower in those given GLP-1 receptor agonists or dipeptidyl peptidase 4 (DPP4) inhibitors compared with individuals prescribed any other oral combination therapy for diabetes [22–24]. Additional evidence comes from the use of bromocriptine, a dopamine receptor agonist, as useful adjunctive agent in the management of T2D. Both bromocriptine and dopamine mediate in vivo tissue glucose uptake in rodents [25]. Furthermore, in the 6-OHDA-induced PD mouse model, metformin itself partially ameliorates akinetic-rigid symptoms [26].
The aim of the present study was to investigate the relationship between PD and T2D by observing a large cohort of consecutive patients presenting at the Parkinson Institute Milan (ASST-Pini-CTO) over a 10-year period. Presently, there are no comparative data on the onset and evolution of PD in subjects who developed it while taking or not taking antidiabetic therapy. We evaluated age at onset of PD and estimated the prevalence of T2D in PD compared with the general population and subjects with essential tremor (ET). We also investigated the effects of T2D occurrence in relation to PD onset on all-cause mortality.
## Patients
For this retrospective study using prospectively collected data, the database of the Parkinson Institute Milan (ASST-Pini-CTO), including consecutive patients followed at our clinical centre over a period of 10 years (from January 2010 to December 2019), was queried. Patients with an established diagnosis of PD and ET were included in the study, while patients with other neurodegenerative diseases, vascular disease or undefined diagnosis as well as patients with type I diabetes were excluded (Fig. 1).Fig. 1Flowchart of patient selection. Abbreviations: PD-noT2D PD patients without T2D, PD-preT2D patients with T2D occurrence before PD onset, PD-postT2D patients with T2D occurrence after PD onset *The diagnosis* of PD or ET was established by a movement disorder specialist, based on established diagnostic criteria for PD [27, 28] and ET [29], with the support of neuroimaging findings (brain magnetic resonance imaging, dopamine transporter single photon emission computed tomography, 18F-fluorodeoxyglucose positron-emission tomography, etc.).
## Clinical evaluation
At baseline (the first visit at the Parkinson Institute Milan, ASST-Pini-CTO), all patients underwent a thorough neurological examination and had self-completed a comprehensive questionnaire including more than 100 items covering a broad range of aspects (e.g., family history, life-style, past medical history, occupational history, detailed history of the neurodegenerative disease, previous and present medication) that was subsequently reviewed by a neurologist. The same questionnaire was updated during each follow-up visit. Patients who reported the occurrence of T2D were selected for a further health interview, with caregiver support if necessary, to confirm their history of diabetes. Information regarding the onset of T2D and current antidiabetic medical treatment was further confirmed through prescriptions from the family doctor, referral diabetologist or by accessing the health data register.
All patients were visited at the outpatient clinic and a subgroup was hospitalized for additional diagnostic and/or therapeutic purposes (herein identified as “inpatients” group). These patients underwent blood glucose and glycosylated hemoglobin after overnight fasting according to consensus criteria [30]. This subgroup was used as an internal control to assess the accuracy of self-reported data on diabetes (onset, therapy, etc.).
The study endpoints were the prevalence of diabetes at the last available visit, age at onset of PD and all-cause mortality. Specifically, age at onset was defined as the age when the first cardinal motor symptoms appeared, as reported by the patient with the aid of a family member or caregiver when necessary. Survival was defined as the time that had elapsed (in years) between the first evaluation at the Parkinson Institute Milan and the date of death or the date of last contact (censoring). Vital status was ascertained by means of active follow-up (in-office visits, telephone or mail enquiries to participants or proxy respondents and linkage to municipal registries) up to February 2020.
## Prevalence
Prevalence of T2D was calculated for PD outpatients, PD inpatients and ET patients at the date of last contact. The direct method was used to standardize raw prevalence based on the most recent data on gender and age distribution of the Italian population (from the 2016 survey by the Italian National Institute of Statistics, https://www.istat.it/it/archivio/202600).
At the last follow-up visit, a total of 673 ($8.4\%$) PD patients and 150 ($14.5\%$) ET patients had T2D. A comparison of the standardized prevalence of T2D identified a lower frequency in PD patients ($3.8\%$) than in the *Italian* general population ($5.3\%$) and ET patients ($6.1\%$, Table 1). Moreover, the standardized rate of T2D was lower in PD patients aged ≥ 65 compared with the Italian population and ET patients (Table 1). The standardized rate of T2D in PD inpatients, who underwent blood glucose and glycosylated haemoglobin evaluation, was comparable to the rate in PD outpatients (Table 1).Table 1T2D prevalence and standardized prevalence in PD outpatients, hospitalized PD patients and ET patientsTotal PD patientsN = 8380PD outpatientsN = 7789PD inpatientsN = 591ET patientsN = 1032Italian populationN = 60,433,962No of patients with T2D673 ($8\%$)629 ($8.1\%$)44 ($7.4\%$)150 ($14.5\%$)3,203,000T2D standardized rate$3.8\%$$3.91\%$$3.14\%$$6.08\%$$5.3\%$T2D standardized rate (≥ 65)$9.01\%$$8.97\%$$10.19\%$$16.44\%$$16.4\%$The raw prevalence rates were standardized based on gender and age distribution of the Italian population using the direct method
## Age at PD onset
At the end of follow-up, PD patients were grouped based on T2D occurrence in relation to PD onset as follows: patients who never developed T2D (PD-noT2D), patients who had T2D before PD onset (PD-preT2D) and patients in whom T2D occurred after PD onset (PD-postT2D, Fig. 1). We compared demographic and clinical data between these three groups.
PD patients with no reliable information regarding the date of T2D diagnosis and those not taking antidiabetic medication were excluded from the study (Fig. 1). T2D occurred before the onset of PD in 413 patients, while 171 patients developed T2D after PD onset. Demographic and clinical features of PD patients based on T2D occurrence in relation to PD onset are reported in Table 2. Age at PD onset was approximately 67 years in PD-preT2D patients, a higher mean onset age of 7 years compared with both PD-postT2D and PD patients without T2D ($p \leq 0.001$). This difference was maintained after adjusting for gender, coffee consumption and smoking: compared with PD-noT2D, patients with T2D prior to the onset of PD were older (a mean difference of 6 years, $95\%$ CI 3.3–8.5, $p \leq 0.001$), while those who developed T2D after PD had a comparable age at PD onset ($$p \leq 0.99$$). A multivariate regression analysis showed that age at PD onset was related to the duration of T2D treatment ($p \leq 0.001$), as patients who had been treated for T2D over a longer time period were older at PD onset (Supplementary Table 2). PD onset in patients who had had T2D for seven years or less was delayed by an average of 4.8 years, while patients with a T2D duration of over 7 years displayed a mean delay of 5.7 years compared with patients without T2D before PD onset (Supplementary Table 3). We also evaluated the effect that metformin treatment had on age of PD onset compared with other antidiabetic treatments, but we found no difference ($$p \leq 0.525$$) (Supplementary Table 4). BMI calculated during the last visit was higher both in the PD-preT2D and PD-postT2D groups compared with PD patients without diabetes ($p \leq 0.001$). The occurrence of heart diseases ($p \leq 0.001$) and hypertension ($p \leq 0.001$) was higher in both groups of PD patients with diabetes compared to non-diabetic PD patients, while the occurrence of strokes was higher only in PD-preT2D patients ($$p \leq 0.002$$). There was no difference between groups regarding occurrence of tumors ($$p \leq 0.099$$). Consumption of coffee and smoking were comparable between the three groups. Table 2Comparison of PD-preT2D, PD-postT2D and PD-noT2D patients at the last contact (total number of patients = 8291)PD-noT2DN = 7707PD-preT2DN = 413PD-postT2DN = 171p valueGender (f%)3258 ($42.3\%$)129 ($31.2\%$)68 ($39.8\%$) < 0.001Age at PD onset (years)60.71 (± 11.01)66.87 (± 8.5)60.58 (± 10.88) < 0.001aAge at T2D onset (years)–56.71 (± 10.96)66.11 (± 9.82) < 0.001Age at last contact (years)72.52 (± 10.33)75.35 (± 8.02)73.74 (± 8.85) < 0.001bPD duration at last contact (years)11.81 (± 7.41)8.47 (± 5.41)13.16 (± 7.28) < 0.001cT2D duration at last contact (years)–18.63 (± 9.98)7.74 (± 5.94) < 0.001BMI at last contact25.22 (± 4.04)27.28 (± 4.31)27.61 (± 4.66) < 0.001dHeart diseases (occurrence%)1433 ($18.6\%$)113 ($27.4\%$)46 ($26.9\%$) < 0.001Hypertension (occurrence%)2824 ($36.6\%$)257 ($62.2\%$)93 ($54.4\%$) < 0.001Stroke (occurrence%)289 ($3.7\%$)30 ($7.3\%$)8 ($4.7\%$)0.002Tumor (occurrence%)844 ($11\%$)54 ($13.1\%$)26 ($15.2\%$)0.099For categorical variables a Chi-squared test was usedFor continuous variables the Mann–Whitney test was used for a comparison between two groups and the Kruskal–Wallis test was used for a comparison among three groups; for post hoc comparisons the Bonferroni correction was used: aPD-preT2D ≠ PD-postT2D = PD-noT2D; bPD-preT2D ≠ PD-noT2D; cPD-preT2D ≠ PD-post-T2D ≠ PD-noT2D; dPD-noT2D ≠ PD-preT2D = PD-postT2D
## Survival
Survival time was defined as the number of years that had elapsed from the first visit to the outcome or to the date of last contact. Outcome was the patient’s death.
Continuous variables were reported as means and standard deviations (SD), categorical variables as frequencies and percentages (%). The normal distribution of continuous variables was checked with the Kolmogorov–Smirnov and Shapiro–Wilk tests. Between-group comparisons were performed using the Chi-squared test, one-way ANOVA or the Kruskal–Wallis test based on the nature and distribution of variables. Post hoc comparisons were performed using the Bonferroni correction. The effect of the duration of T2D treatment on age of PD onset was assessed by a multivariable linear regression analysis. The log rank test was used to compare survival curves between the three groups. The Cox proportional hazards regression model was used to investigate the association between the occurrence of T2D in relation to PD onset on survival time (PD was considered the reference group); covariates included gender, age at baseline, PD duration at baseline and comorbidities (heart disease, hypertension, stroke, and tumor). For all analyses, the statistical level was set at $p \leq 0.05.$ All statistical analyses were performed using the SPSS 25 software package.
Survival analysis included a total of 6,930 PD patients, as 1,327 patients were excluded since they did not attend follow-up and 34 patients experienced T2D onset during the observation period (Fig. 1). The baseline features of the three groups are shown in Table 3. After a median follow-up of 60 months [IQR 36–84], 1877 patients ($27.1\%$) had died. PD-preT2D patients experienced higher mortality rates: 1,742 ($26.9\%$) deaths occurred in patients without T2D, 111 ($32.6\%$) in PD-preT2D patients and 24 ($21.4\%$) in PD-postT2D patients ($p \leq 0.003$; Fig. 2). Cox hazards analysis showed an association between mortality and the occurrence of T2D before PD onset, as this group of patients had an increased hazard ratio compared to PD patients without diabetes (fully adjusted HR = 1.64 [$95\%$ CI 1.33–2.02]; $p \leq 0.001$), while patients who developed T2D after PD onset had a hazard ratio comparable to patients without diabetes (HR = 0.86, [$95\%$ CI 0.53–1.39]; $$p \leq 0.54$$) (Supplementary Table 5). There was no difference in mortality when we compared PD patients with T2D treated with metformin or other antidiabetic drugs ($$p \leq 0.265$$).Table 3Baseline features of PD-preT2D, PD-postT2D and PD-noT2D patients who participated in survival analysis (total number of patients = 6930)PD-noT2DN = 6478PD-preT2DN = 340PD-postT2DN = 112p valueGender (f%)2756 ($42.5\%$)106 ($31.2\%$)43 ($38.4\%$) < 0.001Age at baseline (years)68.1 (± 10.3)71.4 (± 8)69.5 (± 8.8) < 0.001aPD duration (years)7.8 (± 6.7)4.8 (± 4.2)8.1 (± 5.9) < 0.001bT2D duration (years)–14.4 (± 9.5)4.5 (± 5.5) < 0.001Follow-up duration (months)59.6 (± 29.7)55 (± 30.6)60 (± 33.4)0.017 aHeart diseases (occurrence%)1217 ($18.8\%$)93 ($27.4\%$)32 ($28.6\%$) < 0.001Hypertension (occurrence%)2405 ($37.1\%$)212 ($62.4\%$)64 ($57.1\%$) < 0.001Stroke (occurrence%)261 ($4\%$)23 ($6.8\%$)4 ($3.6\%$)0.047Tumor (occurrence%)706 ($10.9\%$)48 ($14.1\%$)14 ($12.5\%$)167For categorical variables, the Chi-squared test was usedFor continuous variables the Mann–Whitney test was used for a comparison between two groups and the Kruskal–Wallis test for a comparison among three groups; for post hoc comparisons Bonferroni’s correction was used: aPD-preT2D ≠ PD-noT2D; bPD-preT2D ≠ PD-postT2D = PD-noT2DFig. 2Kaplan–Meier mortality curves of PD-preT2D, PD-postT2D and PD-noT2D patients (log-rank test: χ[2] = 11.459, $p \leq 0.003$). Abbreviations: PD-noT2D PD patients without T2D, PD-preT2D patients with T2D occurrence before PD onset, PD-postT2D patients with T2D occurrence after PD onset
## Ethics approval
The Ethics Committee of Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milano, Regione Lombardia-Italy approved the study “dated favorable opinion 18.03.2020/opinion 170_2020bis”. The study was performed in accordance with the ethical principles of the Declaration of Helsinki and its later amendments. All subjects provided written informed consent to participate in the study.
## Data sharing
Data sharing is available upon reasonable request to the corresponding author who will evaluate on a case-by-case basis.
## Results
From January 2010 through December 2019, 11,668 patients presented at the Parkinson Institute Milan (ASST-Pini-CTO), of whom 8380 and 1032 had an established diagnosis of PD and ET, respectively (Fig. 1). Five hundred and ninety-one PD patients were hospitalized for diagnostic or therapeutic purposes (Fig. 1) and no undiagnosed case of T2D was identified. The antiparkinsonian and antidiabetic therapies are listed in one of our previous works [17]. The percentage distribution of antidiabetic drugs at the last visit for PD-preT2D and PD-postT2D is further detailed in Supplementary Table 1.
## Discussion
This work stems from a previous case–control design study [17], which showed a worsening trend in motor symptoms in PD-preT2D when compared with PD-noT2D, a finding later confirmed by other authors [31]. Herein, we present data on age at onset in PD-preT2D and PD-postT2D patients, the prevalence of diabetes in PD, and mortality rates across the various groups. Clinical data were collected in a single third-level Italian centre (Parkinson Institute Milan, ASST-Pini-CTO, Milan, Italy). This might be a limitation of the current study, as complex patients are more easily accessed in a third-level centre. Less demanding patients are followed by general neurologists or even by their general practitioners. Besides, data generalizability might be affected by the ethnic homogeneity of our cohort (only 24 non-Caucasians patients). Conversely, there are positive aspects: nine highly experienced neurologists on movement disorders work at the Parkinson Institute Milan providing consistent and comparable follow-up evaluations; diagnoses are more conclusive after a few years of illness.
The result is a retrospective cohort study regarding the onset of PD and T2D, prospective as regards mortality. Collecting mortality data were not a major issue because in addition to family members, the municipality of residence and place of birth of all the patients in the database were consulted, if deemed necessary. Nor was it difficult to establish the age of onset of PD as only motor symptoms were considered. The average age of all PD patients at onset is very similar to those presented by other authors [32]. It was more difficult to establish with relative certainty the onset of diabetes and the beginning of antidiabetic therapy because the disease itself is paucisymptomatic and patients tend to lose long-standing clinical documentation over the years. Therefore, potential bias associated with self-reporting is also acknowledged taking into account that not all patients underwent full screening procedures for diabetes, although in the subgroup of inpatients—used as internal control—no undiagnosed case of T2D was identified. For this reason, we excluded 70 diabetic patients who were unable to document the disease’s onset with any certainty (Fig. 1), and present only the antidiabetic therapy related to the last follow-up visit. We also recognize that glycemic control for each diabetic patient in this study was not available, except for those patients who were hospitalized, in which, however, it was sufficiently controlled but we cannot exclude that these data, along with information on organ damage, would have improved our analyses further. Noteworthy, polymorbid patients are likely to undergo increased medical surveillance and this would have strengthened the reliability of the data analysed herein but we cannot also exclude that some signs of PD might have been interpreted as signs of DM by general practitioners. In Italy, diabetic patients are entitled to free prescriptions, which allows us to exclude bias related to economic status. Besides, the new GLP-1 receptor agonists are provided at no cost only to diabetic patients intolerant to metformin: only few patients are treated with these drugs. As regards gender, the characteristics of the cohort are similar to those reported by other databases [33] with a moderate male prevalence in the total parkinsonian population and even more marked in the diabetic group (Table 2) [34].
This work focuses on age of onset of PD in the non-diabetic population, compared with that of a cohort of diabetic parkinsonian patients. The data are solid with a large sample size. Diabetic patients on any form of antidiabetic therapy were found to report development of PD more than 6 years later than non-diabetic subjects or subjects who go on to develop diabetes after the onset of PD. Until now, there are no known treatments or pathologies that can cause a delay in the onset of PD of this magnitude. The literature already cites studies reporting that diabetic subjects have a delayed onset of PD of the same magnitude reported by Ou and coll [20], but the data have never been evaluated in depth by dividing diabetic patients who developed the disease before or after the onset of PD. This allows us to observe that the age of onset of PD in PD-postT2D patients is identical to that in non-diabetic parkinsonian subjects (Table 2).
Delayed onset of PD by about 6 years in PD-preT2D corresponds to approximately $30\%$ of the average life expectancy of a parkinsonian patient. Therefore, it is reasonable to expect a reduction in the prevalence of diabetes in PD, as the delayed onset means that a certain percentage of diabetic patients will not live long enough to develop PD. Compared with official *Italian data* from 2016, of those who “claim to have diabetes” the prevalence in our groups is about − $30\%$ among all parkinsonian patients. In this regard, data in the literature are rather controversial [14–16]. In particular, out of 18 studies considered in the most recent review by Camargo Maluf and coll [35], the risk of developing PD in patients with T2D was increased in nine studies and decreased or unrelated in nine. Interestingly, the studies that observed a reduced risk of developing diabetes in PD enrolled fewer patients but had better characterized case series. In view of this, we tried to combine a large-sample cohort study with diagnostic accuracy carried out over an average 5-years period, confirmed in virtually all patients by instrumental examinations: MRI or CT scan, DaTScan (i.e., SPECT with Ioflupane) and PET with fluorodeoxyglucose (FDG), and less frequently with a CSF examination. In addition, subjects from the three groups were enrolled in the study between five and nine years (Table 3) after the onset of PD, when symptoms and response to levodopa are generally sufficiently clear and the diagnostic uncertainties of the initial period are overcome. With these methods, ET, a highly frequent disorder, and parkinsonisms with a prevalent vascular component, or other movement disorders of uncertain diagnosis, were differentiated. Mortality data suggested a significant, increased risk in subjects with pre-diabetes compared to subjects without diabetes or with post-diabetes. These results are in line with clinical data available in the literature, especially in PD-preT2D, indicating a more severe and less responsive form of PD than in age- and sex-matched, non-diabetic PD patients [17, 18, 24].
Late onset of PD and early mortality in PD-preT2D subjects can therefore account for the reduced prevalence of diabetes in PD that we observed. It could be argued that treatment of diabetes extends the life of those affected to an age when they (can) develop PD. In addition, people with T2D without PD have higher mortality rates and are likely to die at earlier age than the general population [36].
In the group of patients with PD-preT2D (approximately 20 years of diabetes at death), the risk of death was higher, despite the fact they had only a 12-years history of PD compared to PD-noT2D subjects with an 18-years history of the disease. In the PD-postT2D group, however, with a 10-years history of diabetes before death, no increased risk of mortality was found. We can therefore speculate that diabetes treated with antidiabetic drugs may be associated with late onset of PD. During the first ten years of diabetes, the risk of mortality for PD-postT2D patients is not increased when compared with patients with PD only. This cohort study confirms what many studies now suggest, as well as the opinion of experts [37] in an increasing number of reports in the medical literature, namely, that antidiabetic drugs could have a beneficial effect when used in PD, alpha-synucleinopathies and other neurodegenerative diseases. This hypothesis is difficult to establish and does not find a causal basis in this work. Still, the results presented suggest a correlation of great interest for its potential preventive impact and for the study of novel disease-modifying strategies. In this study, due to limitations in the number of cases of T2D we could not demonstrate any difference between metformin and other antidiabetic agents for delayed disease onset and mortality, but current research mainly focuses on drugs that do not reduce blood sugar and thus can be tested as neurotrophic drugs even in non-diabetic patients. Diabetes alone, with or without drug treatment, could theoretically be fully or partly responsible for this phenomenon. We are unable to report any data in this regard, as only 19 patients in our group are on diet therapy alone. We are also unable to state that this phenomenon is dopaminergic per se as we do not have systematic data such as DaTScan imaging in this patient group. Even if the phenomenon of delayed onset of PD is not related to antidiabetic treatment but is a feature of diabetes itself, it would still be interesting to understand the pathophysiological basis of this phenomenon with future studies. We do believe there are very interesting premises for considering a study addressing the use of antidiabetic drugs at least in those subjects who present what is now considered a preclinical phase of PD, characterized by constipation, hypo/anosmia, REM behaviour-disorder that can occur up to 10–15 years before the onset of motor symptoms [38].
Being able to positively modify the course of PD, even in a prodromal phase by $30\%$, would make the management of this disease much less complex for neurologists and less painful for patients and caregivers. Controlled longitudinal studies are needed to confirm these findings in PD and other neurodegenerative diseases. However, it is objectively difficult to carry out a controlled study able to demonstrate that antidiabetic drugs have the ability to partially prevent PD.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 36 KB)
## References
1. Tysnes OB, Storstein A. **Epidemiology of Parkinson’s disease**. *J Neural Transm* (2017.0) **124** 901-905. DOI: 10.1007/s00702-017-1686-y
2. Dorsey ER, Sherer T, Okun MS, Bloemd BR. **The emerging evidence of the Parkinson pandemic**. *J Parkinson’s Dis* (2018.0) **8** S3-S8. DOI: 10.3233/JPD-181474
3. Cilia R, Akpalu A, Sarfo FS. **The modern pre-levodopa era of Parkinson’s disease: Insights into motor complications from sub-Saharan Africa**. *Brain* (2014.0) **137** 2731-2742. DOI: 10.1093/brain/awu195
4. Lang AE, Espay AJ. **Disease modification in Parkinson’s disease: current approaches, challenges, and future considerations**. *Mov Disord* (2018.0) **33** 660-677. DOI: 10.1002/mds.27360
5. Aviles-Olmos I, Limousin P, Lees A, Foltynie T. **Parkinson’s disease, insulin resistance and novel agents of neuroprotection**. *Brain* (2013.0) **136** 374-384. DOI: 10.1093/brain/aws009
6. Gjerløff T, Fedorova T, Knudsen K. **Imaging acetylcholinesterase density in peripheral organs in Parkinson’s disease with 11 C-donepezil PET**. *Brain* (2015.0) **138** 653-663. DOI: 10.1093/brain/awu369
7. Klinge MW, Borghammer P, Lund S. **Enteric cholinergic neuropathy in patients with diabetes: non-invasive assessment with positron emission tomography**. *Neurogastroenterol Motility* (2020.0). DOI: 10.1111/nmo.13731
8. Beach TG, Adler CH. **Multi-organ distribution of phosphorylated α-synuclein histopathology in subjects with Lewy body disorders**. *Acta Neuropathol* (2010.0) **119** 689-702. DOI: 10.1007/s00401-010-0664-3
9. Martinez-Valbuena I, Amat-Villegas I, Valenti-Azcarate R. **Interaction of amyloidogenic proteins in pancreatic β cells from subjects with synucleinopathies**. *Acta Neuropathol* (2018.0) **135** 877-886. DOI: 10.1007/s00401-018-1832-0
10. Bosco D, Plastino M, Cristiano D. **Dementia is associated with Insulin Resistance in patients with Parkinson’s Disease**. *J Neurol Sci* (2012.0) **315** 39-43. DOI: 10.1016/j.jns.2011.12.008
11. Steneberg P, Bernardo L, Edfalk S. **The Type 2 Diabetes-Associated Gene Ide Is Required for Insulin Secretion and Suppression of a-Synuclein Levels in b-Cells**. *Diabetes* (2013.0) **62** 2004-2014. DOI: 10.2337/db12-1045
12. Cheong JLY, de Pablo-Fernandez E, Foltynie T, Noyce AJ. **The association between Type 2 diabetes mellitus and Parkinson’s disease**. *J Parkinson’s Dis* (2020.0) **10** 775-789. DOI: 10.3233/JPD-191900
13. Cardoso S, Moreira PI, Söderbom G, Esterline R, Oscarsson J. **Chapter two—antidiabetic drugs for Alzheimer’s and Parkinson’s diseases: repurposing insulin, metformin, and thiazolidinediones**. *Metabolic and Bioenergetic Drivers of Neurodegenerative Disease: Treating neurodegenerative diseases as metabolic diseases* (2020.0) 37-64
14. Cereda E, Barichella M, Pedrolli C. **Diabetes and risk of Parkinson’s disease: a systematic review and meta-analysis**. *Diabetes Care* (2011.0) **34** 2614-2623. DOI: 10.2337/dc11-1584
15. Chohan H, Senkevich K, Patel RK. **Type 2 diabetes as a determinant of Parkinson’s disease risk and progression**. *Mov Disord* (2021.0) **36** 1420-1429. DOI: 10.1002/mds.28551
16. Lu L, Fu D, Li H. **Diabetes and risk of Parkinson’s disease: an updated meta-analysis of case-control studies**. *PLoS One* (2014.0) **9** e85781. DOI: 10.1371/journal.pone.0085781
17. Cereda E, Barichella M, Cassani E. **Clinical features of parkinson disease when onset of diabetes came first: a case-control study**. *Neurology* (2012.0) **78** 1507-1511. DOI: 10.1212/WNL.0b013e3182553cc9
18. Pagano G, Polychronis S, Wilson H. **Diabetes mellitus and Parkinson disease**. *Neurology* (2018.0) **90** E1654-E1662. DOI: 10.1212/WNL.0000000000005475
19. Mollenhauer B, Zimmermann J, Sixel-Döring F. **Baseline predictors for progression 4 years after Parkinson’s disease diagnosis in the De Novo Parkinson Cohort (DeNoPa)**. *Mov Disord* (2019.0) **34** 67-77. DOI: 10.1002/mds.27492
20. Ou R, Wei Q, Hou Y. **Effect of diabetes control status on the progression of Parkinson’s disease: a prospective study**. *Ann Clin Transl Neurol* (2021.0) **8** 887-897. DOI: 10.1002/acn3.51343
21. Athauda D, Maclagan K, Clinical C. **Exenatide once weekly versus placebo in Parkinson’s disease: a randomised, double-blind, placebo-controlled trial**. *Lancet* (2018.0) **390** 1664-1675. DOI: 10.1016/S0140-6736(17)31585-4
22. Brauer R, Wei L, Ma T. **Diabetes medications and risk of Parkinson’s disease: a cohort study of patients with diabetes**. *Brain* (2020.0) **143** 3067-3076. DOI: 10.1093/brain/awaa262
23. Mulvaney CA, Duarte GS, Handley J. **GLP-1 receptor agonists for Parkinson’s disease**. *Cochrane Database Syst Rev* (2020.0). DOI: 10.1002/14651858.CD012990.pub2
24. Jeong SH, Chung SJ, Yoo HS. **Beneficial effects of dipeptidyl peptidase-4 inhibitors in diabetic Parkinson’s disease**. *Brain* (2021.0) **144** 1127-1137. DOI: 10.1093/brain/awab015
25. Tavares G, Martins FO, Melo BF. **Peripheral dopamine directly acts on insulin-sensitive tissues to regulate insulin signaling and metabolic function**. *Front Pharmacol* (2021.0) **12** 713418. DOI: 10.3389/fphar.2021.713418
26. Ryu Y-K, Go J, Park H-Y. **Metformin regulates astrocyte reactivity in Parkinson’s disease and normal aging**. *Neuropharmacology* (2020.0) **175** 108173. DOI: 10.1016/j.neuropharm.2020.108173
27. Postuma RB, Berg D, Stern M. **MDS clinical diagnostic criteria for Parkinson’s disease**. *Mov Disord* (2015.0) **30** 1591-1601. DOI: 10.1002/mds.26424
28. Hughes AJ, Ben-Shlomo Y, Daniel SE, Lees AJ. **What features improve the accuracy of clinical diagnosis in Parkinson’s disease: a clinicopathologic study. 1992**. *Neurology* (2001.0) **57** S34-S38. PMID: 11775598
29. Bhatia KP, Bain P, Bajaj N. **Consensus Statement on the classification of tremors from the task force on tremor of the International Parkinson and Movement Disorder Society**. *Movement Disord* (2018.0) **33** 75-87. DOI: 10.1002/mds.27121
30. Chamberlain JJ, Rhinehart AS, Shaefer CFJ, Neuman A. **Diagnosis and management of diabetes: synopsis of the 2016 American Diabetes Association standards of medical care in diabetes**. *Ann Intern Med* (2016.0) **164** 542-552. DOI: 10.7326/M15-3016
31. Kotagal V, Albin RL, Müller MLTM. **Diabetes is associated with postural instability and gait difficulty in Parkinson disease**. *Parkinsonism Relat Disord* (2013.0) **19** 522-526. DOI: 10.1016/j.parkreldis.2013.01.016
32. Olanow CW, Rascol O, Hauser R. **A double-blind, delayed-start trial of rasagiline in Parkinson’s disease**. *N Engl J Med* (2009.0) **361** 1268-1278. DOI: 10.1056/NEJMoa0809335
33. Hirsch L, Jette N, Frolkis A. **The incidence of Parkinson’s disease: a systematic review and meta-analysis**. *Neuroepidemiology* (2016.0) **46** 292-300. DOI: 10.1159/000445751
34. 34.Istat (2017) Il diabete in Italia anni 2000–2016. www.Istat.It
35. Camargo Maluf F, Feder D, Alves De Siqueira Carvalho A. **Analysis of the relationship between type II diabetes Mellitus and Parkinson’s disease: a systematic review**. *Parkinson’s Dis* (2019.0) **2019** 1-14. DOI: 10.1155/2019/4951379
36. Heald AH, Stedman M, Davies M. **Estimating life years lost to diabetes: outcomes from analysis of National Diabetes Audit and Office of National Statistics data**. *Cardiovasc Endocrinol Metab* (2020.0) **9** 183-185. DOI: 10.1097/XCE.0000000000000210
37. Sportelli C, Urso D, Jenner P, Chaudhuri KR. **Metformin as a potential neuroprotective agent in prodromal Parkinson’s disease—viewpoint**. *Front Neurol* (2020.0) **11** 1-10. PMID: 32116995
38. Hustad E, Aasly JO. **Clinical and imaging markers of prodromal Parkinson’s disease**. *Front Neurol* (2020.0) **11** 1-11. PMID: 32116995
|
---
title: White matter hyperintensities are an independent predictor of cognitive decline
3 years following first-ever stroke—results from the PROSCIS-B study
authors:
- Huma Fatima Ali
- Lea Fast
- Ahmed Khalil
- Eberhard Siebert
- Thomas Liman
- Matthias Endres
- Kersten Villringer
- Anna Kufner
journal: Journal of Neurology
year: 2022
pmcid: PMC9971076
doi: 10.1007/s00415-022-11481-5
license: CC BY 4.0
---
# White matter hyperintensities are an independent predictor of cognitive decline 3 years following first-ever stroke—results from the PROSCIS-B study
## Abstract
### Background
White matter hyperintensities (WMH) are the result of cerebral small vessel disease and may increase the risk of cognitive impairment (CI), recurrent stroke, and depression. We aimed to explore the association between selected cerebrovascular risk factors (CVRF) and WMH load as well as the effect of increased WMH burden on recurrent vascular events, CI, and depression in first-ever ischemic stroke patients.
### Methods
431 from the PROSpective Cohort with Incident Stroke (PROSCIS) were included; Age-Related White Matter Changes (ARWMC) score was used to assess WMH burden on FLAIR. The presence of CVRF (defined via blood pressure, body-mass-index, and serological markers of kidney dysfunction, diabetes mellitus, and hyperlipoproteinemia) was categorized into normal, borderline, and pathological profiles based on commonly used clinical definitions. The primary outcomes included recurrent vascular events (combined endpoint of recurrent stroke, myocardial infarction and/or death), CI 3 years post-stroke, and depression 1-year post-stroke.
### Results
There was no clear association between CVRF profiles and WMH burden. High WMH lesion load (ARWMC score ≥ 10) was found to be associated with CI (adjusted OR 1.05 [$95\%$ CI 1.00–1.11]; $p \leq 0.02$) in a mixed-model analysis. Kaplan–Meier survival analysis showed a visible increase in the risk of recurrent vascular events following stroke; however, after adjustment, the risk was non-significant (HR 1.5 [$95\%$ CI 0.76–3]; $$p \leq 0.18$$). WMH burden was not associated with depression 1-year post stroke (adjusted OR 0.72 [$95\%$ CI 0.31–1.64]; $$p \leq 0.44$$).
### Conclusion
Higher WMH burden was associated with a significant decline in cognition 3 years post-stroke in this cohort of first-ever stroke patients.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00415-022-11481-5.
## Introduction
Cerebral small vessel disease (CSVD) is one of the most prevalent pathologies that neurologists and radiologists encounter in routine clinical practice. Not only does the presence of CSVD substantially increase the risk of stroke, but it is known to contribute to cognitive decline and dementia [1–3]. The increased availability of magnetic resonance imaging (MRI) in the clinical routine, has substantially increased the detection rates of CSVD; especially as an incidental finding in asymptomatic cases[2,4,5].
Imaging biomarkers of CSVD include small subcortical infarcts, lacunes, white matter hyperintensities (WMH), enlarged perivascular spaces, cerebral microbleeds, and global cerebral atrophy [4]. The presence of WMH are the most common imaging biomarker of CSVD, and are most easily detected on so-called fluid-attenuated inversion recovery (FLAIR) sequences. Although possible underlying pathologies behind WMH are heterogenous, histological studies suggest that WMH are most likely a result of chronic ischemia leading to demyelination and ultimately axonal loss [6–8]. WMH are so prevalent, that nearly $80\%$ of healthy 60-year-olds and approximately $95\%$ of people aged 90 years or older were found to have developed WMH [9,10]. Within the past decades, numerous modifiable risk factors for the development of WMH have been identified, including smoking, diabetes mellitus, arterial hypertension, hyperlipidemia, chronic kidney disease, and the presence of metabolic syndrome [11–13]. While CSVD and ultimately the presence of WMH can remain asymptomatic and is often merely an incidental finding on routine MRI [2], previous studies suggest that progressive WMH likely leads to increased stroke risk and early cognitive decline depending on distribution and localization within the brai n[3].
Interestingly, recent studies suggest that selected cerebrovascular risk factors lead to different WMH distribution patterns within the brain [3]. For example, a cohort study found that patients with chronic hypo- and hypertension present with different patterns and distributions of WMH (periventricular versus deep white matter) on MRI, with a history of hypertension leading to increased periventricular WMH burden [14]. Moreover, smoking has been linked to low cerebral blood flow, leading to increased WMH volumes across all age groups [15]. Furthermore, a recent study reported a strong association between type 2 diabetes with increased whole-brain WMH volume [16]. Most recently, a large population-based study found that obesity and associated low-grade systemic inflammation led to primarily deep WMH burden and comparatively less periventricular WMH volumes [17]. In other words, the presence or absence of selected risk factors is likely to influence WMH distribution patterns and ultimately may contribute to recurrent stroke risk and risk of dementia.
As the individual risk of stroke recurrence varies considerably among patients [18], patient-specific risk stratification is of enormous clinical importance. This is the case especially in younger patients with first-ever stroke, as early detection and initiation of secondary prevention could subsequently reduce recurrent cerebrovascular events. Therefore, we set out to investigate the relationship between latent or 'borderline' risk profiles (e.g., glucose intolerance without manifest diabetes mellitus) of common and well-known cerebrovascular risk factors and WMH burden in a comprehensive cohort of first-ever stroke patients. More specifically, we aimed to investigate the association between serological markers of selected cerebrovascular risk factors (i.e., HbA1c for diabetes mellitus, low-density lipoprotein ([LDL] for hypolipoproteinemia, and glomerular filtration rate [GFR] for kidney function) as well as clinical parameters (body mass index [BMI] and blood pressure) with WMH lesion load using the Age-related white matter changes (ARWMC) score. Subsequently, we investigated whether high WMH burden (defined as an ARWMC score ≥ 10) was associated with increased recurrent vascular events (including recurrent stroke, MI, and/or death) and cognitive impairment 3 years and depression one year following the first-ever stroke in a prospectively collected cohort of ischemic stroke patients.
## Data availability statement
The data that supports the findings of this study are available upon reasonable request from the corresponding author [AK]. Raw imaging data are not publicly available as they contain information that could compromise patient privacy.
## Cohort characteristics
This is a retrospective analysis of the single-center prospective study PROSpective Cohort with Incident Stroke (PROSCIS) conducted at the Center for Stroke Research Berlin, Charité University Hospital. Patients aged ≥ 18 years with first-ever ischemic stroke were recruited between 2010 and 2013 after providing written informed consent. The exclusion criteria included any prior stroke, patients with a brain tumor or brain metastasis, and/or participation in an intervention study. A detailed description of the inclusion and exclusion criteria of the PROSCIS study can be found in the previously published study protocol [19]. The ethics committee approved the study for all recruiting centers in Berlin according to the Declaration of Helsinki and the study was registered in clinicaltrials.org (NTC01363856). For this study, patients with at least one MRI during standard clinical care following the index event (within 7 days following stroke onset) were included in the analysis.
## MRI assessment
The MRI protocol included the following sequences: T2*, diffusion-weighted imaging (DWI), and fluid-attenuated inversion recovery (FLAIR) images. All MRI examinations were conducted on a 3.0 Tesla or 1.5 Tesla Siemens MRI scanner. Post-processing of MRI images was performed offline with MRIcron Software from the Center for Advanced Brain Imaging (University of South Carolina, Chris Rordan, USA).
The ARWMC visual rating scale developed by Wahlund and team [20] was used to rate WMH on FLAIR images of all patients included in this study. Rating was performed independently by two evaluators (neurologist A.K. and senior neuroradiologist K.V.). The ARWMC score ranges from 0 to 30 with an assessment of both sides of the brain and pre-specified regions of the brain, which includes frontal, parieto-occipital, temporal, basal ganglia, and infra-tentorial. Each region’s grading ranges from 0 to 3 where grade 0 is occasional or non-punctate WMH; grade 1, multiple punctate WMH; grade 2, bridging of punctate WMH into confluent lesions; and grade 3, widespread confluent WMH [20]. For this study, ARWMC score was dichotomized with a cut-off of < 10 representing a lower WMH load and ≥ 10 as a higher WMH load [21].
## Clinical assessment
All patients included in the study were examined and interviewed within 7 days after the onset of stroke symptoms. Clinical parameters included basic patient demographics, education status, smoking status, history of diabetes mellitus, hypertension, atrial fibrillation, and hyperlipoproteinemia. Clinical parameters documented on admission (BMI and systolic and diastolic blood pressure) as well as serological parameters (HbA1c, LDL, and GFR) were used in this analysis.
Clinical examinations included stroke severity and functional outcome on admission. Stroke severity was assessed using the National Institutes of Health Stroke Scale (NIHSS), a 15-item scale ranging from 0 to 42, with higher values reflecting a more severe neurological deficit[[[22]]]. Functional outcome was assessed using the modified Rankin Scale (mRS) score, ranging from 0 to 6. mRS is a measure of disability where 0 reflects no symptoms; 5, severe disability leading to nursing aid; and 6, death[[[23]]].
## Cerebrovascular risk factors assessment
Diabetic status was evaluated by analyzing the HbA1c levels, which were categorized into normal as 4.5–$5.7\%$, borderline as 5.8–$6.4\%$, and pathological as ≥ $6.5\%$ [24]. LDL was categorized into normal as ≤ 70 mg/dL; borderline as 71–139 mg/dL; and pathological as ≥ 140 mg/dL. BMI was divided into normal as ≤ 25, borderline as 26–29, and pathological as ≥ 30 [25]. Blood pressure readings were analyzed separately for systolic and diastolic readings. The mean of three separate blood pressure measurements on admission was calculated and then categorized into the following profiles: systolic blood pressure was categorized into hypotension (≤ 120 mmHg), normal (121–139 mmHg) and hypertension (≥ 140 mmHg). Similarly, diastolic blood pressure readings were categorized into hypotension (≤ 80 mmHg), normal (81–89 mmHg), and hypertension (≥ 90 mmHg) [26]. Kidney function was analyzed using the estimated glomerular filtration rate (eGFR), which was categorized into normal as ≥ 90 mL/min; borderline as 61 mL/min to 89 mL/min; and pathological as ≤ 60 mL/min [27]. The definitions and cut-offs of the selected cerebrovascular risk factors included in this analysis are summarized in Supplementary Table 1.
## Primary endpoints
All three primary endpoints were assessed at 1-, 2-, and 3 years post-stroke.
## Cognitive assessment
Cognitive status was evaluated at baseline using the Mini-Mental State Examination (MMSE) [28]. During follow-up, cognitive status was assessed using the Modified Telephone Interview for Cognitive Status (TICS-M), which is a modified variant of the MMSE which has been validated for use via telephone interviews [29,30]. The MMSE is a 30-point questionnaire for cognitive impairment to screen for dementia where a cut-off of ≥ 24 indicates normal cognition and ≤ 23 indicates cognitive impairment. The TICS-M is an 11-test-item questionnaire yielding a total score of 50 points where a cut-off of ≤ 31 represents cognitive impairment and ≥ 32 represents normal cognition.
## Depression assessment
Depression status was assessed using the Center for Epidemiologic Studies Depression Scale (CES-D) [31]. Depression is an inherently difficult endpoint to assess as it can fluctuate for individual patients over time and is likely influenced by many factors including mobility status, stroke severity etc. For this reason, depression scores at only 1-year follow-up were used in the analysis. CES-D is a 20-item questionnaire comprising six scales that reflect significant facets of depression based on self-reported information on depressive mood, feelings of guilt and worthlessness, helplessness and hopelessness, psychomotor retardation, loss of appetite, and sleep disturbance. The cut-off value of ≥ 16 represents clinical depression [32]. To account for missing scores we applied inverse probability-weighted estimation of death as part of a sensitivity analysis.
## Combined endpoint of recurrent vascular events
The combined endpoint of recurrent vascular events included the first of either recurrent stroke, myocardial infarction (MI), or death by any cause. Incidences of recurrent vascular events were assessed using Rose Angina Questionnaire for cardiovascular events and Stroke Symptom Questionnaire for cerebrovascular events [33,34]. Requested measures were collected at baseline and follow-up of 1-, 2-, and 3 years post-stroke. The follow-up was carried out via a structured telephonic questionnaire or mail. In case of any positive vascular outcome event, the information was validated by the admitting hospital or the treating physician.
## Statistical analysis
Baseline characteristics of categorical data is represented in absolute and relative frequencies, and the distribution of continuous data is described as median and interquartile ranges (IQR). The selection of covariates for our analyses was evaluated by Directed Acyclic Graphs (DAGs). DAGs, a derivate of causal diagrams in epidemiology, provide a graphical representation of rigorous mathematical methodology for mapping all a priori assumptions surrounding a causal question [35]. In our study, a separate DAG was created for each causal relationship of interest to identify the potential variables of confounding while avoiding colliders and intermediate pathways using the online tool, Dagitty [http://dagitty.net/] [36]. All DAGs used to inform our analytic strategy have been uploaded onto an open repository and can be accessed via the following link: [https://doi.org/10.6084/m9.figshare.20152487].
Logistic regression (unadjusted and adjusted analysis) was performed to assess the association of borderline vs. pathological cerebrovascular risk factors with dichotomized ARWMC score; interactions between CVRF categories were not considered in the analysis. The models were adjusted based on DAG graphs created for each variable as described above. The model for HbA1c was adjusted for age, sex, smoking, BMI, and hyperlipidemia; models for systolic and diastolic blood pressures were adjusted for age, sex, diabetes, smoking, BMI, and hyperlipidemia; LDL model was adjusted for age, sex, BMI, and smoking; BMI model for age, sex, and smoking; and GFR model for age, sex, diabetes, BMI, hypertension, hyperlipidemia, smoking, and atrial fibrillation.
Kaplan–Meier survival analysis was performed to analyze the association of WMH lesion load (ARWMC score ≥ 10) on the recurrent vascular events within 3 years post-stroke. We report unadjusted and adjusted hazard ratios (HR) with $95\%$ CI obtained from Cox proportional hazards model, and adjustment was made for age, sex, and hypertension.
The effect of WMH lesion load on cognition status (dichotomized at ≤ 23) 3 years post-stroke was analyzed by (non-linear) mixed-model analysis followed by additional sensitivity analysis which excluded all deaths post-stroke within three years. Adjustments were made for age, sex, education status, smoking, hypertension, diabetes, atrial fibrillation, and hyperlipidemia. All models included subject ID as random effects and WMH and cognition scores as fixed effects.
Depression at 1-year post-stroke relative to WMH severity was analyzed by a logistic regression model followed by weighted analysis for probability of death. An adjustment was made for age, sex, CI at baseline, NIHSS, and mRS status at 1-year post-stroke.
For all models, a 2-sided p-value < 0.05 was considered statistically significant. All analyses were performed using the software STATA IC version 15 (StataCorp, College Station, Texas, USA).
## Cohort description
627 patients in total are included in PROSCIS, of which 431 patients presenting with first-ever ischemic stroke with an available MRI following the index event were analyzed. The mean age of this cohort was 66.8 years (standard deviation [SD] 13.2), of which 244 ($38.9\%$) were females, with baseline NIHSS of (median NIHSS 2 [IQR 01–05]. The 196 patients who did not receive an MRI at the baseline had a mean age of 69.6 years [SD 12.42], and median baseline NIHSS of 3 (IQR 2–5). Baseline characteristics of the entire analyzed cohort are described in Table 1.Table 1Baseline clinical and imaging characteristics of the PROSCIS-B cohort. Total patient cohortDemographics Age, mean (± SD)66.9 (± 13.2) Sex, female, n (%)244 ($38.9\%$) Education level ≥ 10 years, n (%)173 ($28.8\%$)Cardiovascular risk factors Hypertension, n (%)409 ($65.2\%$) Diabetes, n (%)138 ($22\%$) Hyperlipidemia, n (%)128 ($22.7\%$) Atrial fibrillation, n (%)135 ($21.5\%$) BMI, Mean (SD)27.5 (4.9)Current smoking, n (%)173 ($28.0\%$)Stroke etiology Large artery atherosclerosis, n (%)169 ($26.9\%$) Cardioembolic stroke, n (%)149 ($23.8\%$) Small vessel occlusion, n (%)96 ($15.3\%$) Other, n (%)22 ($3.5\%$) Unknown, n (%)191 ($30.5\%$)Baseline clinical characteristics NIHSS at admission, Median[IQR]2 [1–5] mRS at admission, Median [IQR]2 [1–5]Infarct pattern Territorial with subcortical and cortical, n (%)110 ($17.5\%$) Subcortical, n (%)112 ($17.9\%$) Scattered infarct, n (%)118 ($18.8\%$) Lacunar, n (%)2 ($0.3\%$) Infratentorial, n (%)107 ($17.1\%$) Watershed, n (%)33 ($5.3\%$) ARWMC score ARWMC score, Median [IQR]5 [3–9]SD standard deviation, IQR interquartile range, BMI body mass index, NIHSS National Institute of Health Stroke Scale, mRS modified Rankin Score, ARWMC Age-Related White Matter Changes
## Cerebrovascular risk factors and WMH
Logistic regression analysis found no significant association across cerebrovascular risk factor categories (normal, borderline, and pathological as defined in Supplementary Table 1), with ARWMC score ≥ 10 (Table 2). For a graphical depiction of ARWMC score distribution (as a continuous variable) across cerebrovascular risk factor categories, please refer to the violin plots depicted in Supplementary Fig. 1.Table 2Logistic regression analysis for effects of borderline and pathological clinical/serological risk profiles and white matter hyperintensities (ARWMC score ≥ 10)n (%)UnadjustedAdjustedOdds ratio$95\%$ Confidence intervalp-valueOdds ratio$95\%$ Confidence intervalp-valueHbA1ca Normal259 ($43.9\%$)Ref Borderline152 ($25.8\%$)1.500.86–2.560.150.990.52–1.910.99 Pathological179 ($30.3\%$)1.210.69–2.090.490.930.49–1.780.84Systolic BPb Normal116 ($19.3\%$)Ref Borderline205 ($34.1\%$)0.810.39–1.660.570.570.25–1.280.17 Pathological280 ($46.6\%$)1.440.75–2.760.261.030.49–2.160.93Diastolic BPb Normal398 ($66.1\%$)Ref Borderline99 ($16.4\%$)1.070.58–1.950.821.210.58–2.490.60 Pathological105 ($17.4\%$)0.910.49–1.710.781.570.74–3.290.23LDLc Normal49 ($8.2\%$)Ref Borderline362 ($61.2\%$)0.830.35–1.960.671.600.55–4.600.38 Pathological180 ($30.5\%$)1.050.43–2.560.912.250.75–6.680.14BMId Normal261 ($42.3\%$)Ref Borderline206 ($33.4\%$)0.960.58–1.600.881.000.57–1.750.99 Pathological150 ($24.3\%$)0.740.40–1.350.330.990.51–1.900.98eGFRe Normal193 ($32.2\%$)Ref Borderline282 ($47.1\%$)2.351.32–4.180.000.910.43–1.920.79 Pathological124 ($20.7\%$)3.321.66–6.640.000.910.35–2.360.84BP blood pressure, LDL low-density lipoprotein, BMI body mass index, GFR glomerular filtration rateaAdjusted for- age, sex, smoking, BMI, hyperlipidemiabAdjusted for- age, sex, diabetes, smoking, BMI, hyperlipidemiacAdjusted for- age, sex, BMI; smokingdAdjusted for- age, sex, smokingeAdjusted for- age, sex, diabetes, BMI, hypertension, hyperlipidemia, smoking, atrial fibrillation Pathological HbA1c values had an odds ratio (OR) of 1.21; ($95\%$ CI 0.69–2.09) for high WMH load. Pathological diastolic blood pressure values had an adjusted OR of 1.57 ($95\%$ CI 0.74–3.29). Neither systolic blood pressure, high LDL, BMI, nor chronic kidney disease defined by GFR showed a significant association with increased WMH load (Table 2). Due to the fact that systolic and diastolic blood pressure values may vary substantially at the time of acute stroke (irrespective as to whether a true diagnosis of hypertension exists), we performed an additional logistic regression analysis including the history of hypertension (yes/no); here we found a significant association with WMH load and history of hypertension with an adjusted OR of 2.44 ($95\%$ CI 1.30–4.57; $$p \leq 0.00$$).
For a visual depiction of the results of the regression analysis (unadjusted and adjusted OR with $95\%$ CI) for all cerebrovascular risk factors and ARWMC score, please refer to Supplementary Fig. 2.
## WMH and recurrent vascular events
In the Kaplan–Meier survival analysis for the combined vascular endpoint (recurrent stroke, MI, and/or death) patients with high ARWMC scores had visibly lower survival rates when analyzed 3 years post-stroke (Fig. 1). In the cox-regression analysis, patients with a ARWMC score ≥ 10 had an unadjusted HR of 1.6 ($95\%$ CI 0.79–3.13; $$p \leq 0.20$$) and adjusted HR of 1.5 ($95\%$ CI 0.76–3.03; $$p \leq 0.18$$) for the combined vascular endpoint 3 years post-stroke. Fig. 1Kaplan–Meier survival analysis curve showing association of white matter hyperintensities with the incidence of recurrent stroke, myocardial infarction, and/or death on follow-up of 1-, 2, and 3-years post-stroke
## WMH and cognitive impairment
A significant association between higher WMH load and cognitive impairment 3 years post-stroke was found in the mixed model analysis with an unadjusted OR of 2.22 ($95\%$ CI 1.20–4.08; $$p \leq 0.01$$) and an adjusted OR of 1.05 ($95\%$ CI 1.00–1.11; $$p \leq 0.02$$). The sensitivity analysis produced similar results. The models were adjusted for age, sex, education status, smoking, hypertension, diabetes, atrial fibrillation, and hyperlipidemia (Table 3).Table 3Mixed model analysis showing the effect of white matter hyperintensities burden with long-term cognitive functionARWMC ScoreUnadjusted OR[$95\%$ CI]p-valueAdjusted ORa[$95\%$ CI]p-valueCognitive impairment2.221.21–4.080.011.051.00–1.110.02Sensitivity analysis model Cognitive impairment1.880.98–3.560.051.051.00–1.110.03aAdjusted for age, sex, education status, smoking, hypertension, diabetes, atrial fibrillation, and hyperlipidemia
## WMH and depression
In logistic regression analysis, high WMH load had an adjusted OR of 0.83 ($95\%$ CI 0.39–1.77; $$p \leq 0.64$$), following adjustment for age, sex, cognitive impairment at baseline, NIHSS, and mRS at 1-year post-stroke. An additional inverse probability weighted analysis also showed no significant association between increased WMH load and depression 1-year post-stroke with an adjusted OR of 0.72 ($95\%$ CI 0.31–1.64; $$p \leq 0.44$$).
## Discussion
The main finding of this study is that increased WMH load (defined as ARWMC score ≥ 10) was significantly associated with cognitive impairment after 3 years in first-ever stroke patients. In addition, a higher WMH load at the time of stroke is likely associated with recurrent vascular events including recurrent stroke, MI, and/or death in patients with first-ever ischemic stroke. One of the primary aims of the current study was to assess whether selected cardiovascular risk profiles affect the WMH burden in first-ever stroke patients. While a history of hypertension was significantly associated with increased WHH load, the current study found no apparent differences in WMH load across so-called borderline and pathological cerebrovascular risk profiles assessed via clinical and serological markers on admission.
Similar to previous studies, we found that a history of arterial hypertension is associated with an increased burden of WMH [9,14]. However, we found no association between baseline systolic or diastolic blood pressure values and increased WMH load in the current analysis. A previously published meta-analysis of four trials has shown that anti-hypertensive medication could significantly slow the progression of WMH [37]. Unfortunately, in the current study, medication status was not available and could not be considered. This may have affected our categorizations and ultimately our results. For example, patients with a long-term history of hypertension and a recent start of anti-hypertensive medication may have normotensive systolic blood pressure at the time of the index event. The same might be true for serological markers of hypercholesterolemia and diabetes mellitus with LDL cholesterol and HbA1c, respectively. An earlier study found a significant association of diabetes mellitus with increased WMH severity, particularly in terms of increased whole-brain WMH volume in a population-based sample of 99 patients with diagnosed diabetes mellitus type 2 [16]. These patients had an average of nearly 9 years of diabetes duration, therefore generalizability of these findings to a cohort of first-ever stroke patients where diagnostic work-up following stroke often leads to first diagnosis of cerebrovascular risk factors is limited. Although previous population-based studies also suggest high BMI to be significantly associated with the development of increased deep WMH lesions [17], here we also failed to find a clear association between high BMI and WMH burden.
While arterial hypertension, diabetes mellitus, hyperlipoproteinemia, obesity, and poor kidney function are all well-known cerebrovascular risk factors and likely do play an important role in the development of WMH [11,14–18,38,39], we did not see relevant associations with our selected risk profiles and higher ARWMC scores in this cohort of first-ever stroke patients. It is important to note that the aforementioned studies are all population-based studies; in other words, patients were selected based on the presence of the risk factor of interest. Therefore, generalizability to a cohort of first-ever stroke patients is likely limited. Furthermore, the current cohort has a relatively mild WMH burden (median ARWMC score of 5 [IQR 3–9]). The visual assessment of WMH lesion load via the ARWMC score may not be sensitive enough to detect associations of risk factors and very early manifestations of ARWMC in first-ever stroke patients. An analysis using a quantitative assessment of WMH volume is certainly warranted to explore this topic further.
The Kaplan–Meier survival analysis for recurrent vascular events showed a visible increase in risk in patients with ARWMC score ≥ 10 across 3 years following stroke (Fig. 1); however, these results were not statistically significant in adjusted cox-regression analysis (HR 1.52 $95\%$ CI 0.76–3.03; $$p \leq 0.18$$). This may be due to the small sample size in the current study. Literature has shown that the risk of stroke recurrence varies considerably among patients and depends on age, sex, and the presence of co-morbidities [40,41]. In line with our observation, a recent Danish Stroke Registry-based observational study including 832 patients reported a significant association between WMH and increased risk of recurrent stroke (HR 5.28; $95\%$ CI [1.98–14.07]) [42] in a cohort of incident ischemic stroke patients.
We found a significant association of increased WMH lesion load with CI 3 years following first-ever stroke. Additional sensitivity analysis also yielded a similar result (adjusted OR of 1.1 $95\%$ CI 1.0–1.1; $$p \leq 0.03$$). Previous studies on the effect of WMH on cognitive decline have yielded somewhat controversial results; while several studies found no clear association between WMH and cognitive decline [43,44], others have found an increase in WMH volume to be an independent predictor of CI in both stroke cohorts as well as population-based studies [18,45]. Recently, a study reported that increased WMH volume was associated with cognitive decline in patients < 80 years of age following a minor stroke or TIA [44]. Since WMH are very commonly seen on MRI even in healthy subjects, the interpretation of these as a proxy for risk of CI can be complex and should still be reported and applied with caution. However, there is increasing evidence for a causal relationship between increased WMH load and the risk of the development of cognitive decline which could have important clinical implications i.e., early identification of an increased risk of CI could play a major role in initiating secondary prevention strategies in patients at risk.
We found no association between high WMH load and depression post-stroke in this analysis. There are conflicting results reported in the literature on the effect of WMHs on the development of depression. Although several population-based studies have found that WMH progression is associated with depression and antidepressants can even slow WMH progression [46], the effects of WMH burden in stroke patients on post-stroke depression remains unclear. As of yet, different WMH patterns such as peri-ventricular WMH and deep WMH are found to affect depression at baseline or a yearlong follow-up in stroke patients, respectively [47]. Depression is an inherently difficult endpoint to assess in stroke patients, as depression can fluctuate over time for an individual patient and is affected by many factors (medication, functional status post-stroke, and co-morbidities), for which we could not account for in the current analysis. Nonetheless, very few studies have explored the association between WMH and depression in stroke patients, and we recommend more comprehensive studies in independent cohorts to further explore this interesting and clinically highly relevant topic.
To the best of our knowledge, this is the first study to analyze the association of selected clinical and serological cardiovascular risk profiles (borderline vs. pathological) with WMH burden in a comprehensive cohort of first-ever stroke patients. However, our study has several limitations that warrant discussion. First and foremost, this is a retrospective, exploratory analysis of a prospective cohort that was not designed to address our primary research questions, therefore we did not adjust for multiple testing. Furthermore, data on patients’ pre-stroke medication was not available for this cohort, therefore we could not account for the possible influence of pre-stroke medications on baseline clinical and serological risk profiles like blood pressure or HbA1c. This may partially explain why we failed to see a clear association between selected risk profiles and WMH burden. Similarly, we cannot exclude the possible presence of pre-stroke cognitive impairment in this cohort; the use of the mixed-model analysis for cognitive impairment in this analysis only partially compensates for this limitation. Additionally, we could only include patients within PROSCIS that received an MRI, which introduces a potential bias into our cohort because patients with contraindications for an MRI were automatically excluded from the current analyses. This may be reflected by the relatively mild strokes of the patients included in our analysis (median NIHSS at the admission of 2 [IQR 1–5]). Therefore, generalizability to more severely affected stroke cohorts is limited.
A fundamental limitation of this study is that we applied a qualitative visual method for assessing WMH burden (i.e., the ARWMC score). This did not allow us to determine the WMH burden in more detail, i.e., total WMH volume or anatomical distribution (peri-ventricular versus deep white matter). ARWMC score may not be sensitive enough to detect very early manifestations of ARWMC and a ceiling effect at higher scores is to be expected. However, the ARWMC score is a well-known and well-established scoring system for assessing WMH burden in the clinical setting. The advantage of applying the ARWMC score as done in this study is that it increases the clinical applicability of the results. Nonetheless, a more detailed analysis of WMH distribution pattern and volume in terms of cerebrovascular risk profiles and long-term outcomes is undoubtedly warranted to increase our understanding of the prognostic value of early WMH progression. Finally, we can not rule out entirely the possibility that some patients included in the current analysis had other less common causes of WMH – for example, amyloid angiopathy, previous radiation exposure, or genetic disorders. The further assessment of additional imaging biomarkers of CSVD (lacunes, microbleeds, cerebral atrophy, and enlarged perivascular spaces) was beyond the scope of the current study but would also be essential to fully comprehend the causal effect of CSVD on recurrent vascular events, cognitive decline, and depression in first-ever stroke patients with CSVD.
## Conclusion
In summary, our study found no clear associations between selected cerebrovascular risk profiles and WMH burden; this may be due to the small sample size and the methodology applied (i.e., ARWMC score vs. quantitative assessment of WMH including volume and anatomical distribution). However, our study supports the hypothesis that WMH burden leads to a significantly higher risk of cognitive decline and possibly even recurrent vascular events following the first-ever stroke which may allow us to identify patients at risk and initiate early patient-individualized preventative treatment strategies if validated in further studies.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 156 KB)
## References
1. Fisher CM. **Lacunar strokes and infarcts: a review**. *Neurology* (1982) **32** 871-871. DOI: 10.1212/WNL.32.8.871
2. Cannistraro RJ, Badi M, Eidelman BH, Dickson DW, Middlebrooks EH, Meschia JF. **CNS small vessel disease: a clinical review**. *Neurology* (2019) **92** 1146-1156. DOI: 10.1212/WNL.0000000000007654
3. Prins ND, Scheltens P. **White matter hyperintensities, cognitive impairment and dementia: an update**. *Nat Rev Neurol* (2015) **11** 157-165. DOI: 10.1038/nrneurol.2015.10
4. Wardlaw JM, Smith EE, Biessels GJ. **Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration**. *Lancet Neurol* (2013) **12** 822-838. DOI: 10.1016/S1474-4422(13)70124-8
5. Wardlaw JM, Valdés Hernández MC, Muñoz-Maniega S. **What are white matter hyperintensities made of?: relevance to vascular cognitive impairment**. *JAHA* (2015) **4** e001140. DOI: 10.1161/JAHA.114.001140
6. Li Q, Yang Y, Reis C. **Cerebral small vessel disease**. *Cell Transplant* (2018) **27** 1711-1722. DOI: 10.1177/0963689718795148
7. Pantoni L. **Cerebral small vessel disease: from pathogenesis and clinical characteristics to therapeutic challenges**. *Lancet Neurol* (2010) **9** 689-701. DOI: 10.1016/S1474-4422(10)70104-6
8. Launer LJ, Oudkerk M, Nilsson LG. **CASCADE: a European collaborative study on vascular determinants of brain lesions**. *Neuroepidemiology* (2000) **19** 113-120. DOI: 10.1159/000026246
9. de Leeuw FE, de Groot JC, Achten E. **Prevalence of cerebral white matter lesions in elderly people: a population based magnetic resonance imaging study. The Rotterdam Scan Study**. *J Neurol Neurosurg Psychiatry* (2001) **70** 9-14. DOI: 10.1136/jnnp.70.1.9
10. Smith EE, O’Donnell M, Dagenais G. **Early cerebral small vessel disease and brain volume, cognition, and gait**. *Ann Neurol* (2015) **77** 251-261. DOI: 10.1002/ana.24320
11. Hakim AM. **Small vessel disease**. *Front Neurol* (2019) **10** 1020. DOI: 10.3389/fneur.2019.01020
12. Vermeer SE, Longstreth WT, Koudstaal PJ. **Silent brain infarcts: a systematic review**. *Lancet Neurol* (2007) **6** 611-619. DOI: 10.1016/S1474-4422(07)70170-9
13. Wang M, Norman JE, Srinivasan VJ, Rutledge JC. **Metabolic, inflammatory, and microvascular determinants of white matter disease and cognitive decline**. *Amn J Neurodegener Dis* (2016) **5** 171-177
14. Kim JS, Lee S, Suh SW. **Association of low blood pressure with white matter hyperintensities in elderly individuals with controlled hypertension**. *J Stroke* (2020) **22** 99-107. DOI: 10.5853/jos.2019.01844
15. Kim SH, Yun CH, Lee SY, Choi KH, Kim MB, Park HK. **Age-dependent association between cigarette smoking on white matter hyperintensities**. *Neurol Sci* (2012) **33** 45-51. DOI: 10.1007/s10072-011-0617-1
16. Jongen C, van der Grond J. **Automated measurement of brain and white matter lesion volume in type 2 diabetes mellitus**. *Diabetologia* (2007) **50** 1509-1516. DOI: 10.1007/s00125-007-0688-y
17. Lampe L, Zhang R, Beyer F. **Visceral obesity relates to deep white matter hyperintensities via inflammation**. *Ann Neurol* (2019) **85** 194-203. DOI: 10.1002/ana.25396
18. Debette S, Markus HS. **The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis**. *BMJ* (2010) **341** c3666. DOI: 10.1136/bmj.c3666
19. Liman TG, Zietemann V, Wiedmann S. **Prediction of vascular risk after stroke—protocol and pilot data of the prospective cohort with incident stroke (PROSCIS)**. *Int J Stroke* (2013) **8** 484-490. DOI: 10.1111/j.1747-4949.2012.00871.x
20. Wahlund LO, Barkhof F, Fazekas F. **A new rating scale for age-related white matter changes applicable to MRI and CT**. *Stroke* (2001) **32** 1318-1322. DOI: 10.1161/01.STR.32.6.1318
21. Leonards CO, Ipsen N, Malzahn U, Fiebach JB, Endres M, Ebinger M. **White matter lesion severity in mild acute ischemic stroke patients and functional outcome after 1 year**. *Stroke* (2012) **43** 3046-3051. DOI: 10.1161/STROKEAHA.111.646554
22. Wijnhoud AD, Maasland L, Lingsma HF, Steyerberg EW, Koudstaal PJ, Dippel DWJ. **Prediction of major vascular events in patients with transient ischemic attack or ischemic stroke: a comparison of 7 models**. *Stroke* (2010) **41** 2178-2185. DOI: 10.1161/STROKEAHA.110.580985
23. van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJA, van Gijn J. **Interobserver agreement for the assessment of handicap in stroke patients**. *Stroke* (1988) **19** 604-607. DOI: 10.1161/01.str.19.05.604
24. **Diagnosis and classification of diabetes mellitus**. *Diabetes Care* (2010) **33** S62-S69. DOI: 10.2337/dc10-S062
25. Nuttall FQ. **Body mass index: obesity, BMI, and health a critical review**. *Nutr Today* (2015) **50** 117-128. DOI: 10.1097/NT.0000000000000092
26. de la Sierra A. **New American and european hypertension guidelines, reconciling the differences**. *Cardiol Ther* (2019) **8** 157-166. DOI: 10.1007/s40119-019-0144-3
27. Chen TK, Knicely DH, Grams ME. **Chronic kidney disease diagnosis and management: a review**. *JAMA* (2019) **322** 1294. DOI: 10.1001/jama.2019.14745
28. Godefroy O, Fickl A, Roussel M. **Is the Montreal cognitive assessment superior to the mini-mental state examination to detect poststroke cognitive impairment?: a study with neuropsychological evaluation**. *Stroke* (2011) **42** 1712-1716. DOI: 10.1161/STROKEAHA.110.606277
29. Barber M, Stott DJ. **Validity of the telephone interview for cognitive status (TICS) in post-stroke subjects**. *Int J Geriat Psychiatry* (2004) **19** 75-79. DOI: 10.1002/gps.1041
30. Fong TG, Fearing MA, Jones RN. **Telephone interview for cognitive status: creating a crosswalk with the mini-mental state examination**. *Alzheimer’s Dementia* (2009) **5** 492-497. DOI: 10.1016/j.jalz.2009.02.007
31. Radloff LS. **The CES-D scale: a self-report depression scale for research in the general population**. *Appl Psychol Meas* (1977) **1** 385-401. DOI: 10.1177/014662167700100306
32. Parikh RM, Eden DT, Price TR, Robinson RG. **The sensitivity and specificity of the center for epidemiologic studies depression scale in screening for post-stroke depression**. *Int J Psychiatry Med* (1989) **18** 169-181. DOI: 10.2190/BH75-EUYA-4FM1-J7QA
33. Cook DG, Shaper AG, Macfarlane PW. **Using the WHO (Rose) angina questionnaire in cardiovascular epidemiology**. *Int J Epidemiol* (1989) **18** 607-613. DOI: 10.1093/ije/18.3.607
34. Berger K, Hense HW, Rothdach A, Weltermann B, Keil U. **A single question about prior stroke versus a stroke questionnaire to assess stroke prevalence in populations**. *Neuroepidemiology* (2000) **19** 245-257. DOI: 10.1159/000026262
35. Shrier I, Platt RW. **Reducing bias through directed acyclic graphs**. *BMC Med Res Methodol* (2008) **8** 70. DOI: 10.1186/1471-2288-8-70
36. Textor J, van der Zander B, Gilthorpe MS, Liśkiewicz M, Ellison GTH. **Robust causal inference using directed acyclic graphs: the R package ‘dagitty’**. *Int J Epidemiol* (2017). DOI: 10.1093/ije/dyw341
37. van Middelaar T, Argillander TE, Schreuder FHBM, Deinum J, Richard E, Klijn CJM. **Effect of antihypertensive medication on cerebral small vessel disease: a systematic review and meta-analysis**. *Stroke* (2018) **49** 1531-1533. DOI: 10.1161/STROKEAHA.118.021160
38. Zong L, Yao M, Ni J. **Kidney function is associated with severity of white matter hyperintensity in patients with acute ischemic stroke/TIA**. *BMC Neurol* (2016) **16** 193. DOI: 10.1186/s12883-016-0714-0
39. Kim SH, Yun JM, Jeong SM. **Kidney dysfunction impact on white matter hyperintensity volume in neurologically healthy adults**. *Sci Rep* (2019) **9** 8596. DOI: 10.1038/s41598-019-45109-y
40. Yu L, Yang L, Zhang X. **Age and recurrent stroke are related to the severity of white matter hyperintensities in lacunar infarction patients with diabetes**. *CIA* (2018) **13** 2487-2494. DOI: 10.2147/CIA.S184463
41. Putaala J, Haapaniemi E, Kurkinen M, Salonen O, Kaste M, Tatlisumak T. **Silent brain infarcts, leukoaraiosis, and long-term prognosis in young ischemic stroke patients**. *Neurology* (2011) **76** 1742-1749. DOI: 10.1212/WNL.0b013e31821a44ad
42. Andersen SD, Larsen TB, Gorst-Rasmussen A, Yavarian Y, Lip GYH, Bach FW. **White matter hyperintensities improve ischemic stroke recurrence prediction**. *Cerebrovasc Dis* (2017) **43** 17-24. DOI: 10.1159/000450962
43. Vermeer SE, Prins ND, den Heijer T, Hofman A, Koudstaal PJ, Breteler MMB. **Silent brain infarcts and the risk of dementia and cognitive decline**. *N Engl J Med* (2003) **348** 1215-1222. DOI: 10.1056/NEJMoa022066
44. Zamboni G, Griffanti L, Mazzucco S, Pendlebury ST, Rothwell PM. **Age-dependent association of white matter abnormality with cognition after TIA or minor stroke**. *Neurology* (2019) **93** e272-e282. DOI: 10.1212/WNL.0000000000007772
45. Liman TG, Heuschmann PU, Endres M, Flöel A, Schwab S, Kolominsky-Rabas PL. **Changes in cognitive function over 3 years after first-ever stroke and predictors of cognitive impairment and long-term cognitive stability: the Erlangen Stroke Project**. *Dement Geriatr Cogn Disord* (2011) **31** 291-299. DOI: 10.1159/000327358
46. Khalaf A, Edelman K, Tudorascu D, Andreescu C, Reynolds CF, Aizenstein H. **White matter hyperintensity accumulation during treatment of late-life depression**. *Neuropsychopharmacol* (2015) **40** 3027-3035. DOI: 10.1038/npp.2015.158
47. Bae K, Kang H, Kim J. **Associations of white matter hyperintensities with poststroke depression: a 1-year longitudinal study**. *Int J Geriatr Psychiatry* (2019) **34** 162-168. DOI: 10.1002/gps.5005
|
---
title: 'Self-reported causes of cancer among 6881 survivors with 6 tumour types: results
from the PROFILES registry'
authors:
- Carla Vlooswijk
- Olga Husson
- Simone Oerlemans
- Nicole Ezendam
- Dounya Schoormans
- Belle de Rooij
- Floortje Mols
journal: Journal of Cancer Survivorship
year: 2021
pmcid: PMC9971112
doi: 10.1007/s11764-021-00989-w
license: CC BY 4.0
---
# Self-reported causes of cancer among 6881 survivors with 6 tumour types: results from the PROFILES registry
## Abstract
### Objective
Our aim was to describe and compare self-reported causal attributions (interpretations of what caused an illness) among cancer survivors and to assess which sociodemographic and clinical characteristics are associated with them.
### Methods
Data from five population-based PROFILES registry samples (i.e. lymphoma ($$n = 993$$), multiple myeloma ($$n = 156$$), colorectal ($$n = 3989$$), thyroid ($$n = 306$$), endometrial ($$n = 741$$), prostate cancer ($$n = 696$$)) were used. Causal attributions were assessed with a single question.
### Results
The five most often reported causal attributions combined were unknown ($21\%$), lifestyle ($19\%$), biological ($16\%$), other ($14\%$), and stress ($12\%$). Lymphoma ($49\%$), multiple myeloma ($64\%$), thyroid ($55\%$), and prostate ($64\%$) cancer patients mentioned fixed causes far more often than modifiable or modifiable/fixed. Colorectal ($33\%$, $34\%$, and $33\%$) and endometrial ($38\%$, $32\%$, and $30\%$) cancer survivors mentioned causes that were fixed, modifiable, or both almost equally often. Colorectal, endometrial, and prostate cancer survivors reported internal causes most often, whereas multiple myeloma survivors more often reported external causes, while lymphoma and thyroid cancer survivors had almost similar rates of internal and external causes. Females, those older, those treated with hormonal therapy, and those diagnosed with prostate cancer were less likely to identify modifiable causes while those diagnosed with stage 2, singles, with ≥2 comorbid conditions, and those with endometrial cancer were more likely to identify modifiable causes.
### Conclusion
In conclusion, this study showed that patients report both internal and external causes of their illness and both fixed and modifiable causes. This differsbetween the various cancer types.
### Implications for Cancer Survivors
Although the exact cause of cancer in individual patients is often unknown, having a well-informed perception of the modifiable causes of one’s cancer is valuable since it can possibly help survivors with making behavioural adjustments in cases where this is necessary or possible.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s11764-021-00989-w.
## Introduction
Although there are known risk factors for certain types of cancer, the causes of cancer in an individual patient are often unknown. This can cause cancer patients to develop their own theories about the cause of their cancer. One possible motive for forming these theories is to make sense of one’s circumstances. In psychology, the process in which individuals use common sense theories to attribute causes to certain events, in an attempt to understand them, is known as attribution theory [1]. Causal attributions are interpretations of what caused the illness [2]. This is part of the common sense self-regulation model of illness, in which patients respond to their symptoms and signs of illness by forming cognitive representations (beliefs about the illness) and emotional reactions of the illness, known as illness perceptions, that lead to coping responses [3, 4].
*In* general, causal attributions are determined by the extent to which someone sees the cause of their disease as either internal or external, and modifiable or fixed [5]. Patients who attribute causal disease perceptions to external (e.g. environment, chance, or a prior health condition) or fixed (e.g. biological, psychological) causes are more likely to experience less control over their condition and its treatment compared to patients who perceive the cause of their disease as internal (e.g. lifestyle, psychological) or modifiable (e.g. lifestyle, stress). Feeling in control over one’s illness might lead to adjustment of health behaviour [6].
A previous study examining causal attributions among American cancer survivors ($$n = 775$$) of the 10 most common cancers showed that the most common causal attributions were lifestyle (modifiable), biological (fixed), and environmental (fixed) factors [7]. Cancer type was the only characteristic associated with identifying modifiable causes out of an extensive list of sociodemographic, clinical, and psychosocial characteristics. Therefore, the authors urged the need for additional research in larger populations in order to determine whether other characteristics are associated with modifiable attributions. Therefore, our aim was to study this topic in a larger European population–based sample including both solid and non-solid cancers. More specifically, our goal was to [1] describe self-reported causal attributions of cancer survivors with various cancers in a large ($$n = 6881$$) Dutch population–based sample and [2] assess which sociodemographic and clinical characteristics are associated with modifiable causal attributions. A clear picture on the modifiable causal attributions cancer survivors have on their disease can teach us whether improvements in information provision regarding this topic are necessary. Being well-informed on the probable modifiable cause of one’s illness can possibly help survivors with making behavioural adjustments in cases where this is necessary or possible. In addition, being well-informed about a fixed cause is important since patients then know that they had no influence on it. Therefore, we performed secondary data analyses on a pooled cohort of existing studies with a similar design among survivors of lymphoma, multiple myeloma, colorectal, thyroid, endometrial, and prostate cancer.
## Study design and setting
Data from the PROFILES (‘Patient-Reported Outcomes Following Initial Treatment and Long-term Evaluation of Survivorship’) registry were used for secondary analyses [8]. PROFILES is a registry that facilitates studies examining the physical and psychosocial impact of cancer as well as its treatment. PROFILES includes an extensive web-based component and is combined with the clinical data from the Netherlands Cancer Registry (NCR). The PROFILES registry started its first cohort of cancer survivors in 2008 and is still ongoing, including studies on various cancer types.
## Study population
The current analysis includes six study samples from the PROFILES registry, including patients with lymphoma (including chronic lymphocytic leukaemia (CLL)) and multiple myeloma, colorectal, thyroid, endometrial, and prostate cancer [9–16]. Patients were included between May 2008 and May 2013. In all study samples, participants were included if they were older than 18 years at diagnosis and excluded if they were not able to complete a Dutch questionnaire according to their (former) attending specialist (i.e. severe cognitive impairment, non-native speaker, too ill to participate). Ethical approval was obtained for all study samples separately from the local Dutch certified medical ethics committee Maxima Medical Center (CLL and multiple myeloma, #0734; colorectal, #0822; thyroid, this study was reviewed by the Institutional Review Board as deemed non-human subjects research; endometrial #0822 and #NL33429.008.10; and prostate cancer, #0733).
## Data collection
A detailed description of the data collection procedure has been described previously [8]. In brief, in each study, cancer patients were informed about the study via a letter by their (former) attending specialist. This letter contained either an informed consent form and a paper questionnaire, or a secured link to a web-based informed consent form and online questionnaire. In study samples where the secured link was provided, the patient could return a postcard to request a paper-and-pencil questionnaire, if preferred. All participants included informed consent.
## Sociodemographic and clinical data
Sociodemographic (i.e. date of birth and sex) and clinical (i.e. cancer type, disease stage, primary treatments received, and date of diagnosis) data were obtained from the NCR. Cancer type was classified according to the third International Classification of Diseases for Oncology [17], or cancer stage was classified according to TNM [18] or Ann Arbor Code (*Hodgkin lymphoma* and non-Hodgkin lymphoma). TNM 5 was used for patients diagnosed from 2002 to 2003, TNM 6 for patients diagnosed from 2003 to 2010, and TNM 7 for patients diagnosed from 2010. For chronic lymphocytic leukaemia and multiple myeloma, stage was not determined nor registered. Primary treatments received were classified into surgery, systemic therapy (chemotherapy, targeted therapy, and immune therapy), hormonal treatment, radiotherapy, and active surveillance/no treatment. Information on educational level (low/middle/high) and marital status (partner/divorced/widowed/alone) were collected in the questionnaire. Survivors were also asked to identify comorbid conditions present in the past 12 months. Comorbidity was classified according to the adapted Self-Administered Comorbidity Questionnaire (SCQ) [19] and categorized into no comorbidity, 1, or 2 or more comorbid conditions.
## Causal attribution
Causal attribution was assessed with one open-ended item taken from the Dutch version of the Brief Illness Perceptions Questionnaire (BIPQ), an instrument used to assess illness perceptions [20]. We only used the open-ended item that assesses causal beliefs whereby survivors are asked to list the three most important causes for their illness. Our analyses were based upon all three listed causal beliefs combined. Participants’ written responses were coded based on a list of causal attributions derived from the literature [7]. Two authors (CV and OH) coded the listed causes, discussed the coding, and resolved doubtful cases. The responses were condensed into 11 broad categories: lifestyle, biological, environmental, chance/luck, stress, existential, prior health condition, psychological, other, unknown, and missing [7]. The category ‘other’ was chosen if patients wrote down something unclear, unusable, or if they did not understand the question; ‘unknown’ meant that patients themselves indicated that they did not know the cause; and ‘missing’ was specified if they did not answer the question. Each causal attribution was categorized as (a) being either internal (lifestyle, stress, biological, psychological) or external (environmental, chance/luck, existential, prior health condition) to the individual and (b) modifiable (lifestyle and stress) or not modifiable/fixed (biological, environmental, change/luck, existential, prior health condition, and psychological) by an individual (Supplemental Table 1).
## Statistical analyses
Simple descriptive analyses were performed to describe the sociodemographic and clinical characteristics of the sample and to determine the most common causal attributions for the total sample and by cancer type. All variables were described as percentages for categorical data or means and standard deviations for continuous data. Also, the percentage patients who reported only internal, only external, or both internal and external causal attributions and the percentage of patients who reported modifiable, fixed, or both modifiable and fixed causal attributions according to tumour type were shown.
Restricting the total sample to those who identified only fixed or only modifiable causes of their cancer (i.e. excluding those who listed both), we assessed the unadjusted association between demographic and clinical characteristics (including cancer type) and identifying modifiable causal attributions using univariate logistic regression analyses. Backward elimination was used in the multivariate logistic regression; all variables were entered in the model and removed once at a time until all variables in the model were significant. Variables significant at the level of <0.05 were retained in the final multivariate model gender, age at questionnaire, hormonal treatment, stage, marital status, comorbidities, and tumour type. We repeated these analyses for those listing both fixed and modifiable causes of their illness as a sensitivity analyses. Power issues prevented us from performing these analyses separately for each tumour type. Statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, 1999) and two-sided P values of <0.05 were considered statistically significant.
## Sociodemographic and clinical characteristics
In total, data of 6881 patients were used (colorectal, $$n = 3989$$; lymphoma, $$n = 993$$; multiple myeloma, $$n = 156$$; thyroid, $$n = 306$$; endometrial, $$n = 741$$; and prostate, $$n = 696$$). Over half of them ($54\%$) were male, mean age was 68, $77\%$ had a partner, $61\%$ had a medium educational level, $45\%$ had 2 or more comorbid conditions, and $50\%$ were diagnosed >3 years ago (Table 1).Table 1Sociodemographic and clinical data according to tumour typeTotal ($$n = 6881$$), n (%)Colorectal ($$n = 3989$$), n (%)Lymphoma ($$n = 993$$), n (%)*Multiple myeloma* ($$n = 156$$), n (%)Thyroid ($$n = 306$$), n (%)Endometrial ($$n = 741$$), n (%)Prostate ($$n = 696$$), n (%)Gender Male3683 [54]2220 [56]608 [61]84 [54]75 [25]0 [0]696 [100] Female3198 [46]1769 [44]385 [39]72 [46]231 [75]741 [100]0 [0]Mean age at questionnaire (SD)68 [11]69 [10]61 [14]67 [10]56 [15]67 [9]71 [7]Age at questionnaire < 55 years841 [12]312 [8]301 [30]20 [13]146 [48]55 [7]7 [1] 55–64 years1677 [24]926 [23]238 [24]42 [27]77 [25]260 [35]134 [19] 65–74 years2452 [36]1455 [36]264 [27]59 [38]44 [14]299 [40]331 [48] 75+ years1911 [28]1296 [32]190 [19]35 [22]39 [13]127 [17]224 [32]Primary treatment Surgery5186 [75]3946 [99]0 [0]0 [0]302 [99]741 [100]197 [28] Systemic treatment1999 [29]1193 [30]674 [68]124 [79]0 [0]8 [1]0 [0] Hormonal treatment218 [3]3 [0]0 [0]0 [0]6 [2]0 [0]209 [30] Radiotherapy2109 [31]1094 [27]295 [30]59 [38]221[72]167 [23]273 [39] None/active surveillance367 [5]4 [0]210 [21]21 [13]2 [0]0 [0]130 [19]Years since diagnose (mean ± SD)5 [3]5 [3]4 [3]3 [2]10 [5]5 [2]4 [1] < 2 years911 [13]459 [12]240 [24]68 [44]5 [2]122 [16]17 [2] 2–3 years2526 [37]1614 [40]257 [26]46 [29]56 [18]227 [31]326 [47] > 3 years3444 [50]1916 [48]496 [50]42 [27]245 [80]392 [53]353 [51]Stage I2163 [31]1072 [27]231 [23]NA172 [56]686 [93]2 [0] II2336 [34]1515 [38]210 [21]NA59 [19]55 [7]497 [71] III1504 [22]1176 [30]148 [15]NA48 [16]0 [0]132 [19] IV491 [7]187 [5]219 [22]NA20 [7]0 [0]65 [9] Missing387 [6]39 [1]185 [19]bNA7 [2]0 [0]0 [0]*Marital status* Partner5187 [77]2963 [76]772 [79]119 [77]238 [78]515 [72]580 [85] Divorced/widowed/alone1574 [23]955 [24]203 [28]35 [23]68 [22]203 [28]110 [16]*Education levela* Low1181 [18]793 [20]142 [15]27 [17]33 [11]71 [10]115 [17] Middle4087 [61]2341 [60]586 [60]97 [63]192 [63]467 [66]404 [59] High1453 [22]760 [20]241 [25]31 [20]80 [26]174 [24]167 [24]Comorbidities 01592 [25]911 [25]292 [32]27 [19]75 [25]122 [18]165 [25] 11922 [30]1078 [29]292 [32]44 [31]92 [31]198 [29]218 [33] 2 or more2899 [45]1713 [46]338 [37]69 [49]114 [39]362 [53]284 [43]NA, not applicableaEducation levels included the following categories: low=no/primary school, middle=lower general secondary education/vocational training, or high=pre-university education/high vocational training/universitybIncluding patients with chronic lymphatic leukaemia
## Most common causal attributions
The 5 most often reported causal attributions were unknown ($21\%$), lifestyle ($19\%$), biological ($16\%$), other ($14\%$), and stress ($12\%$) (Table 2). Those with colorectal cancer reported lifestyle-related factors ($24\%$), biological factors ($18\%$), unknown ($17\%$), stress ($13\%$), and other ($11\%$) most often. Those with lymphoma most often mentioned unknown ($28\%$), other ($18\%$), stress ($13\%$), lifestyle ($12\%$), or chance/luck ($10\%$) as causes of their cancer. Multiple myeloma patients also reported unknown reasons ($29\%$), other ($21\%$), chance/luck ($11\%$), environmental ($11\%$), or biological ($10\%$) as causes. Furthermore, those with thyroid cancer reported unknown ($29\%$), other ($23\%$), biological ($14\%$), chance/luck ($13\%$), and stress ($12\%$) as factors to be the cause of their cancer. Endometrial cancer survivors mentioned lifestyle ($16\%$), unknown ($16\%$), other ($13\%$), biological ($12\%$), and stress ($10\%$) while prostate cancer survivors reported unknown ($30\%$), other ($21\%$), biological ($20\%$), lifestyle ($10\%$), or chance/luck ($7\%$) as the most common causes of their cancer. Table 2The 3 most important causes of cancer reported by survivors according to tumour type, n(%)Total ($$n = 6$.881$)Colorectal ($$n = 3989$$)Lymphoma ($$n = 993$$)*Multiple myeloma* ($$n = 156$$)Thyroid ($$n = 306$$)Endometrial ($$n = 741$$)Prostate ($$n = 696$$)Lifestyle1,31311 [19]958 [24]123 [12]13 [8]32 [10]115 [16]70 [10] Alcohol drinking100 [1]83 [2]7 [1]0 [0]0 [0]0 [0]10 [1] Smoking/tobacco219 [3]160 [4]26 [3]1 [1]7 [2]7 [1]18 [3] Delay in health care148 [2]107 [3]13 [1]4 [3]3 [1]14 [2]7 [1] Diet596 [9]505 [13]42 [4]6 [4]9 [3]14 [2]20 [3] Use of hormones38 [1]1 [0]3 [0]0 [0]2 [1]29 [4]3 [0] Reproductive history54 [1]2 [0]3 [0]0 [0]1 [0]41 [6]7 [1] (Multiple) harmful behaviour(s)292 [4]221 [6]31 [3]3 [2]7 [2]8 [1]22 [3] Sun exposure3 [0]2 [0]1 [0]0 [0]0 [0]0 [0]0 [0] Less activity113 [2]106 [3]4 [0]0 [0]1 [0]0 [0]2 [0] Work90 [1]59 [1]24 [2]1 [1]3 [1]3 [0]0 [0] Being overweight/obese82[1]57 [1]6 [1]0 [0]2 [1]16 [2]1 [0]Biological1,41084 [16]699 [18]94 [9]16 [10]43 [14]90 [12]142 [20] Age157 [2]71 [2]8 [1]3 [2]2 [1]10 [1]63 [9] Heredity/genetics979 [14]657 [16]86 [9]14 [9]41 [13]82 [11]99 [14]Environmental2,4343 [5]160 [4]85 [9]17 [11]38 [12]16 [2]27 [4] Air pollution68 [1]47 [1]10 [1]2 [1]3 [1]4 [1]2 [0] Asbestos11 [0]4 [0]4 [0]2 [1]0 [0]0 [0]1 [0] Environment151 [2]61 [2]40 [4]8 [5]25 [8]7 [1]10 [1] Household chemicals8 [0]3 [0]3 [0]0 [0]1 [0]1 [0]0 [0] Occupational hazards72 [1]28 [1]23 [2]6 [4]5 [2]2 [0]8 [1] Second-hand smoke12 [0]7 [0]1 [0]0 [0]2 [1]0 [0]2 [0] Toxins46 [1]20 [1]12 [1]2 [1]3 [1]2 [0]7 [1] Health care radiation19 [0]5 [0]3 [0]2 [1]7 [2]1 [0]1 [0]Chance/luck2,4455 [7]213 [5]100 [10]17 [11]40 [13]38 [5]47 [7]Stress1,3815 [12]534 [13]128 [13]13 [8]36 [12]75 [10]29 [4]Existential2,419 [0]7 [0]8 [1]0 [0]1 [0]2 [0]1 [0]Prior health condition2,4393 [6]219 [5]72 [7]14 [9]17 [6]50 [7]21 [3] Infection/bacteria/virus27 [0]5 [0]14 [1]1 [1]0 [0]4 [1]3 [0] Previous medical condition292 [4]182 [5]45 [5]10 [6]11 [0]29 [4]15 [2] Trauma injury4 [0]2 [0]0 [0]1 [1]1 [0]0 [0]0 [0] Medication45 [1]22 [1]9 [1]1 [1]2 [1]9 [1]2 [0] Previous cancer37 [1]12 [0]10 [1]1 [1]1 [1]10 [1]3 [4] Medical treatment22 [0]13 [0]1 [0]1 [1]4 [1]2 [0]1 [0]Psychological1,463 [1]34 [1]7 [1]0 [0]7 [2]10 [1]5 [1]Other958 [14]430 [11]183 [18]32 [21]70 [23]98 [13]145 [21]Unknown1419 [21]685 [17]275 [28]46 [29]88 [29]117 [16]208 [30]Missing4949 [72]2837 [71]701 [71]112 [72]194 [63]601 [81]504 [72]Patients were asked to fill in the 3 most important causes of their cancer, but a large majority filled out only 1 or 2. Therefore, the percentages do not sum up to $100\%$1Internal (lifestyle, stress, biological, psychological) [7]2External (environmental, chance/luck, existential, prior health condition) [7]3Modifiable (lifestyle and stress) [7]4Fixed (biological, environmental, change/luck, existential, prior health condition, and psychological) [7] Colorectal ($68\%$), endometrial ($60\%$), and prostate ($63\%$) cancer survivors reported internal causes of their cancer (i.e. lifestyle, biological, psychological, stress) most often, whereas multiple myeloma survivors more often ($48\%$) reported external causes (i.e. environmental, chance/luck, existential, prior health condition), while lymphoma and thyroid cancer survivors had almost similar rates of internal and external causes (Fig. 1a).Fig. 1a Percentage of only internal ($$n = 1749$$), only external ($$n = 549$$), or both ($$n = 597$$) causal attributions according to tumour type. Internal: lifestyle, biological, stress, psychological [7]; external: environmental, chance/luck, existential, prior health condition [7]. b Percentage modifiable ($$n = 864$$), fixed ($$n = 1161$$), or both ($$n = 870$$) causal attributions according to tumour type. Modifiable: lifestyle, stress [7]; fixed: biological, environmental, change/luck, existential, prior health condition, psychological [7] Lymphoma ($49\%$), multiple myeloma ($64\%$), thyroid ($55\%$), and prostate ($64\%$) cancer patients mentioned fixed causes (i.e. biological, environmental, change/luck, existential, prior health condition, psychological) of their cancer far more often than modifiable causes or a combination of modifiable/fixed (Fig. 1b). Colorectal ($33\%$, $34\%$, and $33\%$) and endometrial ($38\%$, $32\%$, and $30\%$) cancer survivors mentioned causes that were fixed, modifiable (i.e. lifestyle or stress), or modifiable/fixed almost equally.
## Associations with sociodemographic and clinical characteristics
In our sample, 2025 survivors ($29\%$) listed only modifiable ($$n = 864$$; $43\%$) or only fixed ($$n = 1161$$; ($57\%$)) causes of their cancer. Gender, surgery, hormonal treatment, active surveillance/no treatment, stage, marital status, comorbid conditions, and cancer type were associated with reporting modifiable cause of cancer in univariate analyses (Table 3). In multivariate analyses, sex, age, hormonal treatment, stage, marital status, comorbid conditions, and tumour type were associated with reporting modifiable cause of cancer (Table 3). Females, those with a higher age, those treated with hormonal therapy, and those diagnosed with prostate cancer were less likely to identify modifiable causes while those diagnosed with stage 2 disease, those without a partner, with 2 or more comorbid conditions, and those with endometrial cancer were more likely to identify modifiable causes. Table 3Univariate and multivariate associations between survivors’ sociodemographic and clinical characteristics and reporting only modifiable illness attributionsNReported modifiable cause (%)Univariate, odds ratio (CI)Univariate, p valueMultivariable, odds ratio (CI)Multivariable, p valueGender Male1105494 [45]Ref Female920370 [40]0.83 (0.70–0.99)0.0420.60 (0.49–0.75)<.0001Age at questionnaire (mean, SD)86465.3 [11]0.99 (0.99–1.01)0.7850.99 (0.98–0.99)0.007Primary treatment Surgery1516715 [47]0.46 (0.37–0.58)<.0001 Systemic treatment605258 [43]1.00 (0.83–1.21)0.99 Hormonal treatment6018 [30]1.76 (1.01–3.09)0.0470.47 (0.23–0.98)0.044 Radiotherapy626266 [42]1.01 (0.84–1.22)0.915 No treatment/active surveillance12034 [28]1.95 (1.30–2.932)0.001Years since diagnosis (mean, SD)8645.0 [3]1.01 (0.98–1.04)0.652Stage I618269 [44]RefRef II689300 [44]1.06 (0.83–1.35)0.0611.23 (0.95–1.60)0.028 III446200 [45]1.06 (0.83–1.35)0.0261.01 (0.76–1.34)0.799 IV14764 [44]1.00 (0.70–1.44)0.2631.20 (0.79–1.81)0.200Marital status Partner1548636 [41]RefRef No partner465222 [48]1.31 (1.06–1.61)0.0111.41 (1.12–1.77)0.003Education levela Low258108 [42]Ref Middle1210538 [44]1.11 (0.85–1.46)0.101 High536211 [39]0.90 (0.67–1.22)0.164Comorbidities 0513195 [38]RefRef 1583239 [41]1.13 (0.89–1.45)0.7581.29 (0.99–1.67)0.577 2 or more829378 [46]1.37 (1.09–1.71)0.0071.47 (1.15–1.88)0.010Tumour type Colorectal1189603 [51]RefRef Lymphoma29498 [33]0.49 (0.37–0.64)0.6360.47 (0.34–0.65)0.958 Multiple myeloma5412 [22]0.28 (0.15–0.53)0.0740.41 (0.16–1.05)0.721 Thyroid10527 [26]0.34 (0.21–0.53)0.1230.33 (0.20–0.55)0.121 Endometrial18284 [46]0.83 (0.61–1.14)<.00011.11 (0.76–1.62)<.0001 Prostate20140 [20]0.24 (0.17–0.35)0.00010.15 (0.09–0.25)<.0001aEducation levels included the following categories: low=no/primary school, middle=lower general secondary education/vocational training, or high=pre-university education/high vocational training/universityThis table only includes patients who mentioned modifiable illness attributions (i.e. lifestyle and stress) Sensitivity analyses including those with only modifiable causes of cancer in comparison with those reporting both modifiable and fixed causes ($$n = 870$$) showed no significant differences in the abovementioned results (data not shown).
## Discussion
Overall, the most common causal attributions were unknown ($21\%$), lifestyle ($19\%$), biological ($16\%$), other ($14\%$), and stress ($12\%$). A previous smaller study on causal attributions among American survivors with partly overlapping cancers showed a similar percentage of ‘unknown’ ($21.8\%$). Of those who did provide a causal attribution, results were quite similar as well with the three most common causal attributions being lifestyle ($39\%$), biological (including hereditary; $35\%$), and environmental ($24\%$) [7]. The fact that the categories ‘unknown’ and ‘chance/luck’ were reported by patients most often is in accordance with reality since most often, we indeed do not know the exact cause of someone’s cancer and it is often related to chance or luck. If we look more closely at the separate cancer groups, we see that realistic causal attributions are often mentioned (biological, lifestyle, chance/luck). However, most cancers have multiple causes, not all causes are clear at the moment, and the exact cause of cancer in individual patients is often unknown. However, a small part of patients also mentioned various unrealistic ideas about the cause of their cancer and information provision in general can thus be improved. Unrealistic perceptions might prevent patients from making behavioural adjustments in cases where this is necessary or possible.
Internal causes of cancer were most often mentioned by colorectal ($68\%$), endometrial ($60\%$), and prostate ($63\%$) cancer survivors, while multiple myeloma survivors ($48\%$) reported external causes most often, and lymphoma and thyroid cancer survivors mentioned internal and external causes equally often. This is quite realistic in the case of colorectal [21], endometrial [22], and thyroid [23–27] cancer and for lymphoma [28, 29]. However, although the literature on the causes of multiple myeloma is still emergent, we do know that risk factors are not only external (e.g. exposures to chemicals or pesticides, overweight and obesity, patterns of alcohol intake [30]) but also internal, in contrast to what survivors in our study reported. The relative lack of knowledge on the causes of multiple myeloma in the scientific literature is thus likely noticeable in the information provision towards patients.
Fixed causes of cancer were most often mentioned by lymphoma ($49\%$), multiple myeloma ($64\%$), thyroid ($55\%$), and prostate ($64\%$) cancer patients whereas colorectal ($33\%$ and $34\%$) and endometrial ($38\%$ and $32\%$) cancer survivors mentioned both fixed and modifiable respectively. This is not surprising since well-known modifiable risk factors (e.g. obesity) exist for colorectal [31] and endometrial [32] cancer, whereas modifiable risk factors for lymphoma, multiple myeloma, thyroid, and prostate cancer are less clear. Those mentioning a modifiable cause of cancer might be more likely to take action in order to adjust this cause when possible (e.g. lifestyle). However, we do know from the literature that threatening illness perceptions, including causal attributions, are not related to a healthier lifestyle [33].
A meta-analysis of studies on illness perceptions showed that illness perceptions predicted outcomes in various patient groups [34]. Illness perceptions can also influence the process of coping [35, 36] and adherence [37, 38] in a wide range of diseases. Interventions aimed at changing illness perceptions seem to be effective. Two brief in-hospital intervention studies, in which patients had individual in-hospital meetings with a psychologist, changed the perceptions of myocardial infarction patients and this resulted in a faster return to work in the intervention group [3, 39]. Interventions that specifically change causal attributions of cancer survivors in order to improve patient-reported outcomes, health care utilizations, or other outcomes are, is to our knowledge, not currently available.
## Study limitations
The present study has some limitations that are worth mentioning. First, the present study is based upon data from a selection of PROFILES studies, although data collection methods were similar. Also, the included cancer types (colorectal, endometrial, thyroid, and prostate cancer; multiple myeloma; and lymphoma) do not fully represent all cancer survivors. Finally, causal attributions were assessed with a single item from the BIPQ. Therefore, we only have information on the type of casual attribution of patients. Qualitative research is needed to acquire information on why people think something caused their cancer.
Despite these limitations, the present study provides an important contribution to the current literature on the importance of causal attributions of cancer survivors. Having a well-informed perception of the cause of one’s cancer can probably help with making behavioural adjustments in cases where this is necessary or possible. Since our results are based on several large population-based studies with high response rates including survivors with various cancer diagnoses, and including both short- and long-term survivors, extrapolating these results to the larger population of cancer survivors seems justified.
## Clinical implications
In conclusion, this study showed that patients report both internal and external causes of their illness and both fixed and modifiable causes. This differs between the various cancer types included in this study. Although it is almost impossible to know the exact cause of someone’s cancer, having a well-informed perception of the cause of one’s cancer might help with making behavioural adjustments in cases where this is necessary or possible. Future studies should investigate whether unrealistic causal attributions of cancer survivors can be altered or prevented. In addition, they should investigate what effects these changed attributions have on behavioural changes, patient-reported outcomes, health care utilizations, or other outcomes.
## Supplementary information
ESM 1(DOCX 14 kb)
## References
1. Weiner B. *Achievement motivation and attribution theory* (1974.0)
2. Sensky T. **Causal attributions in physical illness**. *J Psychosom Res* (1997.0) **43** 565-573. PMID: 9430070
3. Broadbent E, Ellis CJ, Thomas J, Gamble G, Petrie KJ. **Further development of an illness perception intervention for myocardial infarction patients: a randomized controlled trial**. *J Psychosom Res* (2009.0) **67** 17-23. DOI: 10.1016/j.jpsychores.2008.12.001
4. Leventhal H, Meyer D, Nerenz D. **The common sense representation of illness danger**. *Contrib Med Psychol* (1980.0) **2** 7-30
5. Roesch SC, Weiner B. **A meta-analytic review of coping with illness: do causal attributions matter?**. *J Psychosom Res* (2001.0) **50** 205-219. DOI: 10.1016/S0022-3999(01)00188-X
6. Costanzo ES, Lutgendorf SK, Roeder SL. **Common-sense beliefs about cancer and health practices among women completing treatment for breast cancer**. *Psychooncology.* (2011.0) **20** 53-61. DOI: 10.1002/pon.1707
7. Ferrucci LM, Cartmel B, Turkman YE, Murphy ME, Smith T, Stein KD. **Causal attribution among cancer survivors of the 10 most common cancers**. *J Psychosoc Oncol* (2011.0) **29** 121-140. DOI: 10.1080/07347332.2010.548445
8. van de Poll-Franse LV, Horevoorts N, van Eenbergen M, Denollet J, Roukema JA, Aaronson NK. **The Patient Reported Outcomes Following Initial treatment and Long term Evaluation of Survivorship registry: scope, rationale and design of an infrastructure for the study of physical and psychosocial outcomes in cancer survivorship cohorts**. *Eur J Cancer (Oxford, England : 1990)* (2011.0) **47** 2188-2194. DOI: 10.1016/j.ejca.2011.04.034
9. Lamers RE, Cuypers M, Husson O, de Vries M, Kil PJ, Ruud Bosch JL. **Patients are dissatisfied with information provision: perceived information provision and quality of life in prostate cancer patients**. *Psychooncology.* (2016.0) **25** 633-640. DOI: 10.1002/pon.3981
10. Mols F, Oerlemans S, Vos AH, Koster A, Verelst S, Sonneveld P. **Health-related quality of life and disease-specific complaints among multiple myeloma patients up to 10 yr after diagnosis: results from a population-based study using the PROFILES registry**. *Eur J Haematol* (2012.0) **89** 311-319. DOI: 10.1111/j.1600-0609.2012.01831.x
11. Husson O, Haak HR, Buffart LM, Nieuwlaat WA, Oranje WA, Mols F. **Health-related quality of life and disease specific symptoms in long-term thyroid cancer survivors: a study from the population-based PROFILES registry**. *Acta Oncol* (2013.0) **52** 249-258. DOI: 10.3109/0284186X.2012.741326
12. Mols F, Beijers T, Lemmens V, van den Hurk CJ, Vreugdenhil G, van de Poll-Franse LV. **Chemotherapy-induced neuropathy and its association with quality of life among 2- to 11-year colorectal cancer survivors: results from the population-based PROFILES registry**. *J Clin Oncol* (2013.0) **31** 2699-2707. DOI: 10.1200/JCO.2013.49.1514
13. van de Poll-Franse LV, Mols F, Essink-Bot ML, Haartsen JE, Vingerhoets AJ, Lybeert ML. **Impact of external beam adjuvant radiotherapy on health-related quality of life for long-term survivors of endometrial adenocarcinoma: a population-based study**. *Int J Radiat Oncol Biol Phys* (2007.0) **69** 125-132. DOI: 10.1016/j.ijrobp.2007.02.040
14. Thong MS, Mols F, Lemmens VE, Creemers GJ, Slooter GD, van de Poll-Franse LV. **Impact of chemotherapy on health status and symptom burden of colon cancer survivors: a population-based study**. *Eur J Cancer* (2011.0) **47** 1798-1807. DOI: 10.1016/j.ejca.2011.02.006
15. Thong MS, Mols F, Lemmens VE, Rutten HJ, Roukema JA, Martijn H. **Impact of preoperative radiotherapy on general and disease-specific health status of rectal cancer survivors: a population-based study**. *Int J Radiat Oncol Biol Phys* (2011.0) **81** e49-e58. DOI: 10.1016/j.ijrobp.2010.12.030
16. 16.Oerlemans S, Husson O, Mols F, Poortmans P, Roerdink H, Daniels LA, et al. Perceived information provision and satisfaction among lymphoma and multiple myeloma survivors-results from a Dutch population-based study. Ann Hematol. 2012;91(10):1587–95.
17. 17.Fritz A, Percy C, Jack A, Shanmugaratnam K, Sobin L, Parkin DM, et al.International classification of diseases for oncology , 3rd ed. World Health Organization.; 2000.
18. 18.Sobin LH, Gospodarowicz MK, Wittekind C. TNM Classification of Malignant Tumours, 7th Edition. Wiley & Son; 2011.
19. Sangha O, Stucki G, Liang MH, Fossel AH, Katz JN. **The Self-Administered Comorbidity Questionnaire: a new method to assess comorbidity for clinical and health services research**. *Arthritis Rheum* (2003.0) **49** 156-163. DOI: 10.1002/art.10993
20. Broadbent E, Petrie KJ, Main J, Weinman J. **The Brief Illness Perception Questionnaire**. *J Psychosom Res* (2006.0) **60** 631-637. DOI: 10.1016/j.jpsychores.2005.10.020
21. Brenner H, Kloor M, Pox CP. **Colorectal cancer**. *Lancet.* (2014.0) **383** 1490-1502. DOI: 10.1016/S0140-6736(13)61649-9
22. Morice P, Leary A, Creutzberg C, Abu-Rustum N, Darai E. **Endometrial cancer**. *Lancet.* (2016.0) **387** 1094-1108. DOI: 10.1016/S0140-6736(15)00130-0
23. Iglesias ML, Schmidt A, Ghuzlan AA, Lacroix L, Vathaire F, Chevillard S. **Radiation exposure and thyroid cancer: a review**. *Arch Endocrinol Metab* (2017.0) **61** 180-187. DOI: 10.1590/2359-3997000000257
24. Zhao J, Wang H, Zhang Z, Zhou X, Yao J, Zhang R. **Vitamin D deficiency as a risk factor for thyroid cancer: a meta-analysis of case-control studies**. *Nutrition.* (2019.0) **57** 5-11. DOI: 10.1016/j.nut.2018.04.015
25. Schmid D, Ricci C, Behrens G, Leitzmann MF. **Adiposity and risk of thyroid cancer: a systematic review and meta-analysis**. *Obes Rev* (2015.0) **16** 1042-1054. DOI: 10.1111/obr.12321
26. Han MA, Kim JH. **Diagnostic X-ray exposure and thyroid cancer risk: systematic review and meta-analysis**. *Thyroid.* (2018.0) **28** 220-228. DOI: 10.1089/thy.2017.0159
27. Cao Y, Wang Z, Gu J, Hu F, Qi Y, Yin Q. **Reproductive factors but not hormonal factors associated with thyroid cancer risk: a systematic review and meta-analysis**. *Biomed Res Int* (2015.0) **2015** 103515. DOI: 10.1155/2015/103515
28. Yung L, Linch D. **Hodgkin’s lymphoma**. *Lancet.* (2003.0) **361** 943-951. DOI: 10.1016/S0140-6736(03)12777-8
29. Chihara D, Nastoupil LJ, Williams JN, Lee P, Koff JL, Flowers CR. **New insights into the epidemiology of non-Hodgkin lymphoma and implications for therapy**. *Expert Rev Anticancer Ther* (2015.0) **15** 531-544. DOI: 10.1586/14737140.2015.1023712
30. Sergentanis TN, Zagouri F, Tsilimidos G, Tsagianni A, Tseliou M, Dimopoulos MA. **Risk factors for multiple myeloma: a systematic review of meta-analyses**. *Clin Lymphoma Myeloma Leuk* (2015.0) **15** 563-77 e1–3. DOI: 10.1016/j.clml.2015.06.003
31. Jochem C, Leitzmann M. **Obesity and colorectal cancer**. *Recent Results Cancer Res* (2016.0) **208** 17-41. DOI: 10.1007/978-3-319-42542-9_2
32. Shaw E, Farris M, McNeil J, Friedenreich C. **Obesity and endometrial cancer**. *Recent Results Cancer Res* (2016.0) **208** 107-136. DOI: 10.1007/978-3-319-42542-9_7
33. van Broekhoven M, de Rooij BH, Pijnenborg JMA, Vos MC, Boll D, Kruitwagen R. **Illness perceptions and changes in lifestyle following a gynecological cancer diagnosis: a longitudinal analysis**. *Gynecol Oncol* (2017.0) **145** 310-318. DOI: 10.1016/j.ygyno.2017.02.037
34. Hagger MS, Orbell S. **A meta-analytic review of the common-sense model of illness representations**. *Psychol Health* (2003.0) **18** 141-184. DOI: 10.1080/088704403100081321
35. Dempster M, Howell D, McCorry NK. **Illness perceptions and coping in physical health conditions: a meta-analysis**. *J Psychosom Res* (2015.0) **79** 506-513. DOI: 10.1016/j.jpsychores.2015.10.006
36. Richardson EM, Schuz N, Sanderson K, Scott JL, Schuz B. **Illness representations, coping, and illness outcomes in people with cancer: a systematic review and meta-analysis**. *Psychooncology.* (2017.0) **26** 724-737. DOI: 10.1002/pon.4213
37. Kucukarslan SN. **A review of published studies of patients’ illness perceptions and medication adherence: lessons learned and future directions**. *Res Soc Adm Pharm* (2012.0) **8** 371-382. DOI: 10.1016/j.sapharm.2011.09.002
38. Shahin W, Kennedy GA, Stupans I. **The impact of personal and cultural beliefs on medication adherence of patients with chronic illnesses: a systematic review**. *Patient Prefer Adherence* (2019.0) **13** 1019-1035. DOI: 10.2147/PPA.S212046
39. Petrie KJ, Cameron LD, Ellis CJ, Buick D, Weinman J. **Changing illness perceptions after myocardial infarction: an early intervention randomized controlled trial**. *Psychosom Med* (2002.0) **64** 580-586. DOI: 10.1097/00006842-200207000-00007
|
---
title: Evaluation of the retinal and choroidal microvasculature changes in cases of
sarcoid and tuberculosis-associated posterior uveitis using OCT angiography
authors:
- Lameece Moustafa Hassan
- Ashgan Asaad
- Zeinab ElSanabary
- Maha M. Youssef
journal: International Ophthalmology
year: 2022
pmcid: PMC9971116
doi: 10.1007/s10792-022-02464-6
license: CC BY 4.0
---
# Evaluation of the retinal and choroidal microvasculature changes in cases of sarcoid and tuberculosis-associated posterior uveitis using OCT angiography
## Abstract
### Purpose
Using optical coherence tomography angiography (OCTA) to evaluate retinal microvascular changes in sarcoid and tuberculous (TB) posterior uveitis.
### Methods
Cross-sectional observational study includes 30 eyes. FFA and OCTA images were acquired. OCTA images were analyzed for areas of capillary hypo-perfusion, disorganization of the superficial and deep capillary plexuses (SCP and DCP) and intraretinal cystoid spaces and for measuring the size of the foveal avascular zone and vessel density (VD) in the SCP and DCP.
### Results
A total of 11 eyes were associated with TB and 19 with sarcoidosis. By OCTA, $100\%$ had areas of capillary non-perfusion, $36.7\%$ choroidal voids, $30\%$ disorganization of the SCP and DCP and $26.6\%$ intraretinal cystoid spaces. The VD of the DCP was significantly lower in the TB group. On comparing OCTA and FFA, parafoveal ischemia was detected more frequently on OCTA and macular edema more frequently on FFA (P = < 0.001). The BCVA was not significantly correlated with the VD of the SCP or DCP.
### Conclusion
OCTA can be used in detection of early microvascular changes, segmenting retinal layers and localizing abnormalities. The presence of these changes may aid in the diagnosis of TB and sarcoid uveitis, for prognosis, follow-up and may be the only choice when FFA is contraindicated.
## Introduction
Sarcoid and tuberculous uveitis are common causes of chronic granulomatous inflammation, which can manifest as both anterior and posterior uveitis and result in significant visual disability from their chronic inflammatory sequelae [1]. Granulomatous uveitis most commonly results from sarcoidosis, a multifactorial syndrome stemming mainly from immune dysregulation [2], or a systemic infection, commonly tuberculosis (TB), which may affect the eye only (isolated ocular TB) or is part of a systemic condition [3].
Over many decades, the gold standard modality for the imaging of the retinal and choroidal vasculature and their pathologies, especially in uveitis, has been fundus fluorescein angiography (FFA). However, it is an invasive procedure, requiring the injection of intravenous dyes, which may be poorly tolerated and associated with rare serious side effects. It is also time-and effort-consuming, and leakage with intravenous (IV) FFA can obscure morphological vascular details and window defects can prevent accurate analysis of retinal details. Moreover, this modality provides two-dimensional evaluation of the retina and choroid and from them we are unable to identify the level of vascular abnormalities. Thus, it is impractical to repeatedly use angiography for patients to evaluate disease activity and progression [4] Optical coherence tomography angiography (OCTA), employs amplitude or phase decorrelation to detect blood flow without the need for intravenous dye administration [5]. OCTA allows the study of micro vascular changes, and thus may provide an important tool in the evaluation of inflammatory eye diseases, as the vascular changes in the iris, choroid, and retina play an important role in the pathophysiology of ocular inflammatory diseases [4]. Moreover, OCTA image acquisition is easy, fast and non-invasive, limiting the risk of side effects for the patient. Depth-resolving capability gives important evaluation of the deep retinal capillary plexus, which may be targeted in retinal vascular or inflammatory diseases [6, 7].
The aim of this study is to evaluate the retinal and choroidal microvasculature qualitative and quantitative changes in eyes with sarcoid and tuberculous posterior uveitis using optical coherence tomography angiography (OCTA) and compare these changes with their corresponding fundus fluorescein angiography (FFA) findings and to correlate these changes with the disease type, treatment and activity.
## Methods
This is a cross-sectional observational study including 30 eyes of 30 patients (19 sarcoid and 11 tuberculous patients with posterior uveitis). Patients were recruited from the uveitis subspecialty clinic at Kasr ElAiny Hospital, Cairo University. The study was conducted during the period between December 2019 and July 2020.
## Statement of ethics
Cairo University ethical committee approval was obtained (N-24-2019) and the study followed the tenets of the Declaration of Helsinki. A written consent was taken from all patients participating in the study.
This study included patients with posterior uveitis or panuveitis secondary to sarcoidosis or tuberculosis. We excluded patients with other coexisting retinal diseases, eyes with refractive error of 6 diopters or more and eyes with dense media opacities obscuring imaging or lowering image resolution.
Detailed medical history was taken, and complete ophthalmological examination was conducted including visual acuity (best corrected visual acuity-BCVA, recorded in decimal notation), intraocular pressure measurement and slit lamp examination of the anterior segment, and dilated fundus examination was carried out using a binocular indirect ophthalmoscope and slit-lamp biomicroscopy, as is the routine for any uveitis patient on diagnosis and in follow-up.
The diagnosis of posterior uveitis or pan uveitis in all patients was made according to the SUN (Standardization of Uveitis Nomenclature) classification, grading of activity was done according to SUN classification as well [8]. Eyes with vitreous haze grading of > + 0.5, with or without active choroidal or retinal inflammatory lesions, were considered active and this activity was confirmed by FFA findings. Ocular sarcoidosis was diagnosed according to the international workshop on ocular sarcoidosis (IWOS) criteria [9] and intraocular tuberculosis was diagnosed according to the criteria of classification of intraocular tuberculosis [10]. The consensus on the etiology, anatomical location and state of activity was made by two different uveitis consultants (MY and LH). Laboratory and radiological work up was done guided by the clinical condition of the patients.
Patients were classified according to specific systemic treatment into treated or non-treated (treatment naïve) patients. Treated sarcoid patients were those who received or continue to be under systemic steroid therapy with or without other immunosuppressive drugs and treated tuberculous patients included those who had completed anti-tuberculous treatment (ATT) for at least 6 months with or without steroid therapy.
Fundus photography and FFA were done for all patients using the TOPCON (TRC-50DX, Topcon Medical System Inc., 2015). FFA images were analyzed for vasculitis, focal or multifocal choroiditis, macular edema, optic disc leakage and areas of ischemia (peripheral capillary dropout or macular ischemia).
OCT angiography images were acquired of a 6 × 6 mm area in the central macula from all patients using RTVue XR Avanti (AngioVue, Optovue Inc, Fremont, California, USA). The flow imaging was based on Split-Spectrum Amplitude Decorrelation Angiography (SSADA).
Detection of the following qualitative parameters was done: areas of capillary non-perfusion/hypo perfusion, capillary changes (capillary dilatation, telangiectasia, shunting vessels and areas of rarefied capillaries), disorganization of the superficial and deep capillary network (localized or diffuse loss of the normal architecture of the capillary network), intraretinal cystoid spaces (defined as round black areas without any decorrelation signal and confirmed by both structural OCT and corresponding en face images) and choroidal affection in the form of choroidal voids.
OCTA was used to measure the size of the foveal avascular zone (FAZ) (mm2) and capillary vessel density (VD) in both the superficial and deep capillary plexuses at 9 areas grid-based vessel density (%). The foveal region was defined as the central 1 mm, parafoveal 1–3 mm and perifoveal region 3–6 mm according to the ETDRS Grid. Three captures were taken in the same setting and the mean VD was used.
All images were analyzed by the same consultant (ZS) to avoid interobserver variations. OCTA images were compared with both sequential en face images and SD OCT images to detect areas of signal loss and to differentiate vitreous opacities and artifacts with back shadowing from areas of non-perfusion.
Whole image capillary VD < $50\%$ was considered as ischemia, according to the OCTA normative data for vascular density in the superficial and deep capillary plexuses of healthy adults determined by Coscas et al. [ 11].
## Statistical methods
Microsoft excel 2013 was used for data entry and the statistical package for social science (SPSS version 24) was used for data analysis. Simple descriptive statistics (arithmetic mean, median and standard deviation) were used for summary of normal quantitative data. Bivariate relationship was displayed in cross-tabulations and comparison of proportions was performed using the Chi-square and Fisher’s exact tests where appropriate T-independent was used to compare normally distributed quantitative data and Mann–Whitney for skewed data. P value less than 0.05 was considered statistically significant.
## Results
This study included 30 eyes of 30 patients. In patients with bilateral affection, the eye with clearer media, allowing higher quality of imaging, was chosen.11 out of the 30 eyes were diagnosed as probable TB uveitis ($36.7\%$) and 19 ($63.3\%$) eyes were diagnosed as presumed sarcoid uveitis.
The mean BCVA (in decimal notation) of the TB group patients was 0.36 ± 0.24 SD, while it was 0.4 ± 0.29 SD in the sarcoidosis group, without any statistical difference.
By clinical examination and confirmed by fluorescein angiography, activity was detected in 5 ($45.5\%$) of the eyes with TB uveitis and 6 eyes ($31.6\%$) with sarcoid uveitis (with a statistically insignificant difference) (Tables 1, 2). Phenotypes of both the sarcoid and TB cases were divided between posterior uveitis in the form of multifocal choroiditis and/or retinal vasculitis or panuveitis with associated multifocal choroiditis and/or vasculitis. Three cases of retinal vein occlusion were noted; however no cases of serpiginous choroiditis or large solitary granulomas were encountered. As for the sarcoid associated cases, one case had an associated large nasal peripapillary granuloma (not encroaching on the macula).Table 1Demographic data and clinical findings in patients with TB associated uveitiesCount ($$n = 11$$)%GenderMale327.3Female872.7TreatmentTreated763.6Untreated436.4ActivityInactive654.5Active545.45Clinical patternsVasculitis863.6Multifocal choroiditis436.4Optic disc edema218.2Table 2Demographic data and clinical findings in patients with sarcoidosis associated UveitisCount ($$n = 19$$)%GenderMale00Female19100TreatmentTreated1578.9Untreated421ActivityInactive631.6Active1368.4Clinical patternsVasculitis631.6Multifocal choroiditis526.3Optic Disc edema736.8
## OCTA findings in all patients
OCTA images showed areas of capillary non-perfusion / hypo perfusion in 30 eyes ($100.0\%$), capillary changes in 15 eyes ($50.0\%$), choroidal voids due to ischemia or infiltration in 11 eyes ($36.7\%$), disorganization of the superficial and deep capillary network in 9 eyes ($30.0\%$), and intraretinal cystoid spaces in 8 eyes ($26.6\%$) (Figs. 1, 2, 3, 4).Fig. 1Patient with active tuberculosis. A Infra-red shows peripheral attenuated vessels. B FFA shows peripheral vasculitis, telangiectasia, peripheral choroidal lesions and peripheral ischemia. C Corresponding structural B-scan is normal. D, E SCP and DCP respectively show ischemia more in the deep capillary plexus. F Choriocapillaris layer shows areas of choroidal voids (arrows).Fig. 2Patient with active tuberculosis. A Infra-red photo shows upper temporal ischemia and retinal hemorrhages. B FFA shows upper temporal ischemic branch retinal vein occlusion (RVO) with peripheral phlebitis and telangiectasia. C SCP shows upper temporal non perfusion areas. D DCP shows diffuse ischemia. E, F 8 × 8 mm OCTA image of the SCP and DCP shows capillary changes, capillary network disorganization and upper temporal capillary non perfusion areas corresponding to branch RVO (arrows).Fig. 3Patient with active sarcoidosis A Infra-red photo shows attenuated vessels. B FFA arterial phase. B* FFA late phase shows active vasculitis, optic disc leakage and macular edema. C Structural B-scan shows intra retinal cystoid spaces. D Enface image shows cystoid spaces (arrows) without evidence of back shadowing effect from vitreous floaters. E, F SCP and DCP respectively show ischemia. G Choriocapillaris layer shows areas of choroidal voids (arrows).Fig. 4Patient with active sarcoidosis A Color photo shows lost foveal reflex B FFA shows hot optic disc, macular edema and peripheral vascular leakage. C Structural B-scan shows intra retinal cystoid spaces. D, E SCP and DCP respectively show ischemia more in the deep capillary plexus. F OCTA image of the DCP shows capillary changes and areas of capillary hypoperfusion G Enface image of the DCP shows well-defined black cystoid spaces
## FFA findings in TB versus sarcoid patients
On comparing the FFA findings in both groups, statistical significance was only found when comparing the frequency of optic disc leakage, being higher in the sarcoidosis group (P value = 0.023) (Table 3) (Figs. 3, 4).Table 3FFA findings in TB versus sarcoid patientsTuberculosisSarcoidosisP valueCount%Count%Macular edema654.5526.30.238Optic disk leakage19.11052.60.023Areas of active choroiditis436.4526.30.687Macular ischemia218.2421.11.000Active vasculitis327.3736.80.702Inactive vasculitis545.515.260.016Peripheral ischemia550.7526.30.425 Vasculitis was found in 8 TB patients and 8 sarcoid patients, being occlusive and associated with peripheral ischemia in 5 patients in each group
## OCTA angiography findings in TB versus sarcoid patients
Areas of capillary non-perfusion/hypo perfusion were detected in all 11 eyes ($100.0\%$) of the TB group and in all 19 eyes ($100.0\%$) of the sarcoidosis group, regardless the state of the uveitic activity. Capillary changes (areas of rarefied capillaries) were detected in 6 eyes ($54.5\%$) of the TB group and in 9 eyes ($47.4\%$) of the sarcoidosis group. Choroidal voids due to ischemia or infiltration were detected in 5 eyes ($45.5\%$) of the TB group and in 6 eyes ($31.6\%$) of the sarcoidosis group. Intraretinal cystoid spaces were detected in 4 eyes ($36.4\%$) of the TB group and in 4 eyes ($21.1\%$) of the sarcoidosis group. The aforementioned comparisons were statistically insignificant.
On comparing quantitative parameters determined by OCTA, it was found that vascular densities in the DCP were significantly lower in the TB group in the mean whole image VD, mean inferior VD, mean nasal VD and mean temporal VD (Table 4).Table 4OCTA angiography quantitative changes in TB versus sarcoid patientsTuberculosisSarcoidosisP valueFAZ (mm2) (mean)0.285640.285640.444SCP whole image VD (mean)45.39145.2320.916SCP foveal VD (mean)42.74539.3840.072SCP superior VD (mean)46.02746.6110.759SCP inferior VD (mean)46.00946.4420.804SCP nasal VD (mean)47.96448.0110.981SCP temporal (mean)42.57343.1580.785DCP whole image VD (mean)40.20044.7630.024DCP foveal VD (mean)46.33649.9580.076DCP superior VD (mean)39.91844.4160.085DCP inferior VD (mean)39.68244.7260.020DCP nasal VD (mean)39.84547.3630.015DCP temporal VD (mean)45.09150.3000.016
## Comparing OCTA and FFA findings in all patients
FFA showed macular ischemia in only 6 eyes ($20.0\%$) of all patients while OCTA showed areas of macular non-perfusion/hypo perfusion in all eyes ($100\%$). It also showed macular edema in 11 eyes ($36.7\%$), whereas OCTA showed intraretinal cystoid spaces in only 8 eyes ($26.7\%$) of all patients. These findings were statistically significant (P = < 0.001).
On attempting to associate the presence of parafoveal and perifoveal (the area 1–6 mm from the center of the fovea) ischemia on OCTA in the SCP and DCP with the disease type, treatment, activity and FFA findings, no statistically significant associations were found in either group.
## Correlating VD and BCVA
There was a positive but insignificant correlation between the whole image vascular density in the SCP and the mean BCVA of all patients (R- value = 0.359 and the P-value = 0.052) and between the whole image vascular density in the DCP and the BCVA ($R = 0.143$, $$P \leq 0.451$$).
## Discussion
In this work we analyzed the microvascular qualitative and quantitative changes detected by OCTA in 30 eyes with sarcoid or tuberculous-associated posterior uveitis.
This study showed that, by OCTA, multiple changes could be detected: all eyes had areas of capillary non-perfusion/hypo perfusion (even if activity was not clinically or angiographically detected), while $50\%$ also had capillary changes in the form of areas of rarefied/telangiectatic capillaries. Likewise, Kim et al. [ 12] demonstrated that in contrast with healthy controls, uveitis subjects had distinct areas of qualitatively impaired retinal perfusion in the non-segmented retinal layer (NS-RL), both in the absence and presence of macular edema [12].
In our study we observed changes in the choriocapillaris layer, in $45.5\%$ of the TB group and $31.6\%$ eyes of the sarcoidosis group, in the form of ‘flow-voids’ which may be due to ischemia or infiltration (Figs. 1, 3, 4). This was explained by Cerquaglia et al. [ 13], as active granulomata, chronic tissue damage secondary to previously active granulomata (mechanically), or the presence of focal choroidal arteriolitis [13].
Agarwal et al. [ 14] also reported an increase in the areas of ‘flow-void’ by OCTA, which corresponded to active infiltrates found in 5 patients with tuberculous serpiginous like choroiditis, who had developed paradoxical worsening upon initiation of anti-tuberculous therapy. Thus, they concluded that OCTA may provide a simple, fast, non-invasive and high-resolution alternate imaging method to document progressive or recurrent choriocapillaris hypoperfusion, which is essential in the monitoring and follow up of eyes with choroiditis [14].
When documenting areas of flow voids, we excluded artifacts or loss of transmission by correlating them with their corresponding structural en face images and the cross-sectional OCT scans as recommended by Mahendradas et al. [ 15] When comparing the VD in our patients (in the SCP and DCP), it was less than the normative data determined by Coscas et al. [ 11]. Thus, our study showed macular hypoperfusion in all patients, being even more evident in the DCP (Figs. 1, 2, 3, 4).
Likewise, Emre et al. compared 32 eyes of Behçet uveitis (during the inactive period of the disease as determined by conventional imaging techniques i.e. FFA) and 30 eyes of healthy controls. OCTA revealed microvascular changes such as parafoveal capillary telangiectasia and capillary retinal hypoperfusion despite the absence of activity and showed that the capillary vessel density of the Behçet group was significantly lower than in the control group. In addition, the DCP was affected more than the SCP in these patients. Thus, the authors concluded that OCTA was more reliable than FA in monitoring patients and detecting early risk of central loss of vision [16].
The deep plexus has been shown in prior OCTA-based studies to be more vulnerable to impaired blood flow. This was implicated to be the result of its location in the “water shed region” at the termination of retinal capillary units [13, 13].
On comparing the qualitative and quantitative changes detected by OCTA in the TB group and the sarcoidosis group, our study showed that there was no statistically significant difference between the findings except in the vascular density of the DCP in the whole enface image and in the inferior, nasal and temporal sectors; being lower in the TB group.
Both diseases share many common features as they are both granulomatous and both can cause periphlebitis but in sarcoidosis occlusive periphlebitis was less common than in TB (Fig. 1). Tuberculous retinal periphlebitis is typically an obliterative periphlebitis and tends to cause hemorrhagic infarction of the retina [18]. This may explain the evident significant DCP ischemia detected by OCTA in the TB group and may help in differentiation between the two entities, which often pose a diagnostic dilemma as they share many clinical findings and their investigations (laboratory or radiological) are not always conclusive.
The comparison between the qualitative findings found on OCTA and FFA in this study, revealed that all eyes ($100\%$) showed areas of para and perifoveal non-perfusion/hypo perfusion by OCTA. However, on FFA, only 6 eyes ($20.0\%$ of all patients) showed macular ischemia (highly significant on comparison). Thus, OCTA may be superior to FFA in detection of macular ischemia and also can quantify the ischemia in numerical values (capillary VD %) and localize the abnormalities by segmentation of the retinal and choroidal layers. These advantages are lacking in FFA devices.
In an OCTA study on patients with placoid pattern of TB, areas of choriocapillaris flow deficit detected on OCTA were correlated anatomically with ischemic lesions on FFA and ICGA, but were more extensive [19]. This was similar to our study and correlating with ICGA findings, although useful in these two entities due to their choroidal affection, is unfortunately not available in our region.
In 2017, Khairallah et al. described OCTA findings in eyes with active Behçet’s uveitis. They determined, as we did in our study, that OCTA allowed better visualization and characterization of parafoveal microvascular changes than FA, such as disruption of the capillary arcade, areas of retinal non-perfusion, and capillary abnormalities [20].
On the other hand, in our study FFA proved to be superior in detection of macular edema. FFA images revealed macular edema (leakage from an unhealthy capillary bed) was in 11 eyes ($36.7\%$), whereas OCTA showed intraretinal cystoid spaces (confirmed by sequential en face images and corresponding SD OCT) in only 8 eyes ($26.7\%$) of all patients (P = < 0.001).
FFA remains indispensable in diagnosis and monitoring peripheral vasculitis, as seen in 8 of our TB patients and 8 of the sarcoid patients. Abucham-Neto et al., evaluated nineteen eyes with retinal vasculitis (2 eyes were associated with sarcoidosis) and reported that OCTA is unable to detect clear signs of active inflammation around the affected vessels like signs of vascular sheathing and perivascular leakage on FFA [21].
Our study showed that correlating parafoveal and perifoveal ischemia (detected by OCTA) in the SCP and DCP with the disease type, treatment, and activity was statistically insignificant. This may be related to the chronic and slowly progressive course of the ocular granulomatous inflammation in both groups. This may explain why ischemia is found in the TB and sarcoid eyes regardless whether they are treated or not and whether active or inactive [22].
The correlation between retinal vessel density and visual acuity remain unclear. Prior studies have reported significant negative correlations, while others did not find any correlation between both [17].
Our study found that there is positive, but statistically insignificant correlation between the vascular density in the SCP, DCP and the BCVA. This was contrary to our expectation that the vision would have been significantly correlated to the VD in DCP, as it was more affected than the SCP in our patients. This may be explained by our finding that the hypoperfusion found in the DCP may not necessarily manifest in ischemic damage of the outer retinal layers, as shown in Fig. 3.
Limitations of our study include the lack of differentiation of active/inactive and treated/untreated patients when comparing the FFA, OCTA qualitative and quantitative findings. Likewise, comparing the changes in these findings, following treatment of the active patients, would help further investigate the predictive value of OCTA in posterior uveitis and possibly allow a more significant correlation with the functional outcome (BCVA).
## Conclusion
This study is one of the few studies comparing OCTA findings in two of the most common causes of granulomatous posterior uveitis.
This study confirms the role of OCT angiography in detection of qualitative and quantitative microvascular changes in tuberculous- and sarcoid-associated uveitis and shows that OCTA is superior to FFA in detection of macular ischemia. On the other hand, FFA remains an essential complementary tool to detect activity of posterior uveitis and peripheral retinal vasculitis.
Although FFA is indispensable in demonstrating inflammatory activity in the posterior segment, OCTA can be used in follow up of patients and may be the only choice when FFA is contraindicated or refused. The presence of ischemia and lower vascular density found in the TB group may be a factor aiding in the differentiation between the two often overlapping entities.
## References
1. Mehta H, Sim DA, Keane PA. **Structural changes of the choroid in sarcoid- and tuberculosis-related granulomatous uveitis**. *Eye* (2015.0) **29** 1060-1068. DOI: 10.1038/eye.2015.65
2. Iannuzzi MC, Rybicki BA, Teirstein AS. **Sarcoidosis**. *N Engl J Med* (2007.0) **357** 2153-2165. DOI: 10.1056/NEJMra071714
3. Babu K, Kini R, Mehta R. **Predictors for tubercular uveitis: a comparison between biopsy-proven cases of tubercular and sarcoid uveitis**. *Retina* (2013.0) **32** 1017-1020. DOI: 10.1097/IAE.0b013e31822d3a20
4. Pichi F, Sarraf D, Arepalli S. **The application of optical coherence tomography angiography in uveitis and inflammatory eye diseases**. *Prog Retin Eye Res* (2017.0) **59** 178-201. DOI: 10.1016/j.preteyeres.2017.04.005
5. Gorczynska I, Migacz J, Zawadzki R, Capps A, Werner J. **Comparison of amplitude-decorrelation, speckle-variance and phase-variance OCT angiography methods for imaging the human retina and choroid**. *Biomed Opt Express* (2016.0) **7** 911-942. DOI: 10.1364/BOE.7.000911
6. Leder H, Campbell J, Sepah Y. **Ultra-wide-field retinal imaging in the management of non-infectious retinal vasculitis**. *J Ophthalmic Inflamm Infect* (2013.0) **3** 30. DOI: 10.1186/1869-5760-3-30
7. El-Asrar A, Herbort C, Tabbara K. **A clinical approach to the diagnosis of retinal vasculitis**. *Int Ophthalmol* (2010.0) **30** 149-173. DOI: 10.1007/s10792-009-9301-3
8. Jabs D, Nussenblatt R, Rosenbaum J. **standardization of uveitis nomenclature (sun) working group standardization of uveitis nomenclature for reporting clinical data results of the first international workshop**. *Am J Ophthalmol* (2005.0) **140** 509-516. DOI: 10.1016/j.ajo.2005.03.057
9. 9.Herbort C, Rao N, Mochizuki M, the Members of the Scientific Committee of the first International Workshop on Ocular Sarcoidosis (IWOS) (2009) International criteria for the diagnosis ocular Sarcoidosis. Ocul Immunol Inflam 17:160–169
10. Gupta A, Sharma A, Bansal R, Sharma K. **Classification of intraocular tuberculosis**. *Ocul Immunol Inflamm* (2015.0) **23** 7-22. DOI: 10.3109/09273948.2014.967358
11. Coscas F, Sellam A, Glacet-Bernard A. **Normative data for vascular density in superficial and deep capillary plexuses of healthy adults assessed by optical coherence tomography angiography**. *Invest Ophthalmol Vis Sci* (2016.0) **57** 211-223. DOI: 10.1167/iovs.15-18793
12. Kim A, Rodger D, Shahidzadeh A. **Quantifying retinal microvascular changes in uveitis using spectral domain optical coherence tomography angiography (SD-OCTA)**. *Am J Ophthalmol* (2016.0) **171** 101-112. DOI: 10.1016/j.ajo.2016.08.035
13. Cerquaglia A, Iaccheri B, Fiore T, Fruttini D. **New insights on ocular sarcoidosis: an optical coherence tomography angiography study**. *Ocul Immunol Inflamm* (2019.0) **27** 1057-1066. DOI: 10.1080/09273948.2018.1497665
14. Agarwal A, Aggarwal K, Deokar A. **OCTA study group, optical coherence tomography angiography features of paradoxical worsening in tubercular multifocal serpiginoid choroiditis**. *Ocul Immunol Inflamm* (2016.0) **24** 621-630. DOI: 10.1080/09273948.2016.1207785
15. Mahendradas P, Maruyama K, Mizuuchi K. **Multimodal imaging in ocular sarcoidosis**. *Ocul Immunol Inflamm* (2020.0) **28** 1205-1211. DOI: 10.1080/09273948.2020.1751210
16. Emre S, Guven-Yılmaz S, Ulusoy M, Ates H. **Optical coherence tomography angiography findings in Behcet patients**. *Int Ophthalmol* (2019.0) **39** 2391-2399. DOI: 10.1007/s10792-019-01080-1
17. Dupas B, Minvielle W, Bonnin S. **Association between vessel density and visual acuity in patients with diabetic retinopathy and poorly controlled type 1 diabetes**. *JAMA Ophthalmol* (2018.0) **136** 721-728. DOI: 10.1001/jamaophthalmol.2018.1319
18. 18.Salmon JF (2020) Uveitis. In: Kanski clinical ophthalmology—a systemic approach, 9th edn, pp 428–484
19. Klufas M, Phasukkijwatana N, Iafe N. **Optical coherence tomography angiography reveals choriocapillaris flow reduction in placoid chorioretinitis**. *Ophthalmol Retina* (2017.0) **1** 77-91. DOI: 10.1016/j.oret.2016.08.008
20. Khairallah M, Abroug N, Khochtali S. **Optical coherence tomography angiography in patients with behcet uveitis**. *Retina* (2017.0) **37** 1678-1691. DOI: 10.1097/IAE.0000000000001418
21. Abucham-Neto J, TorricelliLuiA A. **Comparison between optical coherence tomography angiography and fluorescein angiography findings in retinal vasculitis**. *Int J Retina Vitr* (2018.0) **4** 15. DOI: 10.1186/s40942-018-0117-z
22. Bloch-Michel E, Nussenblatt R. **International uveitis study group recommendations for the evaluation of intraocular inflammatory disease**. *Am J Ophthalmol* (1987.0) **103** 234-235. DOI: 10.1016/S0002-9394(14)74235-7
|
---
title: MiR-422a promotes adipogenesis via MeCP2 downregulation in human bone marrow
mesenchymal stem cells
authors:
- Angelica Giuliani
- Jacopo Sabbatinelli
- Stefano Amatori
- Laura Graciotti
- Andrea Silvestrini
- Giulia Matacchione
- Deborah Ramini
- Emanuela Mensà
- Francesco Prattichizzo
- Lucia Babini
- Domenico Mattiucci
- Elena Marinelli Busilacchi
- Maria Giulia Bacalini
- Emma Espinosa
- Fabrizia Lattanzio
- Antonio Domenico Procopio
- Fabiola Olivieri
- Antonella Poloni
- Mirco Fanelli
- Maria Rita Rippo
journal: 'Cellular and Molecular Life Sciences: CMLS'
year: 2023
pmcid: PMC9971129
doi: 10.1007/s00018-023-04719-6
license: CC BY 4.0
---
# MiR-422a promotes adipogenesis via MeCP2 downregulation in human bone marrow mesenchymal stem cells
## Abstract
Methyl-CpG binding protein 2 (MeCP2) is a ubiquitous transcriptional regulator. The study of this protein has been mainly focused on the central nervous system because alterations of its expression are associated with neurological disorders such as Rett syndrome. However, young patients with Rett syndrome also suffer from osteoporosis, suggesting a role of MeCP2 in the differentiation of human bone marrow mesenchymal stromal cells (hBMSCs), the precursors of osteoblasts and adipocytes. Here, we report an in vitro downregulation of MeCP2 in hBMSCs undergoing adipogenic differentiation (AD) and in adipocytes of human and rat bone marrow tissue samples. This modulation does not depend on MeCP2 DNA methylation nor on mRNA levels but on differentially expressed miRNAs during AD. MiRNA profiling revealed that miR-422a and miR-483-5p are upregulated in hBMSC-derived adipocytes compared to their precursors. MiR-483-5p, but not miR-422a, is also up-regulated in hBMSC-derived osteoblasts, suggesting a specific role of the latter in the adipogenic process. Experimental modulation of intracellular levels of miR-422a and miR-483-5p affected MeCP2 expression through direct interaction with its 3′ UTR elements, and the adipogenic process. Accordingly, the knockdown of MeCP2 in hBMSCs through MeCP2-targeting shRNA lentiviral vectors increased the levels of adipogenesis-related genes. Finally, since adipocytes released a higher amount of miR-422a in culture medium compared to hBMSCs we analyzed the levels of circulating miR-422a in patients with osteoporosis—a condition characterized by increased marrow adiposity—demonstrating that its levels are negatively correlated with T- and Z-scores. Overall, our findings suggest that miR-422a has a role in hBMSC adipogenesis by downregulating MeCP2 and its circulating levels are associated with bone mass loss in primary osteoporosis.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00018-023-04719-6.
## Introduction
The methyl-CpG binding protein 2 (MeCP2) is known principally for its ability to inhibit the transcription complex assembly on DNA by binding methylated CpG islands across the genome [1]. Beyond its role in transcriptional repression, more recent studies revealed that MeCP2 may play a complex multifunctional role, coordinating also transcriptional activation, chromatin architecture, and RNA splicing, depending on the molecular context [2–4]. MeCP2 expression is ubiquitous throughout the body, although it is particularly abundant and studied in brain cells. Indeed, in females X-linked mutations of the MECP2 gene cause Rett syndrome (RTT), a neurodevelopmental disorder characterized by loss of acquired motor and language skills, autistic features, and unusual stereotyped movements [5, 6]. However, the variety of phenotypes identified in RTT patients and MeCP2 mutant mouse models points to important roles for MeCP2 in peripheral systems, including altered lipid metabolism, unbalanced adipose tissue endocrine activity [7, 8], and decreased bone mineral density, among others [9, 10].
In bone marrow (BM), mesenchymal stromal cells (BMSCs), the precursors of adipocytes and osteoblasts, are exposed to a plethora of stimuli that determine the balance between adipogenesis and osteogenesis which in turn are competing and reciprocal [11, 12]. The differentiation of MSCs, in fact, is a two-step process: lineage commitment (from BMSCs to lineage-specific progenitors) and maturation (from progenitors to specific cell types). Studying the mechanisms that regulate bone marrow adipogenesis is important because the marrow adipose tissue (MAT) is not only a passive space-filler. Indeed, MAT actively participates in a broad spectrum of physiological functions—e.g. energy homeostasis, immunity, hematopoiesis, coagulation, and regulation of blood pressure—through the release of several molecular mediators [13], including adiponectin, of which BM adipocytes are among the major contributors [14]. Variations in BM adipocyte mass have been reported in primary osteoporosis and other systemic conditions like aging, type 2 diabetes, obesity, myelodysplastic syndrome, cancer therapy, and anorexia nervosa, suggesting that an abnormal differentiation of BMSCs could contribute to pathogenic skeletal manifestations associated with such diseases [15].
Adipogenesis is a finely tuned multi-step process requiring the sequential activation of numerous transcription factors driving the typical physiological and morphological changes observed in the progenitor cells, i.e. cell cycle arrest, metabolic reprogramming, and lipid accumulation [16]. The expression of peroxisome proliferator-activated receptor gamma (PPARγ) is critical to promote fat cell differentiation, and survival of adult adipocytes, inducing the expression of genes involved in insulin sensitivity, lipogenesis, and lipolysis [17–19].
Among the small non-coding RNAs, microRNAs (miRNAs, miRs) represent an additional mechanism for controlling adipogenic gene expression [20]. Given their unique ability to simultaneously regulate multiple protein targets and processes, it has been suggested that miRNAs may play a leading role in BMSC differentiation. Their involvement in adipogenesis has been investigated through experimental and bioinformatic, target-based approaches [21–23]; some identified miRNAs would appear to be involved in lineage commitment while others in maturation. The latter, in some cases, would involve switches that suppress the osteogenic process by activating the adipogenic one [24, 25]. Interestingly, several studies in the nervous system, cancer, and smooth muscle cell differentiation suggest that miRNAs can directly regulate the expression of MeCP2 [26–32].
The role of MeCP2 in the BMSC differentiation process is still unknown but given the relevance of fine control of the transcriptional and epigenetic processes in BMSCs differentiation, it is conceivable that MeCP2 and miRNAs regulating its expression could affect their fate.
Therefore, the aim of our study was to identify miRNAs able to modulate the adipogenic process and their role in the MeCP2-mediated modulation of adipogenesis. For this purpose, we (i) evaluated MeCP2 expression in hBMSCs undergoing differentiation; (ii) screened for miRNAs differentially regulated during hBMSC differentiation in vitro; (ii) validated miRNAs selectively upregulated in adipocytes compared to their precursors; (iii) tested their role in enhancing adipogenesis, and (iv) assessed their ability to modulate MeCP2 expression in hBMSCs. Finally, the levels of circulating miRNAs involved in hBMSC differentiation have been assessed in a cohort of elderly subjects with primary type II osteoporosis, to investigate their association with bone mass loss due to the expansion of the MAT compartment.
## MeCP2 is downregulated in adipose tissue and during adipogenesis of hBMSC in vitro
To determine the role of MeCP2 during adipogenesis, MeCP2 expression was analyzed in human bone marrow MSCs induced to differentiate into adipocytes and osteoblast. As shown in Fig. 1A MeCP2 protein expression was downregulated in BMSC-derived adipocytes compared to undifferentiated cells, whereas an opposite modulation was observed in osteoblasts. Furthermore, immunohistochemical (IHC) staining performed in human bone marrow sections obtained from healthy donors revealed a high expression of MeCP2 in the nuclei of periosteal cells (Fig. 1B(a)) and of several hematopoietic cells (Fig. 1B(b)); on the contrary, MeCP2 expression was not found in adipocytes (Fig. 1B(c)). The same results are confirmed in the bone marrow and adipose tissue (AT) samples derived from rats (Fig. 1C). To strengthen this observation, MeCP2 expression was analyzed in human adipocytes and MSCs, isolated from subcutaneous fat of the same donors, in addition to dedifferentiated adipocytes obtained as previously described [33, 34]. Here, we show that MeCP2 expression is significantly lower in adipocytes compared to MSCs and dedifferentiated cells. PPARγ was used as a control of the differentiation state (Fig. 1D).Fig. 1MeCP2 expression in adipocytes and adipogenic process. A Western blot and densitometric analysis of MeCP2 in human bone marrow mesenchymal stromal cells (hBMSCs) and hBMSC-derived adipocytes (AD) and osteoblasts (OS) after 14 days of pro-differentiating treatment. Data were normalized to β-actin. B Immunohistochemical detection of MeCP2 reactivity in the human femoral bone marrow. 1, Numerous MeCP2-positive nuclei are present. a, Periosteal positive cells (arrows). b, Hematopoietic positive cells. 2, Negative human BM adipose tissue; c, positive hematopoietic cells (arrows) in close contact with adipocyte membrane. Images were taken at 200× and 400× magnification. C Immunohistochemical detection of MeCP2 reactivity in rat. 1, Femoral bone marrow, MeCP2 positive hematopoietic cells in close contact with adipocyte membrane are present (a, arrows), no staining was observed in adipocyte nuclei. 2, Inguinal white adipose tissue (AT), low reactivity in adipocytes was confirmed, whereas interstitial and blood cells were positive (b, arrows). Images were taken at 200× and 400× magnification. D Western blot and densitometric analysis of MeCP2 and PPARγ expression in human adipocytes, mesenchymal stromal cells (AT-MSCs) and dedifferentiated adipocytes (AD dediff) obtained from subcutaneous adipose tissue of the same donor. Western blot image is relative to 1 out of 3 different analyzed donors. Data were normalized to α-Tubulin. Data are mean ± SD of three independent experiments. * t-test $p \leq 0.05$; **t-test $p \leq 0.01$; ***t-test $p \leq 0.001$
## MeCP2 downregulation promotes adipogenic program in hBMSCs
To unravel the role of MeCP2 modulation in adipogenesis, we reduced its expression in hBMSCs by infecting these cells with a pool of three MeCP2-targeting shRNAs (shMeCP2) or control empty-lentiviral vectors (EV). MeCP2 expression was significantly decreased by about $50\%$ both at protein and mRNA levels in undifferentiated shMeCP2-BMSCs compared to EV-BMSCs (Fig. 2A, B). MeCP2 silencing, although showing a not complete abrogation of the protein level, was sufficient to significantly increase PPARγ and PLIN1 mRNA expression in undifferentiated shMeCP2-BMSCs compared to EV-BMSCs (Fig. 2C) and to up-regulate the late adipogenic markers adiponectin and leptin, tested at different time points, once induced to differentiate into adipocytes using complete adipogenic medium (Fig. 2D). Notably, no significant modulation of the osteogenesis-related markers RUNX2, OCN, and BMP2 was observed in shMeCP2-BMSCs (Fig. 2C). Figure 2D shows mRNA levels of genes related to adipogenesis in shMeCP2-BMSCs undergoing adipogenic differentiation. Notably, mRNA of ADIPOQ and LEPT were upregulated after 3 days while LEPT, FABP4, and PLIN1 after 14 days of differentiation in shMeCP2-BMSCs compared to hBMSCs induced to differentiate into adipocytes with empty vector (AD-EV). No significant difference between shMeCP2-BMSCs and AD-EV was observed for PPARγ mRNA expression (Fig. 2D).Fig. 2MeCP2 partial silencing induces adipogenesis. MeCP2 expression in undifferentiated hBMSCs infected with shRNA-containing (sh-MeCP2) or empty lentiviral vectors (EV) was analyzed by A western blot and densitometric analysis, (data were normalized to β-actin) and B by RT-PCR. C Adipogenesis (left)- and osteogenesis (right)-related mRNA fold change in hBMSCs infected with shRNA-containing lentiviral vectors vs cells infected with empty vector (EV). D Adipogenesis-related mRNA fold change in hBMSCs induced to differentiate for 3 and 14 days with shRNA-containing vectors vs hBMSCs induced to differentiate for 3 and 14 days with empty vector (AD-EV). Data are mean ± SD of three independent experiments. * t-test $p \leq 0.05$
## MiR-422a and miR-483-5p are upregulated during adipogenesis and affect MeCP2 expression in hBMSCs and hBMSC-derived adipocytes
The silencing of MeCP2 demonstrates its role in adipogenesis, therefore, we deeply explored the mechanisms underlying MeCP2 protein expression during hBMSC differentiation. We analyze the MeCP2 mRNA levels and the methylation status of CpGs among a genomic region covering the MeCP2 gene in undifferentiated-BMSCs and cells cultured for 14 days in pro-adipogenic or -osteogenic media. The results showed that MeCP2 modulation in differentiated hBMSCs does not depend on the transcription nor on the methylation status of its gene (Fig. 3A, B).Fig. 3miRNA expression in adipogenesis. A MeCP2 mRNA relative expression in arbitrary units (a.u.). Data were normalized to IPO8. B DNA methylation profile of the genomic region encompassing MeCP2 (chrX:153282264–153368188, hg19 assembly). The position of microarray probes along the chromosome is reported in the x-axis, while the y-axis reports DNA methylation levels expressed as beta values, ranging from 0 to 1. C Bar chart reporting the (log2) fold changes of the miRNAs showing a differential expression in adipocytes (AD) compared to hBMSCs. The figure showed miRNAs with absolute (foldchange) ≥ 5. In green and in red are shown miRNA upregulated and downregulated in adipocytes, respectively. D miRNA relative expression validated through Real-time PCR. E Fold-change relative expression of miRNA of adipocytes compared with MSCs from the same donors. F Western blot and densitometric analysis of MeCP2 in hBMSCs transfected with miRVANA miRNA mimic negative control #1 (CTR) and in hBMSCs transfected with miR-422a and miR-483-5p miRNA mimics. Data were normalized to β-actin. G Western blot and densitometric analysis of MeCP2 in adipocytes transfected with miRNA inhibitor negative control #1 (AD-CTR) and in adipocytes transfected with miR-422a and miR-483-5p inhibitors. Data were normalized to β-actin. H Luciferase reporter assay. HEK293 cells were infected with either negative control (NC) or miR-422a mimic, then transfected with the luciferase constructs of the wild-type MeCP2 3′-UTR (MeCP2-miR-422a-WT) or a mutated MeCP2 3′-UTR (MeCP2-miR-422a-mut). The luciferase activity was analyzed. * t-test $p \leq 0.05$; **t-test $p \leq 0.01$; ***t-test $p \leq 0.001$ These observations prompted the hypothesis that the decline in MeCP2 expression during adipogenesis could be related to post-transcriptional mechanisms, i.e. the interaction between miRNAs and MeCP2 mRNA that may inhibit the protein translation. To identify a set of miRNAs potentially involved in BMSC adipogenesis, the differential miRNA expression profiles of hBMSC-derived adipocytes (obtained cultivating hBMSCs in adipogenic medium (AD) for 14 days) and undifferentiated hBMSCs were analyzed by performing a Taqman miRNA qRT-PCR-array. PPARγ and ADIPOQ RT-PCR analysis and fat-soluble staining Oil Red images were used to verify hBMSC differentiation (Supplementary Fig. 1A). Out of 384 tested miRNAs, 15 miRNAs were significantly upregulated (fold change ≥ 5.0), while 14 were downregulated (fold change ≤ −5.0) in AD compared to hBMSCs (Fig. 3C). Complete profiling results have been deposited in NCBI’s Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo) with accession reference GSE189508.
In the attempt to identify miRNAs involved in adipogenesis with a possible role in the post-transcriptional modulation of MeCP2, we first proceeded with qPCR validation of the five most upregulated miRNAs (fold change (log2) > 4), miR-98, -139-3p, -422a, -330-5p, and -483-5p. The analysis confirmed the results of the microarray with miR-422a showing the highest expression among all miRNAs tested (Fig. 3D). However, miR-139-3p validation yielded high Ct values (> 30) in all conditions tested, suggesting its negligible expression (data not shown).
To identify miRNAs endowed with a pro-differentiating effect and those specific for adipogenesis, we also assessed their expression in osteoblasts obtained by culturing hBMSCs for 14 days in an osteogenic medium (OS) (Fig. 3D). All miRNAs were significantly up-regulated in OS, except for miR-422a which did not show a significant modulation in OS compared to hBMSCs. Osteogenesis was assessed by mRNA analysis of Osteocalcin, BMP2 and Runx2, and Alizarin Red S staining (Supplementary Fig. 1B).
To further confirm these results, the expression of the four miRNAs was tested in human BMSCs and AD isolated from the bone marrow of the same donors. Consistently, the expression of all validated miRNAs was significantly upregulated in isolated AD compared to BMSCs (Fig. 3E).
Overall, these results suggest that miR-98, miR-422a, miR-330-5p and miR-483-5p may have a role in hBMSC differentiation, with miR-422a being more specific for adipogenesis.
To test whether miR-98, miR-422a, miR-330-5p, and miR-483-5p may affect hBMSC differentiation by interacting with MeCP2, we performed a search into the ENCORI miRNA–mRNA interaction database, which collects information from five different prediction tools, observing that MeCP2 mRNA is a predicted target of miR-422a (PITA score = −16.08, interaction supported by 2 Ago CLIP-seq experiments) and miR-483-5p (PITA score = −19.01, Pan-cancer score = 6, interaction supported by 2 Ago CLIP-seq experiments), which has been previously demonstrated to downregulate MeCP2 expression by binding the 3′ UTR of its mRNA [32]. Given the upregulation of the former in differentiated hBMSCs and the specific expression of the latter in hBMSC-differentiated into adipocytes, we focused the subsequent experiments on the role of miR-422a and miR-483-5p.
To investigate whether these two miRNAs are responsible for adipocyte MeCP2 down-regulation, hBMSCs and hBMSC-differentiated adipocytes were transfected for 72 h with specific miRNA mimics or inhibitors, respectively. Figure 3F shows that both miRNA mimics strongly reduced MeCP2 expression in undifferentiated cells. In line with these results, miR-422a and miR-483-5p inhibitors induced upregulation of MeCP2 protein in adipocytes (Fig. 3G). On the other hand, miR-422a and miR-483-5p expression were not significantly modulated by MeCP2 silencing (Supplementary Fig. 2). We also performed a luciferase reporter assay, which confirmed that miR-422a binds at least one of the two 3′ untranslated regions of MeCP2 (Fig. 3H).
## MiR-422a and miR-483-5p promote adipogenesis in hBMSCs
To investigate the role of miR-422a in adipogenesis, we analyzed the expression of several specific molecular markers of adipogenesis in hBMSCs induced to differentiate for 14 days in the adipogenic medium in which miR-422a mimic was added instead of indomethacin, i.e. the component of the adipogenic cocktail exerting the greatest impact on the expression of key regulator genes of adipogenesis and lipid accumulation [35]. We also tested the effect of miR-483-5p, which we observed to be significantly up-regulated both in adipogenesis and osteogenesis (Fig. 3B) and whose role in promoting adipogenesis has already been demonstrated by previous reports [36]. We found that supplementation with miR-422a mimic induced a significant expression of the transcripts of all the genes tested (PPARγ, GLUT4, FATP1, FATP4, ACSL1, LEP, ADIPOQ, and PLIN1), in some cases (PPARγ, GLUT4, FATP1, FATP4, ACSL1) with an efficiency comparable to indomethacin (CTR +) (Fig. 4A). MiR-483-5p mimic promoted the upregulation of ACSL1, PLIN1, and ADIPOQ but less efficiently than miR-422a. qPCR analysis of intracellular miR-422a and -483-5p confirmed the efficiency of the transfection (Supplementary Fig. 3).Fig. 4miR-422a and miR-483-5p promote adipogenesis in hBMSCs. A Adipogenesis-related mRNA fold change in hBMSCs transfected with miRNA mimics and negative miRNA mimic controls (CTR− and CTR +). CTR- indicates hBMSCs induced to differentiation in absence of indomethacin, while CTR + with indomethacin. MiRNA mimics were added to the adipogenic medium without indomethacin. * vs CTR−; # vs miR-422a mimic, ° vs miR-483-5p mimic. B Adipogenesis-related mRNA fold change in hBMSCs transfected with miRNA inhibitors. MiRNA inhibitors were added to the complete adipogenic medium. CTR + indicates hBMSCs treated with miRNA inhibitor negative control #1 and induced to adipogenesis with a complete adipogenic medium. * vs CTR +; # vs miR-422a inhibitor. C Representative images and densitometric quantification of cells staining with Oil Red O. For mimic experiments: * vs CTR−; # vs miR-422a mimic, ° vs miR-483-5p mimic. For inhibitor experiments: * vs CTR +; # vs miR-422a inhibitor. D Graph chart represents adiponectin released into the culture medium expressed in ng/ml. For mimic experiments: * vs CTR−; # vs miR-422a mimic, ° vs miR-483-5p mimic. For inhibitor experiments: * vs CTR + (E) Osteogenesis-related mRNA fold change in hBMSCs transfected with miRNA mimics and inhibitors of miR-422a and miR-483-5p. * vs CTR +; # vs miR-422a mimic. Data are mean ± SD of three independent experiments. *, #, ° t-test $p \leq 0.05$; **, ##, °° t-test $p \leq 0.01$; ***, °°°t-test $p \leq 0.001$ To strengthen these data, we further analyzed the effect of miR-422a or miR-483-5p inhibition during hBMSC differentiation by adding their specific antagomiRs to the complete adipogenic medium Notably, after 14 days of culture, the expression of PPARγ, PLIN1, LEP, and ADIPOQ was significantly reduced in miR422a-antagomiR transfected cells compared to adipogenic medium alone (CTR +). MiR-483-5p inhibition yielded similar results only for LEP and ADIPOQ (Fig. 4B).
The effect of both miR-422a and miR-483-5p on adipogenesis was further investigated by Oil Red O staining of hBMSC transfected with miRNA mimics or inhibitors in the same conditions described earlier (Fig. 4C). Lipid content is increased by miR-422a as well as by miR-483-5p mimic transfection compared to the adipogenic medium without indomethacin (CTR-). Accordingly, miR-422a or miR-483-5p inhibition by antagomirs significantly reduced lipid droplet formation induced by complete adipogenic medium (CTR +).
Since adiponectin is an essential factor secreted by mature adipocytes and miR-422a and miR-483-5p mimics alone were able to induce adiponectin (ADIPOQ) mRNA expression in hBMSC cultured in adipogenic medium without indomethacin (Fig. 4A), we assessed the amount of secreted adiponectin in this conditioned medium and compared its concentration with that released in the medium enriched with indomethacin (CTR +) or without miRNA mimics and indomethacin (CTR−). Both mimics of miR-422a and miR-483-5p significantly increased adiponectin release compared to negative control. Accordingly, miRNA inhibitors were able to reduce the amount of secreted adiponectin in the complete adipogenic medium (Fig. 4D).
As previously shown in Fig. 3, miR-422a is not modulated during hBMSC osteogenesis, contrary to miR-483-5p which is strongly up-regulated. Accordingly, forced expression of miR-422a inhibitor did not affect the expression of RUNX2, the master transcription factor for osteogenesis [37], and two other osteogenic-related genes—bone morphogenetic protein 2 (BMP-2) and osteocalcin (OCN)—after 14 days of pro-osteogenic conditions. However, miR-422a mimic caused a significant strong reduction of both BMP-2 and OCN gene expression (Fig. 4E). Interestingly, miR-483-5p inhibition reduced RUNX2 expression.
## MiR-422a and -483-5p are released from differentiating adipocytes and are present in plasma from subjects with osteoporosis
To assess whether miRNAs related to hBMSC adipogenesis can be released in extracellular fluids we evaluated the levels of miR-422a and -483-5p in conditioned media harvested in the last 3 days out of the 14 days of culture in adipogenic and osteogenic medium. We observed that miR-422a level was significantly higher in the AD-conditioned medium compared to both undifferentiated hBMSC and OS-conditioned media, while miR-483-5p was similarly upregulated in both AD and OS-conditioned media compared to hBMSCs one (Fig. 5A).Fig. 5miR-422a has a higher expression in plasma of osteoporotic subjects compared with non-osteoporotic samples. A miRNA fold change in the culture medium of cells induced to differentiate into adipocytes (AD) and osteoblasts (OS) compared to hBMSCs. Data are mean ± SD of three independent experiments. * t-test $p \leq 0.05.$ B Violin plots showing miRNA relative expression in plasma of non-osteoporotic (CTR) and osteoporotic (OP) subjects. C Scatter plot showing correlations between relative miR-422a expression levels (in arbitrary units, a.u.) and T-score or Z-score. Data are expressed as a mean of 2−ΔCt normalized with cel-miR-39 To investigate the value of miR-422a and -483-5p as promising non-invasive candidate biomarkers in osteoporosis, we checked their circulating levels in subjects with primary type II osteoporosis (OP) and age-/gender-matched control subjects (CTR). The clinical and biochemical characteristics of the subjects are listed in Table 1. The levels of circulating miR-422a were higher in OP subjects compared to CTR ($$p \leq 0.0007$$), whereas no significant difference was shown for miR-483-5p (Fig. 5B). Finally, we analyzed the correlation between circulating miR-422a levels and bone mineral density assessed through dual-energy X-ray absorptiometry (DXA) in the whole cohort. Interestingly, we found significant negative correlations between circulating miR-422a and both T-score ($$p \leq 0.002$$) and Z-score ($p \leq 0.001$) (Fig. 5C).Table 1Biochemical and anthropometric characteristics of subjects enrolled for the studyVariablesControl subjects($$n = 16$$)Primary type II osteoporosis ($$n = 17$$)p-ValueAge (years)83.3 (5.9)86.3 (3.5)0.083Gender (males, %)7 ($44\%$)5 ($29\%$)0.392BMI (kg/m2)24.8 (2.0)23.4 (2.5)0.087Creatinine (mg/dL)1.3 (0.5)1.4 (0.7)0.642Sodium (meq/L)141.6 (3.5)141.2 (4.9)0.790Potassium (meq/L)4 (0.4)3.9 (0.3)0.421Calcium (mg/dL)8.9 (0.5)8.6 (0.7)0.169Hemoglobin (g/dL)12.4 (1.7)11.0 (1.3)0.012ESR (mm/h)39.6 (21.1)52.4 (31.8)0.186CRP (mg/L)2.9 (2.7)3.8 (2.7)0.346T-score0.3 (0.7)−2.3 [1] < 0.001Z-score2.4 (0.8)−0.4 (0.8) < 0.001BMD256.3 (485.6)65.1 [249]0.161Data are presented as mean (SD)BMD bone mineral density; CRP c-reactive protein; ESR erythrocyte sedimentation ratep Values for unpaired t test (continuous variables) or Chi-squared test (categorical variables)
## Discussion
Adipogenesis is intimately linked to osteogenesis in the bone marrow milieu. Since bone and adipose tissue share a common origin, the identification of factors driving the hBMSC adipogenic program is of high relevance to human diseases characterized by disruption to the differentiation balance, such as osteoporosis and aging [38]. Interestingly, it has been suggested that MeCP2 plays a role in regulating both subcutaneous adipogenic process and, furthermore, osteogenesis in a rodent model of Rett syndrome [10, 39, 40], therefore, we hypothesized that it may represent a key factor addressing hBMSCs to one or the other differentiation pathway within the bone marrow.
In this work, we show that MeCP2 expression is modulated in an opposite way in adipocytes and osteoblasts both in vitro and in vivo and that its partial silencing in hBMSCs results in the induction of a pro-adipogenic transcriptional program. We demonstrate that MeCP2 modulation does not depend on different expression levels of its mRNA and accordingly on the methylation status of its gene, but on the expression of specific miRNAs. In fact, using a microarray approach, we have identified several differentially expressed miRNAs in hBMSC-derived adipocytes compared to undifferentiated cells. We focused on the most upregulated miRNAs, which could act as post-transcriptional repressors of MeCP2 expression during adipogenesis. Notably, miRNA-mRNA target prediction analysis aimed at identifying which of the most upregulated miRNAs could regulate MeCP2 expression showed that both miR-422a and miR-483-5p are able to target MeCP2 mRNA. However, since miR-422a is the only one modulated in adipogenesis but not in osteogenesis, we thought it may be the right candidate to reduce MeCP2 expression in adipocytes. We found for the first time that forced expression of miR-422a and miR-483-5p in hBMSC can significantly downregulate MeCP2 expression and promote adipogenesis, with miR-422a being more efficient in inducing the expression of adipogenic markers, including adiponectin. Albeit the concentration of adiponectin released by cells treated with miRNA mimics did not reach the levels of positive control (adipogenic medium with indomethacin), we observed a significant > fivefold increase compared to cells treated with the medium without indomethacin. Indeed, indomethacin is the major driver of adiponectin synthesis in in vitro models of adipogenesis [41]. Interestingly, data obtained with miR-483-5p are in agreement with those found by others on its ability to inhibit MeCP2 expression during fetal development [32] and to regulate adipogenesis in the subcutaneous environment [36]. On the contrary, miR-422a and miR-483-5p inhibition rescues MeCP2 expression in adipocytes and reduces adipogenic marker expression. Interestingly, an in vitro study on human visceral preadipocytes showed that metformin inhibits adipogenesis with a concomitant reduction of miR-422a levels [42]. More in general, miR-422a increased during adipogenesis, exerted a greater impact on the adipogenic process compared to miR-483-5p and induced a significant inhibition of the osteogenic markers BMP-2 and osteocalcin. Overall, these observations suggest that miR-422a is specifically boosted in adipogenesis and may possibly inhibit osteogenesis, a hypothesis that deserves future investigations.
The great effect of miR-422a and miR-483-5p on adipogenesis may be related at least in part to the downregulation of MeCP2 expression in differentiating hBMSCs. Indeed, partial silencing of MeCP2 in undifferentiated mesenchymal cells led to the acquisition of an adipogenic profile with upregulation of PPARγ and PLIN1 mRNAs, independently from miR-422a and miR-483-5p. Of note, the expression of miR-422a and miR-483-5p does not depend on MeCP2 itself because they are not modulated by its silencing. Furthermore, this effect is also confirmed by the upregulation of adiponectin, leptin, FABP4, and PLIN1 mRNAs in silenced cells harvested in an adipogenic medium compared to cells transfected with an empty lentiviral vector. However, in this condition, we did not observe an effect on the mRNA levels of the transcription factor PPARγ, which may be covered by the strong induction of the adipogenic cocktail. In agreement with our findings, RTT patients show high levels of circulating leptin [43, 44] and adiponectin [45]. Adiponectin, one of the most extensively studied adipocyte-derived factors, exerts a plethora of beneficial effects through its insulin-sensitizing, anti-atherogenic, and anti-cancer properties [46, 47]. Since MAT appears to be the major contributor to circulating adiponectin, it has been suggested that its increase may have beneficial effects in compromised health conditions such as anorexia nervosa and chemotherapy although this increase is related to a condition of osteoporosis [14].
In fact, one of the proposed mechanisms in the pathogenesis of osteoporosis is a shift of hBMSC differentiation toward adipocyte rather than osteoblast [48]. For this reason, it is important to characterize the broad range of mediators in the bone marrow milieu that can regulate the commitment of MSCs. Studies on RTT patients and different RTT mouse models highlighted the existence of a direct relationship between MeCP2 loss of function and the alteration of bone homeostasis, which contributes to the onset of osteoporosis and to a higher risk of bone fracture [10, 49–51]. Importantly, a mouse model in which MeCP2 has been reactivated specifically in the nervous system but remained silenced elsewhere showed that bone abnormalities are due to a loss of MeCP2 in peripheral tissues [52], confirming a metabolic component in RTT syndrome. Interestingly, a recent study showed that the overexpression of MeCP2 in BMSCs enhanced the expression of osteogenic markers, including RUNX2 and osteocalcin, and promoted calcium deposition in a mouse model of osteoporosis [53]. Here, we showed that not only MeCP2 genetic silencing can affect bone biology but even that specific miRNAs affecting MeCP2 expression, i.e. miR-422a and -483-5p are capable to influence osteogenesis when hBMSCs were cultured under proper conditions. Besides MeCP2, miR-422a significantly reduced mRNA levels of other specific targets involved in bone biology, BMP2, and osteocalcin, while miR-483-5p inhibitor induced a decline in RUNX2 mRNA levels. Accordingly, a recent report showed that intra-articular injection of miR-483-5p inhibitor delays the development of osteoarthritis through the reduction in the number of RUNX2-positive chondrocytes [54]. Nonetheless, miR-483-5p resides in an intron of the IGF2 gene and it has been shown to upregulate the expression of its host gene [55], which is involved in longitudinal and appositional bone growth [56]. Overall, these data suggest that both miRNAs are somehow involved in the commitment of hBMSCs, with miR-422a exerting a more pronounced effect toward the adipogenic lineage. In addition, miR-422a showed a higher expression in the plasma of osteoporotic patients compared with non-osteoporotic controls and was negatively related to T-score and Z-score. A previous study showed that the search for circulating miRNAs as minimally invasive biomarkers for osteoporosis revealed that miR-422a is upregulated in circulating monocytes from low BMD postmenopausal women [57]. De-Ugarte and colleagues showed significant overexpression of miR-483-5p through a microRNA array on bone samples from postmenopausal women with a history of osteoporotic fractures [58]. However, this latter observation was not replicated in our cohort, probably due to the limited number of patients and to the different forms of osteoporosis (type II, age-related vs. type I, postmenopausal) considered.
In conclusion, we showed that miR-422a and miR-483-5p act as pro-differentiation factors in hBMSCs and that these two miRNAs can affect the adipogenesis process by influencing MeCP2 expression. Our findings emphasize the need to unravel MeCP2 expression and regulation in peripheral tissues, especially in bone marrow stromal cells, with a look at the potentially related diseases. A thorough comprehension of the factors capable to affect hBMSC differentiation is important not only in the context of bone mineral disease. Many efforts have been devoted to preventing metabolic complications due to the accumulation and increased secretory activity of visceral adipose tissue. Of note, BM adipose tissue sustains its own integrity through the release of extracellular vesicles (EVs) containing a typical adipocyte signature, as well as anti-osteoblastic miRNAs exerting their effects on the nearby hBMSCs [59]. It is tempting to speculate that such EVs could be secreted also in the bloodstream, affecting adipose tissue homeostasis at the systemic level. In this regard, the study of EVs as mediators of the extracellular dynamics of miR-422a and miR483-5p represents an intriguing future perspective for the present research. Future investigations are warranted to disentangle the roles of miRNAs showing opposite patterns of modulation during adipogenesis and to understand how these miRNAs integrate into a complex network regulating adipose tissue formation and function.
## Cell culture and differentiation
Human bone marrow stromal cells (hBMSCs) were purchased from Lonza (Allendale, NJ, USA) and maintained in α-MEM (Euroclone, 20016, Milano, Italy) supplemented with $10\%$ fetal bovine serum (FBS) (Lonza), 2 mM l-glutamine, 100 U/ml penicillin, 100 mg/ml streptomycin. Experiments were performed using different batches of BMSCs up to the fifth passage.
For adipogenic differentiation, BMSCs were cultured as shown in Rippo et al. [ 60]. Briefly, hBMSCs were seeded at 8 × 103 cells/cm2 on six-well plates in AD containing complete a-MEM, supplemented with 0.5 mM dexamethasone, 5 mg/ml insulin, 0.45 mM isobutylmethylxanthine (IBMX), and 0.2 mM indomethacin (Sigma-Aldrich, St. Louis, MO, USA). In specific miRNA mimics experiments, the adipogenic medium was supplemented with miR-422a and miR-483-5p mimics instead of 0.2 mM indomethacin. In these sets of experiments, we used two different adipogenic cocktails as controls, with (CTR +) or without indomethacin (CTR−).
Dedifferentiation of adipocytes obtained from three different subcutaneous fat tissue human samples was obtained as previously described in Poloni et al. [ 33]. BM-derived mesenchymal cells and adipocytes were collected from patients undergoing hip surgery and maintained in their growth medium until analysis.
All tissue samples were collected in accordance with local ethics committee guidelines (300/DG), and all participants provided their written informed consent to take part in the study. During the dedifferentiation process, mature adipocytes lost their lineage gene expression profile, assumed the typical mesenchymal morphology and immunophenotype, and expressed stem cell genes.
## Adipocyte staining
Adipocyte differentiation was assessed by Oil Red O (ORO) staining. Briefly, cells were washed with PBS and fixed with $4\%$ paraformaldehyde for 5 min. After fixation, samples were washed twice in PBS, followed by incubation with freshly filtered ORO staining solution (six parts Oil Red O stock solution and four parts H2O; Oil Red O stock solution is $0.5\%$ Oil Red O in isopropanol) for 30 min. For quantitative analysis, ORO was extracted from the cells with isopropanol and quantified spectrophotometrically at 520 nm.
## RNA extraction and RNA expression
Total RNA was recovered from hBMSCs using the Total RNA Purification Kit purchased from Norgen (#37500, Norgen Biotek, Thorold, ON, Canada) which allows the isolation of both microRNAs and mRNAs. RNA was used immediately or stored at −80 °C until analysis by Real-Time (RT)-PCR.
## mRNA analysis
1 µg of RNA was transcribed into cDNA using PrimeScript™ RT Reagent Kit with gDNA Eraser (RR047A, Takara Bio) according to the manufacturer’s instructions. One-twentieth of first-strand cDNA was used as a template for RT-PCR amplification. RT-PCR was performed with TB Green Premix Ex Taq (Tli RNase H Plus) (RR420A, Takara Bio) in a reaction volume of 10 µl with specific primers according to the protocol. β-actin or/and IPO8 was used as reference gene. Primers were listed in Supplementary Table 1.
## MicroRNA analysis
MiRNAs were reverse transcribed following the manufacturer’s instructions (#4366596, Thermo Fisher Scientific) using specific stem-loop primers for each miRNA. The RT-PCR reaction mix included TaqMan MicroRNA assay (#4427975, Thermo Fisher Scientific), TaqMan Universal Master mix no UNG (4440040, Thermo Fisher Scientific) and RT product. RNU48 was used as a reference gene. The 2−ΔCT method was used to determine miRNA expression.
## TaqMan MicroRNA array analysis of mature microRNAs
Microarray analysis was performed as previously described [61, 62]. In brief, the previously isolated RNA was reverse transcribed by priming with a mixture of looped primers using the manufacturer’s instructions (Megaplex RT primers Human Pool A v2.1, Thermo Fisher Scientific). Pre-amplification of cDNA was performed using TaqMan Preamp Master Mix (#4384266, Thermo Fisher Scientific) and Megaplex PreAmp Primers (10×), Human Pool A v2.1 (#4399233, Thermo Fisher Scientific). Pre-amplified cDNA was used for mature miRNA profiling by RT-PCR instrument equipped with a 384-well reaction plate (7900 HT, Applied Biosystems) and TaqMan Array Human MicroRNA Cards v2.0 pool A (#4398977, Thermo Fisher Scientific) containing 367 different human miRNA assays in addition to selected small nucleolar RNAs. miRNAs expressed (Ct ≤ 30) in at least one condition (hBMSCs, adipogenesis) were included in the analysis. Data were presented as log2 fold change versus undifferentiated hBMSCs.
## miRNA target prediction analysis
The open-source encyclopedia of RNA interactomes (http://starbase.sysu.edu.cn/index.php) [63] was utilized to locate target genes of miR-422a and miR-483-5p.
## Cell transfection with miRNA mimics and inhibitors
Transfection of miRNA inhibitors and mimics was conducted as previously described [64]. Briefly, 1 × 105 hBMSCs were plated in six-well plates and incubated overnight before transfection with miR-422a and miR-483-5p miRVANA miRNA mimics (MC12541, MC12629), miRVANA miRNA inhibitors (MH12541, MH12629), miRVANA miRNA inhibitor negative control #1 [4464077] or with miRVANA miRNA mimic negative control #1 (4464058, all from Thermo Fisher Scientific, San Jose, CA, USA) at a concentration of 30 nM. Transient transfection was performed using TransIT-2020 transfection reagent (MIR 5404, Mirus Bio LLC, Madison, WI, USA), according to the manufacturer’s protocol. The ratio of transfection reagent (µl)/miR (µg) equal to 2:1 was found to be optimal. Transfection was carried out concomitantly with the induction of differentiation and repeated at every medium replacement. Analyses on adipogenesis induction were performed after 14 days after transfection.
## Luciferase reporter assay
The luciferase reporter assay was performed by Creative Biogene Biotechnology (Shirley, NY, USA). The wild-type MeCP2 reporter (MeCP2-WT) and the mutant MeCP2 reporter (MeCP2-Mut) were generated by subcloning the 3′-UTR sequences of MeCP2 bracketing the predicted miR-422a-5p binding site and the full-length sequences of MeCP2-Mut into the XhoI/XbaI site located at 3′UTR of pmirGLO (Promega) vectors. The MeCP2 3′-UTR sequences were as follows: WT human MECP2 3′UTR sequence (miRNA binding sites in italics), CGACCTTGACCTCACTCAGAAGTCCAGAGTCTAGCGTAGTGCAGCAGGGCAGTAGCGGTAATACTTAGTCAAATGTAATGTGGCTTCTGGAATCATTGTCCAGAGCTGCTTCCCCGTCAC; mutant human MECP2 3′UTR sequence (mutant sites in italics), CGACCTTGACCTCACTCAGATCAGGTCTCAGATCCGTAGTGCAGCAGGGCAGTAGCGGTAATACTTAGTCAAATGTAATGTGGCTTCTGGAATCATTCAGGTCTGCTGCTTCCCCGTCAC. After ligation of the WT/mutant human MECP2 3′UTR fragments into linearized pmirGLO by recombination reaction, HEK293T cells were transfected for 48 h with the pmirGLO-UTR reporter plasmid in combination with Negative Control (NC, sequence UUCUCCGAACGUGUCACGUUU) mimics or miR-422a-5p mimics (sequence: ACUGGACUUAGGGUCAGAAGGC) at a final concentration of 20 nM in 25 μl of pure DMEM. The Firefly luciferase activity, normalized to Renilla luciferase (for transfection efficiency), was determined with the dual-luciferase reporter assay system (Promega), according to the manufacturer’s instructions, and reported as % of the negative control mimic activity.
## Adiponectin production
Cell medium was collected at the end of the experiments, centrifugated, and stored at −80 °C until used in the assay. Adiponectin concentration was measured using a commercially available high-sensitivity enzyme-linked immunosorbent assay (ELISA) (AG-45A-0001YEK-KI01, AdipoGen).
## Von Kossa staining
After osteogenic induction for 21 days, cells were fixed in $4\%$ paraformaldehyde for 20 min before being stained with $5\%$ aqueous silver nitrate solution for 45 min at room temperature under the light. Next, the samples were washed with deionized water and stained with $5\%$ sodium thiosulfate for 10 min.
## Specimen collection and immunohistochemistry
Small fragments of human femoral bone were collected from patients undergoing hip surgery and then fixed in buffered formalin $10\%$ for 24–48 h. All tissue samples were collected in accordance with local ethics committee guidelines 300/DG, and all participants provided their written informed consent to take part in the study.
After a decalcification step in neutral EDTA-sodium hydroxide, a conventional paraffin embedding procedure was performed. Inguinal and omental adipose tissue and femoral BM aspirates were obtained from adult female Sprague–Dawley albino rats ($$n = 3$$, 190–220 g; age, 3 months; Charles River, Milan, Italy). Experiments were carried out in accordance with the Council Directive $\frac{2010}{63}$EU of the European Parliament and the Council of September 22, 2010, on the protection of animals used for scientific purposes and approved by the local veterinary service. Samples were fixed in $4\%$ paraformaldehyde overnight at 4 °C, then paraffin-embedded. Subsequently, 3 µm sections were obtained from all the specimens and used for the detection of MeCP2 reactivity. Briefly, following antigen retrieval, tissues were blocked in $3\%$ H2O2 for 15 min at room temperature, washed, and then probed with rabbit polyclonal anti-MeCP2 antibody (abcam #2828. Cambridge, UK) 1:200 overnight at 4 °C in a humidified chamber. Tissues were washed extensively in PBS and detection was performed using an HRP-conjugated secondary antibody followed by DAB colorimetric detection using a kit (Cell Signalling Technology. MA, USA). Tissues were counterstained with hematoxylin, dehydrated, and mounted. Images were taken using a Nikon Eclipse 80i microscope.
## Methylation analysis
Genomic DNA was extracted from hBMSCs and hBMSC-derived adipocytes and osteoblasts using Qiagen’s QiAmp mini kit following the manufacturer’s recommendations. Each sample type was extracted and assessed for its DNA methylation profile in triplicate. Briefly, 1 µg of DNA was converted with bisulfite using the EZ DNA Methylation Kit (#D5001, Zymo Research) and analyzed using the Infinium HumanMethylationEPIC BeadChip (#20042130, Illumina), which allows assessing the methylation status of more than 850,000 CpG sites across the genome. Raw data files were extracted using the minfi Bioconductor package (CIT Preprocessing, normalization, and integration of the Illumina HumanMethylationEPIC array with minfi). Quality check resulted in the removal of 1536 probes having a detection p-value < 0.05 in more than $1\%$ of the samples and in the removal of 98,855 potentially cross-reactive probes according to Zhou et al. ( CIT Comprehensive characterization, annotation and innovative use of Infinium DNA methylation BeadChip probes). Normalization was performed using the preprocessFunnorm function implemented in minfi and DNA methylation was expressed as beta values ranging from 0 ($0\%$ of methylation) to 1 ($100\%$ of methylation). DNA methylation values in hBMSCs and hBMSC-derived adipocytes and osteoblasts were compared pairwise using the limma package and p-values were adjusted using the Benjamini–Hochberg procedure. Adjusted p-values < 0.01 were retained as significant. For the analysis of MeCP2 DNA methylation, the genomic region encompassing the gene plus 5000 bp upstream and downstream (chrX:153282264–153368188, hg19 assembly) was considered.
## Lentivirus construction and infection
LentiLox 3.7 (pLL3.7) vector system was used to induce RNA interference of MeCP2. Three different short hairpin RNA (shRNA) sequences targeting MeCP2 transcript (MeCP2sh1, MeCP2sh2, and MeCP2sh3—see Supplementary Table 2 for target sequences) were cloned into pLL3.7 as described [65]. Lentiviruses were produced by co-transfecting 293T cells with shRNA-containing pLL3.7 plasmids, or pLL3.7 empty vector, in combination with packaging plasmids as described [66]. Supernatants of transiently transfected 293T cells were recovered after 36 h and two cycles (6 h each) of infection of hBMSCs were performed within 48 h with a pool of the three shRNA-containing lentiviral vectors. EGFP positivity of target cells was monitored to verify the efficiency of infection which approximately reached $90\%$.
## Protein extraction and immunoblotting
Total protein was extracted using RIPA buffer (150 mM NaCl, 10 mM Tris, pH 7.2, $0.1\%$ SDS, $1.0\%$ Triton X-100, 5 mM EDTA, pH 8.0) containing protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN, USA) and quantified using the Bradford method. Proteins were separated on gradient SDS-PAGE gels and transferred to nitrocellulose membranes (Whatman). Membranes were then incubated with the primary antibodies overnight at 4 °C. The following primary antibodies were used: MeCP2 (D4F3) XP Rabbit (#3456, Cell Signaling Technology); β-Actin (8H10D10) Mouse mAb (#3700, Cell Signaling Technology); PPARγ (81B8) Rabbit (#2443, Cell Signaling Technology). After incubation with the specific HRP-conjugated antibody (Vector; 1:10,000 dilution), the chemiluminescent signal was detected using Clarity and/or Clarity Max (Bio-Rad, Italy) and images were acquired with Alliance Mini HD9 (Uvitec, Cambridge, UK). Densitometric analysis was performed with ImageJ software (https://imagej.nih.gov/ij/download.html). Full and uncropped Western Blots were provided as Supplemental Material.
## Plasma samples
Plasma samples were obtained from 16 healthy subjects (CTR) and 17 osteoporotic patients (OS) enrolled in the SAFARI study. Subjects were considered healthy if they did not present osteoporosis, liver diseases, renal failure, history of cancer, neurodegenerative disorders, infectious or autoimmune diseases. Samples were collected at the Ospedali Riuniti Marche Nord (Fano, Italy) hospital facilities. The procedure was approved by the Ethical Committee Regione Marche (CERM). Written informed consent was collected from all participants.
## Statistical analysis
Data are presented as mean ± standard deviation (SD) of at least three independent experiments. The Student’s t-test was applied to determine differences between samples. The correlation between circulating miR-422a levels and the Z-score and T-score score was assessed using Pearson’s correlation coefficient. Probability (p) values lower than 0.05 were considered statistically significant. The reported p-values were two-tailed in all calculations. Data were analyzed with SPSS 25.0 (SPSS Inc., IBM, Chicago, IL, USA).
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (PDF 1821 KB)
## References
1. Kaludov NK, Wolffe AP. **MeCP2 driven transcriptional repression in vitro: selectivity for methylated DNA, action at a distance and contacts with the basal transcription machinery**. *Nucleic Acids Res* (2000) **28** 1921-1928. DOI: 10.1093/nar/28.9.1921
2. Chahrour M, Jung SY, Shaw C, Zhou X, Wong ST, Qin J, Zoghbi HY. **MeCP2, a key contributor to neurological disease, activates and represses transcription**. *Science* (2008) **320** 1224-1229. DOI: 10.1126/science.1153252
3. Schmidt A, Zhang H, Cardoso MC. **MeCP2 and chromatin compartmentalization**. *Cells* (2020) **9** 878. DOI: 10.3390/cells9040878
4. Li R, Dong Q, Yuan X, Zeng X, Gao Y, Chiao C, Li H, Zhao X, Keles S, Wang Z. **Misregulation of alternative splicing in a mouse model of Rett syndrome**. *PLoS Genet* (2016) **12** e1006129. DOI: 10.1371/journal.pgen.1006129
5. Amir RE, Van den Veyver IB, Wan M, Tran CQ, Francke U, Zoghbi HY. **Rett syndrome is caused by mutations in X-linked MECP2, encoding methyl-CpG-binding protein 2**. *Nat Genet* (1999) **23** 185-188. DOI: 10.1038/13810
6. Hagberg B, Aicardi J, Dias K, Ramos O. **A progressive syndrome of autism, dementia, ataxia, and loss of purposeful hand use in girls: Rett’s syndrome: report of 35 cases**. *Ann Neurol* (1983) **14** 471-479. DOI: 10.1002/ana.410140412
7. Kyle SM, Saha PK, Brown HM, Chan LC, Justice MJ. **MeCP2 co-ordinates liver lipid metabolism with the NCoR1/HDAC3 corepressor complex**. *Hum Mol Genet* (2016) **25** 3029-3041. DOI: 10.1093/hmg/ddw156
8. Justice MJ, Buchovecky CM, Kyle SM, Djukic A. **A role for metabolism in Rett syndrome pathogenesis: new clinical findings and potential treatment targets**. *Rare Dis* (2013) **1** e27265. DOI: 10.4161/rdis.27265
9. Shapiro JR, Bibat G, Hiremath G, Blue ME, Hundalani S, Yablonski T, Kantipuly A, Rohde C, Johnston M, Naidu S. **Bone mass in Rett syndrome: association with clinical parameters and MECP2 mutations**. *Pediatr Res* (2010) **68** 446-451. DOI: 10.1203/PDR.0b013e3181f2edd2
10. O’Connor RD, Zayzafoon M, Farach-Carson MC, Schanen NC. **Mecp2 deficiency decreases bone formation and reduces bone volume in a rodent model of Rett syndrome**. *Bone* (2009) **45** 346-356. DOI: 10.1016/j.bone.2009.04.251
11. Chen Q, Shou P, Zheng C, Jiang M, Cao G, Yang Q, Cao J, Xie N, Velletri T, Zhang X. **Fate decision of mesenchymal stem cells: adipocytes or osteoblasts?**. *Cell Death Differ* (2016) **23** 1128-1139. DOI: 10.1038/cdd.2015.168
12. Kang H, Hata A. **The role of microRNAs in cell fate determination of mesenchymal stem cells: balancing adipogenesis and osteogenesis**. *BMB Rep* (2015) **48** 319-323. DOI: 10.5483/bmbrep.2015.48.6.206
13. Horowitz MC, Berry R, Holtrup B, Sebo Z, Nelson T, Fretz JA, Lindskog D, Kaplan JL, Ables G, Rodeheffer MS. **Bone marrow adipocytes**. *Adipocyte* (2017) **6** 193-204. DOI: 10.1080/21623945.2017.1367881
14. Cawthorn WP, Scheller EL, Learman BS, Parlee SD, Simon BR, Mori H, Ning X, Bree AJ, Schell B, Broome DT. **Bone marrow adipose tissue is an endocrine organ that contributes to increased circulating adiponectin during caloric restriction**. *Cell Metab* (2014) **20** 368-375. DOI: 10.1016/j.cmet.2014.06.003
15. Scheller EL, Rosen CJ. **What’s the matter with MAT? Marrow adipose tissue, metabolism, and skeletal health**. *Ann N Y Acad Sci* (2014) **1311** 14-30. DOI: 10.1111/nyas.12327
16. Ghaben AL, Scherer PE. **Adipogenesis and metabolic health**. *Nat Rev Mol Cell Biol* (2019). DOI: 10.1038/s41580-018-0093-z
17. Rosen ED, Hsu CH, Wang X, Sakai S, Freeman MW, Gonzalez FJ, Spiegelman BM. **C/EBPalpha induces adipogenesis through PPARgamma: a unified pathway**. *Genes Dev* (2002) **16** 22-26. DOI: 10.1101/gad.948702
18. Rosen ED, Sarraf P, Troy AE, Bradwin G, Moore K, Milstone DS, Spiegelman BM, Mortensen RM. **PPAR gamma is required for the differentiation of adipose tissue in vivo and in vitro**. *Mol Cell* (1999) **4** 611-617. DOI: 10.1016/S1097-2765(00)80211-7
19. Lefterova MI, Haakonsson AK, Lazar MA, Mandrup S. **PPARgamma and the global map of adipogenesis and beyond**. *Trends Endocrinol Metab* (2014) **25** 293-302. DOI: 10.1016/j.tem.2014.04.001
20. Oskowitz A, McFerrin H, Gutschow M, Carter ML, Pochampally R. **Serum-deprived human multipotent mesenchymal stromal cells (MSCs) are highly angiogenic**. *Stem cell Res* (2011) **6** 215-225. DOI: 10.1016/j.scr.2011.01.004
21. Engin AB. **MicroRNA and Adipogenesis**. *Adv Exp Med Biol* (2017) **960** 489-509. DOI: 10.1007/978-3-319-48382-5_21
22. McGregor RA, Choi MS. **microRNAs in the regulation of adipogenesis and obesity**. *Curr Mol Med* (2011) **11** 304-316. DOI: 10.2174/156652411795677990
23. Hilton C, Neville MJ, Karpe F. **MicroRNAs in adipose tissue: their role in adipogenesis and obesity**. *Int J Obes* (2013) **37** 325-332. DOI: 10.1038/ijo.2012.59
24. Wang J, Guan X, Guo F, Zhou J, Chang A, Sun B, Cai Y, Ma Z, Dai C, Li X. **miR-30e reciprocally regulates the differentiation of adipocytes and osteoblasts by directly targeting low-density lipoprotein receptor-related protein 6**. *Cell Death Dis* (2013) **4** e845. DOI: 10.1038/cddis.2013.356
25. Hamam D, Ali D, Vishnubalaji R, Hamam R, Al-Nbaheen M, Chen L, Kassem M, Aldahmash A, Alajez NM. **microRNA-320/RUNX2 axis regulates adipocytic differentiation of human mesenchymal (skeletal) stem cells**. *Cell Death Dis* (2014) **5** e1499. DOI: 10.1038/cddis.2014.462
26. Zhang XY, Xu YY, Chen WY. **MicroRNA-1324 inhibits cell proliferative ability and invasiveness by targeting MECP2 in gastric cancer**. *Eur Rev Med Pharmacol Sci* (2020) **24** 4766-4774. DOI: 10.26355/eurrev_202005_21165
27. Zhai K, Liu B, Teng J. **MicroRNA-212-3p regulates early neurogenesis through the AKT/mTOR pathway by targeting MeCP2**. *Neurochem Int* (2020) **137** 104734. DOI: 10.1016/j.neuint.2020.104734
28. Zhang N, Wei ZL, Yin J, Zhang L, Wang J, Jin ZL. **MiR-106a* inhibits oral squamous cell carcinoma progression by directly targeting MeCP2 and suppressing the Wnt/beta-Catenin signaling pathway**. *Am J Transl Res* (2018) **10** 3542-3554. PMID: 30662606
29. Yao ZH, Yao XL, Zhang Y, Zhang SF, Hu J. **miR-132 down-regulates methyl CpG binding protein 2 (MeCP2) during cognitive dysfunction following chronic cerebral hypoperfusion**. *Curr Neurovasc Res* (2017) **14** 385-396. DOI: 10.2174/1567202614666171101115308
30. Yan B, Hu Z, Yao W, Le Q, Xu B, Liu X, Ma L. **MiR-218 targets MeCP2 and inhibits heroin seeking behavior**. *Sci Rep* (2017) **7** 40413. DOI: 10.1038/srep40413
31. Zhao H, Wen G, Huang Y, Yu X, Chen Q, Afzal TA, le Luong A, Zhu J, Ye S, Zhang L. **MicroRNA-22 regulates smooth muscle cell differentiation from stem cells by targeting methyl CpG-binding protein 2**. *Arterioscler Thromb Vasc Biol* (2015) **35** 918-929. DOI: 10.1161/ATVBAHA.114.305212
32. Han K, Gennarino VA, Lee Y, Pang K, Hashimoto-Torii K, Choufani S, Raju CS, Oldham MC, Weksberg R, Rakic P. **Human-specific regulation of MeCP2 levels in fetal brains by microRNA miR-483-5p**. *Genes Dev* (2013) **27** 485-490. DOI: 10.1101/gad.207456.112
33. Poloni A, Maurizi G, Leoni P, Serrani F, Mancini S, Frontini A, Zingaretti MC, Siquini W, Sarzani R, Cinti S. **Human dedifferentiated adipocytes show similar properties to bone marrow-derived mesenchymal stem cells**. *Stem Cells* (2012) **30** 965-974. DOI: 10.1002/stem.1067
34. Poloni A, Maurizi G, Anastasi S, Mondini E, Mattiucci D, Discepoli G, Tiberi F, Mancini S, Partelli S, Maurizi A. **Plasticity of human dedifferentiated adipocytes toward endothelial cells**. *Exp Hematol* (2015) **43** 137-146. DOI: 10.1016/j.exphem.2014.10.003
35. Saben J, Thakali KM, Lindsey FE, Zhong Y, Badger TM, Andres A, Shankar K. **Distinct adipogenic differentiation phenotypes of human umbilical cord mesenchymal cells dependent on adipogenic conditions**. *Exp Biol Med* (2014) **239** 1340-1351. DOI: 10.1177/1535370214539225
36. Chen K, He H, Xie Y, Zhao L, Zhao S, Wan X, Yang W, Mo Z. **miR-125a-3p and miR-483-5p promote adipogenesis via suppressing the RhoA/ROCK1/ERK1/2 pathway in multiple symmetric lipomatosis**. *Sci Rep* (2015) **5** 11909. DOI: 10.1038/srep11909
37. Ducy P, Zhang R, Geoffroy V, Ridall AL, Karsenty G. **Osf2/Cbfa1: a transcriptional activator of osteoblast differentiation**. *Cell* (1997) **89** 747-754. DOI: 10.1016/s0092-8674(00)80257-3
38. Kawai M, de Paula FJ, Rosen CJ. **New insights into osteoporosis: the bone-fat connection**. *J Intern Med* (2012) **272** 317-329. DOI: 10.1111/j.1365-2796.2012.02564.x
39. Stachecka J, Lemanska W, Noak M, Szczerbal I. **Expression of key genes involved in DNA methylation during in vitro differentiation of porcine mesenchymal stem cells (MSCs) into adipocytes**. *Biochem Biophys Res Commun* (2020) **522** 811-818. DOI: 10.1016/j.bbrc.2019.11.175
40. Liu C, Wang J, Wei Y, Zhang W, Geng M, Yuan Y, Chen Y, Sun Y, Chen H, Zhang Y. **Fat-specific knockout of Mecp2 upregulates Slpi to reduce obesity by enhancing browning**. *Diabetes* (2020) **69** 35-47. DOI: 10.2337/db19-0502
41. Styner M, Sen B, Xie Z, Case N, Rubin J. **Indomethacin promotes adipogenesis of mesenchymal stem cells through a cyclooxygenase independent mechanism**. *J Cell Biochem* (2010) **111** 1042-1050. DOI: 10.1002/jcb.22793
42. Fujita K, Iwama H, Oura K, Tadokoro T, Hirose K, Watanabe M, Sakamoto T, Katsura A, Mimura S, Nomura T. **Metformin-suppressed differentiation of human visceral preadipocytes: involvement of microRNAs**. *Int J Mol Med* (2016) **38** 1135-1140. DOI: 10.3892/ijmm.2016.2729
43. Blardi P, de Lalla A, D’Ambrogio T, Zappella M, Cevenini G, Ceccatelli L, Auteri A, Hayek J. **Rett syndrome and plasma leptin levels**. *J Pediatr* (2007) **150** 37-39. DOI: 10.1016/j.jpeds.2006.10.061
44. Acampa M, Guideri F, Hayek J, Blardi P, De Lalla A, Zappella M, Auteri A. **Sympathetic overactivity and plasma leptin levels in Rett syndrome**. *Neurosci Lett* (2008) **432** 69-72. DOI: 10.1016/j.neulet.2007.12.030
45. Blardi P, de Lalla A, D’Ambrogio T, Vonella G, Ceccatelli L, Auteri A, Hayek J. **Long-term plasma levels of leptin and adiponectin in Rett syndrome**. *Clin Endocrinol (Oxf)* (2009) **70** 706-709. DOI: 10.1111/j.1365-2265.2008.03386.x
46. Liu Y, Palanivel R, Rai E, Park M, Gabor TV, Scheid MP, Xu A, Sweeney G. **Adiponectin stimulates autophagy and reduces oxidative stress to enhance insulin sensitivity during high-fat diet feeding in mice**. *Diabetes* (2015) **64** 36-48. DOI: 10.2337/db14-0267
47. Parida S, Siddharth S, Sharma D. **Adiponectin, obesity, and cancer: clash of the bigwigs in health and disease**. *Int J Mol Sci* (2019) **20** 2519. DOI: 10.3390/ijms20102519
48. Hu L, Yin C, Zhao F, Ali A, Ma J, Qian A. **Mesenchymal stem cells: cell fate decision to osteoblast or adipocyte and application in osteoporosis treatment**. *Int J Mol Sci* (2018) **19** 360. DOI: 10.3390/ijms19020360
49. Pecorelli A, Cordone V, Schiavone ML, Caffarelli C, Cervellati C, Cerbone G, Gonnelli S, Hayek J, Valacchi G. **Altered bone status in Rett syndrome**. *Life (Basel)* (2021) **11** 521. DOI: 10.3390/life11060521
50. Kamal B, Russell D, Payne A, Constante D, Tanner KE, Isaksson H, Mathavan N, Cobb SR. **Biomechanical properties of bone in a mouse model of Rett syndrome**. *Bone* (2015) **71** 106-114. DOI: 10.1016/j.bone.2014.10.008
51. Blue ME, Boskey AL, Doty SB, Fedarko NS, Hossain MA, Shapiro JR. **Osteoblast function and bone histomorphometry in a murine model of Rett syndrome**. *Bone* (2015) **76** 23-30. DOI: 10.1016/j.bone.2015.01.024
52. Ross PD, Guy J, Selfridge J, Kamal B, Bahey N, Tanner KE, Gillingwater TH, Jones RA, Loughrey CM, McCarroll CS. **Exclusive expression of MeCP2 in the nervous system distinguishes between brain and peripheral Rett syndrome-like phenotypes**. *Hum Mol Genet* (2016) **25** 4389-4404. DOI: 10.1093/hmg/ddw269
53. Ji W, Sun X. **Methyl-CpG-binding protein 2 promotes osteogenic differentiation of bone marrow mesenchymal stem cells through regulating forkhead box F1/Wnt/beta-Catenin axis**. *Bioengineered* (2022) **13** 583-592. DOI: 10.1080/21655979.2021.2012357
54. Wang H, Zhang H, Sun Q, Wang Y, Yang J, Yang J, Zhang T, Luo S, Wang L, Jiang Y. **Intra-articular delivery of antago-miR-483-5p inhibits osteoarthritis by modulating matrilin 3 and tissue inhibitor of metalloproteinase 2**. *Mol Ther* (2017) **25** 715-727. DOI: 10.1016/j.ymthe.2016.12.020
55. Liu M, Roth A, Yu M, Morris R, Bersani F, Rivera MN, Lu J, Shioda T, Vasudevan S, Ramaswamy S. **The IGF2 intronic miR-483 selectively enhances transcription from IGF2 fetal promoters and enhances tumorigenesis**. *Genes Dev* (2013) **27** 2543-2548. DOI: 10.1101/gad.224170.113
56. Uchimura T, Hollander JM, Nakamura DS, Liu Z, Rosen CJ, Georgakoudi I, Zeng L. **An essential role for IGF2 in cartilage development and glucose metabolism during postnatal long bone growth**. *Development* (2017) **144** 3533-3546. DOI: 10.1242/dev.155598
57. Cao Z, Moore BT, Wang Y, Peng XH, Lappe JM, Recker RR, Xiao P. **MiR-422a as a potential cellular microRNA biomarker for postmenopausal osteoporosis**. *PLoS One* (2014) **9** e97098. DOI: 10.1371/journal.pone.0097098
58. De-Ugarte L, Yoskovitz G, Balcells S, Guerri-Fernandez R, Martinez-Diaz S, Mellibovsky L, Urreizti R, Nogues X, Grinberg D, Garcia-Giralt N. **MiRNA profiling of whole trabecular bone: identification of osteoporosis-related changes in MiRNAs in human hip bones**. *BMC Med Genomics* (2015) **8** 75. DOI: 10.1186/s12920-015-0149-2
59. Martin PJ, Haren N, Ghali O, Clabaut A, Chauveau C, Hardouin P, Broux O. **Adipogenic RNAs are transferred in osteoblasts via bone marrow adipocytes-derived extracellular vesicles (EVs)**. *BMC Cell Biol* (2015) **16** 10. DOI: 10.1186/s12860-015-0057-5
60. Rippo MR, Babini L, Prattichizzo F, Graciotti L, Fulgenzi G, TomassoniArdori F, Olivieri F, Borghetti G, Cinti S, Poloni A. **Low FasL levels promote proliferation of human bone marrow-derived mesenchymal stem cells, higher levels inhibit their differentiation into adipocytes**. *Cell Death Dis* (2013) **4** e591. DOI: 10.1038/cddis.2013.115
61. Olivieri F, Lazzarini R, Recchioni R, Marcheselli F, Rippo MR, Di Nuzzo S, Albertini MC, Graciotti L, Babini L, Mariotti S. **MiR-146a as marker of senescence-associated pro-inflammatory status in cells involved in vascular remodelling**. *Age (Dordr)* (2013) **35** 1157-1172. DOI: 10.1007/s11357-012-9440-8
62. Olivieri F, Spazzafumo L, Santini G, Lazzarini R, Albertini MC, Rippo MR, Galeazzi R, Abbatecola AM, Marcheselli F, Monti D. **Age-related differences in the expression of circulating microRNAs: miR-21 as a new circulating marker of inflammaging**. *Mech Ageing Dev* (2012) **133** 675-685. DOI: 10.1016/j.mad.2012.09.004
63. Li JH, Liu S, Zhou H, Qu LH, Yang JH. **starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein–RNA interaction networks from large-scale CLIP-Seq data**. *Nucleic Acids Res* (2014) **42** D92-D97. DOI: 10.1093/nar/gkt1248
64. Giuliani A, Cirilli I, Prattichizzo F, Mensà E, Fulgenzi G, Sabbatinelli J, Graciotti L, Olivieri F, Procopio AD, Tiano L. **The mitomiR/Bcl-2 axis affects mitochondrial function and autophagic vacuole formation in senescent endothelial cells**. *Aging (Albany NY)* (2018) **10** 2855-2873. DOI: 10.18632/aging.101591
65. Rubinson DA, Dillon CP, Kwiatkowski AV, Sievers C, Yang L, Kopinja J, Rooney DL, Zhang M, Ihrig MM, McManus MT. **A lentivirus-based system to functionally silence genes in primary mammalian cells, stem cells and transgenic mice by RNA interference**. *Nat Genet* (2003) **33** 401-406. DOI: 10.1038/ng1117
66. Amatori S, Papalini F, Lazzarini R, Donati B, Bagaloni I, Rippo MR, Procopio A, Pelicci PG, Catalano A, Fanelli M. **Decitabine, differently from DNMT1 silencing, exerts its antiproliferative activity through p21 upregulation in malignant pleural mesothelioma (MPM) cells**. *Lung Cancer* (2009) **66** 184-190. DOI: 10.1016/j.lungcan.2009.01.015
|
---
title: Hypoxia-Inducible Factor-1α Protects Against Intervertebral Disc Degeneration
Through Antagonizing Mitochondrial Oxidative Stress
authors:
- Wen Yang
- Chunwang Jia
- Long Liu
- Yu Fu
- Yawei Wu
- Zhicheng Liu
- Ruixuan Yu
- Xiaojie Ma
- Ao Gong
- Fangming Liu
- Yanni Xia
- Yong Hou
- Yuhua Li
- Lei Zhang
journal: Inflammation
year: 2022
pmcid: PMC9971142
doi: 10.1007/s10753-022-01732-y
license: CC BY 4.0
---
# Hypoxia-Inducible Factor-1α Protects Against Intervertebral Disc Degeneration Through Antagonizing Mitochondrial Oxidative Stress
## Abstract
Intervertebral disc degeneration (IVDD) demonstrates a gradually increased incidence and has developed into a major health problem worldwide. The nucleus pulposus is characterized by the hypoxic and avascular environment, in which hypoxia-inducible factor-1α (HIF-1α) has an important role through its participation in extracellular matrix synthesis, energy metabolism, cellular adaptation to stresses and genesis. In this study, the effects of HIF-1α on mouse primary nucleus pulposus cells (MNPCs) exposed to TNF-α were observed, the potential mechanism was explored and a rabbit IVDD model was established to verify the protective role of HIF-1α on IVDD. In vitro results demonstrated that HIF-1α could attenuate the inflammation, apoptosis and mitochondrial dysfunction induced by TNF-α in MNPCs; promote cellular anabolism; and inhibit cellular catabolism. In vivo results demonstrated that after establishment of IVDD model in rabbit, disc height and IVD extracellular matrix were decreased in a time-dependent manner, MRI analysis showed a tendency for decreased T2 values in a time-dependent manner and supplementation of HIF-1α improved histological and imaginative IVDD while downregulation of HIF-1α exacerbated this degeneration. In summary, HIF-1α protected against IVDD, possibly through reducing ROS production in the mitochondria and consequent inhibition of inflammation, metabolism disorders and apoptosis of MNPCs, which provided a potential therapeutic instrument for the treatment of IVDD diseases.
## INTRODUCTION
As the population ages, intervertebral disc degeneration (IVDD) demonstrates a gradually increased incidence and has developed into a major health problem worldwide [1–3]. In addition to age, the risk factors of IVDD also include inflammatory cytokines, mechanical trauma, genetic susceptibility, lifestyle factors, certain metabolic disorders and so on [4–9]. IVDD is the main cause of disability because it often causes chronic low back pain (LBP) [10, 11]. Current treatments of IVDD mainly include pharmacological and surgical interventions, aiming for managing symptoms and minimizing disability. However, both of them are costly, often result in complications and have questionable efficacy [12]. Thus, more and more studies have focused on new therapies for IVDD [13, 14].
The nucleus pulposus is characterized by the hypoxic and avascular environment, in which hypoxia-inducible factor-1α (HIF-1α) has an important role through its participation in extracellular matrix (ECM) synthesis, energy metabolism, cellular adaptation to stresses and genesis [15, 16]. In the late stage of IVDD, the expression of HIF-1α is significantly decreased, and neovascularization increases the oxygen concentration. A study reports that HIF-1α can attenuate the apoptosis of nucleus pulposus-derived stem cells induced by excessive mechanical load [14]. IVDD is characterized by increased levels of pro-inflammatory cytokines such as TNF, IL-1β and IL-17. These cytokines are secreted by the IVD cells and can promote ECM degradation, changes in IVD cell phenotype and chemokine production, which consequently results in the degeneration of IVD tissues [17]. Tumour necrosis factor-α (TNF-α) can exacerbate the inflammatory process and is demonstrated to be a key regulator in the development of IVDD [18]. However, the effects of HIF-1α on mouse primary nucleus pulposus cells (MNPCs) exposed to TNF-α and the potential mechanism are still not investigated. In this study, the effects of HIF-1α on MNPCs exposed to TNF-α were observed and the potential mechanism was explored, and a rabbit IVDD model was established to verify the protective role of HIF-1α on IVDD.
## HIF-1α Attenuated TNF-α-Induced Inflammation in Primary Mouse Nucleus Pulposus Cells (MNPCs)
Primary mouse nucleus pulposus cells (MNPCs) were isolated and then co-cultured with TNF-α, TNF-α and ML228 or TNF-α and Oltipraz for 24 h to detect the mRNA levels of COX-2 and iNOS and 48 h to detect their protein levels. As shown in Fig. 1a, b, ML228 (HIF-1α activator) could reduce the elevated mRNA expression levels of iNOS and COX-2 induced by TNF-α while Oltipraz (HIF-1α inhibitor) could not. The Western blotting results (Fig. 1c–e) and immunofluorescence (Fig. 1f–i) were completely consistent with the real-time PCR results. These suggested that HIF-1α significantly alleviated the TNF-α-induced inflammatory response in MNPCs. Fig. 1HIF-1α attenuated TNF-α-induced inflammation in primary mouse nucleus pulposus cells (MNPCs). a, b The mRNA levels of iNOS and COX-2 after cells were treated as above were detected using real-time PCR. c–e The protein levels of iNOS and COX-2 were detected using Western blotting and f–i immunofluorescence staining. The scale bar was 50 μm. The values were the mean of at least three independent experiments. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ and ****$P \leq 0.0001.$
## HIF-1α Promoted Cellular Anabolism and Inhibited Catabolism of MNPCs
Figure 2a–c shows that activation of HIF-1α could upregulate the decreased mRNA levels of Col-2 induced by TNF-α and simultaneously downregulate the increased mRNA levels of MMP-13 and ADAMTS-5 induced by TNF-α while inhibition of HIF-1α could not. Figure 2d–h shows that activation of HIF-1α could upregulate the decreased protein levels of Col-2 and Aggrecan induced by TNF-α and simultaneously downregulate the increased protein levels of MMP-13 and ADAMTS-5 induced by TNF-α while inhibition of HIF-1α could not. Immunofluorescence (Fig. 2i, j) was also performed to demonstrate the expression level of MMP-13, and the results were the same as the above. Fig. 2HIF-1α promoted cellular anabolism and inhibited the catabolism of MNPCs. a–c The expression levels of Col-2, MMP-13 and ADAMTS-5 were detected using real-time PCR. d–h The protein levels of Aggrecan, Col-2, MMP-13 and ADAMTS-5 were assayed using Western blotting. i, j The expression levels of MMP-13 were detected by immunofluorescence staining. The scale bar was 50 μm. The values were the mean of at least three independent experiments. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ and ****$P \leq 0.0001$
## HIF-1α Alleviated the TNF-α-Mediated Apoptosis in MNPCs
Figure 3a–c shows that activation of HIF-1α could downregulate the increased mRNA levels of Bax and Caspase-3 induced by TNF-α but upregulate the mRNA level of Bcl-2, while inhibition of HIF-1α could not. Figure 3d–h shows the same results as the above at protein levels by Western blotting, and the inhibitory effect of HIF-1α on cleaved Caspase-3 was particularly significant. Moreover, TUNEL staining of cells (Fig. 3i, j) was performed, which showed that TNF-α promoted cell death while HIF-1α repressed this disorganization. Additionally, flow cytometry (Fig. 3k) was performed to test Annexin/PI to reflect the apoptosis rate of MNPCs, which indicated that TNF-α exaggerated the apoptosis of MNPCs, and this phenomenon could be diminished by HIF-1α. Fig. 3HIF-1α alleviated the TNF-α-mediated apoptosis in MNPCs. a–c The expression levels of Bcl-2, BAX and Caspase-3 were detected using real-time PCR. d–h The protein levels of Bcl-2, BAX and C-Caspase-3 were assayed using Western blotting. i, j TUNEL staining was performed to examine the apoptosis. k Flow cytometry was performed to detect apoptosis. The scale bar was 100 μm. The values were the mean of at least three independent experiments. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ and ****$P \leq 0.0001.$
## HIF-1α Alleviated TNF-α-Induced Mitochondrial Dysfunction in MNPCs
DCFDA assays were used to evaluate ROS synthesis in the mitochondria. As shown in Fig. 4a, b, activation of HIF-1α could decrease the elevated ROS synthesis induced by TNF-α while inhibition of HIF-1α could not. Moreover, JC-1 and Mitotracker assays were performed to detect the membrane potential of the mitochondria (Fig. 4c–f), which showed TNF-α exacerbated the dysfunction of the mitochondria in MNPCs, and activation of HIF-1α could alleviate the mitochondrial dysfunction while inhibition of HIF-1α could not. Fig. 4HIF-1α alleviated TNF-α-induced mitochondrial dysfunction in MNPCs. a, b ROS levels of MNPCs were detected with DCFDA. c, d JC-1 and e, f MitoTracker were performed to detect the mitochondrial membrane potential of MNPCs. The scale bar was 50 μm. The values were the mean of at least three independent experiments. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ and ****$P \leq 0.0001.$
## Exogenous Supplementation of HIF-1α Exhibited a Protective Effect on Degeneration of NP Tissue In Vivo
The rabbit IVDD model was established, and HIF-1α recombinant protein or Oltipraz were locally delivered into the NP tissue. Before the rabbits were executed, an X-ray was performed and the IVD height was demonstrated to be reduced, which could be improved by the application of HIF-1α (Fig. 5a). MRI was performed at 3, 6, 11 and 14 weeks, which showed a higher signal intensity of IVD in the HIF-1α supplementation group compared with the PBS group and Oltipraz group (Fig. 5b), and relative signal intensity at week 14 is shown in Fig. 5c. These results suggested that HIF-1α could alleviate the degeneration phenotype of IVD. The IVD samples were then collected for histological analysis. As shown in Fig. 5d–g, HE, Safranin O and Masson staining results indicated that morphological degeneration score such as the height of IVD in this IVDD model was alleviated by supplementation of HIF-1α in comparison with the PBS group and Oltipraz group. Besides, Safranin O and Masson staining results showed that HIF-1α reduced proteoglycan and collagen loss during the process of IVDD while suppression of HIF-1α expression aggravated that (Fig. 6).Fig. 5Exogenous supplementation of HIF-1α exhibited a protective effect on the regeneration of NP tissue in vivo. a X-ray and b, c MRI to assess the degree of IVDD in rabbits from the sham group, PBS group, HIF-1α group and Oltipraz group. d–g HE, Safranin O and Masson staining indicated morphological degeneration score such as the height of IVD in this IVDD model was alleviated by supplementation of HIF-1α in comparison with the PBS group and Oltipraz group. Safranin O and Masson staining showed that HIF-1α reduced proteoglycan and collagen loss during the process of IVDD, but suppression of HIF-1α expression aggravated that. The histological scores of each indicated group were calculated according to the grading scale previously published. h Photos of the surgical procedure in vivo. The scale bar is 500 or 100 μm. The values are the mean of at least three independent experiments. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ and ****$P \leq 0.0001.$Fig. 6Schematic depicting the proposed model for the role of HIF-1α in IVDD based on this study.
## Isolation and Culture of Mouse Primary Nucleus Pulposus Cells (MNPCs)
In this study, mice were sacrificed by cervical vertebra dislocation and then soaked in $75\%$ ethyl alcohol for 10 min to disinfect the entire body. After the dorsal hair had been shaved, the whole spine was separated from the back. The disc tissue was separated under a microscope, cut into pieces and placed in culture dishes. The cells were digested with $0.2\%$ collagenase type II (Gibco, USA) at 37 °C for 8 h. The cells were then cultured in DMEM/F12 (HyClone, USA) supplemented with $10\%$ foetal bovine serum (FBS, Gibco, USA), $1\%$ penicillin and streptomycin (P1400, SolarBio, China) under standard incubation conditions (37 °C, $5\%$ CO2). The culture medium was replaced every 3 days, and the cells were passaged when they reached 80–$90\%$ confluence. The cells from within five generations were used in all vitro experiments. In subsequent experiments, the control-group and TNF-α (HY-P7058, MCE, USA) (10 ng/ml) group were cultured under standard incubation conditions (37 °C, $5\%$ CO2), while the ML228 (HY-12754, MCE, USA) (1 μM) group and Oltipraz (HY-12519, MCE, USA) (10 μM) group were cultured under hypoxic conditions (37 °C, $1\%$ O2, $5\%$ CO2 and $94\%$ N2).
## Western Blotting Analysis
Total protein was extracted from MNPCs of each group with the precooled RIPA Lysis Buffer (P0013C, Beyotime Biotechnology) containing 1 mM PMSF on ice for 30 min. The collected liquid was centrifuged at 12,000 rpm for 15 min at 4 °C, and the supernatant was retained. Protein concentration was detected with a BCA protein assay kit (PC0020, SolarBio). Then, to destroy the 3-dimensional protein structure, the proteins in the loading buffer were heated at 100 °C for 10 min. An equal amount of protein from each sample was separated by SDS-PAGE on $8\%$, $10\%$ or $12\%$ SDS–polyacrylamide gels and then transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, USA). After being blocked with QuickBlockTM Blocking Buffer (P0252, Beyotime Biotechnology) for 20 min at room temperature, the membranes were incubated with anti-iNOS(1:1000, 18,985–1-AP, Proteintech), anti-COX-2(1:1000, 27,308–1-AP, Proteintech), anti-Tubulin (1:1000, 10,068–1-AP, Proteintech), anti-Aggrecan (1:1000, 13,880–1-AP, Proteintech), anti-Col-2 (1:1000, 28,459–1-AP, Proteintech), anti-ADAMTs-5 (1:1000, ab41037, Abcam), anti-MMP-13 (1:1000, 18,165–1-AP, Proteintech), anti-Bcl-2 (1:1000, ab196495, Abcam), anti-Bax (1:1000, BM3964, Boster) and anti-Caspase-3 (1:1000, 19,677–1-AP, Proteintech) antibodies at 4 °C overnight. The next day, after washing with Tris-buffered saline Tween-20 (TBST), these membranes were incubated with goat anti-rabbit IgG-HRP secondary antibody (1:5000, Jackson ImmunoResearch) at room temperature for 1 h. Bound antibody was visualized using an enhanced chemiluminescence system (Amersham Life Science, Arlington Heights, IL, USA), and the density of protein bands was quantified using the ImageJ software.
## Real-Time PCR
An RNAfast200 Kit (220011, Fastagen) was used to extract the total RNA from the MNPCs of each group according to the recommended procedure. Total RNA (1 µg) was reverse-transcribed to complementary DNA (cDNA) using HiScript II Q RT SuperMix for qPCR (R222-01, Vazyme). Real-time PCR was carried out with RealStar Fast SYBR qPCR Mix (A301, GenStar). The experiment was repeated three times for each target gene of each group. The nucleotide sequences of the primers are listed in Table 1. The expression levels of target genes were normalized to Tubulin and were calculated by the 2 − ΔΔCT method. Table 1Primers Used for Quantitative Real-Time PCRSourceTargetForward primer, 5′-3′Reverse primer, 5′-3′MouseCOX-2AATGCTGACTATGGCTACAAAAAAAACTGATGCGTGAAGTGCTGiNOSACAGGAGGGGTTAAAGCTGCTTGTCTCCAAGGGACCAGGMMP-13ACTTTGTTGCCAATTCCAGGTTTGAGAACACGGGGAAGACADAMTS-5GCATTGACGCATCCAAACCCCGTGGTAGGTCCAGCAAACAGTTACCol-2ACTAGTCATCCAGCAAACAGCCAGGTTGGCTTTGGGAAGAGACBcl-2TGTGGTCCATCTGACCCTCCACATCTCCCTGTTGACGCTCTBaxCTGAGCTGACCTTGGAGCGACTCCAGCCACAAAGATGCaspase-3AGGAGGGACGAACACGTCTCAAAGAAGGTTGCCCCAATCTβ-TubulinCAGCGATGAGCACGGCATAGACCCAGGTTCCAAGTCCACCAGAATG
## Immunofluorescence Staining
The cells were treated as indicated, and after 24 h, the cells were fixed with $4\%$ paraformaldehyde for 20 min. After being permeabilized with $0.2\%$ Triton X-100 for 20 min, the samples were blocked by BSA at 37 ℃ for 1 h. Then, the cells were incubated with anti-iNOS (1:500, 18985–1-AP, Proteintech), anti-COX-2 (1:500, 27308–1-AP, Proteintech) and MMP-13 (1:500, 18,165–1-AP, Proteintech) antibodies at 4 °C overnight. The next day, the cells were incubated with fluorescently labelled goat anti-rabbit IgG (1:100, Abbkine) for 1 h at 37 ℃. The nuclei were stained with DAPI. The images were taken using a fluorescence microscope (ZEISS Vert. A1) and analysed with the ImageJ software.
## TUNEL Staining
To examine the apoptosis of MNPCs in each experimental group, cells were stained with a TMR (red) Tunel Cell Apoptosis Detection Kit (G1502, Servicebio). All the procedures were performed according to the manufacturer’s instructions. The images were captured using a fluorescence microscope (ZEISS Vert. A1).
## Flow Cytometry
The apoptosis of MNPCs from each group was detected by flow cytometry. Cells were stained with propidium iodide (PI) and Annexin V-FITC for 15 min at room temperature in the dark with a FITC Annexin V Apoptosis Detection Kit (E-CK-A211, Elabscience) according to the manufacturer’s instructions. Cell apoptosis was detected with a CytoFLEX S flow cytometer (Beckman Coulter, USA), and the data obtained were analysed with the CtyExpert software.
## Reactive Oxygen Species Assay
To detect intracellular reactive oxygen species (ROS), we used an ROS assay kit (S0033, Beyotime Biotechnology). All the procedures were performed according to the manufacturer’s instructions. Briefly, after washing twice with sterile PBS, cells were stained with 10 μM DCFDA at 37 °C for 20 min in the dark. Then, the cells were washed with a basal culture medium three times. The images were captured using a fluorescence microscope (ZEISS Vert. A1).
## JC-1 Assay
The mitochondrial membrane potential changes of MNPCs after treatment were detected with a JC-1 assay kit (C2006, Beyotime). Based on the manufacturer’s instructions, each group’s cells were stained with the JC-1 staining solution at 37 °C for 20 min to protect them from light. Then, the cells were washed twice with JC-1 staining buffer, and the images were observed and captured using a fluorescence microscope (ZEISS Vert. A1).
## MitoTracker Assay
MitoTracker staining was performed to visualize the mitochondria and detect the mitochondrial membrane potential of each group following the instructions of the Mito-Tracker Red CMXRos (C1049B, Beyotime Biotechnology). The cells were incubated with the culture medium containing 20 nM Mito-Tracker Red CMXRos for 30 min at 37 ℃ in the dark. Then, the images were captured using a fluorescence microscope (ZEISS Vert. A1) after changing the fresh culture medium.
## X-Ray and Magnetic Resonance Imaging (MRI)
The rabbits in each group were performed an X-ray before execution. Radiographs were captured at a collimator-to-film distance of 66 cm, an exposure of 63 mAs and a penetration power of 35 kv. MRI was performed for each group at 3, 6, 11 and 14 weeks, and T2-weighted images (repetition time: 3000 ms; echo time: 80 ms; field of view: 200 mm2; slice thickness: 1.4 mm) were obtained by MRI using a 1.5-T system (GE) in the sagittal plane. The MRI grade of NPs was evaluated as previously reported.
## Histological Staining
The rabbits were sacrificed at 14 weeks after indicated surgery, and the IVD tissues were collected and fixed in $4\%$ paraformaldehyde for 2 days. After decalcification in $10\%$ EDTA (pH 7.2–7.4), the samples were processed, embedded in paraffin and cut into 5-μm sections using a microtome. H&E staining was performed to evaluate the morphological changes of the nucleus pulposus with a H&E Staining Kit (EE0012, Sparkjade), and the histological grading of these samples was evaluated in accordance with the grading scale based on the morphology of AF and the cellularity of NP. Safranin O staining was performed to detect the changes in proteoglycans with a Safranin O staining kit (G1371, SolarBio) according to the manufacturer’s recommended procedure. Masson staining was performed to confirm collagen loss of these samples with a Masson’s Trichrome Stain Kit (G1340, SolarBio) according to the manufacturer’s recommended procedure. The images were captured by a microscope (Leica DMI3000B).
## Surgery Procedure
The protocol of this study was approved by the Institutional Animal Care and Use Committee. Twelve New Zealand white rabbits (female), ranging from 2.5 to 3.0 kg in body weight (Jinfeng Experimental Animal Co. Ltd., Jinan, China), were used in this study. Rabbits were housed in separate cages under standard conditions with a light–dark cycle (12 h-12 h) and dry-bulb room temperature at 22–24 °C and provided ad libitum access to tap water and food pellets daily. Rabbits were anaesthetized by an intravenous injection of $10\%$ chloral hydrate (3 mL/kg). Rabbits were then placed into a left lateral position, and the vertical line for outward $\frac{1}{3}$ of the connecting line between the anterior superior spine and navel or the outer margin of the transverse process was chosen as the incision. The external oblique muscle was outwardly separated from the beginning of its fascia to find the outer margin of the longissimus muscle, the deep fascia was cut open to locate the reference transverse process, the transverse abdominis was separated to expose the psoas major, and the psoas major was retracted toward the abdomen and the position was leaned 20° toward the back to expose the vertebral body. The lumbar spine’s lowest levels (L6–L7) should be avoided to eliminate possible influences of the lumbosacral junction. After the nucleus pulposus was removed, HIF-1α recombinant protein (11977-H07E, Sino Biological) was injected into the target intervertebral disc in the HIF group, and Oltipraz (HY-12519, MCE, USA) was injected into the target intervertebral disc in the HIF inhibitor group. PBS was injected into the target intervertebral disc in the PBS group. The sham group only underwent surgery, but no substance was injected into the intervertebral disc. MRI examinations were performed at 3, 6, 11 and 14 weeks postoperatively, and X-ray examinations were performed before execution. After euthanasia, disc specimens were obtained, and histological analysis was performed.
## Statistical Analysis
Analysis of data was performed with GraphPad Prism (GraphPad Software Inc., USA). Comparisons of various groups were performed using analysis of variance (ANOVA) with Tukey’s post hoc test. Data were presented as “mean ± SEM”. Statistical significance was indicated when two-sided $P \leq 0.05.$
## DISCUSSION
In vitro results demonstrated that HIF-1α could attenuate the inflammation, apoptosis and mitochondrial dysfunction induced by TNF-α in MNPCs, promote cellular anabolism and inhibit cellular catabolism.
TNF-α is an important pro-inflammatory cytokine which can stimulate an inflammatory cascade through binding to the TNFR [17, 19–21]. Its level is significantly elevated in the disc tissue and peripheral serum of patients with IVDD [22–24]. TNF-α has been shown to be associated with a variety of pathological IDD processes such as inflammatory cascade, ROS production, apoptosis, ECM degradation, pyroptosis and proliferation. These indicate their critical role in the development of IVDD [25, 26]. Additionally, it can also induce an inflammatory response in the nucleus pulposus, leading to an increase of iNOS and COX2 and acceleration of intervertebral disc destruction [27]. Therefore, MNPCs stimulated with TNF-α were used to investigate the effects of HIF-1α on protecting against IVDD in this study. Through co-culturing MNPCs with TNF-α, TNF-α and ML228 or TNF-α and Oltipraz, the results showed that activation of HIF-1α could downregulate the expression of iNOS and COX-2 induced by TNF-α while inhibition of HIF-1α could counteract the effect. This suggested that HIF-1α could possibly attenuate the inflammatory response in IVDD.
Degradation of the nucleus pulposus extracellular matrix is an important cause of IVDD [28]. TNF-α can cause IVDD by promoting the expression of MMP-13 and ADAMTS-5 to enhance catabolism and inhibiting the synthesis of Aggrecan and Col-2 to reduce anabolism. Studies have found that HIF-1α is closely related to the synthesis of ECM in chondrocytes [29, 30]. In our study, activation of HIF-1α could inhibit catabolism through downregulating the elevated expression of MMP-13 and ADAMTS-5 induced by TNF-α and enhance anabolism through upregulating the decreased expression of Aggrecan and Col-2 induced by TNF-α while inhibition of HIF-1α could not.
Cellular senescence and death occur widely in various tissues, including apoptosis, necrosis and autophagy, and the loss of nucleus pulposus cells due to apoptosis is one of the important causes of IVDD [31]. Studies have reported that massive death of NP cells and significant degeneration of IVD were observed in mice after conditional knockout of the HIF-1α gene [32, 33]. It is well known that TNF-α can induce the apoptosis of NP cells through increasing the expression of proapoptotic cytokines such as BAX and C-Caspase-3 and reducing the expression of antiapoptotic cytokines such as Bcl-2 [34–36], which has been considered as a potential target for investigation of IVDD. In our study, activation of HIF-1α could downregulate the increased expressions of BAX and C-Caspase-3 induced by TNF-α and upregulate the decreased expression of Bcl-2 induced by TNF-α while inhibition of HIF-1α could counteract the effect. Moreover, TUNEL and flow cytometry also showed the attenuated effect of HIF-1α on the apoptosis of NP cells induced by TNF-α.
The mitochondria, as the centre of cellular energy metabolism, is involved in a variety of signalling pathways and regulates cellular function and survival [37]. After the mitochondria is exposed to adverse stimulation, multiple harmful changes will occur such as increased oxidative stress, swelling and deformation and decreased membrane potential [38, 39]. These changes can lead to inflammasome activation, cell senescence and death, which plays a crucial role in some degenerative diseases [40]. Studies have reported abnormal mitochondrial morphology and dysfunction in ageing NP cells, and mitochondrial dysfunction plays a detrimental role in the development of IVDD [41]. TNF-α can cause damage to the mitochondrial structure, such as swelling and deformation in NP cells [42]. HIF-1α is a transcription factor that responds to the reduction of intracellular oxygen concentration and can enhance cellular resistance to oxidative stress as an endogenous anti-oxidative stress regulator [43]. In our study, activation of HIF-1α reversed the enhanced ROS production and impaired membrane potential in the mitochondria induced by TNF-α while inhibition of HIF-1α was not.
In addition to in vitro experiments, we also evaluated the radiological and histological changes in rabbit IVDs through modulation of local HIF-1α activity in IVD for the first time. After the establishment of the IVDD model in rabbits, disc height and IVD extracellular matrix were decreased in a time-dependent manner, and MRI analysis showed a tendency for decreased T2 values in a time-dependent manner, which is in line with previous findings. Intriguingly, our results indicated that supplementation of HIF-1α improved histological and imaginative IVDD while downregulation of HIF-1α exacerbated. These results further confirmed the conclusions of in vitro experiments.
In summary, HIF-1α protected against IVDD, possibly through reducing ROS production in the mitochondria and consequent inhibition of inflammation, metabolism disorders and apoptosis of NP cells, which provided a new idea for exploring the therapeutic strategies and targeted drugs for IVDD (Fig. 6).
The limitations of this study mainly included two aspects. The first was a small sample size for model rabbits, and the other was that the potential mechanism associated with the protective role of HIF-1α for IVDD was not verified in vivo.
## References
1. Kos N, Gradisnik L, Velnar T. **A brief review of the degenerative intervertebral disc disease**. *Medieval Archaeology* (2019.0) **73** 421-424. DOI: 10.5455/medarh.2019.73.421-424
2. 2.Yurube, T., M. Ito, Y. Ka, R Kuroda, K. Kakutani. 2020. Autophagy and mTOR signaling during intervertebral disc aging and degeneration. JOR Spine 3 (1): e1082. 10.1002/jsp2.1082.
3. Saberi M, Zhang X, Mobasheri A. **Targeting mitochondrial dysfunction with small molecules in intervertebral disc aging and degeneration**. *Geroscience* (2021.0) **43** 517-537. DOI: 10.1007/s11357-021-00341-1
4. Clouet J, Vinatier C, Merceron C, Pot-Vaucel M, Hamel O, Weiss P, Grimandi G, Guicheux J. **The intervertebral disc: From pathophysiology to tissue engineering**. *Joint, Bone, Spine* (2009.0) **76** 614-618. DOI: 10.1016/j.jbspin.2009.07.002
5. Hadjipavlou AG, Tzermiadianos MN, Bogduk N, Zindrick MR. **The pathophysiology of disc degeneration: A critical review**. *Journal of Bone and Joint Surgery British Volume* (2008.0) **90** 1261-1270. DOI: 10.1302/0301-620X.90B10.20910
6. Urban JP, Roberts S. **Degeneration of the intervertebral disc**. *Arthritis Research & Therapy* (2003.0) **5** 120-130. DOI: 10.1186/ar629
7. Board D, Stemper BD, Yoganandan N, Pintar FA, Shender B, Paskoff G. **Biomechanics of the aging spine**. *Biomedical Sciences Instrumentation* (2006.0) **42** 1-6. PMID: 16817576
8. Hanaei S, Abdollahzade S, Khoshnevisan A, Kepler CK, Rezaei N. **Genetic aspects of intervertebral disc degeneration**. *Reviews in the Neurosciences* (2015.0) **26** 581-606. DOI: 10.1515/revneuro-2014-0077
9. Papadakis M, Sapkas G, Papadopoulos EC, Katonis P. **Pathophysiology and biomechanics of the aging spine**. *The Open Orthopaedics Journal* (2011.0) **5** 335-342. DOI: 10.2174/1874325001105010335
10. Choi YS. **Pathophysiology of degenerative disc disease**. *Asian Spine Journal* (2009.0) **3** 39-44. DOI: 10.4184/asj.2009.3.1.39
11. Maher C, Underwood M, Buchbinder R. **Non-specific low back pain**. *Lancet* (2017.0) **389** 736-747. DOI: 10.1016/S0140-6736(16)30970-9
12. Kloppenburg M, Berenbaum F. **Osteoarthritis year in review 2019: Epidemiology and therapy**. *Osteoarthritis Cartilage* (2020.0) **28** 242-248. DOI: 10.1016/j.joca.2020.01.002
13. Sakai D, Andersson GB. **Stem cell therapy for intervertebral disc regeneration: Obstacles and solutions**. *Nature Reviews Rheumatology* (2015.0) **11** 243-256. DOI: 10.1038/nrrheum.2015.13
14. He R, Wang Z, Cui M, Liu S, Wu W, Chen M, Wu Y, Qu Y, Lin H, Chen S, Wang B, Shao Z. **HIF1A Alleviates compression-induced apoptosis of nucleus pulposus derived stem cells via upregulating autophagy**. *Autophagy* (2021.0) **17** 3338-3360. DOI: 10.1080/15548627.2021.1872227
15. Semenza GL. **Hypoxia-inducible factor 1: Oxygen homeostasis and disease pathophysiology**. *Trends in Molecular Medicine* (2001.0) **7** 345-350. DOI: 10.1016/s1471-4914(01)02090-1.15
16. Merceron C, Mangiavini L, Robling A, Wilson TL, Giaccia AJ, Shapiro IM, Schipani E, Risbud MV. **Loss of HIF-1α in the notochord results in cell death and complete disappearance of the nucleus pulposus**. *PLoS ONE* (2014.0) **9** e110768. DOI: 10.1371/journal.pone.0110768
17. Risbud MV, Shapiro IM. **Role of cytokines in intervertebral disc degeneration: Pain and disc content**. *Nature Reviews Rheumatology* (2014.0) **10** 44-56. DOI: 10.1038/nrrheum.2013.160
18. Wang Y, Che M, Xin J, Zheng Z, Li J, Zhang S. **The role of IL-1β and TNF-α in intervertebral disc degeneration**. *Biomedicine & Pharmacotherapy* (2020.0) **131** 110660. DOI: 10.1016/j.biopha.2020.110660
19. Puimège L, Libert C, Van Hauwermeiren F. **Regulation and dysregulation of tumor necrosis factor receptor-1**. *Cytokine & Growth Factor Reviews* (2014.0) **25** 285-300. DOI: 10.1016/j.cytogfr.2014.03.004
20. Walczak H. **TNF and ubiquitin at the crossroads of gene activation, cell death, inflammation, and cancer**. *Immunological Reviews* (2011.0) **244** 9-28. DOI: 10.1111/j.1600-065X.2011.01066.x
21. Borghi A, Verstrepen L, Beyaert R. **TRAF2 multitasking in TNF receptor-induced signaling to NF-κB, MAP kinases and cell death**. *Biochemical Pharmacology* (2016.0) **15** 1-10. DOI: 10.1016/j.bcp.2016.03.009
22. Liang H, Yang X, Liu C, Sun Z, Wang X. **Effect of NF-kB signaling pathway on the expression of MIF, TNF-α, IL-6 in the regulation of intervertebral disc degeneration**. *Journal of Musculoskeletal and Neuronal Interactions* (2018.0) **18** 551-556. PMID: 30511959
23. Zhang QX, Guo D, Wang FC, Ding WY. **Necrosulfonamide (NSA) protects intervertebral disc degeneration via necroptosis and apoptosis inhibition**. *European Review for Medical and Pharmacological Sciences* (2020.0) **24** 2683-2691. DOI: 10.26355/eurrev_202003_20538
24. Wang S, Wei J, Fan Y, Ding H, Tian H, Zhou X, Cheng L. **Progranulin is positively associated with intervertebral disc degeneration by interaction with IL-10 and IL-17 through TNF pathways**. *Inflammation* (2018.0) **41** 1852-1863. DOI: 10.1007/s10753-018-0828-1
25. Zhao Y, Qiu C, Wang W, Peng J, Cheng X, Shangguan Y, Xu M, Li J, Qu R, Chen X, Jia S, Luo D, Liu L, Li P, Guo F, Vasilev K, Liu L, Hayball J, Dong S, Pan X, Li Y, Guo L, Cheng L, Li W. **Cortistatin protects against intervertebral disc degeneration through targeting mitochondrial ROS-dependent NLRP3 inflammasome activation**. *Theranostics* (2020.0) **10** 7015-7033. DOI: 10.7150/thno.45359
26. 26.Liu, K., J. Wei, G. Li, R. Liu, D. Zhao, Y. Zhang, J. Shi, Q. Xie, and L. Cheng. 2021. Fexofenadine protects against intervertebral disc degeneration through TNF signaling. Frontiers in Cell and Development Biology 9. 10.3389/fcell.2021.687024.
27. Liu Y, Lin J, Wu X, Guo X, Sun H, Yu B, Shen J, Bai J, Chen Z, Yang H, Geng D, Mao H. **Aspirin-mediated attenuation of intervertebral disc degeneration by ameliorating reactive oxygen species in vivo and in vitro**. *Oxidative Medicine and Cellular Longevity* (2019.0) **2019** 7189854. DOI: 10.1155/2019/7189854
28. Liu H, Pan H, Yang H, Wang J, Zhang K, Li X, Wang H, Ding W, Li B, Zheng Z. **LIM mineralization protein-1 suppresses TNF-α induced intervertebral disc degeneration by maintaining nucleus pulposus extracellular matrix production and inhibiting matrix metalloproteinases expression**. *Journal of Orthopaedic Research* (2015.0) **33** 294-303. DOI: 10.1002/jor.22732
29. 29.Wang, P., Q. Meng, W. Wang, S. Zhang, X. Xiong, S. Qin, J. Zhang, A. Li, Z. Liu. 2020. Icariin inhibits the inflammation through downregulating NF-κB/HIF-2α signal pathways in chondrocytes. Bioscience Reports 40 (11): BSR20203107. 10.1042/BSR20203107
30. Wang P, Zhu P, Liu R, Meng Q, Li S. **Baicalin promotes extracellular matrix synthesis in chondrocytes via the activation of hypoxia-inducible factor-1α**. *Experimental and Therapeutic Medicine* (2020.0) **20** 226. DOI: 10.3892/etm.2020.9356
31. Chen D, Xia D, Pan Z, Xu D, Zhou Y, Wu Y, Cai N, Tang Q, Wang C, Yan M, Zhang JJ, Zhou K, Wang Q, Feng Y, Wang X, Xu H, Zhang X, Tian N. **Metformin protects against apoptosis and senescence in nucleus pulposus cells and ameliorates disc degeneration in vivo**. *Cell Death & Disease* (2016.0) **7** e2441. DOI: 10.1038/cddis.2016.334
32. Meng X, Zhuang L, Wang J, Liu Z, Wang Y, Xiao D, Zhang X. **Hypoxia-inducible factor (HIF)-1alpha knockout accelerates intervertebral disc degeneration in mice**. *International Journal of Clinical and Experimental Pathology* (2018.0) **11** 548-557. PMID: 31938140
33. Silagi ES, Schoepflin ZR, Seifert EL, Merceron C, Schipani E, Shapiro IM, Risbud MV. **Bicarbonate recycling by HIF-1-dependent carbonic anhydrase isoforms 9 and 12 is critical in maintaining intracellular pH and viability of nucleus pulposus cells**. *Journal of Bone and Mineral Research* (2018.0) **33** 338-355. DOI: 10.1002/jbmr.3293
34. Zhu H, Sun B, Shen Q. **TNF-α induces apoptosis of human nucleus pulposus cells via activating the TRIM14/NF-κB signalling pathway**. *Artificial Cells, Nanomedicine, and Biotechnology* (2019.0) **47** 3004-3012. DOI: 10.1080/21691401.2019.1643733
35. Zhang G, Liao Y, Yang H, Tao J, Ma L, Zuo X. **Irigenin reduces the expression of caspase-3 and matrix metalloproteinases, thus suppressing apoptosis and extracellular matrix degradation in TNF-α-stimulated nucleus pulposus cells**. *Chemico-Biological Interactions* (2021.0) **349** 109681. DOI: 10.1016/j.cbi.2021.109681
36. Lin XL, Zheng ZY, Zhang QS, Zhang Z, An YZ. **Expression of miR-195 and its target gene Bcl-2 in human intervertebral disc degeneration and their effects on nucleus pulposus cell apoptosis**. *Journal of Orthopaedic Surgery and Research* (2021.0) **16** 412. DOI: 10.1186/s13018-021-02538-8
37. Abate M, Festa A, Falco M, Lombardi A, Luce A, Grimaldi A, Zappavigna S, Sperlongano P, Irace C, Caraglia M, Misso G. **Mitochondria as playmakers of apoptosis, autophagy and senescence**. *Seminars in Cell & Developmental Biology* (2020.0) **98** 139-153. DOI: 10.1016/j.semcdb.2019.05.022
38. Jang S, Chapa-Dubocq XR, Parodi-Rullán RM, Fossati S, Javadov S. **Beta-amyloid instigates dysfunction of mitochondria in cardiac cells**. *Cells* (2022.0) **11** 373. DOI: 10.3390/cells11030373
39. Marei WFA, Smits A, Mohey-Elsaeed O, Pintelon I, Ginneberge D, Bols PEJ, Moerloose K, Leroy JLMR. **Differential effects of high fat diet-induced obesity on oocyte mitochondrial functions in inbred and outbred mice**. *Science and Reports* (2020.0) **10** 9806. DOI: 10.1038/s41598-020-66702-6
40. Sun K, Jing X, Guo J, Yao X, Guo F. **Mitophagy in degenerative joint diseases**. *Autophagy* (2021.0) **17** 2082-2092. DOI: 10.1080/15548627.2020.1822097
41. Xu X, Wang D, Zheng C, Gao B, Fan J, Cheng P, Liu B, Yang L, Luo Z. **Progerin accumulation in nucleus pulposus cells impairs mitochondrial function and induces intervertebral disc degeneration and therapeutic effects of sulforaphane**. *Theranostics* (2019.0) **9** 2252-2267. DOI: 10.7150/thno.30658
42. Zhang Z, Xu T, Chen J, Shao Z, Wang K, Yan Y, Wu C, Lin J, Wang H, Gao W, Zhang X, Wang X. **Parkin-mediated mitophagy as a potential therapeutic target for intervertebral disc degeneration**. *Cell Death & Disease* (2018.0) **9** 980. DOI: 10.1038/s41419-018-1024-9
43. Loor G, Schumacker PT. **Role of hypoxia-inducible factor in cell survival during myocardial ischemia-reperfusion**. *Cell Death and Differentiation* (2008.0) **15** 686-690. DOI: 10.1038/cdd.2008.13
|
---
title: Residue-specific binding of Ni(II) ions influences the structure and aggregation
of amyloid beta (Aβ) peptides
authors:
- Elina Berntsson
- Faraz Vosough
- Teodor Svantesson
- Jonathan Pansieri
- Igor A. Iashchishyn
- Lucija Ostojić
- Xiaolin Dong
- Suman Paul
- Jüri Jarvet
- Per M. Roos
- Andreas Barth
- Ludmilla A. Morozova-Roche
- Astrid Gräslund
- Sebastian K. T. S. Wärmländer
journal: Scientific Reports
year: 2023
pmcid: PMC9971182
doi: 10.1038/s41598-023-29901-5
license: CC BY 4.0
---
# Residue-specific binding of Ni(II) ions influences the structure and aggregation of amyloid beta (Aβ) peptides
## Abstract
Alzheimer’s disease (AD) is the most common cause of dementia worldwide. AD brains display deposits of insoluble amyloid plaques consisting mainly of aggregated amyloid-β (Aβ) peptides, and Aβ oligomers are likely a toxic species in AD pathology. AD patients display altered metal homeostasis, and AD plaques show elevated concentrations of metals such as Cu, Fe, and Zn. Yet, the metal chemistry in AD pathology remains unclear. Ni(II) ions are known to interact with Aβ peptides, but the nature and effects of such interactions are unknown. Here, we use numerous biophysical methods—mainly spectroscopy and imaging techniques—to characterize Aβ/Ni(II) interactions in vitro, for different Aβ variants: Aβ(1–40), Aβ(1–40)(H6A, H13A, H14A), Aβ(4–40), and Aβ(1–42). We show for the first time that Ni(II) ions display specific binding to the N-terminal segment of full-length Aβ monomers. Equimolar amounts of Ni(II) ions retard Aβ aggregation and direct it towards non-structured aggregates. The His6, His13, and His14 residues are implicated as binding ligands, and the Ni(II)·Aβ binding affinity is in the low µM range. The redox-active Ni(II) ions induce formation of dityrosine cross-links via redox chemistry, thereby creating covalent Aβ dimers. In aqueous buffer Ni(II) ions promote formation of beta sheet structure in Aβ monomers, while in a membrane-mimicking environment (SDS micelles) coil–coil helix interactions appear to be induced. For SDS-stabilized Aβ oligomers, Ni(II) ions direct the oligomers towards larger sizes and more diverse (heterogeneous) populations. All of these structural rearrangements may be relevant for the Aβ aggregation processes that are involved in AD brain pathology.
## Introduction
Alzheimer’s disease (AD), the leading cause of dementia worldwide, is a progressive, irreversible, and currently incurable chronic neurodegenerative disorder1–3, primarily manifesting as short-term memory loss. Pathological hallmarks of AD include brain atrophy, with extensive brain deposits of amyloid plaques and neurofibrillary Tau tangles occurring years before symptom manifestation3–5. The plaques, which consist mainly of amyloid-β (Aβ) peptides aggregated into insoluble fibrils6, display a characteristic cross-β structure at the core of their constituent fibrils7,8. The plaques are the end-product of an aggregation process involving formation of extra- and intracellular intermediates such as neurotoxic Aβ oligomers9–14. The oligomeric aggregates may spread from neuron to neuron via exosomes15,16. However, the relationship between Aβ aggregation, neurodegenerative mechanisms, cognitive decline, the proposed amyloid cascade hypothesis, and disease progression is not fully understood2,13,14,17.
The 36–43 residues long Aβ peptides found in the plaques are produced by enzymatic cleavage of the membrane-binding amyloid-β precursor protein, APP18. In monomeric form, the Aβ peptides are intrinsically disordered and soluble in water. The central and C-terminal segments are hydrophobic and may interact with membranes or fold into a hairpin conformation that likely is required for aggregation19. The negatively charged N-terminal segment is hydrophilic and readily interacts with metal ions and other cationic molecules20–23.
AD brains typically display altered metal homeostasis17,24,25, and AD plaques accumulate metals such as calcium (Ca), copper (Cu), iron (Fe), and zinc (Zn)26–28. Thus, dysregulated metal chemistry might be part of the AD pathology process29–32. The precursor protein APP is known to bind Cu and Zn ions33, and a possible physiological role of APP (and perhaps its fragments) might be to regulate the Cu(II) and Zn(II) concentrations in the neuronal synaptic clefts, where these ions are released in their free form34 and where Aβ aggregation may be initiated35. Metal ions such as Cu(II), Fe(II), Mn(II), Pb(IV) and Zn(II) have previously been shown to bind to specific Aβ residues and modulate the Aβ aggregation pathways20,29,36–40. Binding of metal ions, and also of other cationic molecules such as polyamines, has furthermore been reported to modulate and sometimes inhibit Aβ toxicity21,22,41. However, it is unclear which possible metal interactions may be relevant for AD pathology, and which exogenous or endogenous metal ions may participate in such interactions30–32.
Nickel (Ni) is a common metal in the industrialized world, where it is used in e.g. stainless steel alloys, Ni–Cd batteries, coins, and jewelry. As a result of low-level exposure, between 10 and $20\%$ of all people have developed some degree of contact allergy towards Ni42,43. It is therefore important to clarify the health effects of long-term Ni exposure, including potential effects on neurodegenerative diseases44. Some studies have demonstrated specific binding between Ni(II) ions and N-terminal Aβ fragments, with possible effects on Aβ structure and toxicity41,45–47. Yet, the interactions between pathologically relevant (i.e., full-length) Aβ peptides and Ni ions are poorly explored.
In this study, we use a range of biophysical spectroscopy and imaging techniques to investigate in vitro interactions between Ni(II) ions and Aβ peptides, with a focus on characterizing binding properties and effects on Aβ structure and aggregation. The different Aβ peptides studied include the pathologically relevant Aβ(1–40), Aβ(4–40), and Aβ(1–42) variants, together with the Aβ(1–40)(H6A, H13A, H14A) mutant. Because the Aβ peptides interact with membranes48, and as membrane-disruption is a possible toxicity mechanism for Aβ oligomers30, the measurements have been carried out in aqueous solution as well as in a membrane-mimetic model consisting of micelles of the SDS (sodium dodecyl sulfate) detergent49. The results are compared to previous studies of the effects of both different chemical environments and metal ion interactions on Aβ peptides37,38,40,48–53.
## Biological relevance of Ni and sources of exposure
Even though Ni function is limited in human biology54, it has been suggested as an essential element in humans55, just as it is for many human-associated bacteria56 and possibly all higher plants57. Elevated Ni concentrations are however toxic to plants58. In animals, inadequate Ni amounts have shown adverse effects on nutrient absorption and metabolism59,60. Due to the abundance of Ni in many plant-based foods58,61, Ni deficiency is unlikely to occur in humans, even though only some $10\%$ of ingested *Ni is* absorbed62. Ni concentrations in potable water vary between 2 and 13 mg/L with a WHO limit of 70 mg/L, which is exceeded in Ni mining regions, where Ni concentrations of 200 mg/L have been found63. Respiratory Ni exposure is related to Ni industries and fossil fuel combustion, which overall are the main global sources of Ni emissions58,62, and to cigarette smoking including e-cigarettes38,64. In addition to Ni–Cd batteries, the main industrial uses of *Ni is* as a whitening agent in Cu alloys for e.g. coins and jewellery, which may cause allergic contact dermatitis58, and as a provider of corrosion resistance in steel alloys for e.g. surgical tools, biomedical implants, and body piercings. One study of Ni-containing hip prosthetic devices found that Ni blood concentration rose about twofold after metal-on-metal hip arthroplasty65. Ni is also present in some formulations for dental amalgam fillings66.
The human health risks of Ni exposure are well-known and widespread, as Ni use and exposure has gradually grown from human prehistory67 to modern times68. Ni is known to be haematotoxic, immunotoxic, neurotoxic, genotoxic, reproductive toxic, pulmonary toxic, nephrotoxic, hepatotoxic and carcinogenic58,69. Other pathological effects of Ni exposure in the occupational settings are rhinitis, asthma, nasal septum perforation, nasal sinus cancer, and respiratory cancer70. Ni can pass the placental barriers and accumulate in the fetus71. It can also pass the blood–brain barrier, and brain accumulation can result from high levels of exposure54. Yet, most people experience low exposure levels66, and the most common health effect is allergic reactions42,43.
The neurotoxic properties of Ni are well documented, but data on Ni in neurodegenerative disorders are scarce44. A case report of skin tissue Ni concentrations in a patient recovering from ALS after metal chelation described a Ni concentration of 850 µg/kg72. A 44 year old ALS patient died after 9 years of heavy metal exposure in a nickel–cadmium battery factory73. Increased blood Ni concentrations have been detected in multiple sclerosis (MS) patients74, and soil Ni was found to be elevated in a Canadian MS cluster75. Recent data also indicate a possible contribution from Ni in the causation of neurodevelopmental dysfunctional states such as autism76.
As a component of tobacco, cigarette smoke, and air pollution, Ni may contribute to environmental risk factors for AD. Mice exposed to a Ni nanoparticle model of air pollution showed doubled brain levels of Aβ40 and Aβ42 within 24 h, even at a permissible limit of nickel hydroxide exposure according to occupational safety and health standards77. Another study reported higher yet not statistically significantly elevated Ni concentrations in post-mortem brain and ventricular fluid of AD patients ($$n = 14$$), compared to healthy controls ($$n = 15$$)25. A recent study reported that Ni(II) ions interfered with aggregation of the Tau protein78.
## Samples and preparations
Ni(II) acetate and 2-(N-Morpholino)ethanesulfonic acid hydrate (MES) buffer were purchased from Sigma (Sigma/Merck KGaA, Darmstadt, Germany). The SDS detergent was bought from ICN Biomedicals Inc (USA). Sodium chloride and sodium hydroxide were purchased from Sigma-Aldrich (St. Louis, MO, USA).
Wild-type (wt) Aβ(1–42) peptides, abbreviated as Aβ42, with the primary sequence DAEFR5HDSGY10EVHHQ15KLVFF20AEDVG25SNKGA30IIGLM35VGGVV40IA, were purchased synthetically manufactured from JPT Peptide Technologies (Germany), while recombinantly produced Aβ42 peptides were purchased from rPeptide LLC (USA). Recombinantly produced wild-type (wt) Aβ(1–40) peptides, abbreviated as Aβ40, as well as N-terminal truncated Aβ(4–40) peptides, were purchased as lyophilized powder from AlexoTech AB (Umeå, Sweden). The Aβ40 peptides were either unlabeled, uniformly 15N-labeled, or uniformly 13C,15N-labeled. A recombinantly produced mutant version of Aβ40, where the three histidine residues H6, H13, and H14 have been replaced with alanines, i.e. Aβ(1–40)(H6A, H13A, H14A) was also purchased from AlexoTech AB. This mutant is here abbreviated as Aβ40(NoHis). All Aβ variants were stored at − 80 °C until use, when they were dissolved to monomeric form before the measurements. The Aβ40 and Aβ(4–40) peptides were then dissolved in 10 mM NaOH to 100 µM concentration, and sonicated for 5 min in an ice-bath to dissolve possible pre-formed aggregates. Finally, buffer was added to the peptide solutions. All preparation steps were performed on ice, and the peptide concentrations were determined by weighing the dry powder and/or by NanoDrop measurements of dissolved material.
## Preparation of Aβ42 oligomers
Monomeric solutions of Aβ42 peptides were prepared via size exclusion chromatography, according to the following procedure. First, lyophilized Aβ42 powder (1 mg) was dissolved in pure dimethyl sulfoxide (DMSO; 250 µL). A solution of 5 mM NaOH (pH = 12.3) was used to equilibrate a Sephadex G-250 HiTrap desalting column (GE Healthcare, Uppsala), which was then washed with 10–15 mL of 5 mM NaOD (pD = 12.7)79. The Aβ42 solution in DMSO was added to the column, followed by 5 mM NaOD (1.25 mL). Peptide fractions in 5 mM NaOD were then collected on ice at a flow rate of 1 mg/mL. Ten fractions of 1 mL were collected in 1.5 mL low-binding reaction tubes. The Aβ42 concentration in each fraction was measured with a NanoDrop instrument (Eppendorf, Germany) at 280 nm, using a molar extinction coefficient of 1280 M−1 cm−1 for the single tyrosine residue in the peptide80. Liquid nitrogen was used to flash-freeze the fractions, which then were topped with argon gas, and stored at − 80 °C until use. Two well-defined sizes of SDS-stabilized Aβ42 oligomers—named according to the SDS concentration used: AβO$0.05\%$SDS (approximately dodecamers) and AβO$0.2\%$SDS (approximately tetramers)—were prepared using an established protocol81 with the following modifications: the preparations were carried out in D2O without the original dilution step, and at a fourfold lower peptide concentration82. The reaction mixtures, consisting of 100 µM Aβ42 peptide in phosphate buffered saline (PBS) buffer containing either $0.05\%$ SDS or $0.2\%$ SDS, which corresponds to 1.7 mM and 6.9 mM SDS, respectively, were incubated at 37 °C for 24 h together with 0–500 µM of Ni(II) acetate. Liquid nitrogen was used to flash-freeze the prepared oligomer solutions, which then were stored at − 20 °C until further use. When thawed at room temperature for experimental analyses, the oligomers were stable for several days.
## NMR spectroscopy measurements of Aβ40 binding to Ni(II) ions
1D and 2D nuclear magnetic resonance (NMR) spectra were recorded on Bruker Avance 500 and 700 MHz spectrometers equipped with cryoprobes. First 42 μM and then 84 μM of Ni(II) acetate was added to 84 μM of monomeric Aβ40 peptides, either 13C,15N-double-labelled or 15N-mono-labelled, in 20 mM sodium phosphate buffer at pH 7.3 or pH 5.6 ($\frac{90}{10}$ H2O/D2O). During the titrations, 2D 1H,15N-HSQC and 2D 1H,13C-HSQC spectra were recorded at 5 °C. Measurements were also conducted in the presence of 50 mM SDS detergent, at 25 °C. As the critical micelle concentration for SDS is 8.2 mM in water at 25 °C83, micelles have clearly formed under these conditions. SDS micelles are simple membrane models suitable for NMR spectroscopy due to their small size, i.e. on average 62 molecules per micelle49. Thus, there is approximately 0.8 mM of SDS micelles in the sample, i.e., around 10× more micelles than Aβ40 peptides, which means that no micelle should harbour multiple Aβ peptides. All NMR data were processed and evaluated using the Topspin software (v. 3.2), employing already published HSQC crosspeak assignments for Aβ40 in buffer84–86 and in the presence of SDS micelles51.
## CD spectroscopy measurements of Ni(II)-induced changes in Aβ secondary structure
Circular dichroism (CD) was carried out in a Chirascan CD spectrometer (Applied Photophysics Ltd., U.K.) using a 2 mm quartz cuvette containing 600 µl of Aβ peptide in 20 mM phosphate buffer, pH 7.3. The studied Aβ variants were Aβ40 (10 µM), Aβ40(NoHis) (10 µM), and Aβ(4–40) (5 µM).
Measurements were conducted either in buffer only or with added SDS micelles. CD spectra were recorded at 25 °C between 190 and 260 nm, using steps of 0.5 nm. After the first recorded spectrum, 50 mM SDS detergent was added to some of the samples. Then, small volumes of Ni(II) acetate (2 mM and 10 mM stock solutions) were titrated to each sample, in steps of 2 µM, 4 µM, 16 µM, 56 µM, 156 µM, 256 µM, and finally 512 µM. All data was processed with an eight points smoothing filter (Savitsky-Golay) using the Chirascan Pro-Data v.4.4.1 software (Applied Photophysics Ltd., U.K.).
## Binding affinity of Ni(II)·Aβ complexes
Binding affinities for Ni(II)·Aβ complexes were estimated by fitting NMR and CD titration data, respectively, to Eq. [ 1] (the Morrison equation87):1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I={I}_{0}+\frac{{I}_{\infty }-{I}_{0}}{2*\left[A\beta \right]}*\left(\left({K}_{D}+\left[Ni\right]+\left[A\beta \right]\right)-\sqrt{{\left({K}_{D}+\left[Ni\right]+\left[A\beta \right]\right)}^{2}-4*\left[Ni\right]*\left[A\beta \right]}\right)$$\end{document}I=I0+I∞-I02∗Aβ∗KD+Ni+Aβ-KD+Ni+Aβ2-4∗Ni∗Aβ This equation assumes a single metal binding site, where [Aβ] is the peptide concentration, [Ni] is the concentration of the titrated Ni(II) ions, I0 is the initial signal intensity, I∞ is the steady-state (saturated) signal intensity at the end of the titration series, and KD is the dissociation constant for the Ni(II)·Aβ complex. As no corrections for buffer conditions are done, the computed dissociation constants should be considered as apparent (KDApp).
For the CD data, a binding curve was generated by plotting the CD intensity at 208 nm vs the Ni(II) concentration. For the NMR data, binding curves were generated by plotting HSQC crosspeak intensities versus Ni(II) concentration. The measured crosspeak intensities were normalized to the intensity of the V40 crosspeak for each step of added Ni(II) ions.
## Aβ40 aggregation kinetics monitored via ThT fluorescence measurements
The kinetics of amyloid aggregation over time for Aβ40 peptides together with Ni(II) ions was monitored via measurements with a 96-well plate reader (FLUOstar Omega) of the fluorescence signal of the dye Thioflavin T (ThT), a molecular probe that displays strong fluorescence intensity when bound to amyloid material88,89. The samples contained 10 μΜ Aβ40 peptide, 10 mM sodium phosphate buffer, pH 7.4, 40 μΜ ThT dye, and 0, 1, 2.5, 5, 7.5, 10, 20, or 50 μM of Ni(II) acetate. The peptide concentration was determined using a NanoDrop microvolume spectrophotometer. Measurements were recorded at three-minute intervals for 24 h at 37 °C, with five replicate samples per condition and excitation and emission wavelengths of 440 and 480 nm, respectively. The samples were continuously shaken (orbital mode) between the measurements. To determine the maximum growth rate and the half-time of Aβ aggregation, the resulting ThT fluorescence data were fitted to a sigmoidal curve using Eq. [ 2]90:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F\left(t \right) = F_{0} + \frac{A}{{1 + \exp \left[{ - r_{\max } \left({t - t_{$\frac{1}{2}$} } \right)} \right]}}$$\end{document}Ft=F0+A1+exp-rmaxt-t$\frac{1}{2}$ where F0 is the baseline fluorescence intensity, A is the total increase in fluorescence intensity, rmax is the maximum growth rate, and t$\frac{1}{2}$ is the time when half of the Aβ monomers have aggregated.
## Atomic force microscopy images of Aβ40 aggregates
Images of Aβ40 aggregates were recorded with a BioScope Catalyst (Bruker Corp., USA) atomic force microscope (AFM), operating in peak force mode in air and using MSLN and SLN cantilevers (Bruker Corp., USA). The scan rate was 0.51 Hz, with a resolution of 512 Å ~ 512 pixels. Samples of 10 μΜ Aβ40 peptide were incubated for 24 h with either 0, 1, 10, or 50 μΜ Ni(II) acetate, using the same conditions as in the ThT experiments described above. At the end of the procedure, 30 μL samples were diluted in 30 μL Milli-Q water and then applied on freshly prepared mica substrates. After 20 min, the mica substrates were washed three times with Milli-Q water and left to air-dry.
## Blue native polyacrylamide gel electrophoresis of Aβ42 oligomers
The Aβ42 oligomer samples prepared with 0–500 µM Ni(II) acetate, as described in the materials section, were analyzed with blue native polyacrylamide gel electrophoresis (BN-PAGE) using the *Invitrogen electrophoresis* system. First, 4–$16\%$ Bis–Tris Novex gels (ThermoFisher Scientific, USA) were loaded with Aβ42 oligomer samples (10 μL) in addition to the Amersham High Molecular Weight Calibration Kit for native electrophoresis (GE Healthcare, USA). The gels were run at 4 °C according to the Invitrogen instructions (ThermoFisher Scientific, USA). Staining was done with the Pierce Silver Staining Kit (ThermoFisher Scientific, USA).
## Infrared spectroscopy
Fourier-transformed infrared (FTIR) spectra of the Aβ42 oligomers prepared with 0–500 µM Ni(II) acetate, as described in the materials section, were recorded on a Tensor 37 FTIR spectrometer (Bruker Optics, Germany) operating in transmission mode at room temperature and equipped with a liquid nitrogen-cooled MeCdTe detector and a sample shutter. During the measurements, the instrument was continuously purged with dry air. 8–10 µL of the 80 µM Aβ42 oligomer samples were placed between two flat CaF2 discs, which were separated by a 50 µm plastic spacer that had been covered with vacuum grease at the periphery. The mounted IR cuvette was put in a holder inside the sample chamber, and was then allowed to sit for at least 20 min after the chamber lid was closed, to remove H2O vapor. FTIR spectra were recorded between 1900 and 800 cm−1, at a resolution of 2 cm−1 and with 6 mm aperture. The IR intensity above 2200 cm−1 was blocked with a germanium filter, and that below 1500 cm−1 with a cellulose membrane, to increase the light intensity in the relevant spectral range91. The OPUS 5.5 software was used for analysis and plotting of the spectra. Second derivatives were computed with a smoothing factor of 17.
## Fluorescence measurements of Ni(II)-induced formation of dityrosine in Aβ40
A Jobin Yvon Horiba Fluorolog 3 fluorescence spectrometer (Longjumeau, France) was used to record fluorescence emission spectra between 330 and 500 nm (excitation at 315 nm) at room temperature of 10 µM Aβ40 peptide dissolved in 20 mM MES buffer, pH 7.3. The samples were put in a quartz cuvette with 4 mm path length (volume 1 mL). To investigate the effect of Ni(II) ions on dityrosine formation, one sample contained 100 µM Ni(II) acetate. The control sample contained 50 µM of the chelator EDTA, to remove possible free metal ions. Spectra were recorded after 0 and 6 h of incubation, where the sample was kept at room temperature without agitation or other treatment. All experiments were conducted in triplicate, and before the final measurement, 300 µM of EDTA was added to the sample to remove metal ions.
## NMR spectroscopy: molecular details of Ni(II) binding to the Aβ40 monomer
High-resolution NMR experiments were conducted to investigate possible residue-specific molecular interactions between Ni(II) ions and monomeric Aβ40 peptides. Figures 1 and 2 show 2D 1H,15N-HSQC spectra for the amide crosspeak region for 84 µM 13C,15N-labeled Aβ40 peptides at either pH 7.3 or pH 5.6, recorded before and after addition of first 42 µM and then 84 µM Ni(II) acetate, i.e. Ni(II):Aβ40 ratios of respectively 1:2 and 1:1. Similar measurements were conducted also for Aβ40 peptides together with SDS micelles (50 mM SDS detergent concentration) at pH 7.3 (Fig. 1C).Figure 1NMR 2D 1H,15N-HSQC-spectra of 84 µM 15N-labeled Aβ40 peptides before (blue peaks) and after addition of first 42 µM (red peaks) and then 84 µM (teal peaks) Ni(II) acetate. Spectra were recorded at 5 °C in 20 mM sodium phosphate buffer at pH 7.3 (A), at pH 5.6 (B), and at pH 7.3 together with 50 mM SDS detergent (C). The peak intensity in the bar charts is given as the ratio between the crosspeak intensity with added Ni(II) ions relative to the intensity before addition of Ni(II) ions, i.e. I/I0.Figure 2NMR 2D 1H,13C-HSQC spectra of 84 µM 13C,15N-labeled Aβ40 peptides in 20 mM phosphate buffer, pH 7.3, before (blue) and after (red) addition of 42 µM Ni(II) acetate. Some Phe crosspeaks could not be assigned to individual residues, and are instead listed as F(I), F(II), and F(III).
For Aβ40 in aqueous pH 7.3 buffer, addition of Ni(II) ions induces a concentration-dependent loss of amide crosspeak intensity, especially in the N-terminal region (Fig. 1A). This indicates specific binding of Ni(II) ions to N-terminal Aβ40 residues. The specific loss of N-terminal crosspeak intensity is likely caused by intermediate or even slow (on the NMR time-scale) chemical exchange between a free and a bound state of the Aβ40 peptides, similar to the effect induced by Cu(II) and Zn(II) ions40,92, probably together with paramagnetic quenching effects of the Ni(II) ions93. When the Ni(II) ions are added to the sample, no new crosspeaks corresponding to Ni(II)-bound Aβ40 peptides are observed (Fig. 1A). This suggests that no single well-defined Ni(II)·Aβ40 complex exists. Instead, a range of Ni(II)-bound states of the Aβ40 peptides are likely present, probably at different stages of aggregation and oligomerization. Each state is then too weakly populated to create distinct NMR crosspeaks. This is in line with earlier NMR studies reporting that Aβ40 peptides are in a dynamic exchange between NMR-observable monomers and heterogeneous NMR-invisible oligomers (a “dark state”)92,94.
There is also a general loss of crosspeak intensity for all Aβ40 residues, including the central and C-terminal amino acids, upon addition of Ni(II) acetate. This effect is likely caused by a combination of non-specific Ni(II) binding interactions, on an intermediate or slow NMR time-scale, and by the Ni(II) ions promoting aggregation of the Aβ40 peptides into complexes that are too large to be observed with HSQC NMR, or they may simply precipitate out of the solution.
At pH 5.6, the loss of crosspeak intensity after added Ni(II) ions is very uniform, i.e., there is no residue-specific binding (Fig. 1B). The main difference at this lower pH is that histidine residues are protonated, as they have pKa values around 6.8 in short peptides95. Naturally, protonated His residues are less prone to bind cationic metal ions. The loss of residue-specific Ni(II) ion binding at pH 5.6 therefore strongly indicates that histidines are involved in the observed residue-specific Ni(II) ion binding at neutral pH. This is supported by the NMR results at pH 7.3 for the Aβ40 aromatic side chains, where the aromatic rings of the N-terminal residues His6, His13, and His14, together with Tyr10, display a somewhat larger loss of crosspeak signal intensity than the aromatic rings of the Phe4, Phe19, and Phe20 residues (Fig. 2).
Specific loss of crosspeak intensity for N-terminal residues, upon addition of Ni(II) acetate, is observed also for Aβ40 peptides positioned in SDS micelles (Fig. 1C), which constitute a simple model for bio-membranes48,49. The central and C-terminal Aβ regions are known to insert themselves as α-helices into SDS micelles48,51, and thus, the 1H,15N-HSQC spectrum for Aβ40 in SDS micelles corresponds to an α-helical conformation of the Aβ peptide. The N-terminal segment is known to remain unstructured outside the micelle surface, where it is available for interactions with e.g. metal ions96. The C-terminal residues are unaffected by the added Ni(II) ions (Fig. 1C), which shows that they are neither affected by direct Ni(II)-binding, nor by Ni(II)-induced aggregation—most likely Aβ peptides cannot aggregate when bound to SDS micelles, at least when there is on average less than one peptide per micelle. Thus, also in the presence of SDS micelles (Fig. 1C), the NMR spectrum reflects Ni(II) ion binding to monomeric Aβ40 peptides.
## CD spectroscopy measurements of Aβ secondary structure
CD spectroscopy was used to investigate possible changes in Aβ secondary structure upon addition of Ni(II) ions, both in aqueous buffer and in a membrane-mimetic environment (i.e., SDS micelles). The three peptide variants Aβ40, Aβ40(NoHis) mutant, and Aβ(4–40) were investigated. In aqueous buffer, the CD spectra for monomers of all three variants display typical random coil signals with minima around 196–198 nm (Fig. 3).Figure 3CD spectra of Ni(II) acetate titrated to Aβ peptides in 20 mM phosphate buffer, pH 7.3, at 25 °C. The titrations were conducted either in the presence of SDS micelles (A–C) or in aqueous buffer only (D–F), for three different peptide variants, i.e. 10 µM Aβ40 (A,D), 10 µM Aβ40(NoHis) (B,E), and 5 µM Aβ(4–40) (C,F). The black spectra show Aβ peptides in buffer only. For samples A–C, 50 mM SDS was then added (red spectra). Next, for all samples, Ni(II) acetate was titrated in steps of 2 µM (orange), 4 µM (yellow), 16 µM (turquoise), 56 µM (green), 156 µM (purple), and finally 256 µM (blue spectra).
Addition of Ni(II) ions to Aβ peptides in aqueous buffer induces concentration-dependent changes in the CD spectra for the monomers of all three peptide variants (Fig. 3D–F). For the Aβ40(NoHis) peptide, these changes correspond to a decrease in intensity, without changing the shape of the spectrum (Fig. 3). This likely means that the Ni(II) ions induce peptide aggregation and precipitation, thereby reducing the effective Aβ concentration in the solution.
For the Aβ40 and Aβ(4–40) variants, addition of Ni(II) ions induces structural transitions between two distinct conformations, as evidenced by the isodichroic points around 210 nm (Fig. 3D,F). Aβ peptides in solution are known to contain some polyproline II (PPII) helix structure, especially at low temperatures50. The loss of signal intensity around 196–198 nm, and the isodichroic points around 210 nm, might be compatible with a conversion of PPII helix into random coil structure50,97,98. However, the difference spectra created by subtracting the CD spectra with no added Ni(II) acetate from those with 256 μM Ni(II) acetate, shown in Supp. Fig. S1, clearly correspond to β-sheet secondary structure99. Formation of β-sheets upon addition of Ni(II) ions is supported also by the IR spectra shown in Supp. Fig. S2. We therefore conclude that Ni(II) ions induce β-sheet structure in Aβ40 and Aβ(4–40) peptides, in aqueous solution at neutral pH.
When SDS micelles were added to the three peptide variants, all of them adopted α-helical secondary structures, producing CD signals with characteristic minima around 208 and 222 nm (Fig. 3A–C). This is consistent with previous studies reporting that Aβ peptides adopt α-helical conformations in membrane-like environments, at least when there is on average less than one peptide per micelle48,49,51,96,100. The Aβ40(NoHis) variant, where the three His residues have been replaced with alanines, displays the strongest α-helical CD signal after addition of SDS, compared to the intensity of the Aβ40(NoHis) random coil signal before adding SDS (Fig. 3). This is reasonable given the strong propensity of alanines to form α-helices101. The Aβ(4–40) peptide also shows a relatively strong α-helical CD signal, compared to the Aβ(4–40) random coil intensity before added SDS. This might be related to a lack of α-helical structure51 in the first three residues of Aβ40, which are missing in the Aβ(4–40) variant.
Addition of Ni(II) ions induces a concentration-dependent loss of CD signal around 208 nm, but not around 222 nm, for the Aβ40 and the Aβ(4–40) peptides, although not for the Aβ40(NoHis) mutant (Fig. 3). As the 222 nm signal intensity remains approximately constant, the observed spectral changes do not correspond to a general loss of α-helicity but may rather be due to a change in helix supercoiling, i.e. when two or more α-helices form coiled coils via hydrophobic interactions102–104. The degree of helix supercoiling is known to be reflected by the [θ222]/[θ208] ratio, where ratios close to 1 reflect large amounts of superhelicity105. During the titrations with Ni(II) acetate the [θ222]/[θ208] ratio increases from 0.70 to 0.89 for Aβ40, and from 0.75 to 1.03 for the Aβ(4–40) variant (Table 1). In both cases this would correspond to a significant increase in superhelicity. Such metal-induced changes of Aβ superhelicity, in a membrane environment, has previously been reported to be induced by Cu(II) ions96. The lack of any Ni(II)-induced structural effects in the Aβ40(NoHis) mutant (Fig. 3; Table 1) appears to support Ni(II) binding to Aβ via the His residues. But it is also possible that the His residues are necessary for forming a coiled-coil Aβ structure. Table 1CD signal intensities at 208 nm and 222 nm for the three Aβ variants Aβ40, Aβ40(NoHis), and Aβ(4–40), as a function of added Ni(II) acetate. Wavelength (nm)0 µM Ni(II)2 µM Ni(II)4 µM Ni(II)16 µM Ni(II)56 µM Ni(II)156 µM Ni(II)256 µM Ni(II)Aβ40208− 11,131− 10,808− 10,557− 9928− 9515− 9064− 9025222− 7793− 7925− 8074− 8241− 8339− 8012− $\frac{8114222}{2080.7000.7330.7650.8300.87640.8840.899}$Aβ40(NoHis)208− 14,062.7− 13,979− 14,112.3− 13,919.9− 13,962.919,802.913,725.2222− 11,184.4− 11,100.6− 11,249.3− 11,138.9− 11,172.6− 11,149.8− 11,$\frac{039.3222}{2080.79530.79410.79710.80020.80020.80780.8043}$Aβ(4–40)208− 5335.7− 5265.9− 5258.7− 4876.8− 4338.2− 4235.2− 4150.7222− 3978.0− 3989.9− 4193.3− 4302.0− 4205.2− $\frac{4176.14268.9222}{2080.7460.7580.7970.8820.9690.9861.028}$
## Estimates of Aβ·Ni(II) binding affinity
Ni(II)·Aβ binding curves were generated by plotting crosspeak intensities from the NMR titrations (Fig. 1A) and 208 nm intensities from the CD titrations (Table 1) versus Ni(II) concentration (Fig. 4). Fitting Eq. [ 1] to these curves yields apparent dissociation constants (KD) for the Ni(II)·Aβ complexes. Figure 4Binding curves for Ni(II)·Aβ complexes, derived from the NMR data in Fig. 1A and the CD data in Fig. 3A,C, respectively. The NMR and CD signal intensities are given as the ratio between the intensity with added Ni(II) ions relative to the intensity before addition of Ni(II) ions, i.e. I/I0. Apparent binding affinities were obtained by fitting Eq. [ 1] to the curves. ( A–C) NMR crosspeak intensity vs Ni(II) concentration for three of the crosspeaks/residues in Fig. 1A, i.e. for 84 µM Aβ40 in 20 mM sodium phosphate buffer, pH 7.3 at 5 °C. ( D,E) CD intensity at 208 nm vs Ni(II) concentration for the CD data in Fig. 3A,C, i.e. for 10 µM Aβ40 (D) and 5 µM Aβ(4–40) (E) in 20 mM phosphate buffer, pH 7.3 at 25 °C.
For the CD data (Fig. 4D,E), the signal intensities have been normalized to the first value in each titration series, i.e. the signal intensity without added Ni(II) ions. The derived KD values are 7.8 µM for binding to Aβ40, and 17.2 µM for binding to the Aβ(4–40) variant. These KD values should however only be considered as approximations, as there may not be a direct correlation between Ni(II) binding and the structural changes observed in the CD spectra (Fig. 3). Our NMR measurements (Fig. 1C) confirm earlier studies showing that the N-terminal Aβ segment is free to interact with metal ions also when the central and C-terminal Aβ segments are inserted into SDS micelles48,51,96. Yet, earlier studies with Cu(II) ions have shown that the binding affinity for metal ions is reduced when the central Aβ segment is bound to some other entity106, although this effect appears to be minor for SDS micelles96.
For the NMR data, each crosspeak in Fig. 1A generates one binding curve. However, the Ni(II)-induced loss of crosspeak intensity is not only related to concentration-dependent chemical exchange, on an intermediate NMR time-scale, but also to Ni(II)-induced peptide aggregation and paramagnetic quenching of the NMR signal. Even though these effects are somewhat mitigated by normalizing the crosspeak intensity values to the V40 crosspeak intensity, for each titration step (thereby obtaining the relative intensity scale used in Fig. 4A–C), the derived KD values should only be considered as rough approximations. The binding curves shown in Fig. 4A–C correspond to the three Aβ40 crosspeaks that display the strongest apparent binding, with KD values of respectively 5.3 µM, 6.7 µM, and 7.0 µM. These values are very similar to the apparent KD value of 7.8 µM derived for Aβ40 from the CD measurements (Fig. 4D), but they should still only be regarded as lower limits for the true KD value. Thus, we conclude that the binding affinity for the Aβ40·Ni(II) complex is in the low µM range.
## Effects of Ni(II) ions on Aβ40 aggregation kinetics
To investigate the influence of Ni(II) ions on Aβ40 aggregation, samples of 10 µM Aβ40 were incubated for 24 h in the absence or presence of increasing concentrations of Ni(II) acetate. The resulting ThT curves are shown in Fig. 5, and the rmax, t$\frac{1}{2}$, and A parameters obtained from fitting the curves to Eq. [ 2] are shown in Table 2. For sub-stoichiometric Ni(II):Aβ40 ratios, the Ni(II) ions slow down the Aβ40 aggregation kinetics in a concentration-dependent manner. The aggregation half-time (t$\frac{1}{2}$) increases with increasing Ni(II) concentrations, i.e. from around 8 h without Ni(II) ions to over 13 h with 10 µM Ni(II) ions. The maximum aggregation rate rmax shows no systematic change with increasing Ni(II) concentration, instead it fluctuates around 0.9 h−1 to 1 h−1 (Table 2). The end-point ThT fluorescence intensities (parameter “A” in Eq. 2) generally decrease with increasing Ni(II) concentrations (Fig. 5), suggesting that less amyloid material (ThT-binding aggregates) is formed when Ni(II) ions are present. Other explanations are however possible, such as binding competition between ThT molecules and Ni(II) ions, or formation of very large Aβ aggregates that may block the transmitted light, or simply precipitate out of the solution. For the samples with high Ni(II) concentrations, i.e. 20 µM and 50 µM, above the stoichiometric Ni(II):Aβ40 ratio, the ThT curves first increase but then decrease back towards the starting value (Fig. 5). Our best explanation for this unusual behaviour is formation of large samples that precipitate, thereby effectively reducing the Aβ40 concentration in the sample. It was not possible to fit Eq. [ 2] to these data curves. Figure 5ThT kinetic time curves for aggregation of 10 µM Aβ40, in 10 mM sodium phosphate buffer, pH 7.4, together with different concentrations of Ni(II) acetate: 0 µM (black), 1 µM (red), 2.5 µM (azure blue), 5 µM (magenta), 7.5 µM (green), 10 µM (navy blue), 20 µM (yellow), and 50 µM (orange). The solid lines show curves fitted with Eq. [ 2] to the ThT data sets. Table 2Parameters rmax, t$\frac{1}{2}$, and A for aggregation of 10 µM Aβ40 in the presence of different concentrations of Ni(II) acetate. Ni(II)0 µM1 µM2.5 µM5 µM7.5 µM10 µM20 µM50 µMrmax [h−1]0.96 ± 0.221.08 ± 0.360.88 ± 0.341.0 ± 0.451.02 ± 0.850.96 ± 0.58n/an/at$\frac{1}{2}$ [h]8.3 ± 1.510.2 ± 0.311.8 ± 1.211.2 ± 0.712.5 ± 0.513.5 ± 1.4n/an/aA4115 ± 4362967 ± 2382085 ± 5751449 ± 4801511 ± 1321803 ± 442n/an/aThe parameters were obtained from sigmoidal curve-fitting (Eq. 2) to the ThT curves shown in Fig. 5.
## AFM imaging: effects of Ni(II) ions on the morphology of Aβ40 aggregates
To further characterize the influence of Ni(II) ions on Aβ40 fibril morphology, AFM images were recorded on Aβ40 aggregates formed after 24 h incubation in the presence and the absence of Ni(II) (Fig. 6). Without Ni(II) acetate, 10 µM Aβ40 formed typical amyloid fibrils with an apparent height around 4–5 nm (Fig. 2A), which is in line with previously published work on Aβ fibrils formed in vitro107–109. A similar apparent height was observed in the presence of 1 µM Ni(II) ions (Fig. 6B). The presence of 10 µM Ni(II) ions, i.e. a 1:1 Ni(II):Aβ40 ratio, significantly reduces fibril formation: only occasional very short Aβ40 fibril fragments were observed, which display the same height as the fibrils formed by Aβ40 alone (Fig. 6C). In the presence of 50 µM Ni(II) ions no fibrils were present, but instead amorphous clumps of Aβ40 aggregates with variable heights around 13 nm (Fig. 6D). These results are consistent with the concentration-dependent effects of Ni(II) acetate on the Aβ40 aggregation process observed with the ThT measurements (Fig. 5).Figure 6Top row: AFM images of the aggregation products obtained after incubation of 10 µM Aβ40 peptides for 24 h together with either 0, 1, 10, or 50 µM Ni(II) acetate. Bottom row: Representative AFM cross-sections of respectively Aβ40 amyloid fibrils (A,B) and Aβ40 unstructured aggregates (C,D), corresponding to the colored lines shown in the AFM images.
## Influence of Ni(II) ions on Aβ42 oligomer formation
PAGE analysis was used to investigate the effect of Ni(II) ions on the formation of Aβ42 oligomers, using a previously published protocol for formation of stable and homogeneous Aβ42 oligomers together with SDS detergent81,82. While most of the current study investigates variants of the Aβ40 peptide, oligomers of Aβ40 are not stable and therefore not suitable model systems. Thus, SEC-purified monomeric solutions of synthetic Aβ42 peptides were mixed with low concentrations of SDS, i.e., below the critical micelle concentration. Incubation of Aβ42 with $0.2\%$ SDS (6.9 mM) leads to formation of mostly tetrameric oligomers (AβO$0.2\%$SDS), while incubation with $0.05\%$ SDS (1.7 mM) produces larger oligomers—predominantly dodecamers (AβO$0.05\%$SDS)81 (Fig. 7, lanes 2 and 6). Figure 7 also shows the effect of different Ni(II) concentrations on oligomer formation. In the absence of Ni(II) ions, the AβO$0.05\%$SDS (Lane 2) and AβO$0.2\%$SDS (Lane 6) oligomers are the dominating species in their respective lanes. With increasing Ni(II) concentration, the band intensity for the major oligomeric structure declines in both cases, while smears towards higher molecular weights appear (Fig. 7, Lanes 3–5 and 7–9). Formation of AβO$0.05\%$SDS is largely disrupted when Ni(II) ions are present at 500 μM concentration (Aβ42:Ni(II) molar ratio of 1:5), while the smear extends over almost the entire length of the lane (Fig. 7, Lane 5). A similar, but less drastic effect is observed for the smaller AβO$0.05\%$SDS (Fig. 7, Lane 9). For both types of SDS-stabilized oligomers, Ni(II) ion concentrations above ~ 100 μM disrupt oligomer formation and more heterogeneous Aβ42 oligomeric solutions containing larger oligomers are produced. A similar smearing effect of Ni(II) ions on the formation of SDS-stabilized Aβ42 oligomers was observed also by SDS-PAGE experiments. These results are shown and discussed in the supplementary information (Fig. S3).Figure 7Effects of Ni(II) ions on formation of SDS-stabilized Aβ42 oligomers (AβO$0.05\%$SDS and AβO$0.2\%$SDS) studied by BN-PAGE. Lane 1: Monomers; Lanes 2–5: AβO$0.05\%$SDS oligomers with respectively 0, 10, 100, and 500 µM Ni(II) ions; Lanes 6–9: AβO$0.2\%$SDS oligomers with respectively 0, 10, 100, and 500 µM Ni(II) ions.
## FTIR spectroscopy of Aβ42 oligomers formed in the presence of Ni(II) ions
FTIR spectroscopy is a powerful technique for studying the secondary structure of proteins110–116, and can be used to characterize the backbone conformation for different aggregation states of amyloid proteins, including Aβ peptides117–119. Here, the effects of Ni(II) ions on the secondary structures of both AβO$0.05\%$SDS and AβO$0.2\%$SDS oligomers were studied with transmission mode FTIR spectroscopy, using a D2O-based buffer. The results are presented in Fig. 8 as second derivatives of IR absorption spectra, where negative bands indicate the component bands of the absorption spectra. The respective absorbance spectra are shown in Fig. S4 of the Supplementary Information. Figure 8Transmission FTIR data for synthetic AβO$0.05\%$SDS (upper row) and synthetic AβO$0.2\%$SDS (lower row) formed in the presence of different concentrations of Ni(II) ions. The spectra show raw data without normalization. Left: Second derivatives of IR absorbance spectra in the amide I range (1700–1600 cm−1) at zero Ni(II) (red); 1 μM Ni(II) (orange); 10 μM Ni(II)—(green); 100 μM Ni(II) (blue); 500 μM Ni(II) (violet). The black spectrum is for Aβ42 monomers. Right: Dependence on Ni(II) ion concentration for the position (in cm−1) of the main amide I band. For a more clear presentation, the data point at 1 µM was omitted.
For both types of Aβ42 oligomers, two bands are resolved in the amide I region (i.e., 1700–1600 cm−1): a high intensity, low wavenumber band around 1630 cm−1 (the main band for β-sheet structure), and a low intensity, high wavenumber band at 1685 cm−1. This pattern with a split double-band in the amide I region is routinely considered as indicative of the anti-parallel β-sheet structure117,118,120. When Ni(II) acetate is introduced during the Aβ42 oligomer formation reactions, the main band is slightly down-shifted: for AβO$0.05\%$SDS (Fig. 8, upper row) from 1629.2 cm−1 in the absence of Ni(II) acetate to 1628.0 cm−1 at 500 μM of Ni(II) acetate (the highest concentration), and for AβO$0.2\%$SDS (Fig. 8, lower row) from 1630.1 to 1629.6 cm−1 upon addition of Ni(II) acetate up to 500 μM. The downshift is smaller for the oligomers prepared at the higher SDS concentration, indicating that they are less sensitive to Ni(II)-induced effects on the oligomer conformation. The spectral changes observed with Ni(II) are in interesting contrast to the absence of spectral effects upon addition of Li(I) ions121. The downshift is mainly observed in the presence of Ni(II) concentrations between 10 and 100 μM, which agrees with the binding affinities estimated from the CD and NMR results. Detailed analysis of the spectra in Fig. 8 shows that the shifts in band position are associated with a widening of the main β-sheet band on its low wavenumber side, indicating a higher abundance of larger oligomers82. Most of this widening occurs between 10 and 100 μM Ni(II). It further increases between 100 and 500 μM Ni(II), which correlates with the appearance of a high molecular weight smear on the BN-PAGE gel (Fig. 7).
Our previous study on the IR characterization of Aβ42 oligomers has revealed a relationship between oligomer size and position of the main band in the amide I region82. According to these findings, the downshift of the main IR band is associated with an increase in average oligomer size and a concomitant extension of their β-sheet structure. However, the Ni(II)-induced size change is rather modest. The band position of AβO$0.2\%$SDS at the highest Ni(II) ion concentration is still higher than the AβO$0.05\%$SDS band position in the absence of Ni(II) ions, indicating that the oligomers contain less than twelve peptides. Also, the band position of AβO$0.05\%$SDS at 500 µM Ni(II) concentration is considerably higher than that of oligomers formed in the absence of SDS (1623.1 cm−1), which had an average molecular weight of ~ 100 kDa according to Western blotting82. Thus, the dominant β-sheet containing oligomer species of AβO$0.05\%$SDS seem to be smaller than ~ 100 kDa.
The interpretation of the current IR results is in good agreement with the results from the gel electrophoresis experiments, particularly with the BN-PAGE data (Fig. 7). Both IR and PAGE results indicate that addition of Ni(II) ions appears to interfere with the SDS-induced conversion of Aβ42 monomers into homogeneous and stable oligomeric structures, instead favoring formation of larger and more diverse (heterogeneous) oligomer populations.
## Ni(II)-induced dityrosine formation
Fluorescence measurements were carried out to investigate if binding of Ni(II) ions could induce formation of covalent dityrosine crosslinks in Aβ peptides, similar to what has been observed for Cu(II) ions122–125. Fluorescence emission spectra for Aβ40 peptides incubated over time with or without Ni(II) acetate are shown in Fig. 9.Figure 9Fluorescence spectra of 10 µM Aβ40 in 20 mM MES buffer, pH 7.3, incubated together with 50 µM EDTA (A) or with 100 µM Ni(II) acetate (B). Black line—0 h; red line—6 h.
The control sample without added Ni(II) ions, which contained 50 µM EDTA to ensure no free metal ions were present, displayed virtually identical spectra before and after 6 h of incubation (Fig. 9A). This is in stark contrast to the sample with 100 µM Ni(II) ions added, where two large new peaks around 410 nm and 435 nm formed during the incubation time (Fig. 9B). The peak around 410 nm is from dityrosine126, while the peak around 435 nm likely is from a different but related system, such as excimers127. These results are not surprising, as nickel is known to be very redox-active. Both the Ni(I)/Ni(II) and Ni(II)/Ni(III) redox pairs could be involved in generating the oxygen radicals required for dityrosine formation128. It should be noted that weak peaks around 410 nm and 435 nm are present in both samples already at time zero (Fig. 9). This shows that some dityrosine cross-links have been generated even before the experiment was initiated, which is consistent with previous reports stating that Aβ peptides can induce oxidative stress on their own, especially in somewhat aggregated states129,130.
## Discussion
Nickel is a well-known neurotoxicant, but its role in neurodegenerative diseases remains unclear44. Several studies have investigated the possible effects of transition metals such as Cu and Zn in AD neuropathology, with an emphasis on interactions with the amyloid-forming Aβ peptides29–32. We therefore interpret our current results on Aβ interactions with Ni(II) ions mainly in the light of earlier work on Aβ binding to Cu(II) and Zn(II) ions.
## Residue-specific binding of Ni(II) ions to Aβ peptides
Our NMR results show that equimolar amounts of Ni(II) ions display residue-specific binding to the N-terminal segment of the Aβ40 peptide (Fig. 1). This is in line with earlier studies showing that Ni(II) ions can bind N-terminal Aβ fragments45–47. The 2D 1H-13C-HSQC NMR data (Fig. 2) suggest that the three histidine residues His6, His13, and His14 are involved as binding ligands, possibly together with the Tyr10 residue. Previous work has established the metal-binding capacity of the Tyr phenol ring131, and in Aβ peptides the Tyr10 residue seems to be involved in binding to Pb(IV) ions38. The weaker Ni(II)/Aβ40 interactions observed at low pH (Fig. 1) further support the histidines being binding ligands, as these residues become protonated at low pH and therefore less prone to interact with cations92,106,132. The CD spectroscopy measurements also support the histidines being involved as binding ligands: addition of Ni(II) ions induces structural changes in Aβ40 peptides, but not in Aβ40(NoHis) mutant peptides (Fig. 3). This indicates that Ni(II) ions do not bind Aβ peptides when the His residues are absent, which is not surprising, given that Ni(II) ions are known to bind His residues such as those in protein His-tags133. Thus, Ni(II) ions seem to belong to a family of metal ions that coordinate to the Aβ N-terminal segment mainly by the His residues, just like Ag(I), Cu(II), Fe(II), Hg(II), Mn(II), Zn(II), and possibly Pb(IV) ions20,37,38,40,47,52,53,92,134,135. The exact binding coordination could not be determined from our measurements, and it is possible that multiple alternating binding conformations exist, as has been shown for Cu(II) ions136.
According to the Irving-Williams series137, the binding affinities of certain divalent metal ions to peptides and proteins should follow the order Mn(II) < Fe(II) < Co(II) < Ni(II) < Cu(II) > Zn(II). Metal binding affinities are however notoriously difficult to quantify, as they tend to vary both with the experimental conditions (buffer, temperature) and the employed measurement technique138. For example, binding affinities varying by several orders of magnitude have previously been reported for the Aβ·Cu(II) complex, with a consensus value in the low nM region for buffer-corrected affinity138. In our earlier studies, we have reported apparent (not buffer-corrected) KD values around 50–100 µM for Mn(II) ions in phosphate buffer, pH 7.3537, around 1–10 µM for Zn(II) ions in phosphate or Hepes buffer, pH 7.240, and around 0.5 – 2.5 µM for Cu(II) ions in phosphate or Hepes buffer, pH 7.2–pH 7.3540,96,106.
Both the CD and the NMR measurements suggest an affinity in the low µM range for Ni(II) binding to Aβ peptides (Fig. 4), i.e. weaker than Cu(II) ions, stronger than Mn(II) ions, and perhaps somewhat similar to Zn(II) binding affinity, which would be consistent with the Irving-Williams series. As the Ni(II) ions bind to the N-terminal Aβ segment, the binding affinity should be rather the same for Aβ40 and Aβ42 peptides, and also for shorter Aβ versions such as Aβ(1–28) and Aβ(1–16). The CD measurements indicate that the Ni(II) binding affinity to the truncated Aβ(4–40) peptide is similar to, or even somewhat weaker than, the affinity to the full-length Aβ40 peptide (Figs. 3 and 4). This is unexpected, as the Aβ(4–40) peptide has been reported to contain an N-terminal binding motif that supposedly provides very strong binding to Cu(II) and Ni(II) ions139, i.e. possibly fM affinity for Cu(II) ions140. Binding of metal ions to truncated Aβ variants is biologically relevant as such variants, and especially Aβ(4–42), are abundant in amyloid plaques from both healthy and AD brain tissues141–144.
## Effects of Ni(II) ions on Aβ structure and aggregation
Similar to e.g. Ag(I), Cu(II), Hg(II), and Zn(II) ions39,40,52,53, Ni(II) ions retard Aβ40 amyloid formation in a concentration-dependent manner by directing the aggregation pathways towards non-fibrillar amorphous aggregates as demonstrated both by ThT fluorescence and AFM imaging (Figs. 5 and 6). Already at a 1:1 Ni(II)/Aβ ratio, Aβ40 fibrillation appears to be completely inhibited. This supports earlier studies reporting that Ni(II) ions can influence protein aggregation145. We have previously shown that low Zn(II) concentrations induce a Zn(II)-bound structure that prevents the Aβ peptides from forming the β-hairpin required for fibrillation19,40. At higher Zn(II) concentrations β-sheet structure was induced92, similar to our current observations with CD spectroscopy that Ni(II) ions induce β-sheet structure in Aβ40 and Aβ(4–40) peptides (Fig. 3D,F). Aβ aggregation is promoted by the direct electrostatic effect of binding cations to the anionic Aβ peptides, thereby reducing repulsion between the Aβ peptides40. Given that Ni(II) and Zn(II) ions have similar charge and binding ligands, they likely affect Aβ aggregation and fibrillation via similar mechanisms.
Although Aβ fibrils, such as those shown in Fig. 6A,B, are the end products of Aβ aggregation, intermediate aggregates known as soluble oligomers are now generally considered to be the main toxic species in AD pathology146,147. The toxic mechanisms are unclear, but may involve membrane disruption30, as some studies have reported that Aβ oligomers can form membrane-spanning “pores” that can induce leakage of e.g. Ca(II) ions148. Interestingly, other studies have reported that this harmful Ca(II) leakage can be inhibited by histidine-associating compounds, such as imidazole, Zn(II), and Ni(II) ions41,149. Because Cu(II), Zn(II), and other divalent ions have been shown to affect the structure, stability, and/or toxicity of Aβ oligomers149–152, it is worth noting that both our FTIR (Figs. 8 and S2) and CD results (Figs. 3 and S1) show that Ni(II) ions induce structural changes in both Aβ oligomers and Aβ monomers. This is in line with earlier studies showing that both SDS-stabilized Aβ42 oligomers81,153,154, and SDS micelle-bound Aβ monomers51,96 contain surface-exposed N-termini which makes it possible for the N-terminal H6, H13, and H14 residues to interact with metal ions. We speculate that Ni(II) ions affect Aβ oligomer structure in similar ways as Cu(II) and Zn(II) ions do, even though the exact mechanisms are not fully understood30. Addition of Ni(II) ions reduces the intensity of the NMR crosspeaks for monomeric Aβ40 peptides in random coil structure (Fig. 1), showing that this conformation becomes less populated. But no new NMR crosspeaks appear (Fig. 1), which shows that the Ni(II) ions do not bind to form a single well-defined Ni(II)·Aβ complex. Instead, a range of heterogeneous β-sheet-containing structures are induced, which most likely exist in different stages of aggregation. They can be observed in CD and FTIR spectra, but not in NMR spectra92,94.
## Effects of Ni(II) ions on Aβ dityrosine formation
Ni(II) ions may affect the Aβ aggregation processes also via formation of reactive oxygen species (ROS). Nickel is well known as a redox-active metal that can adopt a wide range of oxidation states, i.e., from − 1 to + 4155. While this can be usefully employed in engineering contexts such as in Ni–Cd batteries, in biological systems it means that Ni can induce oxidative stress, and this may be one of the main mechanisms of Ni toxicity69. Our experiments show that addition of Ni(II) ions initiates formation of Aβ40 dityrosine cross-links (Fig. 9), which is a common ROS effect. Earlier studies have shown that bound redox-active Cu ions can initiate dityrosine formation, both in Aβ and other peptides and proteins96,122–125,152,156,157. It is therefore not surprising that a similar effect is observed for Ni(II) ions, especially as the NMR results indicate that Tyr10 is one of the Ni(II) binding ligands (Fig. 2). As wt Aβ peptides only contain one Tyr residue, i.e. Tyr10, dityrosine formation must involve two Aβ peptides, which then combine into a covalently linked dimer. Such dityrosine-linked Aβ dimers are of biological significance, as they have been found in amyloid plaques in AD brains158. As these plaques contain elevated levels of bound redox-active Cu and Fe ions26,27, it is likely that the Aβ dityrosine links observed in AD patients are generated by metal-induced ROS. Because dimerization is the first step towards peptide aggregation, dityrosine formation in Aβ peptides is clearly a process that influences aggregation. In vitro studies have shown that dityrosine-linked Aβ dimers undergo rapid aggregation into oligomers that are stable, soluble, and neurotoxic159.
## Conclusions
We here show for the first time that Ni(II) ions bind to the N-terminal segment of biologically relevant (i.e., full-length) Aβ peptides. The Ni(II) binding affinity is in the low µM range, with the three N-terminal His residues and possibly Tyr10 involved as binding ligands. At equimolar amounts, Ni(II) ions impede Aβ fibrillation by directing the aggregation towards amorphous aggregates. The redox-active Ni(II) ions induce dityrosine cross-links via redox chemistry, thereby creating covalent Aβ dimers. Ni(II) ions induce structural alterations in Aβ monomers, both in aqueous buffer (formation of beta sheets) and in membrane-mimicking SDS micelles (likely formation of coil-coil helix), and affect also Aβ oligomerization. Although Ni(II) binding to Aβ is somewhat weaker than Cu(II) binding, the two metal ions induce similar effects on Aβ structure and aggregation. Exposure to stochiometric amounts of Ni(II) ions induces formation of heterogeneous Aβ oligomers, which can be observed with CD and IR but not NMR spectroscopy. These oligomers, which are in a dynamic equilibrium with Aβ monomers, may be important contributors to AD brain pathology.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-29901-5.
## References
1. 1.Prince, M. et al. World Alzheimer Report 2015—The Global Impact of Dementia (Alzheimer’s Disease International (ADI), London, UK, 2015).
2. **2017 Alzheimer's disease facts and figures**. *Alzheimer's Dementia* (2017.0) **13** 325-373. DOI: 10.1016/j.jalz.2017.02.001
3. Fan L. **New insights into the pathogenesis of Alzheimer's disease**. *Front. Neurol* (2019.0) **10** 1312. DOI: 10.3389/fneur.2019.01312
4. Masters CL. **Alzheimer's disease**. *Nat. Rev. Dis. Primers* (2015.0) **1** 15056. DOI: 10.1038/nrdp.2015.56
5. Irvine GB, El-Agnaf OM, Shankar GM, Walsh DM. **Protein aggregation in the brain: The molecular basis for Alzheimer's and Parkinson's diseases**. *Mol. Med.* (2008.0) **14** 451-464. DOI: 10.2119/2007-00100.Irvine
6. Glenner GG, Wong CW. **Alzheimer's disease: Initial report of the purification and characterization of a novel cerebrovascular amyloid protein**. *Biochem. Biophys. Res. Commun.* (1984.0) **120** 885-890. DOI: 10.1016/S0006-291X(84)80190-4
7. Sunde M, Blake CC. **From the globular to the fibrous state: Protein structure and structural conversion in amyloid formation**. *Q. Rev. Biophys.* (1998.0) **31** 1-39. DOI: 10.1017/S0033583598003400
8. Eisenberg D, Jucker M. **The amyloid state of proteins in human diseases**. *Cell* (2012.0) **148** 1188-1203. DOI: 10.1016/j.cell.2012.02.022
9. Haass C, Selkoe DJ. **Soluble protein oligomers in neurodegeneration: Lessons from the Alzheimer's amyloid β-peptide**. *Nat. Rev. Mol. Cell Biol.* (2007.0) **8** 101-112. DOI: 10.1038/nrm2101
10. Walsh DM, Selkoe DJ. **Aβ oligomers—A decade of discovery**. *J. Neurochem.* (2007.0) **101** 1172-1184. DOI: 10.1111/j.1471-4159.2006.04426.x
11. Glabe CG. **Structural classification of toxic amyloid oligomers**. *J. Biol. Chem.* (2008.0) **283** 29639-29643. DOI: 10.1074/jbc.R800016200
12. Sandberg A. **Stabilization of neurotoxic Alzheimer amyloid-β oligomers by protein engineering**. *Proc. Natl. Acad. Sci. USA* (2010.0) **107** 15595-15600. DOI: 10.1073/pnas.1001740107
13. Luo J, Wärmländer SK, Gräslund A, Abrahams JP. **Cross-interactions between the Alzheimer disease amyloid-β peptide and other amyloid proteins: A further aspect of the amyloid cascade hypothesis**. *J. Biol. Chem.* (2016.0) **291** 16485-16493. DOI: 10.1074/jbc.R116.714576
14. Selkoe DJ, Hardy J. **The amyloid hypothesis of Alzheimer's disease at 25 years**. *EMBO Mol. Med.* (2016.0) **8** 595-608. DOI: 10.15252/emmm.201606210
15. Nath S. **Spreading of neurodegenerative pathology via neuron-to-neuron transmission of β-amyloid**. *J. Neurosci.* (2012.0) **32** 8767-8777. DOI: 10.1523/JNEUROSCI.0615-12.2012
16. Sardar Sinha M. **Alzheimer's disease pathology propagation by exosomes containing toxic amyloid-β oligomers**. *Acta Neuropathol.* (2018.0) **136** 41-56. DOI: 10.1007/s00401-018-1868-1
17. Aaseth J. **Copper, iron, selenium and lipo-glycemic dysmetabolism in Alzheimer's disease**. *Int. J. Mol. Sci.* (2021.0). DOI: 10.3390/ijms22179461
18. Owen MC. **Effects of in vivo conditions on amyloid aggregation**. *Chem. Soc. Rev.* (2019.0). DOI: 10.1039/c8cs00034d
19. Abelein A. **The hairpin conformation of the amyloid β peptide is an important structural motif along the aggregation pathway**. *J. Biol. Inorg. Chem.* (2014.0) **19** 623-634. DOI: 10.1007/s00775-014-1131-8
20. Wärmländer S. **Biophysical studies of the amyloid β-peptide: Interactions with metal ions and small molecules**. *ChemBioChem* (2013.0) **14** 1692-1704. DOI: 10.1002/cbic.201300262
21. Luo J. **Endogenous polyamines reduce the toxicity of soluble Aβ peptide aggregates associated with Alzheimer's disease**. *Biomacromol* (2014.0) **15** 1985-1991. DOI: 10.1021/bm401874j
22. Luo J. **Cellular polyamines promote amyloid-beta (Aβ) peptide fibrillation and modulate the aggregation pathways**. *ACS Chem. Neurosci.* (2013.0) **4** 454-462. DOI: 10.1021/cn300170x
23. Wallin C, Luo J, Jarvet J, Wärmländer S, Gräslund A. **The Amyloid-β peptide in amyloid formation processes: Interactions with blood proteins and naturally occurring metal ions**. *Isr. J. Chem.* (2017.0) **57** 674-685. DOI: 10.1002/ijch.201600105
24. Duce JA, Bush AI, Adlard PA. **Role of amyloid-β–metal interactions in Alzheimer’s disease**. *Future Neurol.* (2011.0) **6** 641-659. DOI: 10.2217/fnl.11.43
25. Szabo ST, Harry GJ, Hayden KM, Szabo DT, Birnbaum L. **Comparison of metal levels between postmortem brain and ventricular fluid in Alzheimer's disease and nondemented elderly controls**. *Toxicol. Sci.* (2016.0) **150** 292-300. DOI: 10.1093/toxsci/kfv325
26. Lovell MA, Robertson JD, Teesdale WJ, Campbell JL, Markesbery WR. **Copper, iron and zinc in Alzheimer's disease senile plaques**. *J. Neurol. Sci.* (1998.0) **158** 47-52. DOI: 10.1016/S0022-510X(98)00092-6
27. Miller LM. **Synchrotron-based infrared and X-ray imaging shows focalized accumulation of Cu and Zn co-localized with β-amyloid deposits in Alzheimer's disease**. *J. Struct. Biol.* (2006.0) **155** 30-37. DOI: 10.1016/j.jsb.2005.09.004
28. Beauchemin D, Kisilevsky R. **A method based on ICP-MS for the analysis of Alzheimer's amyloid plaques**. *Anal. Chem.* (1998.0) **70** 1026-1029. DOI: 10.1021/ac970783f
29. Tiiman A, Palumaa P, Tougu V. **The missing link in the amyloid cascade of Alzheimer's disease-metal ions**. *Neurochem. Int.* (2013.0) **62** 367-378. DOI: 10.1016/j.neuint.2013.01.023
30. Wärmländer SKTS. **Metal binding to the amyloid-β peptides in the presence of biomembranes: Potential mechanisms of cell toxicity**. *J. Biol. Inorganic Chem.* (2019.0) **24** 1189-1196. DOI: 10.1007/s00775-019-01723-9
31. Huat TJ. **Metal toxicity links to Alzheimer's disease and neuroinflammation**. *J. Mol. Biol.* (2019.0) **431** 1843-1868. DOI: 10.1016/j.jmb.2019.01.018
32. Kim AC, Lim S, Kim YK. **Metal ion effects on Aβ and Tau aggregation**. *Int. J. Mol. Sci.* (2018.0). DOI: 10.3390/ijms19010128
33. Wild K, August A, Pietrzik CU, Kins S. **Structure and synaptic function of metal binding to the amyloid precursor protein and its proteolytic fragments**. *Front. Mol. Neurosci.* (2017.0) **10** 21. DOI: 10.3389/fnmol.2017.00021
34. Ayton S, Lei P, Bush AI. **Metallostasis in Alzheimer's disease**. *Free Radic. Biol. Med.* (2013.0) **62** 76-89. DOI: 10.1016/j.freeradbiomed.2012.10.558
35. Branch T, Barahona M, Dodson CA, Ying L. **Kinetic analysis reveals the identity of Aβ-metal complex responsible for the initial aggregation of Aβ in the synapse**. *ACS Chem. Neurosci.* (2017.0) **8** 1970-1979. DOI: 10.1021/acschemneuro.7b00121
36. Miller Y, Ma B, Nussinov R. **Metal binding sites in amyloid oligomers: Complexes and mechanisms**. *Coord. Chem. Rev.* (2012.0) **256** 2245-2252. DOI: 10.1016/j.ccr.2011.12.022
37. Wallin C. **Characterization of Mn(II) ion binding to the amyloid-β peptide in Alzheimer's disease**. *J. Trace Elem. Med. Biol.* (2016.0) **38** 183-193. DOI: 10.1016/j.jtemb.2016.03.009
38. Wallin C. **Alzheimer's disease and cigarette smoke components: Effects of nicotine, PAHs, and Cd(II), Cr(III), Pb(II), Pb(IV) ions on amyloid-β peptide aggregation**. *Sci. Rep.* (2017.0) **7** 14423. DOI: 10.1038/s41598-017-13759-5
39. Faller P, Hureau C, Berthoumieu O. **Role of metal ions in the self-assembly of the Alzheimer's amyloid-β peptide**. *Inorg. Chem.* (2013.0) **52** 12193-12206. DOI: 10.1021/ic4003059
40. Danielsson J, Pierattelli R, Banci L, Gräslund A. **High-resolution NMR studies of the zinc-binding site of the Alzheimer's amyloid β-peptide**. *FEBS J.* (2007.0) **274** 46-59. DOI: 10.1111/j.1742-4658.2006.05563.x
41. Arispe N, Diaz JC, Flora M. **Efficiency of histidine-associating compounds for blocking the Alzheimer's Aβ channel activity and cytotoxicity**. *Biophys. J.* (2008.0) **95** 4879-4889. DOI: 10.1529/biophysj.108.135517
42. 42.ATSDR. Agency for Toxic Substances and Disease Registry: Toxicological Profile for Nickel (2005).
43. Uter W, Pfahlberg A, Gefeller O, Geier J, Schnuch A. **Risk factors for contact allergy to nickel—Results of a multifactorial analysis**. *Contact Dermatitis* (2003.0) **48** 33-38. DOI: 10.1034/j.1600-0536.46.s4.29_102.x
44. 44.Ijomone, O. M. Advances in Neurotoxicology (ed Ashner, M.) 263–284 (Academic Press, 2021).
45. Józsa E. **Nickel(II) and mixed metal complexes of amyloid-β N-terminus**. *Dalton Trans.* (2010.0) **39** 7046-7053. DOI: 10.1039/c0dt00189a
46. Grenács A, Sóvágó I. **Copper(II), nickel(II) and zinc(II) complexes of the N-terminal nonapeptide fragment of amyloid-β and its derivatives**. *J. Inorg. Biochem.* (2014.0) **139** 49-56. DOI: 10.1016/j.jinorgbio.2014.06.001
47. Lermyte F. **Metal ion binding to the amyloid β monomer studied by native top-down FTICR mass spectrometry**. *J. Am. Soc. Mass Spectrom.* (2019.0). DOI: 10.1007/s13361-019-02283-7
48. Österlund N. **Amyloid-β peptide interactions with amphiphilic surfactants: Electrostatic and hydrophobic effects**. *ACS Chem. Neurosci.* (2018.0) **9** 1680-1692. DOI: 10.1021/acschemneuro.8b00065
49. Österlund N, Luo J, Wärmländer SKTS, Gräslund A. **Membrane-mimetic systems for biophysical studies of the amyloid-β peptide**. *Biochim. Biophys. Acta Proteins Proteom.* (1867.0) **492–501** 2019. DOI: 10.1016/j.bbapap.2018.11.005
50. Danielsson J, Jarvet J, Damberg P, Gräslund A. **The Alzheimer β-peptide shows temperature-dependent transitions between left-handed 3-helix, β-strand and random coil secondary structures**. *FEBS J.* (2005.0) **272** 3938-3949. DOI: 10.1111/j.1742-4658.2005.04812.x
51. Jarvet J, Danielsson J, Damberg P, Oleszczuk M, Gräslund A. **Positioning of the Alzheimer Aβ(1–40) peptide in SDS micelles using NMR and paramagnetic probes**. *J. Biomol. NMR* (2007.0) **39** 63-72. DOI: 10.1007/s10858-007-9176-4
52. Wallin C. **Mercury and Alzheimer's disease: Hg(II) ions display specific binding to the amyloid-β peptide and hinder its fibrillization**. *Biomolecules* (2019.0). DOI: 10.3390/biom10010044
53. Wallin C. **Metal ion coordination delays amyloid-β peptide self-assembly by forming an aggregation-inert complex**. *J. Biol. Chem.* (2020.0) **295** 7224-7234. DOI: 10.1074/jbc.RA120.012738
54. 54.Klein, C. & Costa, M. Handbook on the Toxicology of Metals, 4th ed. Volume II: Specific Metals (eds Nordberg, G. F. et al.) (Elsevier, 2015).
55. Anke M, Groppel B, Kronemann H, Grun M. *Nickel—An essential element* (1984.0) 339-365
56. Fischer F. **Characterization in**. *PLoS Pathog.* (2016.0) **12** e1006018. DOI: 10.1371/journal.ppat.1006018
57. Eskew DL, Welch RM, Cary EE. **Nickel: An essential micronutrient for legumes and possibly all higher plants**. *Science* (1983.0) **222** 621-623. DOI: 10.1126/science.222.4624.621
58. Genchi G, Carocci A, Lauria G, Sinicropi MS, Catalano A. **Nickel: Human health and environmental toxicology**. *Int. J. Environ. Res. Public Health* (2020.0). DOI: 10.3390/ijerph17030679
59. Nielsen FH, Ollerich DA, Fosmire GJ, Sandstead HH. **Nickel deficiency in chicks and rats: Effects on liver morphology, function and polysomal integrity**. *Adv. Exp. Med. Biol.* (1974.0) **48** 389-403. DOI: 10.1007/978-1-4684-0943-7_18
60. Stangl GI, Kirchgessner M. **Nickel deficiency alters liver lipid metabolism in rats**. *J. Nutr.* (1996.0) **126** 2466-2473. DOI: 10.1093/jn/126.10.2466
61. Mania M, Rebeniak M, Postupolski J. **Exposure assessment of the population in Poland to the toxic effects of nickel from vegetable and their products**. *Rocz. Panstw. Zakl. Hig.* (2019.0) **70** 401-406. DOI: 10.32394/rpzh.2019.0095
62. 62.World Health, O. Air quality guidelines for Europe. WHO Reg Publ Eur Ser, V-X, 1–273 (2000).
63. **Update of the risk assessment of nickel in food and drinking water**. *EFSA J.* (2020.0) **18** e06268. DOI: 10.2903/j.efsa.2020.6268
64. Bernhard D, Rossmann A, Wick G. **Metals in cigarette smoke**. *IUBMB Life* (2005.0) **57** 805-809. DOI: 10.1080/15216540500459667
65. Dahlstrand H, Stark A, Anissian L, Hailer NP. **Elevated serum concentrations of cobalt, chromium, nickel, and manganese after metal-on-metal alloarthroplasty of the hip: A prospective randomized study**. *J. Arthroplasty* (2009.0) **24** 837-845. DOI: 10.1016/j.arth.2008.07.019
66. Setcos JC, Babaei-Mahani A, Silvio LD, Mjor IA, Wilson NH. **The safety of nickel containing dental alloys**. *Dent. Mater.* (2006.0) **22** 1163-1168. DOI: 10.1016/j.dental.2005.11.033
67. Cheng CF, Schwitter CM. **Nickel in ancient bronzes**. *Am. J. Archaeol.* (1957.0) **61** 351-365. DOI: 10.2307/500603
68. Rauch JN, Pacyna JM. **Earth's global Ag, Al, Cr, Cu, Fe, Ni, Pb, and Zn cycles**. *Glob. Biogeochem. Cycles* (2009.0) **23** GB2001. DOI: 10.1029/2008GB003376
69. Das KK, Das SN, Dhundasi SA. **Nickel, its adverse health effects & oxidative stress**. *Indian J. Med. Res.* (2008.0) **128** 412-425. PMID: 19106437
70. Nordberg G, Costa M. *Handbook on the Toxicology of Metals* (2022.0)
71. Hou YP. **The characteristics of placental transfer and tissue concentrations of nickel in late gestational rats and fetuses**. *Placenta* (2011.0) **32** 277-282. DOI: 10.1016/j.placenta.2010.12.021
72. Mangelsdorf I, Walach H, Mutter J. **Healing of amyotrophic lateral sclerosis: A case report**. *Complement. Med. Res.* (2017.0) **24** 175-181. DOI: 10.1159/000477397
73. Bar-Sela S, Reingold S, Richter ED. **Amyotrophic lateral sclerosis in a battery-factory worker exposed to cadmium**. *Int. J. Occup. Environ. Health* (2001.0) **7** 109-112. DOI: 10.1179/107735201800339470
74. Forte G. **Quantification of chemical elements in blood of patients affected by multiple sclerosis**. *Ann. Ist. Super Sanita* (2005.0) **41** 213-216. PMID: 16244395
75. Irvine DG, Schiefer HB, Hader WJ. **Geotoxicology of multiple sclerosis: The Henribourg, Saskatchewan, cluster focus. II. The soil**. *Sci. Total Environ.* (1988.0) **77** 175-188. DOI: 10.1016/0048-9697(88)90054-x
76. Frye RE. **Early life metal exposure dysregulates cellular bioenergetics in children with regressive autism spectrum disorder**. *Transl. Psychiatry* (2020.0) **10** 223. DOI: 10.1038/s41398-020-00905-3
77. Kim SH. **Rapid doubling of Alzheimer's amyloid-β40 and 42 levels in brains of mice exposed to a nickel nanoparticle model of air pollution**. *F1000Res* (2012.0) **1** 70. DOI: 10.12688/f1000research.1-70.v1
78. Gorantla NV, Das R, Balaraman E, Chinnathambi S. **Transition metal nickel prevents Tau aggregation in Alzheimer's disease**. *Int. J. Biol. Macromol.* (2020.0) **156** 1359-1365. DOI: 10.1016/j.ijbiomac.2019.11.176
79. Glasoe PK, Long FA. **Use of glass electrodes to measure acidities in deuterium oxide**. *J. Phys. Chem.* (1960.0) **64** 88-90. DOI: 10.1021/j100830a521
80. Edelhoch H. **Spectroscopic determination of tryptophan and tyrosine in proteins**. *Biochemistry* (1967.0) **6** 1948-1954. DOI: 10.1021/bi00859a010
81. Barghorn S. **Globular amyloid β-peptide1-42 oligomer—A homogenous and stable neuropathological protein in Alzheimer’s disease**. *J. Neurochem.* (2005.0) **95** 834-847. DOI: 10.1111/j.1471-4159.2005.03407.x
82. Vosough F, Barth A. **Characterization of homogeneous and heterogeneous amyloid-β42 oligomer preparations with biochemical methods and infrared spectroscopy reveals a correlation between infrared spectrum and oligomer size**. *ACS Chem. Neurosci.* (2021.0) **12** 473-488. DOI: 10.1021/acschemneuro.0c00642
83. Dominguez A, Fernandez A, Gonzalez N, Iglesias E, Motenegro E. **Determination of critical micelle concentration of some surfactants by three techniques**. *J. Chem. Educ.* (1997.0) **74** 1227-1231. DOI: 10.1021/ed074p1227
84. Danielsson J, Andersson A, Jarvet J, Gräslund A. **15N relaxation study of the amyloid β-peptide: Structural propensities and persistence length**. *Magn. Reason. Chem.* (2006.0) **44** S114-S121. DOI: 10.1002/mrc.1814
85. Roche J, Shen Y, Lee JH, Ying J, Bax A. **Monomeric Aβ(1–40) and Aβ (1–42) peptides in solution adopt very similar Ramachandran map distributions that closely resemble random coil**. *Biochemistry* (2016.0) **55** 762-775. DOI: 10.1021/acs.biochem.5b01259
86. Yamaguchi T, Matsuzaki K, Hoshino M. **Transient formation of intermediate conformational states of amyloid-β peptide revealed by heteronuclear magnetic resonance spectroscopy**. *FEBS Lett.* (2011.0) **585** 1097-1102. DOI: 10.1016/j.febslet.2011.03.014
87. 87.Kuzmic, P. History, variants and usage of the "Morrison equation" in enzyme inhibition kinetics, BioKin Technical Note TN-2015-01 http://www.biokin.com/TN/2015/01 (2015).
88. Biancalana M, Koide S. **Molecular mechanism of Thioflavin-T binding to amyloid fibrils**. *Biochem. Biophys. Acta.* (1804.0) **1405–1412** 2010. DOI: 10.1016/j.bbapap.2010.04.001
89. Gade Malmos K. **ThT 101: A primer on the use of thioflavin T to investigate amyloid formation**. *Amyloid* (2017.0) **24** 1-16. DOI: 10.1080/13506129.2017.1304905
90. Hellstrand E, Boland B, Walsh DM, Linse S. **Amyloid β-protein aggregation produces highly reproducible kinetic data and occurs by a two-phase process**. *ACS Chem Neurosci* (2010.0) **1** 13-18. DOI: 10.1021/cn900015v
91. Baldassarre M, Barth A. **Pushing the detection limit of infrared spectroscopy for structural analysis of dilute protein samples**. *Analyst* (2014.0) **139** 5393-5399. DOI: 10.1039/c4an00918e
92. Ghalebani L, Wahlström A, Danielsson J, Wärmländer SK, Gräslund A. **pH-dependence of the specific binding of Cu(II) and Zn(II) ions to the amyloid-β peptide**. *Biochem. Biophys. Res. Commun.* (2012.0) **421** 554-560. DOI: 10.1016/j.bbrc.2012.04.043
93. Barnhart JL, Berk RN. **Influence of paramagnetic ions and pH on proton NMR relaxation of biologic fluids**. *Investig. Radiol.* (1986.0) **21** 132-136. DOI: 10.1097/00004424-198602000-00009
94. Fawzi NL, Ying J, Torchia DA, Clore GM. **Kinetics of amyloid β monomer-to-oligomer exchange by NMR relaxation**. *J. Am. Chem. Soc.* (2010.0) **132** 9948-9951. DOI: 10.1021/ja1048253
95. Pogostin BH, Malmendal A, Londergan CH, Akerfeldt KS. **pKa determination of a histidine residue in a short peptide using Raman spectroscopy**. *Molecules* (2019.0). DOI: 10.3390/molecules24030405
96. Tiiman A. **Specific binding of Cu(II) ions to amyloid-β peptides bound to aggregation-inhibiting molecules or SDS micelles creates complexes that generate radical oxygen species**. *J. Alzheimers Dis.* (2016.0) **54** 971-982. DOI: 10.3233/JAD-160427
97. Gielnik M. **The prion protein octarepeat domain forms transient β-sheet structures upon residue-specific Cu(II) and Zn(II) binding**. *bioRxiv* (2021.0). DOI: 10.1101/2021.12.12.472308
98. Kjaergaard M. **Temperature-dependent structural changes in intrinsically disordered proteins: Formation of alpha-helices or loss of polyproline II?**. *Protein Sci.* (2010.0) **19** 1555-1564. DOI: 10.1002/pro.435
99. Greenfield N, Fasman GD. **Computed circular dichroism spectra for the evaluation of protein conformation**. *Biochemistry* (1969.0) **8** 4108-4116. DOI: 10.1021/bi00838a031
100. Usachev KS, Filippov AV, Khairutdinov BI, Antzutkin ON, Klochkov VV. **NMR structure of the Arctic mutation of the Alzheimer’s Aβ(1–40) peptide docked to SDS micelles**. *J. Mol. Struct.* (2014.0) **1076** 518-523. DOI: 10.1016/j.molstruc.2014.08.030
101. Lopez-Llano J, Campos LA, Sancho J. **Alpha-helix stabilization by alanine relative to glycine: Roles of polar and apolar solvent exposures and of backbone entropy**. *Proteins* (2006.0) **64** 769-778. DOI: 10.1002/prot.21041
102. Lau SY, Taneja AK, Hodges RS. **Synthesis of a model protein of defined secondary and quaternary structure. Effect of chain length on the stabilization and formation of two-stranded alpha-helical coiled-coils**. *J. Biol. Chem.* (1984.0) **259** 13253-13261. DOI: 10.1016/S0021-9258(18)90686-1
103. Zhou NE, Kay CM, Hodges RS. **Synthetic model proteins: The relative contribution of leucine residues at the nonequivalent positions of the 3–4 hydrophobic repeat to the stability of the two-stranded alpha-helical coiled-coil**. *Biochemistry* (1992.0) **31** 5739-5746. DOI: 10.1021/bi00140a008
104. Zhou NE, Kay CM, Hodges RS. **Synthetic model proteins. Positional effects of interchain hydrophobic interactions on stability of two-stranded alpha-helical coiled-coils**. *J. Biol. Chem.* (1992.0) **267** 2664-2670. DOI: 10.1016/S0021-9258(18)45932-7
105. Barbar E, Nyarko A. **NMR characterization of self-association domains promoted by interactions with LC8 hub protein**. *Comput. Struct. Biotechnol. J.* (2014.0) **9** e201402003. DOI: 10.5936/csbj.201402003
106. Lindgren J. **Engineered non-fluorescent Affibody molecules facilitate studies of the amyloid-beta (Aβ) peptide in monomeric form: Low pH was found to reduce Aβ/Cu(II) binding affinity**. *J. Inorg. Biochem.* (2013.0) **120** 18-23. DOI: 10.1016/j.jinorgbio.2012.11.005
107. Luo J. **Inhibiting and reversing amyloid-β peptide (1–40) fibril formation with gramicidin S and engineered analogues**. *Chemistry* (2013.0) **19** 17338-17348. DOI: 10.1002/chem.201301535
108. Luo J, Wärmländer SK, Gräslund A, Abrahams JP. **Human lysozyme inhibits the in vitro aggregation of Aβ peptides, which in vivo are associated with Alzheimer's disease**. *Chem. Commun.* (2013.0) **49** 6507-6509. DOI: 10.1039/c3cc42325e
109. Luo J, Wärmländer SK, Gräslund A, Abrahams JP. **Reciprocal molecular interactions between the Aβ peptide linked to Alzheimer's disease and insulin linked to diabetes mellitus type II**. *ACS Chem. Neurosci.* (2016.0) **7** 269-274. DOI: 10.1021/acschemneuro.5b00325
110. Barth A. **Infrared spectroscopy of proteins**. *Biochem. Biophys. Acta.* (2007.0) **1767** 1073-1101. DOI: 10.1016/j.bbabio.2007.06.004
111. Barth A, Zscherp C. **What vibrations tell us about proteins**. *Q. Rev. Biophys.* (2002.0) **35** 369-430. DOI: 10.1017/s0033583502003815
112. Yang H, Yang S, Kong J, Dong A, Yu S. **Obtaining information about protein secondary structures in aqueous solution using Fourier transform IR spectroscopy**. *Nat. Protoc.* (2015.0) **10** 382-396. DOI: 10.1038/nprot.2015.024
113. Goormaghtigh E, Cabiaux V, Ruysschaert JM. **Determination of soluble and membrane protein structure by Fourier transform infrared spectroscopy. III. Secondary structures**. *Subcell Biochem.* (1994.0) **23** 405-450. DOI: 10.1007/978-1-4615-1863-1_10
114. Jackson M, Mantsch HH. **The use and misuse of FTIR spectroscopy in the determination of protein structure**. *Crit. Rev. Biochem. Mol. Biol.* (1995.0) **30** 95-120. DOI: 10.3109/10409239509085140
115. Arrondo JL, Muga A, Castresana J, Goni FM. **Quantitative studies of the structure of proteins in solution by Fourier-transform infrared spectroscopy**. *Prog. Biophys. Mol. Biol.* (1993.0) **59** 23-56. DOI: 10.1016/0079-6107(93)90006-6
116. 116.Fabian, H. & Mäntele, W. Handbook of Vibrational Spectroscopy (eds Chalmers, J. M. & Griffiths, P.) 3399–3426 (Wiley, 2002).
117. Sarroukh R, Goormaghtigh E, Ruysschaert JM, Raussens V. **ATR-FTIR: A "rejuvenated" tool to investigate amyloid proteins**. *Biochem. Biophys. Acta.* (1828.0) **2328–2338** 2013. DOI: 10.1016/j.bbamem.2013.04.012
118. Li H, Lantz R, Du D. **Vibrational approach to the dynamics and structure of protein amyloids**. *Molecules* (2019.0). DOI: 10.3390/molecules24010186
119. Moran SD, Zanni MT. **How to get insight into amyloid structure and formation from infrared spectroscopy**. *J. Phys. Chem. Lett.* (2014.0) **5** 1984-1993. DOI: 10.1021/jz500794d
120. Cerf E. **Antiparallel β-sheet: A signature structure of the oligomeric amyloid β-peptide**. *Biochem. J.* (2009.0) **421** 415-423. DOI: 10.1042/BJ20090379
121. Berntsson E. **Lithium ions display weak interaction with amyloid-beta (Aβ) peptides and have minor effects on their aggregation**. *Acta Biochim. Pol.* (2021.0) **68** 169-179. DOI: 10.18388/abp.2020_5493
122. Atwood CS. **Copper mediates dityrosine cross-linking of Alzheimer's amyloid-β**. *Biochemistry* (2004.0) **43** 560-568. DOI: 10.1021/bi0358824
123. Dong X. **Copper ions induce dityrosine-linked dimers in human but not in murine islet amyloid polypeptide (IAPP/amylin)**. *Biochem. Biophys. Res. Commun.* (2019.0) **510** 520-524. DOI: 10.1016/j.bbrc.2019.01.120
124. Gu M, Bode DC, Viles JH. **Copper redox cycling inhibits Aβ fibre formation and promotes fibre fragmentation, while generating a dityrosine Aβ dimer**. *Sci. Rep.* (2018.0) **8** 16190. DOI: 10.1038/s41598-018-33935-5
125. Vazquez G. **Copper, dityrosine cross-links and amyloid-β aggregation**. *J. Biol. Inorg. Chem.* (2019.0) **24** 1217-1229. DOI: 10.1007/s00775-019-01734-6
126. Huggins TG, Wells-Knecht MC, Detorie NA, Baynes JW, Thorpe SR. **Formation of o-tyrosine and dityrosine in proteins during radiolytic and metal-catalyzed oxidation**. *J. Biol. Chem.* (1993.0) **268** 12341-12347. DOI: 10.1016/S0021-9258(18)31395-4
127. Keleti T. **The excimer fluorescence of tryptophan, tyrosine and**. *FEBS Lett.* (1970.0) **7** 280-282. DOI: 10.1016/0014-5793(70)80181-8
128. Zilbermann I, Maimon E, Cohen H, Meyerstein D. **Redox chemistry of nickel complexes in aqueous solutions**. *Chem. Rev.* (2005.0) **105** 2609-2625. DOI: 10.1021/cr030717f
129. Harris ME, Hensley K, Butterfield DA, Leedle RA, Carney JM. **Direct evidence of oxidative injury produced by the Alzheimer's β-amyloid peptide (1–40) in cultured hippocampal neurons**. *Exp. Neurol.* (1995.0) **131** 193-202. DOI: 10.1016/0014-4886(95)90041-1
130. Cenini G. **Generation of reactive oxygen species by β amyloid fibrils and oligomers involves different intra/extracellular pathways**. *Amino Acids* (2010.0) **38** 1101-1106. DOI: 10.1007/s00726-009-0319-7
131. Shimazaki Y, Yajima T, Yamauchi O. **Properties of the indole ring in metal complexes. A comparison with the phenol ring**. *J. Inorg. Biochem.* (2015.0) **148** 105-115. DOI: 10.1016/j.jinorgbio.2015.03.001
132. Zhang S, Casey N, Lee JP. **Residual structure in the Alzheimer's disease peptide: Probing the origin of a central hydrophobic cluster**. *Fold. Des.* (1998.0) **3** 413-422. DOI: 10.1016/S1359-0278(98)00054-6
133. Knecht S, Ricklin D, Eberle AN, Ernst B. **Oligohis-tags: Mechanisms of binding to Ni**. *J. Mol. Recognit.* (2009.0) **22** 270-279. DOI: 10.1002/jmr.941
134. Bousejra-ElGarah F, Bijani C, Coppel Y, Faller P, Hureau C. **Iron(II) binding to amyloid-β, the Alzheimer's peptide**. *Inorg. Chem.* (2011.0) **50** 9024-9030. DOI: 10.1021/ic201233b
135. Faller P, Hureau C. **Bioinorganic chemistry of copper and zinc ions coordinated to amyloid-β peptide**. *Dalton Trans.* (2009.0). DOI: 10.1039/b813398k
136. Faller P. **Copper and zinc binding to amyloid-β: Coordination, dynamics, aggregation, reactivity and metal-ion transfer**. *ChemBioChem* (2009.0) **10** 2837-2845. DOI: 10.1002/cbic.200900321
137. 137.Irving, H. M. N. H. & Williams, R. J. P. The stability of transition-metal complexes. J. Chem. Soc. 3192–3210 (1953).
138. Alies B. **Cu(II) affinity for the Alzheimer's peptide: Tyrosine fluorescence studies revisited**. *Anal. Chem.* (2013.0) **85** 1501-1508. DOI: 10.1021/ac302629u
139. Bar-Or D, Curtis G, Rao N, Bampos N, Lau E. **Characterization of the Co(2+) and Ni(2+) binding amino-acid residues of the N-terminus of human albumin. An insight into the mechanism of a new assay for myocardial ischemia**. *Eur. J. Biochem.* (2001.0) **268** 42-47. DOI: 10.1046/j.1432-1327.2001.01846.x
140. Mital M. **A functional role for Aβ in metal homeostasis? N-truncation and high-affinity copper binding**. *Angew. Chem. Int. Ed. Engl.* (2015.0) **54** 10460-10464. DOI: 10.1002/anie.201502644
141. Masters CL. **Amyloid plaque core protein in Alzheimer disease and Down syndrome**. *Proc. Natl. Acad. Sci. USA* (1985.0) **82** 4245-4249. DOI: 10.1073/pnas.82.12.4245
142. Masters CL. **Neuronal origin of a cerebral amyloid: Neurofibrillary tangles of Alzheimer's disease contain the same protein as the amyloid of plaque cores and blood vessels**. *EMBO J.* (1985.0) **4** 2757-2763. DOI: 10.1002/j.1460-2075.1985.tb04000.x
143. Lewis H. **Quantification of Alzheimer pathology in ageing and dementia: Age-related accumulation of amyloid-β(42) peptide in vascular dementia**. *Neuropathol. Appl. Neurobiol.* (2006.0) **32** 103-118. DOI: 10.1111/j.1365-2990.2006.00696.x
144. Portelius E. **Mass spectrometric characterization of brain amyloid β isoform signatures in familial and sporadic Alzheimer's disease**. *Acta Neuropathol.* (2010.0) **120** 185-193. DOI: 10.1007/s00401-010-0690-1
145. Hedberg YS. **Synergistic effects of metal-induced aggregation of human serum albumin**. *Colloids Surf. B Biointerfaces* (2019.0) **173** 751-758. DOI: 10.1016/j.colsurfb.2018.10.061
146. Lee SJ, Nam E, Lee HJ, Savelieff MG, Lim MH. **Towards an understanding of amyloid-β oligomers: Characterization, toxicity mechanisms, and inhibitors**. *Chem. Soc. Rev.* (2017.0) **46** 310-323. DOI: 10.1039/c6cs00731g
147. Sengupta U, Nilson AN, Kayed R. **The role of amyloid-β oligomers in toxicity, propagation, and immunotherapy**. *EBioMedicine* (2016.0) **6** 42-49. DOI: 10.1016/j.ebiom.2016.03.035
148. Bode DC, Baker MD, Viles JH. **Ion channel formation by amyloid-β42 oligomers but not amyloid-β40 in cellular membranes**. *J. Biol. Chem.* (2017.0) **292** 1404-1413. DOI: 10.1074/jbc.M116.762526
149. Kim M, Son J, Kim Y. **NMR studies of the ion channel-forming human amyloid-β with zinc ion concentrations**. *Membranes* (2021.0). DOI: 10.3390/membranes11110799
150. Lee MC. **Zinc ion rapidly induces toxic, off-pathway amyloid-β oligomers distinct from amyloid-β derived diffusible ligands in Alzheimer's disease**. *Sci. Rep.* (2018.0) **8** 4772. DOI: 10.1038/s41598-018-23122-x
151. Garai K, Sengupta P, Sahoo B, Maiti S. **Selective destabilization of soluble amyloid β oligomers by divalent metal ions**. *Biochem. Biophys. Res. Commun.* (2006.0) **345** 210-215. DOI: 10.1016/j.bbrc.2006.04.056
152. Williams TL, Serpell LC, Urbanc B. **Stabilization of native amyloid β-protein oligomers by copper and hydrogen peroxide induced cross-linking of unmodified proteins (CHICUP)**. *Biochem. Biophys. Acta.* (1864.0) **249–259** 2016. DOI: 10.1016/j.bbapap.2015.12.001
153. Yu L. **Structural characterization of a soluble amyloid β-peptide oligomer**. *Biochemistry* (2009.0) **48** 1870-1877. DOI: 10.1021/bi802046n
154. Gu L, Liu C, Guo Z. **Structural insights into Aβ42 oligomers using site-directed spin labeling**. *J. Biol. Chem.* (2013.0) **288** 18673-18683. DOI: 10.1074/jbc.M113.457739
155. Nag K, Chakravorty A. **Monovalent, trivalent and tetravalent nickel**. *Coord. Chem. Rev.* (1980.0) **33** 87-147. DOI: 10.1016/S0010-8545(00)80405-1
156. Smith DP. **Concentration dependent Cu**. *Biochemistry* (2007.0) **46** 2881-2891. DOI: 10.1021/bi0620961
157. Wang J. **Peptide backbone-copper ring structure: A molecular insight into copper-induced amyloid toxicity**. *Chin. Phys. B.* (2022.0) **31** 108702. DOI: 10.1088/1674-1056/ac8920
158. Al-Hilaly YK. **A central role for dityrosine crosslinking of Amyloid-β in Alzheimer's disease**. *Acta Neuropathol. Commun.* (2013.0) **1** 83. DOI: 10.1186/2051-5960-1-83
159. Kok WM. **Synthetic dityrosine-linked β-amyloid dimers form stable, soluble, neurotoxic oligomers**. *Chem. Sci.* (2013.0) **4** 4449-4454. DOI: 10.1039/c3sc22295k
|
---
title: The UDPase ENTPD5 regulates ER stress-associated renal injury by mediating
protein N-glycosylation
authors:
- Lifen Xu
- Yuxia Zhou
- Guifang Wang
- Li Bo
- Bangming Jin
- Lujun Dai
- Qinli Lu
- Xueni Cai
- Laying Hu
- Lu Liu
- Yixuan Wu
- Xuebing Chang
- Yali Huang
- Lingyu Song
- Tian Zhang
- Yuanyuan Wang
- Ying Xiao
- Fan Zhang
- Lingling Liu
- Mingjun Shi
- Tuanlao Wang
- Bing Guo
journal: Cell Death & Disease
year: 2023
pmcid: PMC9971188
doi: 10.1038/s41419-023-05685-4
license: CC BY 4.0
---
# The UDPase ENTPD5 regulates ER stress-associated renal injury by mediating protein N-glycosylation
## Abstract
Impaired protein N-glycosylation leads to the endoplasmic reticulum (ER) stress, which triggers adaptive survival or maladaptive apoptosis in renal tubules in diabetic kidney disease (DKD). Therapeutic strategies targeting ER stress are promising for the treatment of DKD. Here, we report a previously unappreciated role played by ENTPD5 in alleviating renal injury by mediating ER stress. We found that ENTPD5 was highly expressed in normal renal tubules; however, ENTPD5 was dynamically expressed in the kidney and closely related to pathological DKD progression in both human patients and mouse models. Overexpression of ENTPD5 relieved ER stress in renal tubular cells, leading to compensatory cell proliferation that resulted in hypertrophy, while ENTPD5 knockdown aggravated ER stress to induce cell apoptosis, leading to renal tubular atrophy and interstitial fibrosis. Mechanistically, ENTPD5-regulated N-glycosylation of proteins in the ER to promote cell proliferation in the early stage of DKD, and continuous hyperglycemia activated the hexosamine biosynthesis pathway (HBP) to increase the level of UDP-GlcNAc, which driving a feedback mechanism that inhibited transcription factor SP1 activity to downregulate ENTPD5 expression in the late stage of DKD. This study was the first to demonstrate that ENTPD5 regulated renal tubule cell numbers through adaptive proliferation or apoptosis in the kidney by modulating the protein N-glycosylation rate in the ER, suggesting that ENTPD5 drives cell fate in response to metabolic stress and is a potential therapeutic target for renal diseases.
## Introduction
Renal hypertrophy is a major morphological change in the early stage of diabetic kidney disease (DKD) and is characterized by expanded and enlarged glomeruli that contain more tubular cells than those in healthy kidneys [1–3]. Although renal hypertrophy is initially a compensatory or adaptive change, it eventually contributes to renal maladaptation, resulting in apoptosis, tubular atrophy and renal fibrosis [4, 5]. Therefore, it is important to examine the mechanism of this alteration by focusing on key components involved in pathological DKD progression and searching for new therapeutic strategies to revitalize renal tubules and increase their integrity.
Glucose is a fuel source for energy metabolism and a regulatory signal indicating protein modifications, such as glycosylation [6, 7]. Aberrant glucose metabolism in DKD may lead to abnormal glycosylation, which drives DKD progression [8, 9]. N-glycosylation (N-acetylglucosamine, GlcNAc) crucially regulates the maturation and quality control of protein synthesis and controls receptor translocation to the plasma membrane, where it promotes cell growth and proliferation [7]. Impaired glycosylation often results in the disruption of protein maturation and incorrect protein folding, inducing the unfolded protein response (UPR) in the endoplasmic reticulum (ER) [10, 11]. UPR-mediated ER stress can trigger adaptive survival responses [12] or cell death [13].
Ectonucleoside triphosphate diphosphohydrolase 5 (ENTPD5), a nucleotide hydrolase located in the ER, hydrolyzes UDP to UMP, is mediated by UGGT, and promotes the correct folding of N-glycoproteins in the ER [14]. ENTPD5 promotes tumor proliferation through ATP consumption and favors aerobic glycolysis [14, 15]. The transcription factor SP1 promotes mutp53 binding to the ENTPD5 promoter, which can accelerate tumor progression and metastasis [16]. In addition, ENTPD5 participates in HRD1-mediated (an ER-associated ubiquitin ligase) regulation of liver metabolism [17]. However, the role of ENTPD5 in the kidney has not been examined.
We found that ENTPD5 was highly expressed in the proximal renal tubules but not in the glomerulus, which prompted us to examine the role of ENTPD5 in the kidney. Further examination revealed that ENTPD5 expression in the kidney was closely related to pathological DKD progression in human patients and mouse models. Herein, we present data demonstrating that ENTPD5 overexpression effectively alleviates kidney failure and in contrast, that ENTPD5 downregulation exacerbates kidney failure. Our results suggest that ENTPD5 prevents renal tubular cells from undergoing adaptive proliferation or apoptosis by modulating protein N-glycosylation in the ER.
## Dynamic expression of ENTPD5 in the kidneys of DKD patients and diabetic mice
ENTPD5, a nucleotide hydrolase located in the ER, hydrolyzes UDP to UMP. In this study, we found that ENTPD5 was expressed in kidney and liver tissues but not in the spleen or myocardial tissues in humans (Fig. 1A), and according to the Genecards database (http://www.genecards.org), ENTPD5 is highly expressed in proximal renal tubules. Considering the specific expression of ENTPD5 in the kidney, we examined the role of ENTPD5 in DKD. To determine whether ENTPD5 is associated with DKD pathophysiological progression, immunohistochemical (IHC) staining showed that ENTPD5 was mainly expressed in proximal renal tubules in renal biopsy samples taken from DKD patients. Surprisingly, this expression showed dynamic alterations; that is, pathological diagnosis revealed that ENTPD5 levels were increased in the biopsy samples from patients with class I DKD but decreased in class II, III, and IV samples (Fig. 1B). Dynamic alterations in transcription were observed via FISH (Fig. 1C). We studied other proteinuric nephropathies and found that the protein and transcriptional expression of ENTPD5 was increased in minimal glomerular lesions (minimal change disease, MCD) but was decreased in sclerosing glomerulonephritis (SGN) (Fig. 1D, E). Notably, no correlation was found between ENTPD5 and urinary albumin or lipid levels (Fig. 1F, G); however, in all patients, ENTPD5 levels were negatively correlated with serum creatinine and were positively correlated with eGFR (Fig. 1H, I), which are key indicators of kidney damage. These observations suggest that ENTPD5 may be related to renal disease. Fig. 1Dynamic expression of ENTPD5 in the kidneys of DKD patients and diabetic mice. A Representative IHC images of ENTPD5 expression in healthy adult kidney, liver, spleen, and heart tissues. Scale bar, black 200 μm. B Representative IHC images and quantification of proximal tubular ENTPD5 expression in renal tissues from DKD patients: Normal human kidney tissue ($$n = 4$$), subjects with mild lesion type (class I, $$n = 3$$), thylakoid hyperplasia type (class II, $$n = 4$$), tuberous sclerosis type (class III, $$n = 8$$), and advanced glomerulosclerosis (class IV, $$n = 7$$). Scale bar, black 200 μm. C Representative fluorescence images of ENTPD5 mRNA levels detected by FISH assay in the kidneys of normal human and different histological grades of DKD patients. Scale bar, white 50 μm. D Representative IHC images and quantification of ENTPD5 expression in renal tissues from patients with MCD and SGN, normal human kidney tissue ($$n = 4$$), MCD ($$n = 5$$), SGN ($$n = 6$$). Scale bar, black 200 μm. E Representative fluorescence images of ENTPD5 mRNA levels detected by FISH assay in the kidneys of normal human, MCD, and SGN patients. Scale bar: white 50 μm. F–I Correlation of ENTPD5 expression in renal tubules with 24-h urine protein quantification ($$n = 33$$) (F), serum triglycerides ($$n = 33$$) (G), serum creatinine (SCr, $$n = 33$$) (H), and eGFR ($$n = 33$$) (I), respectively. J Representative western blot and quantification of ENTPD5 expression in the kidneys of db/db mice at 16, 24, 32, and 40 weeks ($$n = 4$$ per group). K, M Representative western blot (WT, $$n = 3$$ and db/db, $$n = 4$$ blots) and quantification of ENTPD5 expression in the kidneys of db/db mice at 16 (K) and 40 weeks (M). L, N Relative mRNA levels of ENTPD5 in the kidneys of db/db mice at 16 weeks (L) and 40 weeks (N) ($$n = 5$$ per group). O, P Representative IHC images and quantification of ENTPD5 in renal tissue of db/db mice at 16 weeks (O) and 40 weeks (P) ($$n = 5$$ per group). Scale bar, blue 100 μm. Q Functional enrichment analysis of ENTPD5 expression using KEGG pathway based on LC-MS/MS data in the kidneys of 16-week-old db/db mice. R, S Representative western blot and quantification of ENTPD5 expression in RTECs exposed to glucose (15 and 30 mmol/L) (R) and palmitic acid (PA, 0.1 and 0.2 mmol/L) (S) ($$n = 3$$ blots). Data are mean ± SD. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
Similarly, ENTPD5 protein levels were significantly higher in the kidneys of db/db diabetic mice at 16 weeks, 24 weeks, and 32 weeks and were significantly lower at 40 weeks than in the wild-type group (Fig. 1J). The mRNA and protein levels of ENTPD5 were higher in the kidney tissues of 16-week-old db/db mice (Fig. 1K, L) and lower in 40-week-old db/db mice (Fig. 1M, N); similar results were obtained by IHC staining of the kidney tissues from db/db mice (Fig. 1O, P). Liquid chromatography with tandem mass spectrometry (LC‒MS/MS) protein profiling of kidney tissue from 16-week-old db/db diabetic mice revealed that ENTPD5 was enriched in multiple pathways (Fig. 1Q), indicating that ENTPD5 may play an important role in regulating metabolic events.
Since ENTPD5 was mainly expressed in proximal renal tubules, the expression of ENTPD5 was further examined in renal tubule epithelial cells (RTECs) under pathophysiological conditions. In vitro, ENTPD5 protein levels were significantly increased in RTECs exposed to low concentrations of glucose or palmitic acid (PA) (15 mM glucose and 0.1 mM PA), while ENTPD5 expression was decreased in RTECs exposed to high concentrations of glucose (HG) or PA (30 mM glucose and 0.2 mM PA) compared with those in the control group after long-term culture (Fig. 1R, S). These results were consistent with those showing dynamic ENTPD5 expression in DKD; ENTPD5 levels were increased in the early stage and then decreased with persistent hyperglycemia and hyperlipidemia in the late stage, suggesting that high ENTPD5 expression may alleviate kidney injury and that the downregulation of ENTPD5 expression is closely associated with the pathophysiological progression of DKD.
## Decreased ENTPD5 in RTECs exacerbates renal injury in diabetic mice
Considering the dynamic expression of ENTPD5 in the DKD kidney, we altered ENTPD5 expression in the early stage of DKD, since ENTPD5 was upregulated in this stage. *We* generated an ENTPD5-specific knockdown mouse model through multipoint injection of adeno-associated virus expressing short hairpin ENTPD5 (AAV-sh-ENTPD5) into the renal cortex in db/db mice subjected to B-mode ultrasound (Fig. 2A). Intense red fluorescence was observed in the renal cortex of the kidney (Fig. 2B), and the mRNA and protein levels of ENTPD5 were significantly reduced compared with those in the control group, as determined by qPCR, western blot analysis and IHC staining (Fig. 2C–E), indicating that AAV-sh-ENTPD5 was specifically delivered to the renal cortex and sufficiently inhibited ENTPD5 expression. The kidneys were smaller in ENTPD5-knockdown mice than in control mice (Fig. 2F). The serum levels of triglycerides and creatinine were increased in ENTPD5-knockdown db/db mice (Fig. 2G, H). Masson’s staining and periodic acid Schiff (PAS) staining revealed exacerbated renal tubular damage, as indicated by multifocal atrophy in renal tubules, multifocal fibrosis in the renal interstitium, and chronic infiltration of inflammatory cells in ENTPD5-knockdown db/db mice (Fig. 2I, J). The minimum shrinkage area of the renal tissue was ~$13.59\%$, and the maximum area was ~$25.52\%$ compared with those in the control group (Fig. 2K), as calculated by PAS staining. Fig. 2Decreased ENTPD5 in RTECs exacerbates renal injury in diabetic mice. A Beginning at 20 weeks of age, male db/db mice were followed multipoint injection in situ of ENTPD5 knockdown adeno-associated virus (AAV-Sh-ENTPD5) or AAV-vector by using ultrasound and continue feeding to 28 weeks to sacrifice. B Representative fluorescent images of adeno-associated in the kidney of db/db mice after injection AAV-Sh-ENTPD5. Scale bar, blue 100 μm. C, D Relative mRNA expression levels ($$n = 3$$ per group) and representative western blot and quantification ($$n = 3$$ blots) of ENTPD5 in the kidney of db/db mice with AAV-Sh-ENTPD5 virus. E Representative IHC images and quantification of ENTPD5 in renal tissue of db/db mice with AAV-Sh-ENTPD5 virus ($$n = 5$$ per group). Scale bar, blue 100 μm. F Kidney morphology of db/db mice injected with AAV-Sh-ENTPD5 virus at 28 weeks ($$n = 5$$ per group). G, H Serum triglycerides (G) and creatinine (H) in db/db mice (WT, $$n = 5$$; db/db, $$n = 6$$) injected with AAV-Sh-ENTPD5 virus at 28 weeks. I, J Representative PAS (I) and Masson staining (J) images in the kidneys of db/db mice injected with AAV-Sh-ENTPD5 virus ($$n = 3$$ per group). Scale bar, blue 100 μm. K *Atrophic area* of renal tubules in the kidneys of db/db mice injected with AAV-Sh-ENTPD5 virus, calculated by PAS staining. L Representative western blot and quantification of EMT- and ECM-related proteins (E-cadherin, Collagen III, FN, and Vimentin) expression in the kidneys of ENTPD5 knockdown db/db mice ($$n = 3$$ blots). M Representative transmission electron microscopy images of kidney from ENTPD5 knockdown db/db mice ($$n = 6$$ images). Scale bar, red 100 nm. N Representative TUNEL staining images and quantification of apoptosis in the kidney of ENTPD5 knockdown db/db mice ($$n = 3$$ per group). Scale bar, white 50 μm. O Representative western blot and quantification of apoptosis-related protein (BCL-2, Bax and Caspase-3) in the kidneys of ENTPD5 knockdown db/db mice ($$n = 3$$ blots). Data are mean ± SD. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
Furthermore, the levels of epithelial-mesenchymal transition (EMT)- and extracellular matrix (ECM)-related proteins in the kidney tissues of ENTPD5-knockdown db/db mice were increased (Fig. 2L). Notably, the number of mitochondria and endoplasmic reticula in RTECs was significantly reduced in the kidneys of ENTPD5-knockdown db/db mice, as indicated by transmission electron microscopy (TEM) (Fig. 2M). Importantly, TUNEL staining indicated that the apoptosis rates of RTECs were dramatically increased (Fig. 2N), and similar results were determined by Western blot analysis of proteins extracted from the kidneys of ENTPD5-knockdown db/db mice; specifically, the levels of proapoptotic proteins (Bax, cleaved-caspase-3) were increased, and the levels of antiapoptostic proteins (BCL-2) were decreased (Fig. 2O).
These results suggested that ENTPD5 downregulation in the early stage of DKD exacerbated renal tubular damage and induced cell apoptosis, thus accelerating tubular atrophy and renal interstitial fibrosis in the kidney.
## ENTPD5 drives proliferation or apoptosis in RTECs under diabetic conditions
Based on the observation that ENTPD5 was closely related to DKD progression, functional investigations were carried out to examine the manner in which ENTPD5 is involved in renal injury in DKD. Gene expression in RTECs was measured by RNA sequencing (RNA-Seq) and the results revealed 750 upregulated genes and 920 downregulated genes in ENTPD5-overexpressing cells. *These* genes are primarily involved in ER stress and apoptosis signaling pathways (Fig. 3A). Therefore, we hypothesized that ENTPD5 may play an important role in regulating ER stress and apoptosis. Fig. 3ENTPD5 drives proliferation or apoptosis in RTECs under diabetic conditions. A Pathway enrichment analysis in control or ENTPD5 overexpressing RTECs, based on RNA-*Seq data* ($$n = 3$$ per group). B, C Representative western blot and quantification of ER stress-related proteins (p-PERK, p-IRE1, ATF6) and apoptosis-related protein (Caspase12, CHOP, BCL-2, Bax, and Caspase-3) in the kidneys of 16-week-old db/db mice (B) and 40-week-old db/db mice (C) ($$n = 3$$ blots). D, E Representative western blot and quantification of apoptosis-related protein expression (BCL-2, Bax, and Caspase-3) in ENTPD5 knockdown (D) and overexpressing (E) RTECs exposed to HG (30 mmol/L) and PA (0.2 mmol/L) for 48 h, respectively ($$n = 3$$ blots). F, G Annexin V-PE/7-AAD double staining images (F) and apoptosis rate (G) in ENTPD5 overexpressing RTECs exposed to HG (30 mmol/L) or PA (0.2 mmol/L) for 48 h, using flow cytometry assay ($$n = 3$$ per group). H, I Annexin V-PE/7-AAD double staining images (H) and quantification (I) of apoptosis rate in ENTPD5 knockdown RTECs exposed to HG (30 mmol/L) and PA (0.2 mmol/L) for 48 h, respectively ($$n = 3$$ per group). J Representative TUNEL images of kidney tissues from DKD patients and quantification of renal tubular apoptosis rate. Scale bar, white 50 μm. Data are mean ± SD. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
The compensatory proliferation of renal tubular cells has been reported to contribute to hypertrophy, and the loss of compensatory proliferation has been suggested to result in apoptotic atrophy in DKD [4, 5]. We found that the cell proliferation rate was increased and that the number of apoptotic cells was decreased, as determined by TUNEL and hematoxylin and eosin (HE) and Masson staining, in the early stage of 16-week-old db/db mice (Supplementary Fig. 1A, B). We examined pathological changes in glomeruli, and increased glomerular volume, narrowed renal capsule lumen, irregular thickening of the glomerular basement membrane, and fusion of the foot process basement membrane were observed in 16-week-old db/db mice, as indicated by PAS staining and TEM (Supplementary Fig. 1C, D). In addition, ER stress was activated in renal tubular cells (Fig. 3B), as characterized by decreased protein levels of the apoptosis-regulated protein C/EBP homologous protein (CHOP) (Fig. 3B). While ER stress and apoptosis pathways were activated in the kidneys of 40-week-old db/db mice in end-stage DKD (Fig. 3C), and the cell proliferation rate decreased while the number of apoptotic cells increased (Supplementary Fig. 1E); in addition, the kidney exhibited tubular atrophy and interstitial fibrous tissue hyperplasia (Supplementary Fig. 1F). However, PAS staining and TEM analysis showed heavy hyperplasia in focal segments, marked widening of the thylakoid zone and thickening of the basement membrane, and marked fusion of podocyte peduncles in db/db mice at 40 weeks compared to WT mice (Supplementary Fig. 1G, H). TEM showed that the numbers of mitochondria and endoplasmic reticula in RTECs were significantly reduced in the kidneys of 40-week-old db/db mice (Supplementary Fig. 1I), while the numbers of mitochondria and endoplasmic reticula were increased in the kidneys of 16-week-old db/db mice (Supplementary Fig. 1J). In vitro, low concentrations of glucose (15 mmol/L) and PA (0.1 mmol/L) promoted proliferation and inhibited apoptosis in RTECs, while HG (30 mmol/L) and PA (0.2 mmol/L) led to the opposite effects (Supplementary Fig. 1K–M), suggesting that compensatory proliferation and decompensatory apoptosis in RTECs contributed to DKD progression.
Combining the dynamic expression pattern of ENTPD5 in DKD with the RNA-Seq data, we sought to determine whether ENTPD5 was involved in the process that begins with hypertrophy and progresses to RTEC apoptosis in DKD. *We* generated stable cell lines with ENTPD5 overexpression and knockdown via lentivirus. The efficiency of lentiviral-induced ENTPD5 overexpression and knockdown were significant (Supplementary Fig. 1N). Western blot analysis showed that the apoptosis pathway was significantly activated (as indicated by the increased Bax and cleaved caspase-3 levels and reduced BCL-2 levels) in ENTPD5-knockdown RTECs exposed to HG or high concentrations of PA (Fig. 3D). In contrast, ENTPD5 overexpression inhibited apoptosis pathway activation (Fig. 3E). Cell apoptosis was further analyzed by flow cytometry, and the results showed that fewer RTECs overexpressing ENTPD5 underwent apoptosis, and a higher number of ENTPD5-knockdown cells underwent apoptosis (Fig. 3F–I), suggesting that ENTPD5 participates in apoptosis in DKD. Furthermore, we found that the apoptosis rate of RTECs was decreased in class I DKD patients, while the apoptosis rate was increased in class II, III, and IV DKD patients, which negatively correlated with ENTPD5 expression in the kidneys of DKD patients (Fig. 3J).
Overall, ENTPD5 may regulate RTEC proliferation in a compensatory manner to adapt to the metabolic environment in the early stage of DKD. As the duration and size of the lesions increase, however, a reduction in ENTPD5 levels leads to increased RTEC apoptosis, causing renal tubular injury, tubular atrophy, and interstitial fibrosis in the kidney.
## ENTPD5 regulates ER stress-mediated cell proliferation and apoptosis through protein N-glycosylation
Further analysis of the RNA-*Seq data* of RTECs overexpressing ENTPD5 revealed that CHOP, an ER stress-specific transcription factor, transcriptionally regulates death receptor 5 (DR5) expression, thereby activating exogenous apoptotic pathways to mediate apoptosis [18, 19]. Western blot analysis showed that the protein levels of CHOP and DR5 were reduced in ENTPD5-overexpressing RTECs, generally in response to HG- or PA-induced CHOP and DR5 expression, indicating that ENTPD5 was negatively correlated with CHOP and DR5 expression (Fig. 4A), while CHOP and DR5 levels were elevated in ENTPD5-knockdown RTECs in the presence or absence of HG or PA (Fig. 4B).Fig. 4ENTPD5 regulates ER stress-mediated cell proliferation and apoptosis through protein N-glycosylation. A Representative western blot and quantification of protein expression (N-GlcNAc, CHOP, DR5, and EGFR) in ENTPD5 overexpressing RTECs exposed to HG (30 mmol/L) and PA (0.2 mmol/L) for 48 h ($$n = 3$$ blots). B Representative western blot and quantification of protein expression (N-GlcNAc, CHOP, DR5, and EGFR) in ENTPD5 knockdown RTECs exposed to HG (30 mmol/L) and PA (0.2 mmol/L) for 48 h ($$n = 3$$ blots). C Representative western blot and quantification of protein expression (N-GlcNAc, CHOP, DR5, and EGFR) in the kidneys of ENTPD5 knockdown db/db mice ($$n = 3$$ blots). D Representative western blot and quantification of protein expression (ENTPD5, N-GlcNAc, DR5, CHOP, and EGFR) in RTECs exposed to NG (5.5 mmol/L) and HG (30 mmol/L) plus N-glycosylated substrate UDP-GlcNAc (10 mmol/L) for 48 h ($$n = 3$$ blots). Data are mean ± SD. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
Proteins modified by N-GlcNAc play key roles in the maturation and quality control of proteins, and N-glycosylation deficiency induces UPR and cell apoptosis [6]. In addition, N-GlcNAc is critical for growth factor receptor translocation to the plasma membrane [7]. To clarify the specific mechanism by which ENTPD5 regulates ER stress, we performed a lectin-binding assay using concanavalin A (Con A), which specifically detects proteins with the N-GlcNAc modification by binding to Asn-linked glycans in the candidate protein. HG and PA treatment reduced the levels of proteins modified with N-GlcNAc in RTECs. In contrast, ENTPD5 significantly increased the level of protein glycosylation (Fig. 4A). In addition, the level of EGFR, a representative cell surface receptor that regulates cell proliferation, was significantly increased (Fig. 4A). Conversely, ENTPD5 knockdown inhibited protein glycosylation and reduced EGFR expression levels in RTECs (Fig. 4B). Importantly, the abundance of protein glycosylation and EGFR expression levels were greatly reduced in the kidneys of ENTPD5-knockdown db/db mice in vivo, but CHOP and DR5 expression levels were significantly increased (Fig. 4C).
Moreover, we used UDP-GlcNAc, an essential substrate of N-GlcNAc, to treat RTECs cultured with normal glucose (NG, 5.5 mmol/L) and HG (30 mmol/L) and found that the expression levels of ENTPD5, N-GlcNAc, and EGFR were decreased, while the expression of CHOP and DR5 was increased. When RTECs were treated with UDP-GlcNAc plus HG, these outcomes were amplified (Fig. 4D). These results indicated that high concentrations of UDP-GlcNAc, particularly in a HG environment, inhibited protein glycosylation and activated ER stress.
Because protein N-GlcNAc is tightly regulated by the amount of UDP-glucose transported to the ER, UMP/CMP kinase1 requires UMP to produce the UDP required for UDP-glucose generation. However, UMP production is mediated by ENTPD5 primarily through UDP hydrolyzation. As mentioned previously, even in the presence of a high concentration of UDP-GlcNAc, the level of proteins with the N-GlcNAc modification was inhibited (Fig. 4D), which was probably due to a decrease in UDP-glucose transported to the ER. Specifically, UDP-glucose transport to the ER is facilitated by an antiporter in conjunction with UMP export from the ER lumen; however, decreased ENTPD5 levels result in insufficient UMP, which inhibits protein glycosylation. Taken together, these data suggested that ENTPD5 strictly regulates the levels of proteins with the N-GlcNAc modification, triggering CHOP-induced apoptosis and EGFR-induced proliferation of RTECs in the kidney.
## SP1 regulates the expression of ENTPD5 under diabetic conditions
To clarify the mechanism by which ENTPD5 is dynamically expressed in RTECs in DKD, we focused on the transcription factor SP1, which regulates the transcription of ENTPD5. IHC staining showed that SP1 was distributed in the cytoplasm and nucleus in RTECs, and interestingly, increased SP1 expression in the cytoplasm and nucleus was observed in the kidneys of 16-week-old db/db mice (Fig. 5A). In contrast, decreased SP1 expression was observed in the kidneys of 40-week-old db/db mice (Fig. 5B). More importantly, an increase in the SP1 level was observed in the cytoplasm and nucleus of renal tubules in class I DKD patients (Fig. 5C). However, with the progression of the lesion, the expression of SP1 in the cytoplasm and nucleus gradually decreased in class II, III and IV DKD patients (Fig. 5C). In addition, the expression of SP1 was increased in RTECs cultured with lower concentrations of glucose or PA but was decreased in RTECs cultured with HG or high concentrations of PA (Supplementary Fig. 2A). Thus, the dynamic expression of SP1 in the kidney tissue of DKD paralleled the expression pattern of ENTPD5; that is, the levels of both proteins first increased and then gradually decreased. Moreover, SP1 knockdown decreased the mRNA and protein levels of ENTPD5 (Fig. 5D, E) while SP1 overexpression increased ENTPD5 mRNA and protein levels (Fig. 5F, G). Chromatin immunoprecipitation (ChIP)-qPCR analysis clearly demonstrated that SP1 bound to the promoter region of the ENTPD5 gene (Fig. 5H) and this binding was confirmed with a dual-luciferase reporter assay system (Fig. 5I), suggesting that SP1 directly regulates the transcription of ENTPD5.Fig. 5SP1 regulates the expression of ENTPD5 under diabetic conditions. A, B Representative IHC images and quantification of SP1 expression in the kidneys of 16-week-old db/db mice (A) and 40-week-old db/db mice (B) ($$n = 5$$ per group). Scale bar, blue 100 μm. C Representative IHC images and quantification of proximal tubular SP1 expression in the kidney of DKD patients (grading and the number of cases as in Fig. 1B). Scale bar, blue 100 μm. D, E Relative mRNA levels (D) and representative western blot ($$n = 3$$ blots) and quantification of ENTPD5 expression (E) in SP1 knockdown RTECs. F, G Relative mRNA levels (F) and representative western blot ($$n = 3$$ blots) and quantification of ENTPD5 expression (G) in SP1 overexpressing RTECs. H, I ChIP assay (H) and dual-luciferase assay (I) to verify the regulation of transcript levels of ENTPD5 by SP1 ($$n = 3$$ per group); normal rabbit IgG as a negative control antibody and rabbit Histone H3 as a positive antibody. J, K Representative western blot and quantification of protein expression (GFAT, O-GlcNAc) in the kidneys of 16-week-old db/db mice (J) and 40-week-old db/db mice (K) ($$n = 3$$ blots). L Representative western blot and quantification of protein expression (GFAT, O-GlcNAc expression in RTECs exposed to HG (15 and 30 mmol/L) and PA (0.1 and 0.2 mmol/L) ($$n = 3$$ blots). M Representative western blot and quantification of protein expression (GFAT, O-GlcNAc and SP1) in RTECs exposed to HG (30 mmol/L) plus UDP-GlcNAc (10 mmol/L) for 48 h. ($$n = 3$$ blots). Data are mean ± SD. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
Next, we examined how SP1 dynamically regulates the expression of ENTPD5. SP1 activity is mediated via O-glycosylation (with O-GlcNAc), which determines the nuclear translocation and stability of SP1 [20, 21]. Low levels of O-glycosylated SP1 are preferentially degraded via the proteasome to inhibit transcriptional activity. Hyperglycemia has been shown to increase the degree of SP1 modification with O-GlcNAc through the hexosamine biosynthesis pathway (HBP) [22–24].
Consistent with this finding, the levels of O-GlcNAc and glutamine-fructose-6-phosphate aminotransferase (GFAT), which is the rate-limiting enzyme in the HBP [25, 26], were increased in the kidneys of 16-week-old db/db mice (Fig. 5J) and decreased in db/db mice at 40 weeks (Fig. 5K). Similarly, O-GlcNAc abundance and GFAT expression were increased in RTECs cultured with low concentrations of glucose or PA, while high concentrations of HG or PA led to the opposite outcomes (Fig. 5L). However, UDP-GlcNAc, an end-product of HBP, negatively regulates GFAT expression [25]. Our experiments revealed that a low concentration of UDP-GlcNAc exerted no inhibitory effect on GFAT expression, while a high concentration of UDP-GlcNAc significantly inhibited GFAT expression (Supplementary Fig. 2B). Importantly, the level of O-GlcNAc and the expression levels of GFAT and SP1 were decreased in RTECs cultured with HG or NG plus high concentrations of UDP-GlcNAc, and this effect was more obvious in RTECs exposed to UDP-GlcNAc plus HG (Fig. 5M).
Taken together, these results suggest that ENTPD5 expression is regulated by O-GlcNAc modification of SP1. Hyperglycemia promotes glucose use in the HBP, thus increasing the level of the end-product UDP-GlcNAc via the rate-limiting enzyme GFAT and increasing the rate of SP1 glycosylation with O-GlcNAc to promote SP1-induced transcription of ENTPD5. In hyperglycemia and DKD conditions, the continual increase in UDP-GlcNAc inhibits GFAT expression via a negative feedback mechanism, resulting in a decrease in UDP-GlcNAc levels, inhibiting SP1 modification with O-GlcNAc and, therefore, downregulating ENTPD5 expression.
## ENTPD5 negatively regulates renal injury in UUO mice
To examine whether ENTPD5 is involved in renal injury in other kidney diseases, we used a murine model of unilateral ureteral obstruction (UUO)-induced nephropathy, which recapitulates human SGN with typical pathological changes in renal interstitial fibrosis in the end stage of chronic kidney disease. UUO mice were administered AAV-ENTPD5 and AAV-sh-ENTPD5 separately via renal cortex multipoint injection to upregulate and downregulate ENTPD5 expression, respectively (Fig. 6A). The results showed that the protein levels of ENTPD5 were significantly decreased in the kidneys of UUO mice compared with those in the control group after ENTPD5 knockdown 3 days or 7 days after UUO surgery (Fig. 6B and Supplementary Fig. 3A) and were significantly increased in UUO mice with ENTPD5 overexpression (Fig. 6C and Supplementary Fig. 3B).Fig. 6ENTPD5 negatively regulates renal injury in UUO mice. A Beginning at 8 weeks of age, the renal cortex of male C57BL mice was followed multipoint injection in situ of ENTPD5 knockdown or ENTPD5 overexpressing adeno-associated virus (AAV-Sh-ENTPD5, AAV-ENTPD5) or AAV-vector by using ultrasound. To operate UUO surgery at 16 weeks after the injection and the mice were executed at 3 and 7 days postoperatively. B, C Representative western blot and quantification of ENTPD5 expression in the kidneys of UUO mice with ENTPD5 knockdown (B) or overexpression (C) at 3 days postoperatively ($$n = 3$$ blots). D Kidney morphology of UUO mice at 3 days postoperatively after injection of AAV-Sh-ENTPD5 and AAV-ENTPD5 virus. E, F The level of serum creatinine (E) and blood urea nitrogen (BUN) (F) from UUO mice with ENTPD5 knockdown or overexpression at 3 and 7 days postoperatively (vector, $$n = 3$$ and AAV-ENTPD5 or AAV-sh-ENTPD5, $$n = 4$$, respectively). G, H Representative IHC images of ENTPD5 expression, PAS and Masson staining of kidney tissue from UUO mice with ENTPD5 knockdown (G) or overexpression (H) at 3 days postoperatively (vector, $$n = 3$$ and AAV-ENTPD5 or AAV-sh-ENTPD5, $$n = 4$$, respectively). Scale bar: green 500 μm, blue 100 μm and white 50 μm. I, J Representative western blot and quantification of EMT- and ECM-related proteins (E-cadherin, Collagen III, FN and Vimentin) expression in the kidney of UUO mice with ENTPD5 knockdown (I) or overexpression (J) at 3 days postoperatively ($$n = 3$$ blots). K, L Representative IHC images of ENTPD5 expression, PAS and Masson staining of kidney tissue from UUO mice with ENTPD5 knockdown (K) or overexpression (L) at 7 days postoperatively (vector, $$n = 3$$ and AAV-ENTPD5 or AAV-sh-ENTPD5, $$n = 4$$, respectively). Scale bar: green 500 μm, blue 100 μm, and white 50 μm. M, N Representative western blot and quantification of EMT- and ECM-related proteins (E-cadherin, Collagen III, FN, and Vimentin) expression in the kidney of UUO mice with ENTPD5 knockdown (M) or overexpression (N) at 3 days postoperatively ($$n = 3$$ blots).
Notably, kidney size in mice with ENTPD5 knockdown was smaller than that in the control group but was increased in ENTPD5-overexpressing mice 3 days after the UUO operation (Fig. 6D), and similar results were observed in mice 7 days after the operation (data not shown). Serum creatinine and blood urea nitrogen levels were significantly increased in ENTPD5-knockdown UUO mice but were significantly reduced in ENTPD5-overexpressing UUO mice on the 3rd and 7th days after the UUO operation (Fig. 6E, F). PAS and Masson’s staining revealed great improvements in renal morphology and reduced renal interstitial fibrosis after ENTPD5 was overexpressed (Fig. 6H), while pathological renal morphology and increased renal interstitial fibrosis were observed in the kidneys of ENTPD5-knockdown UUO mice 3 days after the UUO operation (Fig. 6G). Similar results were observed in mice 7 days after the operation (Fig. 6K, L). In addition, the levels of EMT- and ECM-related proteins in the kidney tissues of the UUO mice expressing AAV-ENTPD5 were reduced (Fig. 6J, N), while these protein levels were increased in UUO mice with ENTPD5 knockdown on the 3rd and 7th days after the UUO operation (Fig. 6I, M). These results suggested that ENTPD5 plays an important role in renal fibrosis in chronic kidney disease.
In addition, the TUNEL assay showed an increase in the number of apoptotic renal tubular cells in UUO mouse kidneys after ENTPD5 knockdown on the 3rd and 7th days after surgery compared to that in the control group at the same time points (Supplementary Fig. 3C, D). The apoptosis rates of renal tubular cells on the 3rd and 7th days after surgery were attenuated in UUO mice overexpressing ENTPD5 (Supplementary Fig. 3E, F). Western blot analysis demonstrated that proapoptotic protein levels were significantly decreased and antiapoptotic protein levels were increased in the kidneys of UUO mice after ENTPD5 was overexpressed (Supplementary Fig. 3G, I), and the opposite outcomes were observed in UUO mice with ENTPD5 knockdown on the 3rd and 7th days after surgery (Supplementary Fig. 3H, J). These results suggest that ENTPD5 alleviated renal injury in UUO mice and may be a potential therapeutic target to protect RTECs against injury.
## Discussion
Renal tubules exhibited hypertrophy in the early stage of DKD in mice, and the numbers of proximal and distal tubules were increased in the kidneys of mice treated with streptozotocin (STZ) to induce diabetes at 13 weeks [5], as well as in 12-week-old db/db mice [27, 28]. In our study, we also observed that the numbers of proximal and distal tubules were increased in the kidneys of 16-week-old db/db mice, indicating that renal tubule hypertrophy and hyperplasia contributed to the increased kidney size in DKD, which is considered an early pathological change in DKD. With continuous hyperglycemia, renal tubule lengths were initially increased with tubule lumen enlargement, and eventually, renal tubule cells underwent apoptosis [29, 30], which ultimately led to tubular atrophy and interstitial fibrosis; however, the observed phenomena might arise from secondary effects of the Lepr mutation in db/db mice. However, the mechanisms by which renal tubule pathology progresses from tissue hypertrophy to cell apoptosis and then to atrophy in diabetes mellitus are still unclear.
ENTPD5 is the only identified intracellular ENTPDase [31]. In this study, we confirmed the role of ENTPD5 in mediating ER stress to regulate adaptation and induce damage to renal tubules in DKD. Specifically, ENTPD5 was differentially expressed in the kidneys of diabetic mice, as determined by LC‒MS analysis. Functional experiments demonstrated that ENTPD5 was mainly expressed in the renal tubules of the kidney and that the levels of ENTPD5 were altered under pathological conditions, initially increasing and then decreasing in the late stage of DKD in diabetic mice and patients. Therefore, ENTPD5 can be used as a diagnostic marker to determine the pathological stage of DKD in the clinic. Importantly, ENTPD5 downregulation in RTECs significantly exacerbated kidney injury, inhibiting RTEC proliferation and promoting apoptosis in diabetic mice and UUO mice, while ENTPD5 overexpression attenuated kidney injury, suggesting that ENTPD5 is pivotal in preventing renal tubule pathology from progressing from tissue hypertrophy and cell apoptosis to atrophy and that it plays an important role in protecting renal tubules from injury.
Through mechanistic investigations, we found that ENTPD5 upregulation promoted RTEC proliferation and inhibited RTEC apoptosis, while ENTPD5 downregulation led to opposite results. In addition, we found that ENTPD5 expression was transcriptionally regulated by the transcription factor SP1 after SP1 O-glycosylation, which prevented SP1 from undergoing proteasomal degradation [24]. Glucose is metabolized through the HBP, and only 1–$3\%$ of intracellular glucose is converted to glucosamine 6-phosphate via the HBP. Notably, glucose into the HBP has been previously shown to be significantly increased with a continuous supply of intracellular glucose [32] or when glucose/lipid metabolism is dysregulated. Hyperactivity of the HBP, which is regulated by the GFAT rate-limiting enzyme, leads to an increase in the final product UDP-GlcNAc, which is the substrate for protein glycosylation with N-GlcNAc and O-GlcNAc. We confirmed that excessive UDP-GlcNAc in RTECs exerts an inhibitory effect on GFAT, which decreases the level of O-GlcNAc-modified SP1, as mediated via a feedback mechanism, and subsequently reduces ENTPD5 transcription. As previously reported, reduced ENTPD5 expression leads to reduced UDP hydrolyzation into UMP [26, 33].
ENTPD5 hydrolyzes UDP to UMP by UGGT, which relieves the inhibitory effect of UDP on protein modification with N-GlcNAc in the ER and enables greater antiporter-mediated influx of the glucose carrier UDP-glucose, which is necessary for protein modification with N-GlcNAc into the ER [34]. Moreover, the expression of the growth factor receptor EGFR is regulated by ENTPD5-mediated protein N-glycosylation and promotes cell proliferation [16, 35–37]. This finding suggested that the increase in ENTPD5 expression participated in RTEC proliferation and renal hypertrophy in the early stage of DKD. Moreover, reducing ENTPD5 levels reduced the pool of UMP-glucose available for antiporter influx of UDP-glucose into the ER, limiting the amount of substrate available for protein N-glycosylation, which led to the accumulation of unfolded or misfolded proteins in the ER and ultimately initiated the ER stress-associated apoptosis pathway. Thus, the decrease in ENTPD5 expression participated in RTEC apoptosis in the end stage of DKD.
In summary, our study first showed that ENTPD5 was important for the regulation of ER stress in RTECs from early renal hypertrophy to late apoptotic atrophy. Mechanistically, as summarized in Fig. 7, in DKD, hyperglycemia activates the HBP to promote or inhibit SP1 O-glycosylation via a negative feedback mechanism, thus regulating ENTPD5 expression at the transcriptional level. ENTPD5 regulates unfolded protein N-glycosylation in the ER to promote cell proliferation or apoptosis. Because ENTPD5 is pathophysiologically related to renal tubule injury, this work provides a new therapeutic strategy to mediate ENTPD5 expression to protect the kidney against injury in DKD or other forms of chronic kidney disease. The study also suggests that ENTPD5 may be a diagnostic marker of progressive DKD. It also explains, in part, the possible mechanism underlying the pathological changes in RTEC in DKD.Fig. 7Mechanism diagram of ENTPD5 regulating protein N-glycosylation in DKD.In DKD, hyperglycemia activates the hexosamine biosynthesis pathway (HBP) to promote or inhibit SP1 O-glycosylation via a negative feedback mechanism, thus regulating ENTPD5 expression at the transcriptional level. ENTPD5 regulates unfolded protein N-glycosylation in the ER to promote cell proliferation or apoptosis of renal tubular epithelial cell.
## Human renal biopsy samples
Human kidney biopsies were collected as part of routine clinical diagnostic and are shown in Supplementary Tables 1 and 2. Kidney tissue samples were obtained from the Department of Pathology, Affiliated Hospital of Guizhou Medical University. Normal control samples were obtained from healthy renal poles from patients who had undergone tumor nephrectomy without diabetes or renal disease. The investigations were conducted in accordance with the principles of the Declaration of Helsinki and were approved by the Research Ethics Committee of Guizhou Medical University (Document No.2020209). All renal biopsy specimens diagnosed with DKD were classified according to the new pathological classification provided by the Society of Renal Pathology [38].
## Animal studies
All animal experimental protocols were approved by the Ethics Committee of Guizhou Medical University (Document No.2200782) and carried out in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All mice (3–5 per cage) were randomly grouped and housed under standard laboratory conditions (12 h on/off; lights on at 9 am; temperature (24 °C)) with free access to water and diets.
## Establishment of db/db mice model
Homozygote BKS db/db male mice were purchased from GemPharmatech (Guangzhou, China). WT mice were used as genetic control and 10-week-old db/db mice were continued to be maintained on normal diets for another 6 weeks, 14 weeks, 22 weeks, and 30 weeks, respectively. At the end of the study, blood samples were collected for biochemical analysis and the kidney samples were harvested for histopathological analysis. The renal cortex was dissected for protein or RNA extraction for subsequent analysis. The physical and biochemical parameters of experimental animals are shown in Supplementary Table 3.
## UUO mice
After the 12-week-old C57BL/6 male mice (GuangDong medical laboratory animal center, China) were anesthetized and fixed, an operation of unilateral ureteral obstruction (UUO) was performed. In brief, the right side of the abdomen was opened in the middle position, and the ureter was searched along the renal hilum. The ureter was separated from the surrounding fatty tissue at the renal hilum, and the proximal and distal end of the ureter were ligated with silk threads and cut from the middle with scissors, and finally sutured.
## Generation of ENTPD5 knockdown and overexpression mice
*To* generate ENTPD5-specific knockdown mice model, 20-week-old male db/db male mice and 8-week-old C57BL/6 male mice were injected adeno-associated virus expressing short hairpin ENTPD5 (AAV-sh-ENTPD5) through renal cortex multipoint injection subjected to B-mode ultrasound with the help of ultrasound doctors. For ENTPD5-specific overexpression mice model, 8-week-old C57BL/6 male mice were injected AAV8-ENTPD5 with the above operation. 100 μL of each AAV (1 × 1011 pfu mL−1) was aspirated with a 29 G insulin syringe (Becton Dickinson, USA). After injection AAV, 20-week-old db/db mice were fed on normal chow diets until to 28 weeks (Supplementary Table 4). Eight-week-old C57BL were fed on normal chow diets to 16 weeks and then performed the UUO surgery.
## Cell culture and handling
Mouse proximal tubule epithelial cells RTEC were obtained from ATCC and cultured in DMEM (1.0 g/L glucose) containing $10\%$ FBS and penicillin/streptomycin. All cells requiring intervention were synchronously quiescent after 6 h in a serum-free medium and then treated with different stimuli as follows: high glucose (HG, 15 and 30 mmol/L D-glucose in the medium, 5.5 mmol/L glucose as a control), palmitic acid (PA, 0.1 and 0.2 mmol/L) or UDP-GlcNAc (10 mmol/L). Tool cells HEK293T cells were obtained from ATCC and were cultured in DMEM (4.5 g/L glucose) containing $10\%$ FBS.
## Histological analysis of kidney tissue
Renal biopsy and the mouse kidneys were embedded in paraffin and cross-sectioned (3 μm) for histology examination. Hematoxylin–eosin staining (HE) and immunohistochemistry (IHC) analysis and were performed according to the routine procedures. Periodic acid schiff (PAS) and Masson staining were performed according to manufacturers’ instructions by using their staining kits (Solarbio, China). Slices were photographed with Olympus BX53 microscope (Olympus, Japan) and the staining of positive areas in the renal tubules was quantified with ImageJ software.
## RNA in situ hybridization
The mRNA expression of ENTPD5 in the kidney was detected by the FISH detection kit (GenePharma, China), according to the manufacturer’s instruction. In brief, paraffin slice of kidney tissues was dewaxed, digested, denatured and hybridized. The images were acquired by a FV3000 laser scanning confocal microscopy.
## RNA isolation and real-time qRT-PCR
RNA of tissues or cells was extracted with TRIzol reagent (Invitrogen, USA), and then reverse-transcribed into cDNA using the PrimeScriptRT Master Mix (Yeasen, China) according to the manufacturer’s instructions. qRT-PCR was performed using Hieff UNICON Universal Blue qPCR SYBR Green Master Mix (Yeasen, China). Fold change in gene expression normalized to GAPDH was calculated by the ∆∆CT method using Equation 2-∆∆CT. The results were shown as fold changes compared to the control group. The primers for target genes in this study are shown in Supplementary Table 5.
## Western blot
Tissues or cultured cells were lysed with RIPA buffer containing protease inhibitors, PMSF and phosphatase inhibitors. The tissues were lysed with homogenizer. The tissue or cell lysate were incubated on ice for 40 min with shaking and centrifuged. The soluble supernatant was carefully transferred to fresh EP tubes for protein assay using the BCA protein assay. Proteins were separated by $10\%$ or $15\%$ SDS-PAGE and transferred to PVDF membranes. The selected proteins are detected with antibodies summarized in Supplementary Table 6.
## Detection of glycoprotein with concanavalin A (Con A-HRP)
The protein transferred to the PVDF membrane was blocked with $0.5\%$ Tween 20-PBS for 5 min, then Con A-HRP complex with a final concentration of 5 μg/mL was added, incubated at 4 °C for 16 h, and washed with PBS twice for 5 min each time. The bands were incubated with ECL luminescent solution and exposed by Tanon chemiluminescent imaging system.
## Biochemical analysis of serum samples
Serum creatinine, urea nitrogen, and triglycerides were determined by BS-240VET veterinary biochemical automatic analyzer (Mindray, China).
## Transmission electron microscopy
Electron microscopic sample handling and detection were performed by the affiliated Hospital of Guizhou Medical University. The images were collected and analyzed under transmission electron microscope (H-7500, Japan).
## TUNEL assay
Paraffin-embedded tissue sections of the kidney were stained in situ with a detection kit (Kgi Biotechnology, China) following the manufacturer’s protocols. The apoptosis rate was made by randomly counting TUNEL-positive cells in each renal cortex.
## Flow cytometry
Cell apoptosis was determined with the kit according to the manufacturer’s instructions (BD, USA). The cells were collected and labeled by fluorescein isothiocyanate (FITC)-conjugated Annexin V and propidium iodide (PI) staining.
## Lentivirus-mediated gene interference and overexpression
Gene overexpression was achieved by pHJLV004-CMV-MCS-EF1-ZsGreen-T2A-puro vector and the silencing of genes was conducted by hU6-MCS-CMV-GFP-SV40 vector, both mediated by lentivirus expression system. For virus preparation, 293T cells were transfected with lentiviral skeleton and helper plasmids (pMD2.G and psPAX2) for 72 h, and the medium supernatant was collected and concentrated. The sequences of shRNA oligonucleotides were as follows: mouse ENTPD5 (GATGGGTCCTATGAAGGCATA).
## Knockdown of siRNA mediated
Cells were cultured in medium without antibiotics. Short interfering RNA (siRNA) or control of the target gene was transfected into cells by FAM-siRNA kit (Sangon Biotech, China) according to the manufacturer’s protocol.
## Mass spectrometry analysis
Kidney tissues of 16-week db/db mice and wild-type mice were collected to perform proteomic analysis. A tandem mass spectrometry (MS/MS) analysis was performed at Xiamen University (Fujian, China). Protein was extracted, digested, and labeled using TMT reagent according to the manufacturer’s instructions (Thermo Scientific). The MS raw data were searched using the MASCOT engine (Matrix Science, London, UK; version 2.2) embedded into Proteome Discoverer 1.4 software for identification and quantitation analysis.
## RNA-sequencing analysis
Lysis of RTEC cells overexpressing ENTPD5 or control were collected to perform RNA-sequencing analysis. The library construction and sequencing were performed at Beijing Novogene (Beijing, China).
## Chromatin immunoprecipitation (ChIP) assay
A ChIP assay was performed using a kit (Cell Signaling Technology, USA) according to the manufacturer’s instructions. In brief, DNA-protein complexes were cross‐linked using $1\%$ formaldehyde for 15 min. The cross‐linked chromatin samples were isolated from the cell lysates by nuclease digestion (37 °C for 20 min) and ultrasound ($20\%$ W, over 5 s, stop for 10 s, over 30–40 times in total). SP1 was immunoprecipitated using a SP1 antibody or control antibody (rabbit IgG) and then DNA was extracted. For quantitative PCR, ChIP DNA was amplified using primers of ENTPD5 promoter (Supplementary Table 5) by qPCR SYBR Green Master Mix.
## Dual-luciferase reporter assay
In brief, the ENTPD5 promoter was cloned into pGL3-Basic vector as a luciferase reporter plasmid and Rinilla luciferase (phRL-TK) was used as a reference gene. Luciferase reporter plasmid and phRL-TK vector were transfected into 293T cells for 48 h, then cell lysis was measured by the kit (Promega, USA) according to the manufacturer’s instructions. Values represent the ratio of firefly luciferase reaction intensity to internal reference Renilla luciferase reaction intensity.
## Statistical analysis
Experimental data were presented as mean ± standard deviation (SD) with GraphPad Prism 8.0. Statistical significance between two groups was assessed using Student’s t tests or among multiple groups using two-way ANOVA. $P \leq 0.05$ was considered statistically significant.
## Supplementary information
aj-checklist Supplementary Tables Supplementary Fig.1 Supplementary Fig.2 Supplementary Fig.3 Supplementary Figure legends Original Data File The online version contains supplementary material available at 10.1038/s41419-023-05685-4.
## References
1. Bonventre JV. **Can we target tubular damage to prevent renal function decline in diabetes?**. *Semin Nephrol* (2012) **32** 452-62. DOI: 10.1016/j.semnephrol.2012.07.008
2. Tang SC, Lai KN. **The pathogenic role of the renal proximal tubular cell in diabetic nephropathy**. *Nephrol Dial Transpl* (2012) **27** 3049-56. DOI: 10.1093/ndt/gfs260
3. Vallon V, Thomson SC. **Renal function in diabetic disease models: the tubular system in the pathophysiology of the diabetic kidney**. *Annu Rev Physiol* (2012) **74** 351-75. DOI: 10.1146/annurev-physiol-020911-153333
4. Wolf G, Ziyadeh FN. **Molecular mechanisms of diabetic renal hypertrophy**. *Kidney Int* (1999) **56** 393-405. DOI: 10.1046/j.1523-1755.1999.00590.x
5. Nyengaard JR, Flyvbjerg A, Rasch R. **The impact of renal growth, regression and regrowth in experimental diabetes mellitus on number and size of proximal and distal tubular cells in the rat kidney**. *Diabetologia* (1993) **36** 1126-31. DOI: 10.1007/BF00401056
6. Reily C, Stewart TJ, Renfrow MB, Novak J. **Glycosylation in health and disease**. *Nat Rev Nephrol* (2019) **15** 346-66. DOI: 10.1038/s41581-019-0129-4
7. Inagi R. **Organelle stress and glycation in kidney disease**. *Glycoconj J* (2021) **38** 341-6. DOI: 10.1007/s10719-021-09989-5
8. Bermingham ML, Colombo M, McGurnaghan SJ, Blackbourn LAK, Vuckovic F, Pucic Bakovic M. **N-glycan profile and kidney disease in type 1 diabetes**. *Diabetes Care* (2018) **41** 79-87. DOI: 10.2337/dc17-1042
9. Dennis JW, Nabi IR, Demetriou M. **Metabolism, cell surface organization, and disease**. *Cell* (2009) **139** 1229-41. DOI: 10.1016/j.cell.2009.12.008
10. Amen OM, Sarker SD, Ghildyal R, Arya A. **Endoplasmic reticulum stress activates unfolded protein response signaling and mediates inflammation, obesity, and cardiac dysfunction: therapeutic and molecular approach**. *Front Pharm* (2019) **10** 977-98. DOI: 10.3389/fphar.2019.00977
11. Wang Y, Chen Y, Zhou Q, Xu J, Qian Q, Ni P. **Mild endoplasmic reticulum stress protects against lipopolysaccharide-induced astrocytic activation and blood-brain barrier hyperpermeability**. *Front Cell Neurosci* (2018) **12** 222-35. DOI: 10.3389/fncel.2018.00222
12. Bhattarai KR, Riaz TA, Kim HR, Chae HJ. **The aftermath of the interplay between the endoplasmic reticulum stress response and redox signaling**. *Exp Mol Med* (2021) **53** 151-67. DOI: 10.1038/s12276-021-00560-8
13. Ricciardi CA, Gnudi L. **The endoplasmic reticulum stress and the unfolded protein response in kidney disease: implications for vascular growth factors**. *J Cell Mol Med* (2020) **24** 12910-9. DOI: 10.1111/jcmm.15999
14. Fang M, Shen Z, Huang S, Zhao L, Chen S, Mak TW. **The ER UDPase ENTPD5 promotes protein N-glycosylation, the Warburg effect, and proliferation in the PTEN pathway**. *Cell* (2010) **143** 711-24. DOI: 10.1016/j.cell.2010.10.010
15. Israelsen WJ, Vander Heiden MG. **ATP consumption promotes cancer metabolism**. *Cell* (2010) **143** 669-71. DOI: 10.1016/j.cell.2010.11.010
16. Vogiatzi F, Brandt DT, Schneikert J, Fuchs J, Grikscheit K, Wanzel M. **Mutant p53 promotes tumor progression and metastasis by the endoplasmic reticulum UDPase ENTPD5**. *Proc Natl Acad Sci USA* (2016) **113** E8433-42. DOI: 10.1073/pnas.1612711114
17. Wei J, Yuan Y, Chen L, Xu Y, Zhang Y, Wang Y. **ER-associated ubiquitin ligase HRD1 programs liver metabolism by targeting multiple metabolic enzymes**. *Nat Commun* (2018) **9** 3659-73. DOI: 10.1038/s41467-018-06091-7
18. Chen P, Hu T, Liang Y, Li P, Chen X, Zhang J. **Neddylation inhibition activates the extrinsic apoptosis pathway through ATF4-CHOP-DR5 axis in human esophageal cancer cells**. *Clin Cancer Res* (2016) **22** 4145-57. DOI: 10.1158/1078-0432.CCR-15-2254
19. Lu M, Lawrence DA, Marsters S, Acosta-Alvear D, Kimmig P, Mendez AS. **Opposing unfolded-protein-response signals converge on death receptor 5 to control apoptosis**. *Science* (2014) **345** 98-101. DOI: 10.1126/science.1254312
20. Majumdar G, Harrington A, Hungerford J, Martinez-Hernandez A, Gerling IC, Raghow R. **Insulin dynamically regulates calmodulin gene expression by sequential o-glycosylation and phosphorylation of sp1 and its subcellular compartmentalization in liver cells**. *J Biol Chem* (2006) **281** 3642-50. DOI: 10.1074/jbc.M511223200
21. Du XL, Edelstein D, Rossetti L, Fantus IG, Goldberg H, Ziyadeh F. **Hyperglycemia-induced mitochondrial superoxide overproduction activates the hexosamine pathway and induces plasminogen activator inhibitor-1 expression by increasing Sp1 glycosylation**. *Proc Natl Acad Sci USA* (2000) **97** 12222-6. DOI: 10.1073/pnas.97.22.12222
22. Springhorn C, Matsha TE, Erasmus RT, Essop MF. **Exploring leukocyte O-GlcNAcylation as a novel diagnostic tool for the earlier detection of type 2 diabetes mellitus**. *J Clin Endocrinol Metab* (2012) **97** 4640-9. DOI: 10.1210/jc.2012-2229
23. Su K, Roos MD, Yang X, Han I, Paterson AJ, Kudlow JE. **An N-terminal region of Sp1 targets its proteasome-dependent degradation in vitro**. *J Biol Chem* (1999) **274** 15194-202. DOI: 10.1074/jbc.274.21.15194
24. Lim K, Yoon BH, Ha CH. **O-Linked N-acetylglucosaminylation of Sp1 interferes with Sp1 activation of glycolytic genes**. *Biochem Biophys Res Commun* (2015) **468** 349-53. DOI: 10.1016/j.bbrc.2015.10.096
25. Ruegenberg S, Horn M, Pichlo C, Allmeroth K, Baumann U, Denzel MS. **Loss of GFAT-1 feedback regulation activates the hexosamine pathway that modulates protein homeostasis**. *Nat Commun* (2020) **11** 687-703. DOI: 10.1038/s41467-020-14524-5
26. Ruegenberg S, Mayr F, Atanassov I, Baumann U, Denzel MS. **Protein kinase A controls the hexosamine pathway by tuning the feedback inhibition of GFAT-1**. *Nat Commun* (2021) **12** 2176-90. DOI: 10.1038/s41467-021-22320-y
27. Rong G, Weng W, Huang J, Chen Y, Yu X, Yuan R. **Artemether alleviates diabetic kidney disease by modulating amino acid metabolism**. *BioMed Res Int* (2022) **2022** 7339611. DOI: 10.1155/2022/7339611
28. Maekawa M, Maekawa T, Sasase T, Takagi K, Takeuchi S, Kitamoto M. **Pathophysiological analysis of uninephrectomized db/db mice as a model of severe diabetic kidney disease**. *Physiol Res* (2022) **71** 209-17. DOI: 10.33549/physiolres.934784
29. Habib SL. **Kidney atrophy vs hypertrophy in diabetes: which cells are involved?**. *Cell Cycle* (2018) **17** 1683-7. DOI: 10.1080/15384101.2018.1496744
30. Wolf G, Neilson EG. **Molecular mechanisms of tubulointerstitial hypertrophy and hyperplasia**. *Kidney Int* (1991) **39** 401-20. DOI: 10.1038/ki.1991.52
31. Robson SC, Sevigny J, Zimmermann H. **The ENTPDase family of ectonucleotidases: Structure function relationships and pathophysiological significance**. *Purinergic Signal* (2006) **2** 409-30. DOI: 10.1007/s11302-006-9003-5
32. Hirschberg CB, Robbins PW, Abeijon C. **Transporters of nucleotide sugars, ATP, and nucleotide sulfate in the endoplasmic reticulum and Golgi apparatus**. *Annu Rev Biochem* (1998) **67** 49-69. DOI: 10.1146/annurev.biochem.67.1.49
33. Kroef V, Ruegenberg S, Horn M, Allmeroth K, Ebert L, Bozkus S. **GFPT2/GFAT2 and AMDHD2 act in tandem to control the hexosamine pathway**. *eLife* (2022) **11** 1-29. DOI: 10.7554/eLife.69223
34. Shen Z, Huang S, Fang M, Wang X. **ENTPD5, an endoplasmic reticulum UDPase, alleviates ER stress induced by protein overloading in AKT-activated cancer cells**. *Cold Spring Harb Symp Quant Biol* (2011) **76** 217-23. DOI: 10.1101/sqb.2011.76.010876
35. Xu C, Ng DT. **Glycosylation-directed quality control of protein folding**. *Nat Rev Mol Cell Biol* (2015) **16** 742-52. DOI: 10.1038/nrm4073
36. Hebert DN, Molinari M. **In and out of the ER: protein folding, quality control, degradation, and related human diseases**. *Physiol Rev* (2007) **87** 1377-408. DOI: 10.1152/physrev.00050.2006
37. Lau KS, Partridge EA, Grigorian A, Silvescu CI, Reinhold VN, Demetriou M. **Complex N-glycan number and degree of branching cooperate to regulate cell proliferation and differentiation**. *Cell* (2007) **129** 123-34. DOI: 10.1016/j.cell.2007.01.049
38. Tervaert TW, Mooyaart AL, Amann K, Cohen AH, Cook HT, Drachenberg CB. **Pathologic classification of diabetic nephropathy**. *J Am Soc Nephrol* (2010) **21** 556-63. DOI: 10.1681/ASN.2010010010
|
---
title: mcPGK1-dependent mitochondrial import of PGK1 promotes metabolic reprogramming
and self-renewal of liver TICs
authors:
- Zhenzhen Chen
- Qiankun He
- Tiankun Lu
- Jiayi Wu
- Gaoli Shi
- Luyun He
- Hong Zong
- Benyu Liu
- Pingping Zhu
journal: Nature Communications
year: 2023
pmcid: PMC9971191
doi: 10.1038/s41467-023-36651-5
license: CC BY 4.0
---
# mcPGK1-dependent mitochondrial import of PGK1 promotes metabolic reprogramming and self-renewal of liver TICs
## Abstract
Liver tumour-initiating cells (TICs) contribute to tumour initiation, metastasis, progression and drug resistance. Metabolic reprogramming is a cancer hallmark and plays vital roles in liver tumorigenesis. However, the role of metabolic reprogramming in TICs remains poorly explored. Here, we identify a mitochondria-encoded circular RNA, termed mcPGK1 (mitochondrial circRNA for translocating phosphoglycerate kinase 1), which is highly expressed in liver TICs. mcPGK1 knockdown impairs liver TIC self-renewal, whereas its overexpression drives liver TIC self-renewal. Mechanistically, mcPGK1 regulates metabolic reprogramming by inhibiting mitochondrial oxidative phosphorylation (OXPHOS) and promoting glycolysis. This alters the intracellular levels of α-ketoglutarate and lactate, which are modulators in Wnt/β-catenin activation and liver TIC self-renewal. In addition, mcPGK1 promotes PGK1 mitochondrial import via TOM40 interactions, reprogramming metabolism from oxidative phosphorylation to glycolysis through PGK1-PDK1-PDH axis. Our work suggests that mitochondria-encoded circRNAs represent an additional regulatory layer controlling mitochondrial function, metabolic reprogramming and liver TIC self-renewal.
Metabolic reprogramming plays vital roles in tumorigenesis. Here, Chen et al. reveal that mitochondria-encoded mcPGK1 drives the mitochondrial translocation of PGK1, promoting liver tumorigenesis and TIC self-renewal by switching energy production from OXPHOS to glycolysis.
## Introduction
Liver cancer is a common tumor type and many liver cancer patients have a very poor prognosis, which is largely due to tumor heterogeneity1. Accumulating researches have proved that tumor heterogeneity originates from the hierarchic organization of tumor cells that are derived from a small population of cells, termed as tumor initiating cells (TICs) or cancer stem cells (CSCs)1. Several markers of liver TICs (or CSCs) have been identified, such as CD133, CD13 and ZIC22–4. Unlike differentiated cancer cells, TICs are resistant to traditional radiotherapy and chemotherapy, and increasing studies demonstrate that TICs are also insensitive to CAR-T and immune checkpoint therapies5,6. Liver TICs is regulated by Wnt/β-catenin7,8, Notch9, Hedgehog10 and Hippo/Yap signaling pathways11, and these pathways are further accurately modulated. However, the molecular mechanisms of liver TIC function remain elusive.
Circular RNAs (circRNAs) are newly identified regulatory RNA molecules that have emerged as critical modulators in multiple biological processes12. Several circRNAs are identified as microRNA sponges, such as ciRS-7/CDR1as13,14. Fusion circRNAs and rtcircRNA are involved in tumorigenesis and drug-resistance15,16. We have identified circPan3 and circKcnt2 as regulators in intestinal stem cell (ISC) and colitis, respectively17,18. Moreover, we also revealed some circRNAs involved in the self-renewal regulation of tumor cells and TICs, including cia-MAF19, cis-HOX20 and circREEP321. Recently, mitochondria-encoded circRNAs have been identified, which are involved in communication between mitochondria and the nucleus22. Nonalcoholic steatohepatitis (NASH)-related mitochondrial circRNA SCAR (abbreviated for Steatohepatitis-associated circRNA ATP5B Regulator) interacts with ATP5B and inhibits mitochondrial reactive oxygen species (mROS) production and fibroblast activation23. However, the functions and regulatory mechanisms of mitochondria-encoded circRNAs in tumorigenesis and TICs are hitherto unclear.
Mitochondria are the key energy factories in almost all eukaryote cells. Mitochondria contain their own DNA, which encodes mitochondria-specific proteins and noncoding RNAs, such as 16 S ribosomal RNA, some transfer RNAs and circRNAs24. Mitochondria contain 1000–3000 proteins, most of which are encoded by nuclear DNA and transported from the cytoplasm by mitochondrial translocases, such as the TOM40 complex25,26. Many metabolic processes, including oxidative phosphorylation (OXPHOS), fatty acid β-oxidation and the urea cycle, occur in mitochondria27,28. These metabolic processes are regulated by various intracellular and extracellular factors. Recently, we have revealed that a mitochondria-located methyltransferase, Mettl4, inhibits mitochondrial transcription, OXPHOS, glycolysis, and mROS production29. Morphological and functional alterations of mitochondria are also driven by a variety of external factors, including hypoxia, starvation, infection, and tumorigenesis30,31. A hallmark of tumorigenesis is metabolic reprogramming, in which the main metabolic pathway switches from OXPHOS to glycolysis, a process also known as the Warburg effect32. One possible necessary of metabolic reprogramming is that some intermediate products of glycolysis provide a material basis for the rapid propagation of tumor cells. Indeed, the activities of glycolysis and OXPHOS are adjusted in cells within various stages of cell cycle. Cells in G1 stage prefer OXPHOS, whereas cells in S stage prefer glycolysis33. In addition to energy production, glycolysis and OXPHOS produce various metabolites that regulate multiple intracellular and extracellular biological processes. For example, the production of lactic acid during glycolysis inhibits the activity of T cells, enabling the immune escape of tumor cells34. In the present study, we have characterized a mitochondrial circRNA, termed mcPGK1 (mitochondrial circRNA for translocating phosphor- glycerate kinase 1), which is highly expressed in liver TICs and liver tumors. We found that mcPGK1 promotes the mitochondrial localization of PGK1, contributes to the metabolic reprogramming from OXPHOS to glycolysis.
## McPGK1 is highly expressed in liver TICs
Mitochondrial DNA-encoded circular RNAs (mecircRNAs) emerge as a new type of circRNAs22, and have been identified as critical modulators in NASH23, but their functions in tumorigenesis and TIC self-renewal are unknown. Here, we sorted CD133+ cells from primary liver cancer and proved these cells as liver TICs (Supplementary Fig. 1A–C), and isolated mitochondria from liver TICs and non-TICs for circRNA sequencing. There were 54 mecircRNAs identified in liver cancer cells. Among them, 9 mecircRNAs were differentially expressed (FC > 2, P-value < 0.05) in liver TICs and non-TICs, and named mecirc1-9 according to their corresponding locus of mitochondrial DNA (Fig. 1A). Then we designed circRNA-specific primers (Supplementary Fig. 1D) and detected their expression levels in six pairs of liver tumor and peri-tumor samples, and eight mecircRNAs were differently expressed in liver tumors (Supplementary Fig. 1E). Four pairs of TICs and non-TICs were used to further analyze mecircRNA expression, and finally six mecircRNAs (mecirc4, mecirc5, mecirc6, mecirc7, mecirc8, mecirc9) were screened out (Supplementary Fig. 1F, G). RNase R digestion, PCR and DNA-sequencing also confirmed that these six mecircRNAs are circular RNA (Supplementary Fig. 1H, I).Fig. 1High expression of mcPGK1 in liver tumor and liver TICs. A TICs and non-TICs were used for mitochondria isolation, and mitochondrial encoded circRNAs (mecirc) were identified through circRNA sequencing. Nine mecircRNAs with fold change |FC | > 2 and P-value < 0.05 were labeled as mecirc1-9 according to their locus on mitochondrial DNA. B Kaplan–Meier survival analysis of mcPGK1 high-expressing (mcPGK1high) and mcPGK1 low-expressing (mcPGK1low) patients, which were clustered according to the average level of mcPGK1 intensity in 90 liver tumor tissues (109.6). C Percentage distribution of CD133+ TICs in mcPGK1low (left) and mcPGK1high (right) samples. D RNAscope detection of 50 peri-tumor, 20 early HCC (eHCC) and 30 advanced HCC (aHCC) samples. Typical RNAscope images, CD133 FACS images and scatter plots were shown. Gating strategy was shown in supplementary Fig. 4E. Scale bars, 30 μm. E Quantitative real-time PCR analysis for mcPGK1 expression in CD133+ TICs and spheres. $$n = 4$$ independent experiments. F Northern blot for mcPGK1 expression in CD133+ TICs and CD133- non-TICs (left), or non-spheres and spheres (right). 18 S rRNA is a loading control. G Fluorescence in situ hybridization of mcPGK1 in non-sphere cells and spheres. Typical images were shown in the left panel and $$n = 10$$ images were taken for statistical analysis of mcPGK1 intensity (right). Scale bars, 10 μm. H RNAscope for mcPGK1 detection in TICs and non-TICs. Representative images and statistical intensity from $$n = 10$$ fields were shown. Scale bars, 10 μm. In all panels, data are shown as mean + s.d. ** $P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$, by CHISQ Test (C) or two-tailed Student’s T-test (for D, E, G, H). Source data are provided as a Source Data file.
To identify functional mecircRNA, we generated mecircRNA overexpressing cells (Supplementary Fig. 2A, B), followed by TIC detection. mecirc4, mecirc5 and mecirc8 drove expression of TIC marker CD133, whereas mecirc5 and mecirc8 promoted c-Myc expression (Supplementary Fig. 2C). Moreover, sphere formation assay showed that mecirc5, mecirc6 and mecirc8 were involved in liver TIC self-renewal (Supplementary Fig. 2D), and we focused on mecirc8 (hereafter termed as mcPGK1) for further analysis. mcPGK1 was generated from CYTB locus of mitochondrial DNA (Supplementary Fig. 2E). Then mcPGK1 specific probe was screened out by RNase H digestion assay and real-time PCR detection (Supplementary Fig. 2F), and confirmed via fluorescence in situ hybridization (Supplementary Fig. 2G). About 400–800 copies of mcPGK1 were detected in each liver TIC or sphere cell (Supplementary Fig. 3A). Then the subcellular location of mcPGK1 was measured. mcPGK1 was enriched in cytosol factions (including mitochondria) but not enriched in nuclear fractions (Supplementary Fig. 3B). Then mitochondria were isolated from cytosol fractions, and mcPGK1 was predominantly localized in mitochondria and also detectable in cytoplasm (Supplementary Fig. 3C–E). Moreover, total and mitochondrial mcPGK1 were increased in liver tumorigenesis (Supplementary Fig. 3F). We also examined the exact location of mcPGK1 in mitochondrial fractions isolated with APEX labeling, and found mcPGK1 was enriched in outer mitochondrial membrane and matrix, and also detectable in intermembrane space (Supplementary Fig. 3G, H). Fraction separation of mitochondria confirmed these results (Supplementary Fig. 3I). These data demonstrated that mcPGK1 was a mitochondrial DNA-encoded circular RNA and preferentially located in mitochondria.
Tox further evaluate the expression signature of mcPGK1 in liver tumors, we performed mcPGK1 in situ hybridization and the results showed that mcPGK1 was highly expressed in liver tumors (Supplementary Fig. 4A). The expression of mcPGK1 was correlated with clinical stages, tumor volumes, relapse and survival (Fig. 1B and Supplementary Fig. 4B, C). mcPGK1 was co-expressed with CD133, a marker of TICs (Supplementary Fig. 4D). Samples with low mcPGK1 expression harbored fewer TICs, whereas samples with high mcPGK1 expression harbored more TICs (Fig. 1C). We also validated the upregulation of mcPGK1 in liver tumors through RNAscope (Fig. 1D). Moreover, mcPGK1 RNAscope signals were positively related to CD133 ratios, confirming the microarray data (Fig. 1D and Supplementary Fig. 4E, F). The high expression of mcPGK1 in spheres and CD133+ TICs was confirmed by real-time PCR (Fig. 1E), Northern blotting (Fig. 1F), fluorescence in situ hybridization (Fig. 1G and Supplementary Fig. 4H), and RNAscope (Fig. 1H). Cytoplasmic and mitochondrial fractions were also separated from non-TICs, TICs, non-sphere and sphere cells, and mcPGK1 was proved to be enriched in mitochondrial fractions, especially in mitochondria of sphere cells (Supplementary Fig. 4G). Of note, the mitochondrial levels were comparable between liver TICs and non-TICs (Supplementary Fig. 4I). Taken together, mitochondria-encoded mcPGK1 is upregulated in liver TICs.
## McPGK1 drives the self-renewal of liver TIC
We then evaluated the function of mcPGK1 in liver TICs. First, we detected mcPGK1 expression levels across several HCC cell lines and primary samples, and found that it was differentially expressed among these cell lines and tissue samples (Supplementary Fig. 5A, B). We then constructed mcPGK1-knockdown cells using mcPGK1 high-expressing cells and shmcPGK1 was screened out to specifically target mcPGK1 but not linear RNA (Fig. 2A and Supplementary Fig. 5C, D). Moreover, mcPGK1 knockdown decreased both cytoplasmic and mitochondrial mcPGK1 levels, although knockdown efficiency of mitochondrial mcPGK1 was lower than cytoplasmic mcPGK1 (Supplementary Fig. 5E, F). McPGK1 silenced cells contained fewer liver TICs, and expressed lower levels of TIC markers and TIC-associated genes (Fig. 2B, C). The mcPGK1 silenced cells also displayed impaired sphere formation and proliferation capacities, but cell apoptosis was not influenced by mcPGK1 knockdown (Fig. 2D–F, and Supplementary Fig. 5G, H). Moreover, decreased tumor initiation and propagation capacity was detected in mcPGK1 silenced cells (Fig. 2G–I). We then obtained shmcPGK1 CD133+ liver TICs and shmcPGK1 sphere cells for self-renewal detection, and revealed that mcPGK1 promotes self-renewal in liver TICs (Fig. 2J and Supplementary Fig. 5I). To further target mitochondrial mcPGK1 efficiently, we also constructed mitochondria-targeting nanoparticles25, and found these nanoparticles can efficiently target mitochondrial mcPGK1 (Supplementary Fig. 5J). shmcPGK1 cells established by mitochondria-targeting nanoparticles also showed impaired sphere formation capacity (Fig. 2K).Fig. 2McPGK1 is required for liver TIC self-renewal. A Quantitative real-time PCR to confirm the knockdown efficiency in mcPGK1 silenced cells. B FACS detection of mcPGK1 silenced cells for CD133 expression. C Quantitative real-time PCR analysis of TIC markers and TIC-associated TFs in mcPGK1 silenced and control cells. D Sphere formation of mcPGK1 silenced and control cells. Typical sphere images were in the left panel and sphere diameters of $$n = 30$$ spheres from three independent experiments were measured in the right panel. D3, day 3. Scale bars, 100 μm. E Sphere formation of mcPGK1 silenced and control HCC #2 sample cells. Scale bars, 500 μm. F McPGK1 fluorescence in situ hybridization and Ki67 staining in mcPGK1 silenced and control Huh7 cells. Typical images and Ki67+ ratios were shown. For each group, $$n = 10$$ fields were observed. Scale bars, 10 μm. G, H Tumor initiation assay of gradient numbers of mcPGK1 silenced and control cells derived from spheres derived from primary #2 cells. mcPGK1 detection for knockdown efficiency was shown in G. $$n = 7$$ 6-week-old male BALB/c nude mice were used for H. TIC ratios and P-value were calculated by extreme limiting dilution analysis (ELDA) (http://bioinf.wehi.edu.au/software/elda/). Scale bars, 30 μm. I 1 × 106 mcPGK1 knockdown and control luciferase labeled primary #2 cells were used for in vivo propagation. Typical liver images and calculated results were shown. $$n = 6$$ mice per group. J Sphere formation with 1000 CD133+ control and shmcPGK1 Huh7 cells, which were sorted via FACS assay. K Sphere formation with mcPGK1 knockdown and control cells, for which shmcPGK1 and control plasmids were delivered into primary HCC cells derived from sample #2 with mitochondria-targeting nanoparticles (Mito-NP). Scale bars, 500 μm. For A, B, C, E, J, K, $$n = 4$$ independent experiments. In all panels, data are shown as mean + s.d. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
We also constructed mcPGK1 overexpressing cells, which also showed mitochondrial location of mcPGK1 (Supplementary Fig. 6A–C). These cells harbored increased TIC ratios and TIC marker expression (Supplementary Fig. 6D, E). Furthermore, mcPGK1 overexpression enhanced TIC self-renewal, tumor initiation but not apoptosis (Supplementary Fig. 6F–I). mcPGK1 was then overexpressed using mitochondria-targeting nanoparticles, and mcPGK1 overexpression in mitochondria also promoted liver TIC self-renewal (Supplementary Fig. 6J, K). Moreover, mcPGK1 overexpression also drove tumor propagation in vivo (Supplementary Fig. 6L). Overall, these findings indicate that mcPGK1 is required for liver TIC self-renewal.
## McPGK1 reprograms metabolism from OXPHOS to glycolysis
As the central function of mitochondria is energy metabolism, we evaluated the effects of mcPGK1 on OXPHOS and glycolysis. OXPHOS activity was enhanced and glycolytic activity was attenuated in mcPGK1-silenced cells, whereas the opposite occurred in mcPGK1-overexpressing cells, indicating critical roles of mcPGK1 in metabolic reprogramming shifting from OXPHOS to glycolysis (Fig. 3A, B). Supporting these findings, the OXPHOS metabolite levels were increased and glycolytic metabolite levels were decreased in mcPGK1-silenced cells, whereas mcPGK1-overexpressing cells contained increased levels of glycolytic metabolites (Fig. 3C). Interestingly, divergent levels of OXPHOS and glycolysis metabolites were detected in primary cells with high or low mcPGK1 expression (Fig. 3D). These results demonstrate the essential role of mcPGK1 in metabolic reprogramming of liver TICs. Fig. 3McPGK1 drives the metabolic reprogramming from OXPHOS to glycolysis. A, B Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of mcPGK1 silenced (A) and overexpressing (B) cells. Huh7 cells were used for mcPGK1 knockdown (A) and Hep3B cells were used for mcPGK1 overexpression (B). $$n = 3$$ independent experiments for each cell. C Intracellular levels of the indicated metabolites in mcPGK1 silenced (left, sample #2) and overexpressing (right, sample #4) panels. $$n = 4$$ independent experiments. D Five mcPGK1 high-expressing (mcPGK1high, sample #2, #3, #12, #10, #13) and mcPGK1 low-expressing (mcPGK1low, sample #4, #6, #7, #1, #15) samples were used for metabolite detection. E, F The abundance of lactate (E) and α-KG (F) in the indicated medium supernatant was measured at the indicated time points. Huh7 and Hep3B were used for mcPGK1 knockdown and overexpression. $$n = 4$$ independent experiments. G Acidification of the culture medium in mcPGK1 silenced (Huh7) and overexpressing (Hep3B) cells, as indicated by the color change of the phenol red indicator in the medium to orange/yellow. Typical images were shown representative of $$n = 3$$ independent experiments. In all panels, data are shown as mean + s.d. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
We then determined the time-course levels of lactic acid and α-KG, two of the main metabolites of glycolysis and OXPHOS, respectively. Lactic acid accumulated rapidly in mcPGK1-overexpressing cells, but not in mcPGK1-knockdown cells (Fig. 3E). In contrast, α-KG was accumulated rapidly in mcPGK1-knockdown cells, but not in mcPGK1-overexpressing cells (Fig. 3F). Similarly, the medium of mcPGK1-overexpressing cells tended to become orange/yellow, whereas the medium of mcPGK1-silenced cells tended to stay red, confirming the role of mcPGK1 in the acidification of culture medium (Fig. 3G). These data demonstrated that mcPGK1 was involved in metabolic reprogramming.
## Metabolic reprogramming drives liver TIC function via Wnt pathway
The functions of metabolic reprogramming in liver TICs are unknown. Among the metabolites we evaluated, lactic acid and α-KG functioned as stemness modulators in liver TICs (Fig. 4A). Lactic acid drove liver TIC self-renewal, whereas α-KG had an opposite effect (Fig. 4B, C). Lactic acid treatment increased the expression of TIC-associated genes, whereas α-KG elicited opposite effects (Fig. 4D). Then the roles of lactic acid and α-KG in liver tumor propagation and liver TIC self-renewal were examined in vivo. FX-11, an inhibitor of lactic acid production, inhibited liver tumor propagation (Fig. 4E and Supplementary Fig. 7A), decreased the ratios of CD133+ liver TICs (Fig. 4F), and impaired the tumor initiation capacity (Fig. 4G), indicating that lactic acid was a driver of liver TIC self-renewal. In contrast with lactic acid, α-KG showed inhibitory effects on liver tumor propagation and liver TIC maintenance (Fig. 4E–G). The modulation of liver tumor propagation by lactic acid and α-KG was confirmed by in vivo luciferase assay (Fig. 4H). These results confirmed that metabolic reprogramming was involved in liver TIC maintenance. Fig. 4Lactate and α-KG are involved in liver TIC self-renewal. A CD133+ TIC ratios after 2 days’ treatment with the indicated metabolites were detected via FACS. DM-αKG, cell permeable α-KG. $$n = 4$$ independent experiments. B Typical images and sphere diameters in lactate and DM-αKG treated sample #2 cells. $$n = 30$$ spheres from three independent experiments were measured. D3, day 3. Scale bars, 100 μm. C Sphere formation of sample #2 cells supplemented with indicated levels of lactate and DM-αKG. $$n = 4$$ independent experiments. Scale bars, 500 μm. D Quantitative real-time PCR detection for the indicated gene expression in lactate and DM-αKG treated primary sample #2 cells. $$n = 3$$ independent experiments. E Xenograft tumors were established in BALB/c nude mice and treated with FX-11 or DM-αKG after tumors reached about 250mm3, and tumor volume was detected every 3 days. $$n = 6$$ 6-week-old male BALB/c nude mice per group. F CD133 immunohistochemistry in Vehicle (Veh), lactic acid production inhibitor FX-11 and DM-αKG treated tumors. $$n = 6$$ tumors per group, and for each tumor, $$n = 10$$ images were detected. Scale bars, 50 μm. G 10, 1 × 102, 1 × 103, 1 × 104 and 1 × 105 Huh7 sphere cells were injected into BALB/c nude mice for tumor initiation assay. $$n = 6$$ 6-week-old male BALB/c nude mice per group. TIC ratios and P-values calculated via ELDA were shown in the right panels. H 1 × 106 luciferase labeled primary #2 cells were used for in vivo propagation, and mice were administered with FX-11 (2 mg/kg) or DM-αKG (500 mg/kg). Typical liver images and calculated results were shown. $$n = 6$$ mice per group. For A, C, $$n = 4$$ independent experiments. In all panels, data are shown as mean + s.d. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
We evaluated several signaling pathways and found that lactic acid and α-KG both targeted Wnt/β-catenin pathway, a central signaling pathway for liver TIC function (Fig. 5A). The enhanced sphere-formation capacity by mcPGK1 was blocked upon Wnt/β-catenin inhibition with Wiki4 or LF3, further confirming that mcPGK1 exerted its role via Wnt/β-catenin pathway (Fig. 5B and Supplementary Fig. 7B). Lactic acid promoted Wnt/β-catenin activation and α-KG inhibited Wnt/β-catenin activation (Fig. 5C–F). Interestingly, β-catenin was increased at protein level but not mRNA level upon lactic acid treatment, and decreased at mRNA level upon DM-αKG treatment (Fig. 5C, F). As expected, lactic acid promoted β-catenin protein stability (Fig. 5G), and α-KG inhibited the activation of β-catenin promoter and β-catenin transcription (Fig. 5H, I). We then generated β-catenin silenced cells, and revealed that mcPGK1 had an impaired function in liver TIC self-renewal and in vivo propagation of liver tumor cells, confirming that mcPGK1 exerted its role mainly via a β-catenin-dependent manner (Supplementary Fig. 7C, D).Fig. 5Lactate and α-KG are involved in Wnt/β-catenin activation. A FACS detection for the activity of indicated signaling pathways. $$n = 4$$ independent experiments for each detection. B mcPGK1 overexpressing (oemcPGK1) and control (oeVec) sample #2 cells were used for sphere formation, supplemented with Wnt/β-catenin inhibitors Wiki4 and LF3. Typical sphere images and calculated ratios were shown. Scale bars, 500 μm. C Immunoblot to evaluate the activation of Wnt/β-catenin signaling pathway in indicated treated sample #2 cells. D Immunofluorescent staining to detect the expression levels of TIC marker CD133 and Wnt/β-catenin target gene CCND1 in lactate and DM-αKG treated spheres, which were from sample #2 cells. Typical immunofluorescent images and statistical analysis of $$n = 10$$ fields were shown. Scale bars, 10 μm. E β-CATENIN, c-MYC and Ki67 immunohistochemistry in FX-11 treated, DM-αKG treated and control tumors. $$n = 6$$ tumors were detected per group. Typical β-CATENIN (β-CAT) immunohistochemistry results were shown. Scale bars, 20 μm. F Quantitative real-time PCR to detect the expression of Wnt/β-catenin-related genes in lactic acid and α-KG treated sample #2 cells. $$n = 4$$ independent experiments. G Western blot for β-catenin levels in cycloheximide (ChX) treated cells, which were cultured in 10 mM lactic acid or control medium. β-catenin levels were normalized to its level at 0 min. H Quantitative real-time PCR for the enrichment of indicated regions of CTNNB1 promoter in ChIP eluate, in which H3K4me3 antibody and DM α-KG treated sample #2 cells were used. I Click-it EU labeling assay to detect the nascent CTNNB1 (β-CATENIN mRNA) levels. In all panels, data are shown as mean + s.d. For A–C, F–I, $$n = 4$$ independent experiments. For A, B, D, E, I, *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
We then explored the molecular mechanism of lactic acid regulation of β-catenin protein stability. We previously identified lnc-β-catm as a modulator as β-catenin stability in liver cancer and TICs, via promoting the methylation of β-catenin35, and here we investigated whether lactic acid promoted the expression of lnc-β-catm. We revealed that lactic acid promoted lnc-β-catm expression (Supplementary Fig. 7E), and subsequent β-catenin methylation (Supplementary Fig. 7F). Moreover, lactic acid displayed a limited role on β-catenin stability and sphere formation in lnc-β-catm knockout cells, indicating the critical role of lnc-β-catm in lactic acid-driven β-catenin stability (Supplementary Fig. 7G, H). We then explored the molecular mechanism of lactic acid in lnc-β-catm expression. Considering the direct effect of lactic acid on histone lactylation36, we firstly detected the lactylation of lnc-β-catm promoter, and found that lactic acid promoted the histone lactylation of lnc-β-catm promoter at −700~−500 fragment (Supplementary Fig. 7I). We also deleted this region through CRISPR/Cas9 approach, and lactic acid didn’t promote lnc-β-catm expression upon lnc-β-catm promoter deletion (Supplementary Fig. 7J). These results proved that lactic acid promoted β-catenin stability largely through lnc-β-catm expression, which depends on histone lactylation of lnc-β-catm promoter.
Previous works have revealed that α-KG is a cofactor of H3K4me3 demethylase JARID1B37,38, thus we investigated whether α-KG regulates the transcription of β-catenin through JARID1B-mediated H3K4me3 modification. We found α-KG inhibited H3K4me3 levels (Supplementary Fig. 8A). Moreover, α-KG inhibited the chromatin accessibility at CTNNB1 (β-catenin mRNA) promoter (Supplementary Fig. 8B). We also generated JARID1B silenced cells, and revealed that α-KG showed impaired roles in the H3K4me3 and accessibility of CTNNB1 promoter, as well as nascent CTNNB1 mRNA expression, upon JARID1B knockdown, indicating that α-KG inhibited β-catenin transcription through H3K4me3 demethylase JARID1B (Supplementary Fig. 8C–E). We also evaluated the function of α-KG with 2-HG, a competitive inhibitor of α-KG-dependent dioxygenases38, and found that 2-HG largely attenuated the functions of α-KG in H3K4me3 enrichment, accessibility of CTNNB1 promoter and nascent CTNNB1 mRNA, further confirming that α-KG functions as a modulator of H3K4me3 demethylase (Supplementary Fig. 8F–H). These findings indicate that both lactic acid and α-KG function in liver TICs through Wnt/β-catenin pathway.
## McPGK1 interacts with PGK1
To analyze the molecular mechanisms of mcPGK1, we performed an RNA pulldown assay, which identified PGK1, TOM40 and TOM70 as the partners of mcPGK1 in liver TICs (Fig. 6A). Immunoblot assay confirmed that mcPGK1 interacted with PGK1, TOM40 and TOM70 (Fig. 6B). We focused on PGK1, which is a critical modulator in glycolysis and OXPHOS39. RNA immunoprecipitation proved PGK1-mcPGK1 interaction (Fig. 6C). We further analyzed the interaction of mcPGK1 and PGK1. Considering the critical role of stem-loop structures in RNA–protein interactions40,41, we analyzed the structure of mcPGK1 and identified seven loops, and found that the second loop (HR#2) was required for the interaction between mcPGK1 and PGK1 (Fig. 6D, E, and Supplementary Fig. 8I). These results confirmed that mcPGK1 interacts with PGK1 in liver TICs. Fig. 6McPGK1 functions as PGK1 partner. A Silver staining of RNA pulldown eluate, for which mcPGK1 probes, control probes, and Huh7 sphere lysate were used. The specific bands indicated by black arrows were identified as PGK1, TOM70 and TOM40 via mass spectrum. B Immunoblot of RNA pulldown eluate to confirm the interaction of mcPGK1 with PGK1, TOM70 and TOM40. Huh7 spheres were used for RNA pulldown. C Quantitative real-time PCR for mcPGK1 enrichment in eluate sample from RNA immunoprecipitation, in which IgG and PGK1 antibodies were used. mcPGK1 enrichment levels were normalized to IgG group. D WT and indicated mutant mcPGK1 were used for TRAP assay, and the enrichment of PGK1 in TRAP eluate was evaluated through immunoblot. E Quantitative real-time PCR for the enrichment of mcPGK1 in PGK1 RIP eluate. For RIP, mcPGK1 lowly expressed cells (Hep3B) were overexpressed with indicated mutant mcPGK1 transcripts. F Sphere formation of indicated Hep3B cells. mcPGK1-WT, wide type mcPGK1; mcPGK1-Mut#2, mcPGK1 mutation losing the second stem-loop region. Scale bars, 500 μm. G Immunoblot to detect the activation of Wnt/β-catenin signaling in Hep3B cells overexpressing WT mcPGK1 and mutant#2 mcPGK1. β-actin is a loading control. H The ratios of tumor-free mice initiated from indicated cells were shown in left panel, TICs ratios were shown in right panel. $$n = 6$$ 6-week-old male BALB/c nude mice per group. For all panels, $$n = 4$$ independent experiments, and data are shown as mean + s.d. For C, E, F, **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001$; ns, not significant, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
We then examined whether the interaction with PGK1 was necessary for mcPGK1’s function. *We* generated cells that were overexpressing mutant mcPGK1, which lost the ability to interact with PGK1. Compared with wild-type mcPGK1, mutant mcPGK1 did not promote Wnt/β-catenin activation and liver TIC maintenance, further confirming the essential role of PGK1-mcPGK1 interaction (Fig. 6F, G). Moreover, mutant mcPGK1 showed an impaired role in tumor initiation (Fig. 6H). Altogether, mcPGK1 interacts with PGK1 and functions through mcPGK1-PGK1 interaction.
## McPGK1 promotes the interaction of PGK1 and TOM40 complex
We then evaluated the combination between PGK1 and TOM40/TOM70, which are core components of TOM40 mitochondria importing complex42. We found that PGK1 interacted with TOM40 and TOM70, and their interactions were impaired in mcPGK1 silenced cells (Fig. 7A). On the contrary, enhanced interactions between PGK1 and TOM40/TOM70 were detected in mcPGK1 overexpressing cells (Fig. 7B). Enhanced interactions between PGK1 and TOM40/TOM70 were confirmed by co-immunoprecipitation assay supplemented with gradient mcPGK1 (Fig. 7C, D). Split-APEX2 assay confirmed the assembly of PGK1-TOM40/TOM70-mcPGK1 complex at outer mitochondrial membrane (Fig. 7E). Moreover, attenuated assembly of PGK1-TOM40/TOM70 complex was detected upon mcPGK1 knockdown (Fig. 7F). These results demonstrated that mcPGK1 promoted the interaction between PGK1 and TOM40/TOM70. Using mutant mcPGK1 transcripts, we found that the seventh loop (HR#7) was required for the interaction between mcPGK1 and TOM40/TOM70 (Fig. 7G). HR#2 mutant and HR#7 mutant mcPGK1 transcripts were overexpressed, which lost the ability to interact with PGK1 and TOM40/TOM70, respectively. Both mutant mcPGK1 transcripts (mut#2 and mut#7) weren’t involved in the regulation of PGK1-TOM40/TOM70 interaction, indicating that mcPGK1 served as a scaffold of PGK1-mcPGK1-TOM40/TOM70 complex (Fig. 7H). Moreover, #2 mutant and #7 mutant mcPGK1 weren’t involved in liver TIC self-renewal and metabolic reprogramming (Fig. 7I, J). HR#2 and HR#7 mutant mcGPK1 transcripts had impaired roles in PGK1-TOM40 interaction, TIC self-renewal and metabolic reprogramming, whereas WT mcGPK1 displayed these roles, thus we concluded that HR#2 and HR#7 were required for PGK1-TOM40/TOM70 interaction and mcPGK1-driven TIC self-renewal. Altogether, mcPGK1 promotes the binding of PGK1 to TOM40 mitochondrial importing complex. Fig. 7McPGK1 promotes the binding of PGK1 and TOM40 complex. A, B Immunoblot of TOM70/TOM40 in PGK1 immunoprecipitation (IP) eluate from mcPGK1 knockdown (A) or overexpressing (B) sphere lysate. $1\%$ Input, $50\%$ IgG IP eluate and PGK1 IP eluate were used for immunoblot. Typical results in left panel and protein quantitative results in right panel. C, D Immunoblot to evaluate TOM70/TOM40 levels in PGK1 IP eluate (C), or PGK1 levels in TOM70/TOM40 IP eluate (D). Primary #4 and #6 cells were used for sphere formation, and sphere lysates supplemented with gradient doses of mcPGK1 transcript were used for IP assay. E Split-APEX2 assay were established (left), followed by real-time PCR for mcPGK1 detection (middle) and Western blot for another outer mitochondrial membrane protein TOM70 (right). TOM40, PGK1 and mcPGK1 are assembled together in outer mitochondrial membrane. F mcPGK1 silenced cells were used for Split-APEX2, followed by Western blot with TOM40 and TOM70 antibodies. G Hep3B cells overexpressing WT and indicated mutant mcPGK1 were used for TRAP assay, and the enrichment of TOM70/TOM40 in TRAP eluate was evaluated via immunoblot. H Immunoblot of TOM70/TOM40 in PGK1 IP eluate. WT and mutant mcPGK1 cells were used for sphere formation, followed by IP assay with IgG control or PGK1 antibodies. $$n = 3$$ independent experiments. I, J Liver tumor cells (sample #4) overexpressing WT and mutant mcPGK1 were used for sphere formation (I) and metabolic analysis (J). Scale bars, 500 μm. For A–D, $$n = 3$$ independent experiments; for E, $$n = 5$$ independent experiments; for I, J, $$n = 4$$ independent experiments. For D, F, G, $$n = 3$$ independent experiments with similar results. Data are shown as mean + s.d. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
## McPGK1 drives the mitochondrial entry of PGK1
PGK1 is expressed in the cytoplasm, but is often translocated to the mitochondria during tumorigenesis, where it phosphorylates PDK1 at T338. In turn, PDK1 phosphorylates and inhibits the PDH complex, and thus inhibits OXPHOS and promotes glycolysis43. Here, we demonstrated that mcPGK1 promoted the binding of PGK1 to TOM40 mitochondria importing complex. Therefore we examined the involvement of mcPGK1 in the translocation of PGK1 to mitochondria and found that this translocation was suppressed in mcPGK1 silenced cells, and increased in mcPGK1 overexpressing cells (Fig. 8A). However, PGK1 expression wasn’t influenced by mcPGK1 (Supplementary Fig. 9A). Immuno-electron microscopy also revealed that mcPGK1 was involved in mitochondrial translocation of PGK1 (Fig. 8B). Mitochondrial isolation and immunoblot confirmed that mcPGK1 was essential for PGK1 mitochondrial translocation (Fig. 8C). The positive role of mcPGK1 in PGK1 mitochondrial entry was further confirmed by mitochondrial fraction separation assay (Fig. 8D, E). We then overexpressed WT and mutant mcPGK1 transcripts and detected the mitochondrial translocation of PGK1 via Matrix-APEX and OMM-APEX. Overexpression of WT mcPGK1 promoted the mitochondrial translocation of PGK1, whereas HR#2 and HR#7 mutant transcripts had no such effect, further confirming the essential role of HR#2/#7 in mcPGK1-dependent mitochondrial translocation of PGK1 (Fig. 8F).Fig. 8McPGK1 promotes the mitochondrial entry of PGK1.A Confocal microscopy to detect the intracellular location of PGK1 in mcPGK1 silenced and control cells. Scale bars, 10 μm. $$n = 10$$ fields per group. B Immuno-electron microscopy for the mitochondrial entry of PGK1. $$n = 5$$ independent experiments with similar results (upper panel) and $$n = 10$$ mitochondria images for statistical analysis (lower panel). Scale bars, 200 nm. C Immunoblots for the mitochondrial entry of PGK1. Voltage-dependent anion channel (VDAC) (mitochondrial) and histone 3 (H3) (nuclear) were used to detect the purity of mitochondria. Huh7 cells were used for mcPGK1 knockdown and Hep3B cells were used for mcPGK1 overexpression. $$n = 3$$ independent experiments. D The indicted fractions of mitochondria were isolated and PGK1 levels were detected with immunoblot. E OMM-APEX, IMM-APEX and matrix-APEX were established in mcPGK1 silenced and overexpressed cells, followed by streptavidin pulldown and PGK1 immunoblot was performed to detect the enrichment of PGK1 in mitochondria. F Hep3B cells expressing OMM-APEX (OMM-AP) system for outer mitochondrial membrane labeling, or expressing Matrix-APEX (Matrix-AP) system for mitochondrial matrix labeling, were used for mcPGK1 or mutant (HR#2, HR#7) mcPGK1 overexpression. G Immunoblot for PDH and PDH inactivating phosphorylation (p-PDH1) levels in mitochondrial fractions, which were isolated from the indicated cells. H PGK1 mitochondria-locating #2 cells were established via mitochondria-targeting nanoparticle (mito-NP) and then mcPGK1 was silenced, followed by sphere formation assay. $$n = 4$$ independent experiments. I Sphere formation of mcPGK1 overexpressing (oemcPGK1) and control (oeVec) sample #4 cells, supplemented with 100 nM PDK1 inhibitor AZD7545. $$n = 4$$ independent experiments. For D–G, $$n = 3$$ independent experiments with similar results. In all panels, data are shown as mean + s.d. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$, ns, not significant, by two-tailed Student’s T-test. Source data are provided as a Source Data file.
We then examined and found that mcPGK1 knockdown decreased the inactivating phosphorylation of PDH (Fig. 8G). As expected, mcPGK1-PGK1-PDK1 mediated PDH phosphorylation inhibited PDH’s function of converting pyruvate to acetyl-CoA, and subsequently drove a metabolic reprogramming from OXPHOS to glycolysis (Supplementary Fig. 9B). These data confirmed the role of mcPGK1-PGK1-PDK1-PDH axis in metabolism reprogramming from OXPHOS to glycolysis (Supplementary Fig. 9C). Finally we evaluated the role of PGK1-PDK1-PDH axis in mcPGK1 function. Nanoparticle-delivered PGK1 mitochondrial translocation largely diminished the role of mcPGK1 knockdown, indicating that mcPGK1 functions through PGK1 mitochondria-entry (Fig. 8H and Supplementary Fig. 9D). Moreover, mcPGK1 overexpression had a limited role in sphere formation upon PDK1 blockade (Fig. 8I). Similarly, mcPGK1 overexpression also showed impaired roles in sphere formation and in vivo propagation upon PDK1 was silenced, further confirming that mcPGK1 exerted its functions through PDK1-PDH pathway (Supplementary Fig. 9E, F). Taken together, these results indicate that mcPGK1 drives mitochondrial translocation of PGK1, inhibits OXPHOS and promotes glycolysis via the PGK1–PDK1–PDH pathway.
## Discussion
In this work, we identified mitochondria-encoded mcPGK1 promotes liver TIC self-renewal via metabolic reprogramming from OXPHOS to glycolysis. McPGK1 interacts with PGK1, promotes the binding of PGK1 to TOM40 mitochondrial importing complex, and drives the mitochondrial translocation of PGK1. Metabolic reprogramming switches the metabolites from α-KG to lactic acid, activates Wnt/β-catenin and liver TIC function. Our work reveals an additional layers to circRNA function, TIC self-renewal and metabolism regulation.
circRNAs, generated by back-splicing of the 3' and 5' ends of RNA, have emerged as critical modulators in a variety of biological processes. With recent advances in RNA sequencing, several new types of circRNAs have been identified, including read-through, virus-encoded, and mitochondria-encoded circRNAs22. Here, we focus on mitochondria-encoded circRNA in liver TICs. Mitochondria contain a copy of circular double-stranded DNA, about 16.5 kb long, from which circRNAs are generated. Very recently, the mitochondria-encoded circRNA SCAR was discovered to alleviate NASH by reducing mROS output23. Here, we identified TIC-regulatory function of mcPGK1, a newly identified mitochondria-encoded circRNA, adding an additional layer to the function of circRNAs and regulation of TICs.
Mitochondria contain 1000–3000 proteins, most of which are encoded by nuclear genes. The blockade of mitochondrial translocation triggers various disorders, including obesity44. Here we revealed that a dysregulated mitochondrial translocation of PGK1 drives liver tumorigenesis, TIC self-renewal and metabolic reprogramming. Because each cell contains many copies of mitochondrial DNA, it is difficult to manipulate the expression of mitochondrial genes, which hinders research aimed at investigating their biological roles. Several studies have revealed that siRNA and shRNA can be used to silence gene expression in mitochondria45, and a CRISPR-free mitochondrial base editing system has been created that can change C:G to T:A in mitochondrial DNA in an efficient and specific manner46. A mitochondria-targeting nanoparticle has also been constructed and used to deliver genes to mitochondria23. In this study, we used shRNA to silence mcPGK1 expression. Nuclear overexpression of mcPGK1 was implicated in liver TIC self-renewal and metabolic reprogramming. Indeed, while mcPGK1 is mainly localized to the mitochondria, it is also present in the cytoplasm. We also demonstrated that mcPGK1 was required for the mitochondrial translocation of PGK1, which was normally localized to the cytoplasm.
Mitochondrial metabolism is closely related to stemness regulation. Actually, increased mitochondrial biogenesis and OXPHOS induce the differentiation of various stem cells, indicating that the loss of mitochondrial function is very important for stemness maintenance47. Mitochondrial biogenesis, fission and metabolic plasticity are involved in asymmetric division and prostate TIC self-renewal48. Some stemness factors, such as Nanog, reduce OXPHOS activity and decrease mROS production to maintain the self-renewal capacity49. Here, we revealed that in liver TICs, a highly expressed circRNA drives metabolic reprogramming from OXPHOS to glycolysis by modulating the mitochondrial distribution of PGK1. Metabolic reprogramming from OXPHOS to glycolysis may provide a material basis for the rapid propagation of tumor cells, and the metabolites may also play key roles in the regulation of stemness. Indeed, here we found that lactic acid and α-KG are involved in Wnt/β-catenin activation and liver TIC function.
In addition to TIC self-renewal, metabolic reprogramming contributes to drug resistance and immune escape. The reduction in OXPHOS results in reduced mROS production, promotes a cellular quiescent state, and maintains the genomic stability of stem cells50. The reduction in mROS production via metabolic reprogramming in TICs is thought to play a key role in resistance to chemotherapeutic drugs51. Glycolysis also inhibits anti-tumor immune activity52. In particular, TICs increase the production of lactic acid through glycolysis, thus maintaining an acidic tumor microenvironment that inhibits the function of anti-tumor immune cells such as T effector and natural killer cells53. Therefore, the metabolic reprogramming from OXPHOS to glycolysis may inhibit immune surveillance during tumorigenesis. Hence, mcPGK1 might contribute to tumor immune escape and the therapeutic effect of immune checkpoint therapy of liver tumors.
## Ethics statement, mice and cells
This work was approved by the ethics committee of Zhengzhou University (ZZUIRB202054 and ZZUIRB202055). For all mouse experiments, 6-week-old male BALB/c nude mice were purchased were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd., and housed in the animal facility at School of Life Sciences, Zhengzhou University. Mice were housed in SPF condition, with 4-7 mice per cage in 12 h light/dark cycle (7:00-19:00 light, 19:00-7:00 dark), with controlled room temperature (23 ± 2 °C) and humidity (40-$60\%$). All mice were randomly grouped and no mice were excluded from analyses. The maximal tumor burden over 2500 mm3 is forbidden by the ethics committee, and this limit was not exceeded in all experiments. All efforts to minimize animal suffering were made. Liver cancer tissues used in this work were obtained from The First Affiliated Hospital of Zhengzhou University. Hep3B cells were obtained from ATCC (catalog no, HB-8064), Huh7 cells were obtained from iCellbioscience (catalog no. iCell-h080), 293 T, PLC and Hep-1 cells were from Zusen Fan lab (Institute of Biophysics, Chinese Academy of Sciences).
## Antibodies and Reagents
Anti-β-Catenin (catalog no. 610153) and anti-CD133 antibody (catalog no. 566598) was purchased from BD Bioscience. Anti-PGK1 (catalog no. 68540 S), anti-EEA1 (catalog no. 3288 S), anti-β-actin (catalog no. 4970), anti-H3 (catalog no. 4499) and anti-H3K4me3 (catalog no. 9751 S) antibodies were from Cell Signaling Technology. Anti-ZIC2 (catalog no. ARP35821_P050) antibody was purchased from Aviva Systems Biology. Anti-TOM40 (catalog no. 18409-1-AP), anti-TOM70 (catalog no. 14528-1-AP), anti-c-MYC (catalog no. 10828-1-AP) and anti-AXIN2 (catalog no. 20540-1-AP) antibodies were from Proteintech Group, Inc. Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 594 antibody (catalog no. A-11005), Goat anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488 (catalog no. A-11008), Goat anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 647 (catalog no. A-21244) were purchased from Invitrogen. HRP-conjugated Affinipure Goat Anti-Mouse IgG(H + L) antibody (catalog no. SA00001-2) and HRP-conjugated Affinipure Goat Anti-Rabbit IgG(H + L) antibody (catalog no. SA00001-2) were purchased from Proteintech Group, Inc. Polymer HRP and AP detection kits were from Beyotime Biotechnology. Biotin labeled RNA mix (catalog no. 11685597910) was from Roche.
## Tumor initiation assay
For tumor initiation assay, 10, 1 × 102, 1 × 103, 1 × 104, and 1 × 105 mcPGK1 knockdown, overexpressing cells were subcutaneously transplantated into 6-week-old male BALB/c nude mice and tumor initiation was detected after 3 months. Online-available Extreme Limiting Dilution Analysis tool (http://bioinf.wehi.edu.au/software/elda/)54 was used for TIC ratio calculation.
## Sphere formation
For sphere formation, 1000 Huh7 and PLC single cells were cultured in Ultra Low Attachment 6-well plates, and incubated with Dulbecco’s modified Eagle’s medium/F12 (Life Technologies) supplemented with N2, B27, 20 ng/ml EGF and 20 ng/ml bFGF (Millipore) for 2 weeks, sphere initiating ratio = (sphere number)/1000 × $100\%$. For primary cells, 5000 cells were used for each well, and sphere initiating ratio = (sphere number)/5000 × $100\%$.
## Separation of mitochondria and mitochondrial fractions
mcPGK1 silenced, overexpressing and control cells were harvested for mitochondria isolation. Isolation buffer (225 mM mannitol, 20 mM MOPS, 75 mM sucrose, 1 mM EGTA, $0.1\%$ BSA, pH 7.2) and lysis buffer (100 mM sucrose, 10 mM MOPS, 1 mM EGTA, $0.1\%$ BSA, pH 7.2) were used for mitochondria isolation.
For separation of mitochondrial fractions, mitochondria were incubated with 1 ml digitonin buffer (mitochondria isolation buffer containing 0.5 mg/ml digitonin) for 15 min, followed by 10,000 x g (10 min, 4 °C) centrifuge. The pellet contained mitoplast and the supernatant contained OMM and IMS fractions. Then OMM fraction was obtained from the precipitate of 10,000 x g (30 min, 4 °C) centrifuge. The mitoplast pellet was re-suspended into 0.2 ml mitochondria isolation buffer and gently disrupted with ultrasonication, followed by 10,000 x g (30 min, 4 °C) centrifuge. The IMM fraction was in pellet and matrix fraction was in supernatant.
## Absolute quantification of mcPGK1
For absolute quantification, total cells or cell fractions from peri-tumor, tumor, TIC and sphere cells were used for RNA extraction, followed by 1U/mg RNase R treatment at 37 °C for 15 min. RNA samples were then reversely transcribed to cDNA and quantitative real-time PCR was performed. In vitro transcribed mcPGK1 was used for standard curve by series dilution.
## APEX submitochondrial fractions
For APEX assay, liver cancer cells were transfected with Matrix-APEX2 (Cat# 72480, Addgene)55, IMS-APEX2 (Cat# 79058, Addgene)56, or OMM-APEX2 plasmids (Cat# 79056, Addgene)55 for submitochondrial labeling. The cells were treated with 500 mM biotin-phenol at 37 °C for 30 min, and then treated with 1 mM H2O2 at room temperature for 1 min. Samples were then treated with 2 mL azide-free quenching solution and 5 mM Trolox for 1 min. Streptavidin-conjugated magnetic beads were washed twice with RIPA lysis buffer (150 mM NaCl, $1\%$ NP-40, $0.5\%$ sodium deoxycholate, $0.1\%$ SDS, 1 mM EDTA, 50 mM Tris, pH 8.0), and subjected into whole cell lysate for 2 h incubation. After washing with RIPA buffer four times, beads were boiled for 15 min and subjected into Western blot.
## Preparation of Mito-nanoparticle
The mito-nanoparticles were designed and synthesized as discribed23,57. Mitochondria-targeting peptide was synthesized by CHENPEPTIDE Biotechnology Co Ltd (Nanjing, China). PSiCoR (Cat# 12084, Addgene) was used for shRNA expression, and modified PCDNA4 plasmid was used for mcPGK1 overexpression. The sequence of the mcPGK1 shRNA and overexpression was confirmed by Sanger sequencing.
## Lentivirus generation and cell infection
pSiCoR was used for knockdown. Sequences of shRNAs targeting the junction sequence of mcPGK1 were cloned into pSiCoR vector (Cat no. 12084, Addgene). For lentivirus packaging, we transfected 293 T cells with pSiCoR and package plasmids (4 mg pSiCoR vector, 1 mg VSVG, 1 mg RRE and 2 mg RSV-REV were used for 10 cm dish). PLC, Huh7 and HCC primary cells were infected by virus supernatants or PEG5000 (Sigma)-enriched precipitates. mcPGK1 overexpressing cells were established similarly. shRNA sequences for PGK1, PDK1 and PDH used in this study were listed in Supplementary Table 1.
## RNA extraction and RT-PCR analyses
Total RNA samples were isolated with TRIzol method. 1 μg RNAs were reverse-transcribed into cDNA and then subjected to quantitative real-time PCR analysis with ABI QuantStudio5 Q5. Relative changes in expression levels were calculated. RT-qPCR primers are listed in Supplementary Table 2.
## RNA pulldown
Spheres were crushed with RIPA buffer supplemented with protease inhibitor cocktail and RNase inhibitor, and pre-cleared with streptavidin beads for 1 h. Biotin labeled RNA probes and cell lysis were mixed together in 4 °C for 3 h, and biotin-enriched components were separated and the binding proteins were detected with silver staining or immunoblot.
## Silver staining and mass spectrometry analysis
Pulldown samples from spheres by mcPGK1 probes and antisense probes were boiled for 15 min, separated through $15\%$ SDS-PAGE, and observed by sliver staining. The variant bands in mcPGK1 eluate were identified through mass spectrometry analysis (LTQ Orbitrap XL).
## Immunoblot
For immunoblot, samples were crushed and boiled in 1×SDS-loading buffer for 15 min, and then proteins were separated by electrophoresis. Proteins were then transferred to nitrate cellulose (NC) membrane, followed by detection with primary antibody and HRP-conjugated antibodies, finally the HRP signals were visualized by ultra-sensitive enhanced chemiluminescent (ECL) substrate58.
## Northern blot
Total RNA from CD133high, CD133low, sphere and non-sphere samples was extracted with standard TRIzol method, separated with electrophoresis and transferred to positively charged NC film (Beyotime Biotechnology), and then cross-linked by UV exposure. RNA samples on NC membranes were detected with digoxin-labeled RNA probes, which were generated through in vitro transcription. Finally RNA signals were detected with HRP-conjugated anti-digoxin antibody.
## RNA immunoprecipitation
Spheres were lyzed in RNase-free RIPA buffer supplemented with RNase inhibitor and protease-inhibitor cocktail, centrifuged and supernatants were collected for preclear with Protein A/G. PGK1 and control antibodies were mixed with Protein A/G, followed by 4 h incubation with precleared sphere lysates. Finally RNA samples in eluate were extracted and mcPGK1 enrichment was detected through quantitative real-time PCR.
## Signaling pathway activity reporter system
Wnt/β-catenin, Notch, Hedgehog, mTOR, NF-kB, P38, JNK, ERK and PKA reporter plasmids were overexpressed liver cancer cells, and treated with lactic acid or DM-αKG59. The activity levels of each signaling pathway were detected by FACS. For example, Wnt activity level = (TOP-GFP intensity)/(FOP-GFP intensity). The plasmids used in this assay are: TOP-GFP (addgene no. 35489), FOP-GFP (addgene no. 35490), 12XCSL-d1EGFP (addgene no. 47684), 7Gli:GFP (addgene no.110494), TORCAR (addgene no. 64927), NF-kB-eGFP (addgene no. 118093), TORCAR(T/A) (addgene no. 64928), p38KTRmCerulean3 (addgene no. 59155), JNKKTRmRuby2 (addgene no. 59154), PKAKTRClover (addgene no.59153), ERKKTRClover (addgene no. 59150).
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36651-5.
## Source data
Source Data
## Peer review information
Nature Communications thanks Shicheng Su and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
## References
1. Losic B. **Intratumoral heterogeneity and clonal evolution in liver cancer**. *Nat. Commun.* (2020.0) **11** 291. DOI: 10.1038/s41467-019-14050-z
2. Zhu PP, Fan ZS. **Cancer stem cell niches and targeted interventions**. *Prog. Biochem Biophys.* (2017.0) **44** 697-708
3. Chen ZZ. **LncSox4 promotes the self-renewal of liver tumour-initiating cells through Stat3-mediated Sox4 expression**. *Nat. Commun.* (2016.0) **7** 12598. DOI: 10.1038/ncomms12598
4. Zhu P. **ZIC2-dependent OCT4 activation drives self-renewal of human liver cancer stem cells**. *J. Clin. Invest* (2015.0) **125** 3795-3808. DOI: 10.1172/JCI81979
5. Miao Y. **Adaptive immune resistance emerges from tumor-initiating stem**. *Cells Cell* (2019.0) **177** 1172-1186.e1114. DOI: 10.1016/j.cell.2019.03.025
6. Perry JM. **Overcoming Wnt-beta-catenin dependent anticancer therapy resistance in leukaemia stem cells**. *Nat. Cell Biol.* (2020.0) **22** 689-700. DOI: 10.1038/s41556-020-0507-y
7. Chen ZZ. **LncFZD6 initiates Wnt/beta-catenin and liver TIC self-renewal through BRG1-mediated FZD6 transcriptional activation**. *Oncogene* (2018.0) **37** 3098-3112. DOI: 10.1038/s41388-018-0203-6
8. Chen Z, Yao L, Liu Y, Zhu P. **LncTIC1 interacts with beta-catenin to drive liver TIC self-renewal and liver tumorigenesis**. *Cancer Lett.* (2018.0) **430** 88-96. DOI: 10.1016/j.canlet.2018.05.023
9. Zhu P. **C8orf4 negatively regulates self-renewal of liver cancer stem cells via suppression of NOTCH2 signalling**. *Nat. Commun.* (2015.0) **6** 7122. DOI: 10.1038/ncomms8122
10. Wu JY. **The long non-coding RNA LncHDAC2 drives the self-renewal of liver cancer stem cells via activation of Hedgehog signaling**. *J. Hepatol.* (2019.0) **70** 918-929. DOI: 10.1016/j.jhep.2018.12.015
11. Zhu PP. **LncBRM initiates YAP1 signalling activation to drive self-renewal of liver cancer stem cells**. *Nat. Commun.* (2016.0) **7** 13608. DOI: 10.1038/ncomms13608
12. Li X, Yang L, Chen LL. **The biogenesis, functions, and challenges of circular RNAs**. *Mol. Cell* (2018.0) **71** 428-442. DOI: 10.1016/j.molcel.2018.06.034
13. Hansen TB. **Natural RNA circles function as efficient microRNA sponges**. *Nature* (2013.0) **495** 384-388. DOI: 10.1038/nature11993
14. Memczak S. **Circular RNAs are a large class of animal RNAs with regulatory potency**. *Nature* (2013.0) **495** 333-338. DOI: 10.1038/nature11928
15. Guarnerio J. **Oncogenic role of fusion-circRNAs derived from cancer-associated chromosomal translocations**. *Cell* (2016.0) **165** 289-302. DOI: 10.1016/j.cell.2016.03.020
16. Chen Z. **rtcisE2F promotes the self-renewal and metastasis of liver tumor-initiating cells via N(6)-methyladenosine-dependent E2F3/E2F6 mRNA stability**. *Sci. China Life Sci.* (2022.0) **65** 1840-1854. DOI: 10.1007/s11427-021-2038-5
17. Liu B. **An inducible circular RNA circKcnt2 inhibits ILC3 activation to facilitate colitis resolution**. *Nat. Commun.* (2020.0) **11** 4076. DOI: 10.1038/s41467-020-17944-5
18. Zhu P. **IL-13 secreted by ILC2s promotes the self-renewal of intestinal stem cells through circular RNA circPan3**. *Nat. Immunol.* (2019.0) **20** 183-194. DOI: 10.1038/s41590-018-0297-6
19. Chen Z. **circular RNA cia-MAF drives self-renewal and metastasis of liver tumor-initiating cells via transcription factor MAFF**. *J. Clin. Invest* (2021.0) **131** e148020. DOI: 10.1172/JCI148020
20. Chen Z. **Identification of cis-HOX-HOXC10 axis as a therapeutic target for colorectal tumor-initiating cells without APC mutations**. *Cell Rep.* (2021.0) **36** 109431. DOI: 10.1016/j.celrep.2021.109431
21. Chen ZZ. **circREEP3 drives colorectal cancer progression via activation of FKBP10 transcription and restriction of antitumor immunity**. *Adv. Sci.* (2022.0) **9** e2105160. DOI: 10.1002/advs.202105160
22. Liu X. **Identification of mecciRNAs and their roles in the mitochondrial entry of proteins**. *Sci. China Life Sci.* (2020.0) **63** 1429-1449. DOI: 10.1007/s11427-020-1631-9
23. Zhao QY. **Targeting mitochondria-located circRNA SCAR alleviates NASH via reducing mROS output**. *Cell* (2020.0) **183** 76-86. DOI: 10.1016/j.cell.2020.08.009
24. Cavalcante GC, Magalhaes L, Ribeiro-dos-Santos A, Vidal AF. **Mitochondrial epigenetics: non-coding RNAs as a novel layer of complexity**. *Int J. Mol. Sci.* (2020.0) **21** 1838. DOI: 10.3390/ijms21051838
25. Demishtein-Zohary K, Azem A. **The TIM23 mitochondrial protein import complex: function and dysfunction**. *Cell Tissue Res.* (2017.0) **367** 33-41. DOI: 10.1007/s00441-016-2486-7
26. Rapaport D, Neupert W. **Biogenesis of Tom40, core component of the TOM complex of mitochondria**. *J. Cell Biol.* (1999.0) **146** 321-331. DOI: 10.1083/jcb.146.2.321
27. Spinelli JB, Haigis MC. **The multifaceted contributions of mitochondria to cellular metabolism**. *Nat. Cell Biol.* (2018.0) **20** 745-754. DOI: 10.1038/s41556-018-0124-1
28. Li S, Xiong GJ, Huang N, Sheng ZH. **The cross-talk of energy sensing and mitochondrial anchoring sustains synaptic efficacy by maintaining presynaptic metabolism**. *Nat. Metab.* (2020.0) **2** 1077-1095. DOI: 10.1038/s42255-020-00289-0
29. Hao Z. **N(6)-deoxyadenosine methylation in mammalian mitochondrial DNA**. *Mol. Cell* (2020.0) **78** 382-395.e388. DOI: 10.1016/j.molcel.2020.02.018
30. Trevisan T. **Manipulation of mitochondria dynamics reveals separate roles for form and function in mitochondria distribution**. *Cell Rep.* (2018.0) **23** 1742-1753. DOI: 10.1016/j.celrep.2018.04.017
31. Sprenger HG, Langer T. **The good and the bad of mitochondrial breakups**. *Trends Cell Biol.* (2019.0) **29** 888-900. DOI: 10.1016/j.tcb.2019.08.003
32. DeBerardinis RJ, Chandel NS. **We need to talk about the Warburg effect**. *Nat. Metab.* (2020.0) **2** 127-129. DOI: 10.1038/s42255-020-0172-2
33. 33.Liu, J. et al. Skp2 dictates cell cycle-dependent metabolic oscillation between glycolysis and TCA cycle. Cell Res.31, 80–93 (2020).
34. Cascone T. **Increased tumor glycolysis characterizes immune resistance to adoptive T cell therapy**. *Cell Metab.* (2018.0) **27** 977-987.e974. DOI: 10.1016/j.cmet.2018.02.024
35. Zhu P. **lnc-beta-Catm elicits EZH2-dependent beta-catenin stabilization and sustains liver CSC self-renewal**. *Nat. Struct. Mol. Biol.* (2016.0) **23** 631-639. DOI: 10.1038/nsmb.3235
36. Zhang D. **Metabolic regulation of gene expression by histone lactylation**. *Nature* (2019.0) **574** 575-580. DOI: 10.1038/s41586-019-1678-1
37. Xiang Y. **JARID1B is a histone H3 lysine 4 demethylase up-regulated in prostate cancer**. *P NATL ACAD SCI USA* (2007.0) **104** 19226-19231. DOI: 10.1073/pnas.0700735104
38. Xu W. **Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of alpha-ketoglutarate-dependent dioxygenases**. *Cancer cell* (2011.0) **19** 17-30. DOI: 10.1016/j.ccr.2010.12.014
39. Nie H. **O-GlcNAcylation of PGK1 coordinates glycolysis and TCA cycle to promote tumor growth**. *Nat. Commun.* (2020.0) **11** 36. DOI: 10.1038/s41467-019-13601-8
40. Zhu P. **LncGata6 maintains stemness of intestinal stem cells and promotes intestinal tumorigenesis**. *Nat. Cell Biol.* (2018.0) **20** 1134-1144. DOI: 10.1038/s41556-018-0194-0
41. Liu B. **A cytoplasmic NF-kappaB interacting long noncoding RNA blocks IkappaB phosphorylation and suppresses breast cancer metastasis**. *Cancer Cell* (2015.0) **27** 370-381. DOI: 10.1016/j.ccell.2015.02.004
42. Hill K. **Tom40 forms the hydrophilic channel of the mitochondrial import pore for preproteins [see comment]**. *Nature* (1998.0) **395** 516-521. DOI: 10.1038/26780
43. Li XJ. **Mitochondria-translocated PGK1 functions as a protein kinase to coordinate glycolysis and the TCA cycle in tumorigenesis**. *Mol. Cell* (2016.0) **61** 705-719. DOI: 10.1016/j.molcel.2016.02.009
44. An YA. **Dysregulation of amyloid precursor protein impairs adipose tissue mitochondrial function and promotes obesity**. *Nat. Metab.* (2019.0) **1** 1243-1257. DOI: 10.1038/s42255-019-0149-1
45. 45.Gao, K. X. et al. Active RNA interference in mitochondria. Cell Res.31, 219–228 (2020).
46. Mok BY. **A bacterial cytidine deaminase toxin enables CRISPR-free mitochondrial base editing**. *Nature* (2020.0) **583** 631-636. DOI: 10.1038/s41586-020-2477-4
47. Xu XL. **Mitochondrial regulation in pluripotent stem cells**. *Cell Metab.* (2013.0) **18** 325-332. DOI: 10.1016/j.cmet.2013.06.005
48. Civenni G. **Epigenetic control of mitochondrial fission enables self-renewal of stem-like tumor cells in human prostate cancer**. *Cell Metab.* (2019.0) **30** 303-318.e306. DOI: 10.1016/j.cmet.2019.05.004
49. Chen CL. **NANOG metabolically reprograms tumor-initiating stem-like cells through tumorigenic changes in oxidative phosphorylation and fatty acid metabolism**. *Cell Metab.* (2016.0) **23** 206-219. DOI: 10.1016/j.cmet.2015.12.004
50. Kim HM. **Increased CD13 expression reduces reactive oxygen species, promoting survival of liver cancer stem cells via an epithelial-mesenchymal transition-like phenomenon**. *Ann. Surg. Oncol.* (2012.0) **19** S539-S548. DOI: 10.1245/s10434-011-2040-5
51. 51.Garcia-Heredia, J. M. & Carnero, A. Role of mitochondria in cancer stem cell resistance. Cells9, 1693 (2020).
52. Renner K. **Metabolic hallmarks of tumor and immune cells in the tumor microenvironment**. *Front. Immunol.* (2017.0) **8** 248. DOI: 10.3389/fimmu.2017.00248
53. Ganapathy-Kanniappan S. **Linking tumor glycolysis and immune evasion in cancer: emerging concepts and therapeutic opportunities**. *Biochem. Biophys. Acta Rev. Cancer* (2017.0) **1868** 212-220. DOI: 10.1016/j.bbcan.2017.04.002
54. Hu Y, Smyth GK. **ELDA: extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays**. *J. Immunol. Methods* (2009.0) **347** 70-78. DOI: 10.1016/j.jim.2009.06.008
55. Lam SS. **Directed evolution of APEX2 for electron microscopy and proximity labeling**. *Nat. Methods* (2015.0) **12** 51-54. DOI: 10.1038/nmeth.3179
56. Hung V. **Proteomic mapping of the human mitochondrial intermembrane space in live cells via ratiometric APEX tagging**. *Mol. Cell* (2014.0) **55** 332-341. DOI: 10.1016/j.molcel.2014.06.003
57. Su S, Cao S, Liu J, Xu X. **Mitochondria-specific delivery system for targeted regulation of mitochondrial gene expression**. *STAR Protoc.* (2021.0) **2** 100275. DOI: 10.1016/j.xpro.2020.100275
58. Zhu P. **Gut microbiota drives macrophage-dependent self-renewal of intestinal stem cells via niche enteric serotonergic neurons**. *Cell. Res.* (2022.0) **32** 555-569. DOI: 10.1038/s41422-022-00645-7
59. Zhu PP. **5-hydroxytryptamine produced by enteric serotonergic neurons initiates colorectal cancer stem cell self-renewal and tumorigenesis**. *Neuron* (2022.0) **110** 2268-2282. DOI: 10.1016/j.neuron.2022.04.024
|
---
title: 'Development of evaluation system for cerebral artery occlusion in emergency
medical services: noninvasive measurement and utilization of pulse waves'
authors:
- Takuma Shimada
- Kazumasa Matsubara
- Daisuke Koyama
- Mami Matsukawa
- Miho Ohsaki
- Yasuyo Kobayashi
- Kozue Saito
- Hiroshi Yamagami
journal: Scientific Reports
year: 2023
pmcid: PMC9971203
doi: 10.1038/s41598-023-30229-3
license: CC BY 4.0
---
# Development of evaluation system for cerebral artery occlusion in emergency medical services: noninvasive measurement and utilization of pulse waves
## Abstract
Rapid reperfusion therapy can reduce disability and death in patients with large vessel occlusion strokes (LVOS). It is crucial for emergency medical services to identify LVOS and transport patients directly to a comprehensive stroke center. Our ultimate goal is to develop a non-invasive, accurate, portable, inexpensive, and legally employable in vivo screening system for cerebral artery occlusion. As a first step towards this goal, we propose a method for detecting carotid artery occlusion using pulse wave measurements at the left and right carotid arteries, feature extraction from the pulse waves, and occlusion inference using these features. To meet all of these requirements, we use a piezoelectric sensor. We hypothesize that the difference in the left and right pulse waves caused by reflection is informative, as LVOS is typically caused by unilateral artery occlusion. Therefore, we extracted three features that only represented the physical effects of occlusion based on the difference. For inference, we considered that the logistic regression, a machine learning technique with no complex feature conversion, is a reasonable method for clarifying the contribution of each feature. We tested our hypothesis and conducted an experiment to evaluate the effectiveness and performance of the proposed method. The method achieved a diagnostic accuracy of 0.65, which is higher than the chance level of 0.43. The results indicate that the proposed method has potential for identifying carotid artery occlusions.
## Introduction
Cerebrovascular disease (CVD) encompasses all disorders that can lead to temporary or permanent impairment of brain function. CVD is primarily caused by ischemia or bleeding and is currently a major cause of morbidity and mortality worldwide. According to the World Health Organization, approximately 6 million people die annually from CVD1. In Japan, the number of such fatalities per year is approximately 110,0002. Ischemic strokes are typically caused by stenosis or occlusion of the cerebral arteries. Acute cerebral large vessel occlusion leads to large infarctions and can result in serious sequelae or death. Rapid reperfusion therapy, including intravenous thrombolysis and mechanical thrombectomy, can reduce disability in patients with large vessel occlusion strokes (LVOS)3–5. As early treatment is crucial for a successful outcome in LVOS patients, rapid transport to a comprehensive stroke center is essential. Therefore, it is necessary to develop a simple evaluation device that can easily and accurately identify LVOS in emergency medical services.
Currently, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasonography are the main methods for diagnosing cerebral artery occlusions6. The effectiveness of mobile stroke units, which are ambulances equipped with a CT scanner, has been reported in Germany and the United States, however, they are not widely available in Japan and other parts of the world7,8. Additionally, portable X-ray systems can also lead to issues with exposure to harmful radiation and high costs. Therefore, a safe, inexpensive, and compact screening method that is convenient for use by paramedics in emergency medical services is needed.
Considering the aforementioned aspects, we focus on simple pulse wave measurements to evaluate occlusion in the main artery of the anterior cerebral circulation in this study. The pulse wave is a temporal variation in the displacement of the skin surface caused by the pressure waves propagating in the artery. Thus, the pulse wave at the carotid artery is a superposition of incident and reflected waves9. The incident wave at the carotid artery is generated by a pressure wave and is referred to as the forward wave; it is caused by blood flow ejected from the heart. The reflected wave, referred to as the backward wave, is primarily generated by the reflection of the pressure wave from the vascular bed10,11. Acute ischemic stroke is generally caused by occlusion of a cerebral blood vessel by atherosclerosis or embolic thrombus. Such an occlusion also constitutes another reflector of the pulse waves, as shown in Fig. 1. These physical characteristics of pulse waves enable us to measure and analyze pulse waveforms and detect occlusion. Figure 1The concept of superposition of flow velocity and pressure waves near the stenosis. F and P imply the flow and pressure waves, respectively. The subscripts i, r, and t denote incident, reflected, and transmitted waves, respectively. A part of the transmitted wave reflects at the vascular bed in brain. If the actual pulse wavelength is very long, then, these waves generally superpose.
Measurement techniques that determine the pulse wave velocity (PWV) and cardio-ankle vascular index (CAVI) are widely used to evaluate arterial stiffness in the body. PWV is obtained from the time difference between two pulse waves measured at a known distance between two pressure sites in the arms and legs. As pulse waves propagate faster in stiff vessels, the measured value increases with the progression of atherosclerosis. However, PWV and CAVI devices measure pulse waves from pressure cuffs mounted on the human body and not those propagating to the brain. Another technique, the augmentation index (AI), is also used to evaluate arterial wall stiffness without occlusion. Similar to the method used to determine PWV, the AI is calculated using the observed pulse wave to extract factors from the peak values of the observed waves. However, pulse waveform analysis of AI does not incorporate various wave characteristics12,13.
In our previous study on the assessment of artery stiffness, we constructed a simple and inexpensive pulse-wave measurement device using a commercial piezoelectric sensor for ultrasonic ranging. This sensor could sense displacement at low frequencies, and in our previous study, we reported the in vivo pulse wave data measured in the carotid artery in healthy elderly individuals using this sensor14. We also compared the results obtained by pressure wave propagation in an artificial vascular model. The study indicated that pressure waves reflected at the vascular bed of the cerebral arteries could be perceived in the form of pulse waveforms at the carotid artery15. We envision that this device can be used to detect occlusions.
In this study, we focus on the reflection of the pulse wave at occlusion and propose a screening system to evaluate carotid artery occlusion using a simple pulse wave measurement technique. As occlusion rarely occurs simultaneously in bilateral cerebral arteries, we focus on the difference in pulse waveforms observed on both sides. To infer occlusion, we use a simple machine-learning technique with a few dynamic features extracted from pulse waveforms, which are only related to the reflection phenomenon at the occlusion. Here, other factors of the data are excluded to ensure the explainability and reliability of the outcome from a physical perspective. We then use a simple classifier for occlusion inference.
## Contribution of this work to pulse wave analysis
From the perspective of pulse wave analysis, readers of this paper may question the significance of this research. We would like to clarify the contribution of this work. Other studies have used pulse waves to assess cardiovascular diseases such as arterial stiffness16–18. To the best of our knowledge, these studies do not focus on inferring cerebral artery occlusion in emergencies. Our current study is the first attempt to address this issue using predictive machine learning.
Unlike conventional studies, we focus on the irregular, non-stationary change in the pulse wave caused by the superposition of waves reflected from an occlusion. Our idea is that the difference between left and right pulse waves contains crucial information about such changes and contributes to occlusion inference. The idea is simple yet novel, and it is embodied in the features we extract.
From a technical perspective, the classifier used in this study is the basic one, the logistic regression, not deep learning techniques. The reason for using such a classifier is to achieve both explainability and performance with limited data. The simplest classifier (in other words, a clear box) makes the effects of features explainable. It is often difficult to collect a large amount of data on occlusion patients, especially in emergencies. Furthermore, the position, size, and shape of occlusions vary greatly among patients, making repetitive measurements of similar symptoms challenging. Given these considerations, we selected a classifier that performs well even with limited data.
As a first attempt, this study makes a contribution to the inference of cerebral artery occlusion by demonstrating the fundamental effectiveness of the proposed method. Collecting more data and improving the classifier will be the focus of our next study.
## Ethical approval and availability of data and materials
All the experiments and data analyses in this study were performed in accordance with the Declaration of Helsinki. They were also approved by the medical ethics committees of Doshisha University (No. 18016, Mar. 5th, 2019) and Nara Medical University (No. 2111, Jan. 21st, 2019). Experiments and analyses were conducted properly following the guidelines set by the medical ethics committees, in addition to the guidelines set by our research group. The data generated/analyzed in this study are available from the corresponding author if the ethical committees of Doshisha university and Nara medical university allow the applicants to use the data. It is worth noting that none of the materials used in this study needed approval.
## Proposed method
Typically, LVOS of anterior cerebral circulation is caused by unilateral artery occlusion. Therefore, for effective occlusion detection, we proposed pulse wave measurement in both the left and right carotid arteries of a person. The proposed method comprises three elemental ideas and techniques, namely, pulse wave measurement, feature extraction with preprocessing, and occlusion inference by classification. The procedures of the proposed method are illustrated in Fig. 2. The information on subjects selected for in vivo studies as well as the methodology that was followed is detailed in subsequent sections. Figure 2Procedures for the analysis of the measured pulse waves.
## Selection of subjects
In vivo studies were conducted on healthy subjects consisting of 15 men and 15 women in the age group of 20s to 90s. The subjects had no history of cardiovascular disease and were not consuming medication for hypertension. Among the patients, 16 men and 7 women with ages in the range of 51 to 96 exhibited an occlusion of the main cerebral artery in the anterior circulation. Table 1 lists ages and characteristics of the subjects. The stage of the disease was acute in 16 patients and chronic in 7 of them. Occlusion was diagnosed by digital subtraction angiography (DSA), computed tomography angiography (CTA), and magnetic resonance angiography (MRA). It was observed that 21 patients had internal carotid artery occlusion (ICO).Table 1Information of subjects. Healthy (negative)Healthy (negative)Patients (positive)AgeSexAgeSexAgeSexOcclusion sideICO (neck)ICO (intra)MCO21Male54Female51MaleR1––21Female60Male58MaleR1––22Male61Male60MaleL1––22Male60Male63FemaleR–1–22Female62Male68MaleL1––22Female62Female70MaleL1––22Female63Male71MaleL1––23Male65Female72FemaleR1––23Female67Male75FemaleL1––23Female70Female78MaleR–1–23Female70Female80MaleR1––41Female75Male82MaleL1––48Female82Male82FemaleR1––50Male84Male82MaleR–1–52Male90Female82FemaleR1––––––83MaleR1––––––84MaleL1––––––84MaleL1––––––86FemaleL––1––––87MaleR–1–––––87MaleL––1––––89MaleL1––––––96FemaleR–1– Among them, 16 patients had extracranial ICO and 5 had intracranial occlusion, termed as ICO (neck) and ICO (intra) respectively. Furthermore, two patients had occlusion of the main trunk of the middle cerebral artery, termed as MCO. The point at which the common carotid artery branches into external and internal carotid arteries is termed as carotid bifurcation. An informed consent for the participation in the study was obtained from each participant.
## Pulse wave measurement
Occlusion has a significant effect on the characteristics of pulse waves at the carotid artery. However, these characteristics may also be influenced by several factors such as the condition of the patient. To understand the relationship between occlusion and the characteristics of pulse waves, other contributing factors were precluded from this study. For example, prior to commencing the measurements, all subjects avoided eating, exercising, and smoking for over 2 h. This was followed by resting in the supine position for 15 min in a quiet room at 25 °C. The cardiovascular function and vasomotor tone in the resting conditions were thus obtained19,20.
The schematic of the measurement condition is illustrated in Fig. 3. The pulse wave was measured at the skin surface by placing a piezoelectric ceramic transducer (MA40E7R, Murata Corp.) at the upper edge of the thyroid cartilage, the position where the strongest pulse wave could be sensed by a finger. We measured pulse waves in both the right and left common carotid arteries. The observed signal was amplified by 40 dB using a preamplifier (NF 5307) and was subsequently digitized using a 14-bit analog-to-digital converter (Keyence NR-500, NR-HA08, or using our prototype measurement system manufactured in collaboration with Proassist. Ltd.) with a sampling frequency of 1.0 kHz21. In accordance with the characteristics of the sensor and the circuit system, the measured pulse wave corresponded to the differential (velocity) waveform in the low-frequency range. Thereafter, an average of the observed waves was obtained, and the DC component was eliminated to obtain the averaged differential pulse waveforms. Differential pulse waves (not integrated pulse waves) were consistently used in the measurement, feature extraction, and classification of the proposed method. Hereinafter, we use the word "pulse wave" to indicate "differential pulse wave" for simplicity. Figure 3Pulse wave measurement.
## Feature extraction
The following text explains the preprocessing performed to acquire a single differential pulse wave using the raw time series data consisting of multiple waves. First, we selected 5–10 cycles from the stable portion of the differential pulse waves. In each cycle, a positive peak was followed by a negative peak. The heart-rate interval was calculated by measuring the time difference between the positive peaks. From the entire period, a section commencing 0.1 s prior to the positive peak was segmented from the raw time series and considered as a cycle. To reduce measurement noise, five such cycles were averaged and normalized, with the maximum amplitude of the positive peak set to one. This resulted in a single differential pulse wave for each subject. As mentioned earlier, occlusions in the main artery of the anterior circulation usually do not occur on both sides simultaneously. Therefore, we measured pulse waves in the left and right carotid arteries of patients with occlusion and healthy subjects. To infer occlusion, we extracted features that represent the difference between the left and right pulse waves. On the side with occlusion, blood flow is obstructed, causing the forward and reflected waves to overlap, resulting in the observed pulse wave. Based on this physical mechanism, we propose three dynamic features, MCC, DNP, and DPA. These features were extracted from differential pulse waves for better understanding of the dynamics.
## Maximum value of the cross-correlation function (MCC)
The first of our proposed features x1 is the maximum value of the cross-correlation function. It is formulated as in Eq. [ 1]. In this equation, l(n) and r(n) refer to the left and right differential pulse waves in which time n is discrete and ranges from 1 to N. C(τ) is the cross-correlation function parameterized by the shift time τ. The maximum of C(τ) is found through a search over the range τ. This is considered as the first feature, x1.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_{1} = \mathop {\max }\limits_{\tau } C(\tau) = \mathop {\max }\limits_{\tau } \frac{1}{N}\mathop \sum \limits_{$$n = 1$$}^{N} l(n)r(n + \tau)$$\end{document}x1=maxτC(τ)=maxτ1N∑$$n = 1$$Nl(n)r(n+τ)
## Left–right difference in the total number of small positive and negative peaks (DNP)
The second proposed feature x2 is the absolute value of the difference in the number of small peaks between the left averaged differential pulse wave and the right one. This feature is formulated as in Eq. [ 2]. The small peaks are ascertained between the first positive and last negative peaks. For instance, in the case of the left pulse wave, small negative and positive peaks are identified between the foremost positive and hindmost negative peaks. The condition for a peak is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{dl(t)}{dt}{}\right)_{t=n}=0$$\end{document}dl(t)dtt=$$n = 0$.$ The identified peaks are aggregated into the set Slp. for the left side. The same applies to Srp for the right side. # S denotes the cardinality of set S, which is the number of set members. The absolute value of the subtraction of #Srp from #Slp yields the second feature, x2.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{aligned} x_{2} & = \left| {\# S_{lp} - \# S_{rp} } \right| \\ & = abs\left(\# \left\{ {l\left(n \right) | \left({\frac{dl\left(t \right)}{{dt}}} \right)_{t = n} = 0} \right\} \right. \\ & \quad\left. - \# \left\{ {r\left(m \right) | \left({\frac{dr\left(t \right)}{{dt}}} \right)_{t = m} = 0} \right\}\right) \\ \end{aligned} $$\end{document}x2=#Slp-#Srp=abs#ln|dltdtt=$$n = 0$$-#rm|drtdtt=$m = 0$
## Left–right difference in the peak amplitude after the dicrotic notch (DPA)
A dicrotic notch is the secondary upstroke caused by the closure of the aortic valve. It appears under the effect of the existence of occlusion. The strong negative peak in a differential pulse wave indicates the upcoming dicrotic notch. Hence, we focused on the positive peak adjacent to the last negative peak. The third proposed feature x3 is the absolute value of the difference between the amplitudes of this peak on the left and right sides. x3 is formulated in Eq. [ 3]. Tnp denotes the time of the last negative peak. The point n is identified based on the zero gradient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\frac{dl\left(t \right)}{{dt}}\right)_{t = n} = 0$$\end{document}dltdtt=$$n = 0$$ and the maximum amplitude l(n) in the time segment ranged by Tnp. This point is regarded as the peak after the dicrotic notch in the left differential pulse wave. The points m and r(m) are determined similarly on the right side. The absolute value of the subtraction of r(m) from l(n) computes the third feature, x3.3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{aligned} x_{3} & = \left| {l(n) - r(m)} \right| \\ & = abs \left(l(n) | \left({\frac{dl(t)}{{dt}}} \right)_{t = n} = 0, \quad l(n) = \max l(i),\quad T_{np} \le i \right.\\ &\left. \quad - r(m) | \left({\frac{dr(t)}{{dt}}} \right)_{t = m} = 0, \quad r(m) = \max r(j),\quad T_{np} \le j \right) \\ \end{aligned} $$\end{document}x3=l(n)-r(m)=absl(n)|dl(t)dtt=$$n = 0$$,l(n)=maxl(i),Tnp≤i-r(m)|dr(t)dtt=$m = 0$,r(m)=maxr(j),Tnp≤j Figure 4 depicts the differential pulse waveform of a patient with occlusion. Here, the points A and B in the waveform illustrate the first positive and last negative peaks, respectively. The small positive and negative peaks between them are also indicated using arrows. The three dynamic features represent the different aspects of a pulse waveform. MCC focuses on the similarity of the entire waveform whereas DNP computes the internal reflection of pressure waves that occur when the heart ejects blood. DPA measures the blood flow velocity at the end of the diastole. Figure 4Differential pulse waves of patient (ICO neck, 82 years old, female). A: positive peak, B: negative peak. Between A and B, small peaks were observed.
## Occlusion inference
As an initial attempt to infer occlusion, we applied the logistic regression (LOGR) to our three dynamic features22. LOGR is a fundamental machine learning (ML) technique that is widely used for classification. Compared to advanced ML techniques, the inference result of LOGR is easy to understand as the original features are not transformed in a complex way. Due to its simplicity, LOGR does not require large datasets or complex model selection using a validation set.
*In* general, LOGR assumes a linear regression function for each class Ck, where $k = 1$, 2, …, K. By constraining the function values to a range between 0 to 1, LOGR estimates the conditional probability P(Ck|x) that an instance x belongs to class Ck. Then it classifies x to Ck, which has the highest probability among all the classes. The model is trained using a training set to maximize its objective function, which is the log-likelihood. Finally, the performance of the LOGR model is evaluated using a test set to ensure generalizability.
While the above is a multiclass classification task, the present study aims to infer the existence of occlusion, which is a binary class classification task. Therefore, we formulated LOGR with one class for the latter task. The input is a feature vector consisting of the three dynamic features for a subject, i.e., x = [x1, x2, x3]. A linear regression function for class C has the following two parameters to be optimized: the intercept b0 and weight vector consisting of the weight coefficients b = [b1, b2, b3] corresponding to x1, x2, and x3, respectively. By substituting this function into a sigmoid function, the conditional probability P (C = “Positive”|x) is estimated using Eq. [ 4]. This signifies that the subject will have an occlusion at a confidence level of P (C = “Positive”|x). Consequently, 1 − P (C = “Positive”|x) is the probability of “Negative, ” meaning that no occlusion exists at this level.4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(C = {\text{"}}Positive{\text{"}}|{\varvec{x}}) = \frac{1}{{1 + \exp \left(-({b_{0} + {\varvec{b}} \cdot {\varvec{x}}} \right))}}.$$\end{document}P(C="Positive"|x)=11+exp-(b0+b·x).
After training LOGR to optimize b0 and b, it can accurately estimate P(C = “Positive”|x). LOGR classifies the subject as “Positive” if P(C = “Positive”|x) is greater than a preset threshold value. In our experiments, the ratio of positive subjects was 23/(23 + 30) = 0.43 (refer to Table 1). This value of 0.43 is the chance level, achieved by assuming that all the subjects are positive, and serves as the baseline performance. To avoid excessive false positives, the threshold of P(C = “Positive”|x) is set to 0.50, which is higher than the baseline value and is appropriate for a binary classification task.
To ensure an unbiased performance estimation, we designed the estimation process as follows: the data points in the original dataset were randomly shuffled and then split into a training set and a test set. In the experiments discussed later, the sizes of the training and test sets were 40 and 13, respectively. A LOGR model was created for the training dataset, and its performance is estimated using the test dataset. This set of procedures was repeated under different data randomizations 20 times. The performance of occlusion inference and the weights of the features were estimated each time and the average performance was also calculated.
## Measured pulse waves
Figures 4 and 5 illustrate typical examples of differential pulse waveforms at the common carotid arteries of patients with occlusion and healthy subjects. Waveforms were normalized using their positive, and maximum peak values. For healthy subjects, the differential pulse waveforms on the left and right sides were similar. As mentioned above, pulse waveforms are affected by various factors such as standing/sitting position, blood pressure, and artery stiffness. These factors affect both left and right pulse waveforms but can be disregarded when comparing characteristic features of left and right waveforms. In contrast, in patients with occlusion, the left and right waveforms are often dissimilar because of occlusion. That brings the advantage of our proposal to utilize the left and right difference. Pulse waves measured at the carotid artery include reflected waves from the end of the cerebral arteries and vascular bed7–9. In the case of an occlusion in the cerebral artery, the wave reflected from the occlusion will be superposed at the carotid artery23. Shimada et al. reported that reflected waves from occlusion in smaller arteries in the brain may be observed at the carotid artery using an artificial cerebral model24. In the event of an occlusion close to the measurement site of the carotid artery, the reflected pressure wave may be observed clearly. This reflection is pronounced and appears earlier than that from the vascular bed. Taking the above knowledge and findings into account, the results for each feature are discussed in the following sections. Figure 5Differential pulse waves of healthy subject (54 years old, female). A: positive peak, and B: negative peak.
## MCC
In some cases, differences in the left- and right-side waveforms were clearly observed in patients with occlusions. Being the first proposed feature that represents the left and right differences, the MCC of each subject was estimated. The results are depicted in Fig. 6. The mean of MCC and their standard deviation (SD) were 0.83 and 0.21, respectively for the patient group with occlusion. Those were 0.92 and 0.05, respectively for the healthy group. Some patients with occlusion had values that were much smaller than the mean value. They were found to be afflicted with ICO (neck). The mean and SD for ICO (neck) were 0.78 and 0.22. The corresponding values for ICO (intra) and MCO patients were 0.94, 0.03, 0.96 and 0.04, respectively. Considering the average values, the cross-correlation was deemed as a good feature to select an ICO occlusion. As the ICO (neck) was closer to the measurement point, the reflection from the occlusion returned clearly. Additionally, owing to the MCC traversing at the entire pulse waveform, the shape of the positive and negative peaks was found to be an influencing factor. These peaks are represented as A and B in the waveforms shown in Fig. 4. It was perceived that this aspect could potentially lead to missing the effects of small variations in between these peaks. Figure 6MCC for each subject.
## DNP
To focus on the small change due to reflection, we calculated the difference in the number of small positive and negative peaks as the second proposed feature. As indicated by the arrows in Fig. 4, small peaks were observed only on the occlusion side. These resulted from repetitive reflections occurring between the occlusion site and the heart. The results for all the subjects are shown in Fig. 7. The DNP appeared larger in patients with occlusion than in all healthy subjects. Figure 7DNP for each subject.
## DPA
The third feature of our proposed method is the DPA. As expected, this difference was observed in the differential pulse waveforms of the majority of the patients. The peak amplitudes after the dicrotic notch of the averaged differential pulse waves were often larger on the occlusion side than on the unaffected side. The results for all subjects are shown in Fig. 8. Yasaka et al. reported that the end-diastolic velocity of the unaffected side was faster than that of the affected side in the ICO and MCO groups25,26. In the occlusion group, the average and SD of DPA were 0.10 and 0.08, respectively. In healthy subjects, the mean and SD were 0.07 and 0.06, respectively. The DPA of the patients with occlusion was slightly larger than that of the healthy subjects. We consider that the end-diastolic blood flow velocity influences the DPA. The peak amplitude after the dicrotic notch of the affected side was larger than that of the unaffected side. Yasaka reported that the blood flow velocity around the occlusion is small on the affected side25. The low velocity indicates reflection of pressure at the occlusion, which is consistent with our DPA data. Figure 8DPA for each subject.
## Accuracy of inference by logistic regression
We examined the independence of the three features and the results are presented in Table 2. To determine the independence, we calculated the correlation between each feature. As we focused on different parts of the differential pulse waveform, a weak correlation was observed between the features. The experimental results suggest that the left–right difference is caused by the influence of the reflection from the occlusion site. However, the pulse waves in both sides also include reflections from the vascular bed. These reflections may result in an unexpectedly weaker correlation between each feature and the existence of occlusion. Although each feature does not have a high correlation with the existence of occlusion, it is expected that the three features complement each other and contribute to occlusion inference as a whole. This is confirmed in the next paragraph. Table 2Correlations for each combination of the three dynamic features and the existence of occlusion for magnetic properties. MCCDNPDPACMCC–− 0.25− 0.43− 0.30DNP––0.540.44DPA–––0.33 By applying LOGR, we obtained the accuracies of the proposed method for a test dataset in trials with randomized data orders. Note that accuracy is the ratio of correct inferences for both positive and negative subjects to all subjects. The definition of accuracy is given in Eq. [ 5], where TP, FP, TN, and FN are the numbers of true positives, false positives, true negatives, and false negatives, respectively27. As shown in Fig. 9, the accuracy fluctuated depending on the randomization, so we calculated its mean and SD. We also obtained the inference performances for positive subjects, precision and recall, as it is important to correctly and completely detect existing occlusions. Precision is the ratio of correct inferences for positive subjects to subjects inferred to be positive, and is defined in Eq. [ 6]. Recall, defined in Eq. [ 7], is the ratio of those to subjects that actually have an occlusion. Table 3 summarizes the means and SDs of accuracy, precision, and recall.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Accuracy = \frac{TP + TN}{{TP + FP + TN + FN}}.$$\end{document}Accuracy=TP+TNTP+FP+TN+FN.6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Precision = \frac{TP}{{TP + FP}}.$$\end{document}Precision=TPTP+FP.7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Recall = \frac{TP}{{TP + FN}}.$$\end{document}Recall=TPTP+FN.Figure 9Accuracies of inference of a logistic regression model for a test dataset under different data randomizations. Table 3Performances of occlusion inference. The mean and standard deviation are provided for each of accuracy, precision, and recall. ** Indicates that the accuracy is statistically higher than the chance level at the significant level of 0.01.AccuracyPrecisionRecallMean0.65**0.650.53Standard deviation0.200.200.20 The mean of accuracy was 0.65, which was statistically higher than the chance level of 0.43 with $p \leq 0.01$, as determined by the t test. The mean precision was 0.65, and the mean recall was 0.53 as shown in Table 3. While the inference performance for negative subjects is not the primary focus, it must be considered in terms of clinical practicality. Therefore, we also calculated the mean and SD of the negative predictive value (NPV), which were 0.69 and 0.18 respectively. *In* general, there is a trade-off among precision, recall, and NPV. If one of them is extremely low, the others become extremely high, leading to a false high accuracy. In contrast, the proposed method achieved a balance of precision, recall, NPV, and accuracy. Overall, it was confirmed that the proposed method, including the three dynamic features and the classifier LOGR, worked well to some extent in inferring the existence of occlusion.
Let us focus on the importance of the three dynamic features MCC, DNP, and DPA. Individually, they did not have a strong correlation with the positive class C (see Table 2), but their combination was effective in occlusion inference. To identify the importance of each feature (in other words, how much it contributed to occlusion inference), we investigated the weights given by LOGR. Figure 10 plots the weights in the trials with different random data orders and Table 4 provides the mean and SD of weights for each of MCC, DNP, and DPA.Figure 10Weights on the three dynamic features, which were obtained via training a logistic regression model under different data randomizations. Table 4Final weights on the three dynamic features, which were obtained via training a logistic regression model under different data randomizations. MCCDNPDPAMean0.450.850.21Standard deviation0.400.280.19 The mean weight of DNP (0.85) was the largest and significantly larger than 0.00. It is worth noting that DNP is the left–right difference in the number of small peaks, counted in the area between the first positive and last negative peaks of the differential pulse waveform. The large weight of DNP suggests that small fluctuations in the waveforms due to reflections are sufficiently represented by DNP and are important for occlusion inference. The mean weight of MCC 0.45 was the second largest. MCC is the cross-correlation representing the left–right difference in the entire waveforms, suggesting that such a global feature is also informative about occlusion. The mean weight of DPA 0.21 was not so large, but still had some contribution to occlusion inference. DPA is the left–right difference in the peak amplitude after the dicrotic notch. In the present experiment, the high importance of DPA was not suggested.
Here is the summary of results and discussion. It was experimentally confirmed that the proposed method, consisting of pulse wave measurement, feature extraction, and classification, can infer the existence of cerebral artery occlusion. The proposed method provided an explainable way of understanding how certain features contribute to the inference. Although its inference performance was not perfect, it could achieve results even with a small amount of data.
In this study, we focused on the effects of the reflected wave at the occlusion on the pulse wave observed at the carotid artery, resulting in differences between the left and right pulse waveforms. In the future, it will be important to consider the effects of pulse waveform changes due to age; adding data of both healthy and patient individuals aged 40–90 years will improve the study.
Additionally, owing to the electrical nature of the measurement system, the current system measures the differential waves of the pulse wave. With more careful integration of observed data, future studies of the pulse wave may provide additional information such as wave amplitude, which depends on the reflection condition at the occlusion.
## Conclusion
Toward the ultimate goal of establishing an occlusion diagnosis support system in emergency medical services, we proposed a method that included a noninvasive measurement of both left and right pulse waves at the carotid artery using a piezoelectric sensor system, the three dynamic features extracted from these pulse waves, and the occlusion inference using these features. In the experiments, we measured both left and right carotid artery pulse waveforms of patients with occlusion as well as healthy subjects. We then extracted the cross-correlation between the left and right waveforms, the number of small positive and negative peaks, and the left and right differences in peak amplitude after dicrotic notch. By applying the logistic regression to these features, we inferred the existence of occlusion. Finally, the accuracy of occlusion inference was estimated as 0.65, which was higher than the chance level of 0.43. This study utilized the logistic regression, which is one of the basic machine learning methods, to comprehend the effect of each feature. As mentioned above, this study focused on three simple features which only resulted from the reflection phenomenon at the occlusion. In the forthcoming research, the accuracy of the occlusion inference will be improved by employing high performance machine learning methods, using other features of waveforms, and information of each data such as age, sex etc. To further enhance the accuracy, we plan to increase the number of subjects and identify the most suitable dynamic features for detecting occlusion. It is expected that our present system will work well not only for occlusion inference but also for screening other diseases with vascular deformation, such as aneurysms.
## References
1. 1.WHO Global Health Estimates. https://www.who.int/data/global-health-estimates (2020).
2. 2.“Vital statistics in Japan, Ministry of Health, Labor and Welfare”. https://www.mhlw.go.jp/english/database/db-hw/vs01.html (Ministry of Health, 2020).
3. Emberson J. **Effect of treatment delay, age, and stroke severity on the effects of intravenous thrombolysis with alteplase for acute ischaemic stroke: A meta-analysis of individual patient data from randomised trials**. *Lancet* (2014.0) **384** 1929-1935. DOI: 10.1016/S0140-6736(14)60584-5
4. Saver JL. **Time to treatment with endovascular thrombectomy and outcomes from ischemic stroke**. *JAMA* (2016.0) **316** 1279-1289. DOI: 10.1001/jama.2016.13647
5. Rajajee V, Saver J. **Prehospital care of the acute stroke patient**. *Tech. Vasc. Interv. Radiol.* (2005.0) **8** 74-80. DOI: 10.1053/j.tvir.2005.03.004
6. Ringelstein EB, Zeumer H, Angelou D. **The pathogenesis of strokes from internal carotid artery occlusion. Diagnostic and therapeutical implications**. *Stroke* (1983.0) **14** 867-875. DOI: 10.1161/01.STR.14.6.867
7. Ebinger M. **Association between dispatch of mobile stroke units and functional outcomes among patients with acute ischemic stroke in Berlin**. *JAMA* (2021.0) **325** 454-466. DOI: 10.1001/jama.2020.26345
8. Grotta JC. **Prospective, multicenter, controlled trial of mobile stroke units**. *N. Engl. J. Med.* (2021.0) **385** 971-981. DOI: 10.1056/NEJMoa2103879
9. Murgo JP, Westerhof N, Giolma JP, Altobelli SA. **Aortic input impedance in normal man: Relationship to pressure wave forms**. *Circulation* (1980.0) **63** 105-116. DOI: 10.1161/01.CIR.62.1.105
10. Murgo JP, Westerhof N, Giolma JP, Altobelli SA. **Manipulation of ascending aortic pressure and flow wave reflections with the Valsalva maneuver: Relation to input impedance**. *Circulation* (1981.0) **63** 122-132. DOI: 10.1161/01.CIR.63.1.122
11. Nichols WW, O’Rourke MF. *McDonald’s Blood Flow in Arteries, Chaps. 2, 3, 10, and 19* (2005.0)
12. Saiki A, Ohira M, Yamaguchi T, Nagayama D, Shimizu N, Shirai K, Tatsuno I. **New horizons of arterial stiffness developed using cardio ankle vascular index (CAVI)**. *J. Atheroscler. Thromb.* (2020.0) **27** 732-748. DOI: 10.5551/jat.RV17043
13. Nichols WW, Edwards DG. **Arterial elastance and wave reflection augmentation of systolic blood pressure: Deleterious effects and implications for therapy**. *J. Cardiovasc. Pharmacol. Ther.* (2001.0) **6** 5-21. DOI: 10.1177/107424840100600102
14. Saito M, Matsukawa M, Asada T, Watanabe Y. **Noninvasive assessment of arterial stiffness by pulse wave analysis**. *IEEE Trans. Ultrason. Ferroelectr. Freq. Control* (2012.0) **59** 2411-2419. DOI: 10.1109/TUFFC.2012.2473
15. Saito M. **One-dimensional model for propagation of a pressure wave in a model of the human arterial network: Comparison of theoretical and experimental results**. *J. Biomech. Eng.* (2011.0) **133** 121005. DOI: 10.1115/1.4005472
16. Li G. **Research on arterial stiffness status in type 2 diabetic patients based on pulse waveform characteristics**. *Comput. Model. Eng. Sci.* (2018.0) **117** 143-155
17. Li G. **Pulse-wave-pattern classification with a convolutional neural network**. *Sci. Rep.* (2019.0) **9** 14930. DOI: 10.1038/s41598-019-51334-2
18. Wang S. **A machine learning strategy for fast prediction of cardiac function based on peripheral pulse wave**. *Comput. Methods Programs Biomed.* (2022.0) **216** 106664. DOI: 10.1016/j.cmpb.2022.106664
19. Avolio AP. **Effects of aging on changing arterial compliance and left ventricular load in a Northern Chinese urban community**. *Circulation* (1983.0) **68** 50-58. DOI: 10.1161/01.CIR.68.1.50
20. Wilkinson IB. **The influence of heart rate on augmentation index and central arterial pressure in humans**. *J. Physiol.* (2000.0) **525** 263-270. DOI: 10.1111/j.1469-7793.2000.t01-1-00263.x
21. Saito M. **Simple analysis of the pulse wave for blood vessel evaluation**. *Jpn. J. Appl. Phys.* (2009.0) **48** 07GJ09. DOI: 10.1143/JJAP.48.07GJ09
22. Pregibon D. **Logistic regression diagnostics**. *Ann. Stat.* (1981.0) **9** 705-724. DOI: 10.1214/aos/1176345513
23. Yamamoto Y. **Experimental study on the pulse wave propagation in human artery model**. *Jpn. J. Appl. Phys.* (2011.0) **50** 07HF12. DOI: 10.1143/JJAP.50.07HF12
24. Shimada S. **Experimental study on the pressure wave propagation in the artificial arterial tree in brain**. *Jpn. J. Appl. Phys.* (2018.0) **57** 7S1. DOI: 10.7567/JJAP.57.07LC06
25. Yasaka M, Omae T, Tsuchiya T, Yamaguchi T. **Ultrasonic evaluation of the site of carotid axis occlusion in patients with acute cardioembolic stroke**. *Stroke* (1992.0) **23** 420-422. DOI: 10.1161/01.STR.23.3.420
26. **Carotid ultrasound examination**. *Neurosonology* (2006.0) **19** 49-69. DOI: 10.2301/neurosonology.19.49
27. Ohsaki M. **Confusion-matrix-based kernel logistic regression for imbalanced data class**. *IEEE Trans. Knowl. Data Eng.* (2017.0) **29** 1806-1819. DOI: 10.1109/TKDE.2017.2682249
|
---
title: Wnt pathway inhibitors are upregulated in XLH dental pulp cells in response
to odontogenic differentiation
authors:
- Elizabeth Guirado
- Cassandra Villani
- Adrienn Petho
- Yinghua Chen
- Mark Maienschein-Cline
- Zhengdeng Lei
- Nina Los
- Anne George
journal: International Journal of Oral Science
year: 2023
pmcid: PMC9971210
doi: 10.1038/s41368-022-00214-z
license: CC BY 4.0
---
# Wnt pathway inhibitors are upregulated in XLH dental pulp cells in response to odontogenic differentiation
## Abstract
X-linked hypophosphatemia (XLH) represents the most common form of familial hypophosphatemia. Although significant advances have been made in the treatment of bone pathology, patients undergoing therapy continue to experience significantly decreased oral health-related quality of life. The following study addresses this persistent oral disease by further investigating the effect of DMP1 expression on the differentiation of XLH dental pulp cells. Dental pulp cells were isolated from the third molars of XLH and healthy controls and stable transduction of full-length human DMP1 were achieved. RNA sequencing was performed to evaluate the genetic changes following the induction of odontogenic differentiation. RNAseq data shows the upregulation of inhibitors of the canonical Wnt pathway in XLH cells, while constitutive expression of full-length DMP1 in XLH cells reversed this effect during odontogenic differentiation. These results imply that inhibition of the canonical Wnt pathway may contribute to the pathophysiology of XLH and suggest a new therapeutic strategy for the management of oral disease.
## Introduction
X-linked hypophosphatemia (XLH) represents the most common form of familial hypophosphatemia occurring in 1–5 per 100,000 annual births.1–3 Defective dentin, cementum, and alveolar bone contribute to the disease’s significant morbidity.4–7 Dental pulp necrosis in the absence of trauma or caries remains a significant long-term side-effect in individuals receiving therapy, with prevalence as high as $75\%$ reported.6,8–10 These lesions present as spontaneous dental abscesses and can lead to more severe infections, tooth loss, occlusal disharmonies, and poor alveolar-dental development. The disorganized odontoblast cell layer and abnormal accumulation of non-collagenous extracellular matrix proteins in the XLH tooth suggest that defects in odontogenic differentiation may also be present in the disease.7,11,12 Odontoblast differentiation requires cell polarization and the formation of membrane domains and cell junctions that ensure the segregation and the unidirectional trafficking of molecules for mineralization.13 Canonical Wnt signaling is involved in tooth initiation and morphogenesis, correlating with odontoblast differentiation and dentin deposition.14–16 *Despite a* gradual decline in Wnt signaling with age, the conditional stabilization of beta-catenin in the adult pulp leads to dentin formation.17,18 The structural changes that accompany cytodifferentiation and tooth morphogenesis directly affect cell signaling and vice versa.19 E-cadherin is one component of adherens junctions necessary for palisade formation that is transcriptionally regulated by the Wnt pathway but also sequesters beta-catenin limiting its downstream Wnt pathway functions.20 The importance of the Wnt pathway in tooth development and regeneration has been well established; however, the status of Wnt signaling within the context of XLH remains unclear.21 Indication for deregulation of the Wnt pathway in XLH is implied from that seen in autosomal recessive hypophosphatemic rickets, a disorder phenotypically similar to XLH resulting from dentin matrix protein 1 (DMP1) loss-of-function.22 Expression of canonical Wnt pathway inhibitors, such as the secreted frizzled-related protein 4 (sFRP-4), have been reported in Dmp1 knockout mice.23 sFRP-4 has been associated with Wnt Family Member 5A (WNT5A) expression and noncanonical Wnt signaling pathway activity, as well as, activation of bone morphogenic protein (BMP) signaling and sclerostin (SOST) gene expression, contributing to decreased bone formation.24 Indeed, patients with XLH are reported to have higher concentrations of circulating sclerostin.25 Our group previously reported impaired matrix mineralization in XLH dental pulp cell cultures that were corrected by the constitutive expression of the full-length human DMP1 gene.26 The following study sought to identify the genetic pathways affected by the induction of odontogenic differentiation in XLH and XLH cells expressing DMP1 in an effort to explain how DMP1 contributed to enhanced matrix mineralization in our initial studies.
## Differentiation significantly upregulates inhibitors of the canonical Wnt pathway in XLH cells
Transcription profiles of XLH dental pulp cells cultured for eight hours in differentiation media were analyzed. ANOVA multi-group and multi-factor analyses revealed that disease status affected the expression of 3832 genes, while constitutive DMP1 expression affected the expression of 3205 genes (Fig. 1a). When compared to control (Ctrl) patients, XLH patients presented with significantly higher expression of sclerostin (SOST), WNT Inhibitory Factor 1 (WIF1), dickkopf 3 (DKK3) a Wnt signaling pathway inhibitor, and Wnt family members 5A and 16 (WNT5A and WNT16) (Fig. 1b).Fig. 1Genomic profiles of transgenic cells in response to differentiation. a ANOVA multi-group and multi-factor analysis were conducted on EdgeR to prioritize genes affected by disease and DMP1 status. Venn diagram represents transcripts with significant interaction and individual main effects combined (false discovery rate, FDR < 0.01). Disease status affected the expression of 3832 genes, while constitutive DMP1 expression affected the expression of 3205 genes. K-means clustering, gene ontology, and pathway analyses were performed to identify interesting biological processes affected by disease and DMP1 expression. b Volcano plots to present the distribution of differentially expressed genes. Dots in gray are those genes that did not meet the criteria of being significantly expressed with a twofold change or greater. Thresholds appear as red dashed lines on the y-axis for significance (FDR < 0.01), y-intercept at −Log10(FDR) = 2, and on the x-axis for fold-change (FC), x-intercepts at Log2(FC) = −1 and 1 (twofold decrease or increase, respectively). Dots in green denote downregulated genes, and dots in red denote upregulated genes K-means clustering ($k = 9$), gene ontology (GO), and pathway analyses were performed to identify interesting biological processes affected by disease and DMP1 expression. Cluster 4 genes were significantly associated with GO terms of interest in odontoblast differentiation, namely collagen fibril organization (GO: 0030199), positive regulation of the Wnt signaling pathway (GO: 0030177), and angiogenesis (GO:0001525). The heatmap representing cluster 4 genes highlights regions where DMP1 expression normalized gene expression (Supplementary Fig. 1).
## DMP1 reverses the expression of Wnt pathway inhibitors in XLH cells
A post hoc pairwise comparison of differentially expressed genes (DEGs) expressing at least twofold changes between Ctrl, XLH, and XLHDMP1 was conducted. Out of the 778 DEGs, 336 genes exhibited a reversal in expression pattern and have been highlighted in blue (e.g., genes significantly downregulated in XLH were now found to be upregulated in XLHDMP1) (Fig. 2). The top DEGs have been labeled with their corresponding names. WIF1 and SOST are among the highly expressed XLH genes whose expression declined upon DMP1 expression. Fig. 2Effect of DMP1 expression on XLH DPSCs. A post hoc pairwise comparison was conducted between the samples. First, significantly differentially expressed genes (DEGs)(FDR < 0.01) between Ctrl and XLH samples were identified. This list of genes was further restricted to those genes that were differentially expressed between XLHDMP1 and XLH samples. A total of 778 DEGs are plotted (red and blue dots). The dotted lines represent Log2(FC) = −1 and 1 threshold (twofold decrease or increase, respectively). In blue, 336 genes are highlighted which exhibited a reversal in expression pattern with DMP1 expression (e.g., in the upper left quadrant are genes significantly downregulated in XLH vs. Ctrl cells, that were found to be upregulated in XLHDMP1 vs. XLH cells) A total of 778 DEGs between XLH and Ctrl cells were uploaded to PANTHER for GO enrichment analysis. The chord diagram presents a subset of highly enriched GO Biological Processes, their constituent genes, and each gene’s corresponding expression pattern as log fold-change (Fig. 3). Among the enriched GO terms were those for collagen fibril organization (GO:0030199), osteoblast differentiation (GO:0001649), odontogenesis (GO:0042476), negative regulation of Wnt signaling pathway (GO:0030178), and regulation of angiogenesis (GO:0045765).Fig. 3Top DEGs in XLH and Corresponding GO Terms. The list of 778 DEGs between Ctrl, XLH, and XLHDMP1 samples was submitted through GO enrichment analysis in PANTHER. The complete GO biological process annotation data set was used, including both manually curated and electronic annotations (GO Ontology database 10.5281/zenodo.5228828 Released 2021-08-18). A total of 748 had uniquely mapped IDs (30 gene IDs were unmapped, 16 were redundant and counted only once). The Fisher’s Exact Test with FDR correction, significance threshold set to FDR < 0.05. A subset of highly enriched GO Biological Processes was identified and further analyzed. These GO processes contained 180 unique genes. Their membership to each GO term and their differential expression in XLH vs. Ctrl cells are presented in this chord diagram. Only GO processes with at least three members are included Real-time PCR was used to assess the expression pattern of the validated genes WNT5A, DKK3, and WNT16 in response to DMP1 expression (Fig. 4a, b). Gene expression was determined at 0, 4, 8, 12, 24, and 48-h timepoints. WNT5A and DKK3 gene expression increased significantly with time in XLH cells. DMP1 expression in XLHDMP1 cells resulted in a decrease in both markers to levels comparable to Ctrl cells. No significant differences were observed between Ctrl and XLH DKK3 levels at 0 h ($$P \leq 0.9998$$) or between Ctrl and XLHDMP1 WNT5A levels at 12 h ($$P \leq 0.0725$$). WNT16 was not consistently expressed by all cell types across timepoints and was undetectable in Ctrl cells at 4- and 48-h timepoints and in XLHDMP1 cells at the 0-h timepoint (Fig. 4b). WIF1 and SOST were undetectable at the 8-h time point using real-time PCR (data not shown). Further optimization of primers and PCR conditions is needed to validate these two important markers. Fig. 4DMP1 reverses the expression of Wnt pathway Inhibitors in XLH cells. a Real-time PCR validation of RNA-seq data using a second biological sample, 8-h timepoint induction. XLH cells had significantly ($P \leq 0.0001$) higher levels of WNT5A, DKK3, and WNT16 transcripts when compared to Ctrl cells. Fold-change in dCT values between XLH and Ctrl cells is presented. Two-way ANOVA, alpha = 0.05, with Sidak’s multiple comparison test. b Time series (0, 4, 8, 12, 24, and 48-h timepoints) of validated genes WNT5A, DDK3, and WNT16. Within the group, significance is denoted by asterisks of the corresponding color. The graph presents fold change (2−ΔΔCT) in gene expression; each timepoint was normalized to 0 h (except for XLHDMP1 WNT16, which was undetectable at 0 h and was normalized to 4 h). WNT16 was not consistently expressed by all cell types across timepoints, therefore we were unable to report statistical significance. Two-way ANOVA, alpha = 0.05, with Tukey’s multiple comparisons. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$ c, d Representative Western blot showing analysis of E-cadherin and Beta-Catenin expression in response to differentiation, normalized to beta-actin loading control. c Under standard growth conditions (No Min, black bars), beta-catenin protein levels were higher in CtrlDMP1 and CtrlGFP cells than in XLH and XLHDMP1 cells. Beta-catenin protein levels decreased with the induction of differentiation (Min, pink bars) in both Ctrl and CtrlDMP1 cells, but increased in XLH and XLHDMP1 cells. Interaction plots representing the RNAseq multigroup analysis. The multigroup analysis revealed that beta-catenin expression differences between Ctrl and XLH cells (main effect $Q = 1.23$E−05) depended on DMP1 status (interaction effect $Q = 1.79$E−03). Under Min conditions, Ctrl cells expressed lower transcript counts than XLH cells. This pattern was also observed in protein expression. DMP1 transduction resulted in greater beta-catenin transcript levels in XLH cells and decreases in Ctrl cells. This pattern was not observed in CtrlDMP1 and XLHDMP1 protein levels. d Under standard growth conditions (black bars), E-cadherin protein levels were highest in Ctrl cells. E-cadherin protein levels increased with the induction of differentiation (Min, pink bars) in both Ctrl and CtrlDMP1 cells, but remained absent or decreased in the remaining cell types. Interaction plots representing the RNAseq multigroup analysis. The RNAseq multigroup analysis revealed significant individual main effects for *Disease status* ($Q = 8.54$E−03), not dependent on DMP1 status or Phosphate source. We observe higher transcript counts in Ctrl cells when compared to XLH cells. E-cadherin gene expression did not differ between CtrlDMP1 and XLHDMP1 cells, although protein levels were lower in XLHDMP1 cells. CPM, counts per million, in log2-scale, with a pseudo-count added to prevent taking the log of 0. Negative numbers indicate lower expression. Min, mineralization/differentiation conditions. No Min, standard growth conditions. Western blots for the second set of experiments can be found in Supplementary Materials
## Inhibition of E-cadherin and activation of beta-catenin in response to XLH differentiation
E-cadherin is one component of adherens junctions necessary for palisade formation that is transcriptionally regulated by the Wnt pathway but also sequesters beta-catenin limiting its downstream Wnt pathway functions.20 Beta-catenin protein levels decreased with the induction of differentiation (Min, pink bars) in both Ctrl and CtrlDMP1 cells but increased in XLH and XLHDMP1 cells (Fig. 4c). E-cadherin protein levels increased with the induction of differentiation (Min, pink bars) in both Ctrl and CtrlDMP1 cells but remained absent or decreased in the remaining cell types (Fig. 4d). Under standard growth conditions (No Min, black bars), protein levels were highest in Ctrl cells, and higher in CtrlDMP1 and CtrlGFP cells than in XLH and XLHDMP1 cells. Corresponding interaction plots from the RNA-seq multigroup analysis revealed that beta-catenin expression differences between Ctrl and XLH cells (main effect $Q = 1.23$E−05) depended on DMP1 status (interaction effect $Q = 1.79$E−03). Under odontogenic differentiation culture medium conditions, Ctrl cells expressed lower beta-catenin transcript counts than XLH cells. This pattern was also observed with the protein expression of beta-catenin. DMP1 expression resulted in greater beta-catenin transcript levels in XLH cells and decreases in Ctrl cells. This pattern was not observed in CtrlDMP1 and XLHDMP1 protein levels.
## Discussion
Transcriptomic analysis of XLH dental pulp cells has not been previously reported. The following study proposes a mechanism by which dentin formation and mineralization are affected in XLH individuals. That is, a defect in the Wnt signaling pathway responsible for odontogenic differentiation is present in the disease. XLH is an inherited metabolic disorder of fibroblast growth factor 23 (FGF23) excess that creates an antagonistic environment to bone formation. Such an environment would reasonably result in Wnt signaling pathway suppression, as this pathway is intractably associated with bone formation.27 Despite an extremely limited sample size and a lack of age-, sex-matching available, the similarities found in the Wnt profiles of these patients suggest disruptions independent of these parameters. Complete penetrance of the genotype without differences between males and females may explain this observation.28 Validation of the RNA sequencing data in the second XLH patient suggests that further study should follow to understand the effects of Phex dysfunction on the Wnt pathway Table 1.Table 1Real-time PCR primers for RNA sequencing validationTargetAccession numberForward primer sequenceReverse primer sequenceGAPDHNM_002046.7ATCCCATCACCATCTTCCAGGAGTCCTTCCACGATACCAAACTBNM_001101AAACTGGAACGGTGAAGGTGAGAGAAGTGGGGTGGCTTTTWNT5ANM_003392GCCAGTATCAATTCCGACATCGTCACCGCGTATGTGAAGGCDKK3NM_013253ATGTGTGCAAGCCGACCTTCCTCAGCGCCATCTCTTCAWNT16NM_016087GCAGAGAATGCAACCGTACATCACATGGGTGTTGTAACCTCG We showed that XLH pulp cells upregulate inhibitors of the canonical Wnt pathway in response to the induction of odontogenic differentiation. *These* genes included SOST, WIF1, WNT16, WNT5A, and DKK3, the latter three of which have been validated (Table 2). Time course experiments revealed that WNT16, WNT5A, and DKK3 expression was highest in XLH cells, peaking at 24-h (Fig. 4b). Despite this 24-h peak, which is also seen in Ctrl cells, it is important to note that sufficiently detectable differences in expression levels were observed at baseline and with DMP1 expression in XLH cells. DMP1 was able to suppress the transcription of these genes up until the 48-h timepoint, at which point expression returned to XLH levels. The return to baseline in XLHDMP1 cells may offer an explanation for the failure to rescue the XLH phenotype in vivo using DMP1.26 Future experiments should assess the time-dependent expression of these proteins relative to their unique roles in odontoblast differentiation. Despite increases in WNT5A, WNT16, and DDK3, the accumulation of beta-catenin in XLH cells in response to induction may suggest either faulty inhibition or communication between established pathways leading to canonical pathway activation (Fig. 4c). Future experiments should differentiate between nuclear and cytoplasmic, active and inactive, beta-catenin to better understand what was observed in XLH cells since only nuclear beta-catenin can mediate transcription. Table 2Negative regulators of the canonical Wnt signaling pathwayGene nameXLH/Ctrl: logFCaXLH/Ctrl: Q valuebXLHDMP1/XLH: logFCaXLHDMP1/XLH: Q valuebSOST5.672.07E−65−5.022.19E−60WIF14.363.79E−04−6.694.13E−05WNTl62.664.22E−1210.334.01E−04WNT5A0.675.04E−53−0.48l.32E−12DKK31.091.18E−1660.415.30E−10aLog2 Fold-change (e.g., 0 = no change, 2 = 4-fold increase, −2 = 4-fold decrease, etc). To reverse the order of the comparison, reverse the sign (+2 becomes −2; e.g., logFC is calculated as Disease/Control, but you want to see Control/Disease)bCorrected P-value (i.e., false discovery rate) Alternatively, a positive correlation between WNT5A activity, Notch signaling, and dental pulp stem cell differentiation suggests that other pathways may interconnect and play equally important roles.29 We have previously shown that calcium-binding proteins, such as DMP1, can activate the serine-threonine Ca2+/calmodulin-dependent protein kinase II (CaMKII) and mediate odontoblast differentiation.30–32 WNT5A has also been linked to Notch signaling activation via CaMKII activity.33 Calcium ion homeostasis, another putative biological process involved in XLH pathology (Fig. 3), along with its role in non-canonical Wnt signaling and pathways such as Notch signaling, must be considered in future studies.
Previous reports of small interfering RNA (siRNAs) silencing of the PHEX gene have revealed a subsequent downregulation of the Wnt pathway upon WNT3A stimulation. Furthermore, genome-wide RNA interference (RNAi) screens for Wnt/beta-catenin pathway components identified PHEX as a positive regulator of this pathway.34,35 Canonical Wnt signaling is important for the survival of undifferentiated dental pulp cells and promotes odontoblast differentiation and mineral formation, in vitro.36 Disruption of canonical Wnt signaling results in defects in dentin apposition, root and molar cusp development, and even tooth agenesis.37,38 WNT5A antagonizes canonical Wnt/beta‐catenin signaling and stimulates non-canonical WNT siganling.39,40 Elevated levels of other canonical Wnt pathway inhibitors, namely sclerostin (SOST), have been identified in XLH patients.41 Immunotherapies neutralizing sclerostin activity have, in fact, proven successful in improving bone mass, formation rate, and strength in Hyp mice.42,43 The effects of suppressing these Wnt signaling inhibitors in the tooth may also prove a useful model for understanding the pathophysiology of XLH.
Despite reports of downregulation of canonical Wnt pathway inhibitors, DKK1, sFRP-2, sFRP-4, and WIF1, during osteoblastic differentiation, absolute depletion of sFRP2 has been associated with the inhibition of odontogenic differentiation in mesenchymal stem cells.44,45 The upregulation of sFRP-2 was observed during odontogenic differentiation of stem cells of the apical papilla, resulting in increased DMP1 gene expression among other markers of differentiation.45 In fact, studies in periodontal ligament stem cells have shown that inhibition of Wnt signaling is required for the maintenance of the osteogenic potential of these cells.46 Meanwhile, increased Wnt signaling, such as in klotho-deficient mice, results in accelerated cellular senescence.47 Furthermore, constitutive activation of Wnt signaling, such as in NOTUM knockout mice, manifests as dentin dysplasia, periodontal inflammation, and periapical abscess formation.6,8–10 These studies highlight the need for further investigation of the temporal regulation of these pathways. The significance of our observations would likely lie within the context of temporal regulation.
Cadherins play a role in the cell–cell junctions of epithelial cells. In vivo studies have shown that differentiating odontoblasts express high levels of N-cadherin and no E-cadherin, while functional odontoblasts express low levels of E-cadherin and high levels of N-cadherin.48 In vitro studies, on the other hand, have shown that induction with 10 mM beta-glycerophosphate results in a gradual increase in E-cadherin and a gradual decrease in N-cadherin.49 Our Ctrl cells corroborate the latter findings better than the former.
E-cadherin is associated with the polarized epithelial phenotype. E-cadherin protein levels were highest in Ctrl cells under standard conditions (No Min) and increased further with the induction of differentiation in both Ctrl and CtrlDMP1 cells. By contrast, a decrease in E-cadherin was observed in XLH and XLHDMP1 cells with the induction of differentiation. This pattern was observed in transcript numbers, as well. Through RNA-seq, we find that N-cadherin (CDH2) is also downregulated in XLH cells. Given the poorly polarized, disorganized odontoblast layer in XLH teeth, it is possible that the observed reduction in E-cadherin may be affecting XLH cell odontoblast layer formation.7 While E-cadherin decreases in XLH and XLHDMP1 cells with differentiation, beta-catenin protein levels increased in XLH and XLHDMP1 cells when compared to standard growth conditions. By contrast, beta-catenin protein levels decreased after induction in Ctrl and CtrlDMP1 cells, concurrent with the observed increase in E-Cadherin protein. These changes could have downstream effects on cell attachment, Wnt signaling, and cell differentiation.
Induction of odontogenic differentiation resulted in the upregulation of inhibitors of the canonical Wnt pathway in XLH cells, while constitutive expression of full-length DMP1 in XLH cells reversed this effect (Fig. 5). The question that arises is that of DMP1’s role in restoring Wnt signaling in these cells. The answer to this question may be more challenging than we would like. The markers implicated in this disease, namely FGF23, Vitamin D, parathyroid hormone, and the sodium–phosphate co-transporters, are all part of a bigger network for calcium and phosphate metabolism. Defining PHEX function and its interaction with DMP1 would thus require a thorough understanding of the physiology of mineral metabolism and its relationship to Wnt signaling. Fig. 5Constitutive expression of DMP1 promotes canonical Wnt signaling. XLH dental pulp cells exhibited impaired differentiation due to the upregulation of inhibitors of the canonical Wnt pathway, such as WNT5a, WNT16, WIF1, and SOST. Constitutive expression of full-length DMP1 (fl-DMP1) resulted in the downregulation of these Wnt inhibitors, restoring differentiation potential. Constitutive DMP1 expression in XLH dental pulp cells resulted in improved mineralization. BMP1 bone morphogenetic protein 1, DKK3 Dickkopf-related protein 3, MMP3 matrix metalloproteinase 3, WIF1 Wnt inhibitory factor 1, SOST sclerostin. Created with BioRender.com
## Cell culture
Dental pulp cells were isolated from the third molars of XLH and healthy controls ($$n = 2$$ per genotype) and stable transduction of full-length human DMP1 gene was achieved, as previously described, producing control (Ctrl) and XLH cells overexpressing DMP1 (CtrlDMP1 and XLHDMP1).26 Empty vectors were transduced as controls, producing CtrlGFP cells. Cells (under seven passages) were plated at a density of 3.125 × 104 cells per cm2 and cultured in odontogenic differentiation media (Dulbecco’s Modified Eagle Medium 1 g·L−1 d-Glucose (DMEM; Invitrogen, Grand Island, NY, USA) supplemented with $10\%$ fetal bovine serum (Invitrogen), $1\%$ antibiotic–antimycotic 100× (Gibco/Invitrogen, Cat. 15240062), ascorbic acid (0.50 mmol·L−1), β-glycerophosphate (10 mM), and dexamethasone (10 nmol·L−1)) for 8 h, 37 °C, $5\%$ CO2. Conditions were repeated in duplicates. After 8 h, RNA was isolated with the miRNeasy Mini Kit (Cat. No. 217004). No DNase treatment was performed. One microgram of RNA was submitted to the RNA-sequencing core facility. Real-time PCR validation of RNA-sequencing data was performed using a second patient sample. Eight hours was the earliest timepoint at which gene expression changes occurred, per our preliminary studies. Where applicable, a series of collection timepoints were used to evaluate changes in gene expression over time (Fig. 6).Fig. 6Experimental design. Dental pulp cells were isolated from the third molars of XLH and healthy controls. The calcium phosphate transfection method was used to transfect full-length human DMP1 cDNA into low-passage 293FT cells using a lentivirus plasmid (pLenti-DMP1-GFP-2A-Puro), together with the psPAX2 (Addgene), pMD2.G (Addgene), and pHPV17 plasmids. Stable transduction of the full-length human DMP1 gene was achieved by producing control (Ctrl) and XLH cells overexpressing DMP1 (CtrlDMP1 and XLHDMP1)(Guirado et al., 2020). Odontogenic differentiation of the cells was performed, and RNA was isolated at 4, 8, 12, 24, and 48 h of culture. Eight-hour samples were chosen for RNA sequencing as this was the earliest time point at which gene expression changes were observed. Protein was isolated at the 48-h timepoint
## Protein isolation and Western blot analyses
Cell pellets were resuspended in 500 μl RIPA buffer (10× RIPA buffer with protease inhibitors). Lysates were incubated on a shaker for one hour at 4 °C, after which they were centrifuged for 30 min at 19 467 × g to remove cell debris. Supernatant protein concentration was assessed using the Bradford assay, 20 μg total protein was resolved using a $10\%$ SDS-PAGE gel at 180 V for 50 min and then transferred onto PVDF membranes at 22 V overnight. Membranes were blocked in $5\%$ dried milk in phosphate-buffered saline (PBS). Primary antibodies against beta-catenin (Sigma-Aldrich No. 04-958; dilution 1:1 000) and E-cadherin (Santa Cruz No. sc-8426; dilution 1:200), and HRP-conjugated secondary antibodies were resuspended at the appropriate concentrations in $5\%$ dried milk in PBS. Western blot was developed using Pierce enhanced chemiluminescence (ECL) Plus western blotting substrate (ThermoFisher, Cat. No. 32106). Western blot analysis was performed on ImageJ.50 Scanned western blot film images were uploaded to ImageJ, image type was changed to 8-bit to allow for light background subtraction. Lanes were plotted using the Gel Analysis Tool, and the area under the curve was calculated. Standardization of each lane was done accordingly to their corresponding loading control (β-actin). The second experiment can be found in supplementary materials.
## RNA sequencing quality control and quantification
RNA sequencing was conducted at the UIC Research Resources Center (GEO accession GSE201313). RNA integrity was assessed using Agilent TapeStation 4200 (all samples had RIN scores above nine). Library construction was based on Universal Plus mRNA-seq chemistry by NuGEN. Sequencing was performed on the NovaSeq 6000 instrument with SP flow cell (2 × 50 reads), 380+ million reads per lane, and approximately 23 million clusters/sample.
Raw sequencing reads were aligned to the human reference genome (HG38) using the STAR aligner and ENSEMBL gene and transcript annotations.51 Gene expression levels were quantified using FeatureCounts52 as raw read counts and as normalized reads-per-million. Normalized expression (in counts per million) accounts for differences in sequencing depth across libraries, allowing expression levels to be directly compared between samples.
Quality control was performed to confirm the depth and quality of the raw sequencing data and the absence of sequencing artifacts and to confirm that the number of reads aligning to the reference genome mapping to coding sequences was sufficient for expression estimates. Prior to differential expression analysis, principal component analysis (PCA) was performed to identify biological outliers that should be removed or further investigated. PCA plots and RNA integrity information can be found in Supplementary Materials.
## Bioinformatics analysis
Additional normalization with TMM (trimmed mean of M-values) scaling was performed in edgeR. TMM normalization is more robust to outlier features and seeks to ensure that the average log-fold change across samples is 0. Pseudo-counts were added to prevent taking the log of 0. Negative numbers simply indicate lower expression. Differential expression statistics (fold-change and p-value) were computed from raw expression counts using edgeR.53,54 Multi-group and multi-factor analyses and post-hoc pairwise analyses were performed. The false discovery rate (FDR) correction of Benjamini and Hochberg was used to correct for multiple comparisons.55 *Significant* genes were determined based on an FDR threshold of $5\%$ (0.05) in the multi-group comparison.
## GO analysis
GO enrichment analysis was conducted on PANTHER.56 Analysis type utilized PANTHER Overrepresentation Test (Released 20210224). The complete GO biological process annotation data set was used, including both manually curated and electronic annotations (GO Ontology database 10.5281/zenodo.5228828 Released 2021-08-18). All Homo sapiens genes in the database were used as the reference list.
The Fisher’s exact test with FDR correction (FDR-adjusted P-value < 0.05) was used to identify the top three significantly enriched GO biological processes. Fold enrichment is presented as the number of genes in the cluster divided by the expected number of genes based on the reference list. Fold enrichment greater than one indicates that the GO term is overrepresented in the cluster.
## Pathway analyses
Qiagen Ingenuity Pathway Analysis software was utilized.57 Pairwise comparisons were matched to the Ingenuity Pathway Analysis library of canonical pathways. A Fisher’s Exact test (alpha = 0.01) was performed, generating a −log(P-value), and a cutoff of 2 was chosen. DEGs from our data (FDR < 0.01) that were associated with a canonical pathway in the Ingenuity Knowledge Base were considered for the analysis.
## Supplementary information
RNA Integrity Principal Component Analysis Supplementary Figure 1 The online version contains supplementary material available at 10.1038/s41368-022-00214-z.
## References
1. Beck-Nielsen SS, Brock-Jacobsen B, Gram J, Brixen K, Jensen TK. **Incidence and prevalence of nutritional and hereditary rickets in southern Denmark**. *Eur. J. Endocrinol.* (2009.0) **160** 491-497. DOI: 10.1530/EJE-08-0818
2. Endo I. **Nationwide survey of fibroblast growth factor 23 (FGF23)-related hypophosphatemic diseases in Japan: prevalence, biochemical data and treatment**. *Endocr. J.* (2015.0) **62** 811-816. DOI: 10.1507/endocrj.EJ15-0275
3. Rafaelsen S, Johansson S, Ræder H, Bjerknes R. **Hereditary hypophosphatemia in Norway: a retrospective population-based study of genotypes, phenotypes, and treatment complications**. *Eur. J. Endocrinol.* (2016.0) **174** 125-136. DOI: 10.1530/EJE-15-0515
4. Foster BL. **Rare bone diseases and their dental, oral, and craniofacial manifestations**. *J. Dent. Res.* (2014.0) **93** 7S-19S. DOI: 10.1177/0022034514529150
5. Foster BL, Nociti FH, Somerman MJ. **The rachitic tooth**. *Endocr. Rev.* (2014.0) **35** 1-34. DOI: 10.1210/er.2013-1009
6. Hanisch M, Bohner L, Sabandal MMI, Kleinheinz J, Jung S. **Oral symptoms and oral health-related quality of life of individuals with x-linked hypophosphatemia**. *Head. Face Med.* (2019.0) **15** 8. DOI: 10.1186/s13005-019-0192-x
7. Salmon B. **Abnormal osteopontin and matrix extracellular phosphoglycoprotein localization, and odontoblast differentiation, in X-linked hypophosphatemic teeth**. *Connect. Tissue Res.* (2014.0) **55** 79-82. DOI: 10.3109/03008207.2014.923864
8. Baroncelli GI. **Prevalence and pathogenesis of dental and periodontal lesions in children with X-linked hypophosphatemic rickets**. *Eur. J. Paediatr. Dent.* (2006.0) **7** 61-66. PMID: 16842025
9. Baroncelli GI. **Pulp chamber features, prevalence of abscesses, disease severity, and PHEX mutation in X-linked hypophosphatemic rickets**. *J. Bone Miner. Metab.* (2021.0) **39** 212-223. DOI: 10.1007/s00774-020-01136-8
10. Lo SH, Lachmann R, Williams A, Piglowska N, Lloyd AJ. **Exploring the burden of X-linked hypophosphatemia: a European multi-country qualitative study**. *Qual. Life Res.* (2020.0) **29** 1883-1893. DOI: 10.1007/s11136-020-02465-x
11. Salmon B. **MEPE-derived ASARM peptide inhibits odontogenic differentiation of dental pulp stem cells and impairs mineralization in tooth models of X-linked hypophosphatemia**. *PLoS ONE* (2013.0) **8** e56749. DOI: 10.1371/journal.pone.0056749
12. Coyac BR. **Tissue-specific mineralization defects in the periodontium of the Hyp mouse model of X-linked hypophosphatemia**. *Bone* (2017.0) **103** 334-346. DOI: 10.1016/j.bone.2017.07.026
13. Hoshino M. **Claudin rather than occludin is essential for differentiation in rat incisor odontoblasts**. *Oral. Dis.* (2008.0) **14** 606-612. DOI: 10.1111/j.1601-0825.2007.01427.x
14. 14.Järvinen, E., Shimomura-Kuroki, J., Balic, A., Jussila, M. & Thesleff, I. Mesenchymal Wnt/β-catenin signaling limits tooth number. Development145, dev158048 (2018).
15. Zhao Y, Yuan X, Bellido T, Helms JA. **A correlation between Wnt/Beta-catenin signaling and the rate of dentin secretion**. *J. Endod.* (2019.0) **45** 1357-1364.e1. DOI: 10.1016/j.joen.2019.07.014
16. Zhu X. **Intra-epithelial requirement of canonical Wnt signaling for tooth morphogenesis ***. *J. Biol. Chem.* (2013.0) **288** 12080-12089. DOI: 10.1074/jbc.M113.462473
17. Kim T-H. **Constitutive stabilization of ß-catenin in the dental mesenchyme leads to excessive dentin and cementum formation**. *Biochem. Biophys. Res. Commun.* (2011.0) **412** 549-555. DOI: 10.1016/j.bbrc.2011.07.116
18. Zhao Y, Yuan X, Liu B, Tulu US, Helms JA. **Wnt-responsive odontoblasts secrete new dentin after superficial tooth injury**. *J. Dent. Res.* (2018.0) **97** 1047-1054. DOI: 10.1177/0022034518763151
19. Palacios J. **Differential spatiotemporal expression of E- and P-cadherin during mouse tooth development**. *Int. J. Dev. Biol.* (1995.0) **39** 663-666. PMID: 8619966
20. Hermans F, Hemeryck L, Lambrichts I, Bronckaers A, Vankelecom H. **Intertwined signaling pathways governing tooth development: a give-and-take between canonical Wnt and Shh**. *Front. Cell Dev. Biol.* (2021.0) **9** 758203. DOI: 10.3389/fcell.2021.758203
21. Kornsuthisopon C, Photichailert S, Nowwarote N, Tompkins KA, Osathanon T. **Wnt signaling in dental pulp homeostasis and dentin regeneration**. *Arch. Oral. Biol.* (2022.0) **134** 105322. DOI: 10.1016/j.archoralbio.2021.105322
22. Martin A. **Bone proteins PHEX and DMP1 regulate fibroblastic growth factor Fgf23 expression in osteocytes through a common pathway involving FGF receptor (FGFR) signaling**. *FASEB J.* (2011.0) **25** 2551-2562. DOI: 10.1096/fj.10-177816
23. Lu Y. **The biological function of DMP-1 in osteocyte maturation is mediated by Its 57-kDa C-terminal fragment**. *J. Bone Miner. Res.* (2011.0) **26** 331-340. DOI: 10.1002/jbmr.226
24. Simsek Kiper PO. **Cortical-bone fragility—insights from sFRP4 deficiency in Pyle’s disease**. *N. Engl. J. Med.* (2016.0) **374** 2553-2562. DOI: 10.1056/NEJMoa1509342
25. Palomo T, Glorieux FH, Rauch F. **Circulating sclerostin in children and young adults with heritable bone disorders**. *J. Clin. Endocrinol. Metab.* (2014.0) **99** E920-E925. DOI: 10.1210/jc.2013-3852
26. Guirado E. **Disrupted protein expression and altered proteolytic events in hypophosphatemic dentin can be rescued by dentin matrix protein 1**. *Front. Physiol* (2020.0) **11** 82. DOI: 10.3389/fphys.2020.00082
27. Baron R, Kneissel M. **WNT signaling in bone homeostasis and disease: from human mutations to treatments**. *Nat. Med.* (2013.0) **19** 179-192. DOI: 10.1038/nm.3074
28. 28.Ruppe, M. D. X-Linked Hypophosphatemia. in GeneReviews(®) (eds Adam, M. P. et al.) (University of Washington, Seattle, 1993).
29. Kornsuthisopon C. **Non-canonical Wnt signaling participates in Jagged1-induced osteo/odontogenic differentiation in human dental pulp stem cells**. *Sci. Rep.* (2022.0) **12** 7583. DOI: 10.1038/s41598-022-11596-9
30. Eapen A. **Calcium-mediated stress kinase activation by DMP1 promotes osteoblast differentiation***. *J. Biol. Chem.* (2010.0) **285** 36339-36351. DOI: 10.1074/jbc.M110.145607
31. Eapen A. **Dentin phosphophoryn activates smad protein signaling through Ca2+-calmodulin-dependent protein kinase II in undifferentiated mesenchymal cells**. *J. Biol. Chem.* (2013.0) **288** 8585-8595. DOI: 10.1074/jbc.M112.413997
32. Narayanan K. **Dual functional roles of dentin matrix protein 1 implications in biomineralization and gene transcription by activation of intracellular Ca2+ store**. *J. Biol. Chem.* (2003.0) **278** 17500-17508. DOI: 10.1074/jbc.M212700200
33. Ann E-J. **Wnt5a controls Notch1 signaling through CaMKII-mediated degradation of the SMRT corepressor protein**. *J. Biol. Chem.* (2012.0) **287** 36814-36829. DOI: 10.1074/jbc.M112.356048
34. Tang W. **A genome-wide RNAi screen for Wnt/beta-catenin pathway components identifies unexpected roles for TCF transcription factors in cancer**. *Proc. Natl Acad. Sci. USA* (2008.0) **105** 9697-9702. DOI: 10.1073/pnas.0804709105
35. Schmidt EE. **GenomeRNAi: a database for cell-based and in vivo RNAi phenotypes, 2013 update**. *Nucleic Acids Res.* (2013.0) **41** D1021-D1026. DOI: 10.1093/nar/gks1170
36. Vijaykumar A, Root SH, Mina M. **Wnt/β-catenin signaling promotes the formation of preodontoblasts in vitro**. *J. Dent. Res.* (2021.0) **100** 387-396. DOI: 10.1177/0022034520967353
37. Xu M. **WNT10A mutation causes ectodermal dysplasia by impairing progenitor cell proliferation and KLF4-mediated differentiation**. *Nat. Commun.* (2017.0) **8** 15397. DOI: 10.1038/ncomms15397
38. Bae CH. **Wntless regulates dentin apposition and root elongation in the mandibular molar**. *J. Dent. Res.* (2015.0) **94** 439-445. DOI: 10.1177/0022034514567198
39. Kikuchi A, Yamamoto H, Sato A, Matsumoto S. **Wnt5a: its signalling, functions and implication in diseases**. *Acta Physiol.* (2012.0) **204** 17-33. DOI: 10.1111/j.1748-1716.2011.02294.x
40. Oishi I. **The receptor tyrosine kinase Ror2 is involved in non-canonical Wnt5a/JNK signalling pathway**. *Genes Cells Devoted Mol. Cell. Mech.* (2003.0) **8** 645-654. DOI: 10.1046/j.1365-2443.2003.00662.x
41. Ni X. **Low levels of serum sclerostin in adult patients with tumor-induced osteomalacia compared with X-linked hypophosphatemia**. *J. Clin. Endocrinol. Metab.* (2022.0) **107** e361-e371. DOI: 10.1210/clinem/dgab579
42. Carpenter KA. **Sclerostin antibody improves phosphate metabolism hormones, bone formation rates, and bone mass in adult Hyp mice**. *Bone* (2022.0) **154** 116201. DOI: 10.1016/j.bone.2021.116201
43. Carpenter KA, Ross RD. **Sclerostin antibody treatment increases bone mass and normalizes circulating phosphate levels in growing Hyp mice**. *J. Bone Miner. Res.* (2019.0) **35** 596-607. DOI: 10.1002/jbmr.3923
44. Mashhadikhan M, Kheiri H, Dehghanifard A. **DNA methylation and gene expression of sFRP2, sFRP4, Dkk 1, and Wif1 during osteoblastic differentiation of bone marrow derived mesenchymal stem cells**. *J. Oral. Biosci.* (2020.0) **62** 349-356. DOI: 10.1016/j.job.2020.08.001
45. Yu G. **Demethylation of SFRP2 by histone demethylase KDM2A regulated osteo-/dentinogenic differentiation of stem cells of the apical papilla**. *Cell Prolif.* (2016.0) **49** 330-340. DOI: 10.1111/cpr.12256
46. Liu H. **Augmented Wnt signaling in a mammalian model of accelerated aging**. *Science* (2007.0) **317** 803-806. DOI: 10.1126/science.1143578
47. Liu Q. **DKK1 rescues osteogenic differentiation of mesenchymal stem cells isolated from periodontal ligaments of patients with diabetes mellitus induced periodontitis**. *Sci. Rep.* (2015.0) **5** 13142. DOI: 10.1038/srep13142
48. Heymann R. **E- and N-cadherin distribution in developing and functional human teeth under normal and pathological conditions**. *Am. J. Pathol.* (2002.0) **160** 2123-2133. DOI: 10.1016/S0002-9440(10)61161-3
49. Lee H-K. **Nuclear factor I-C (NFIC) regulates dentin sialophosphoprotein (DSPP) and E-cadherin via control of Krüppel-like factor 4 (KLF4) during dentinogenesis***. *J. Biol. Chem.* (2014.0) **289** 28225-28236. DOI: 10.1074/jbc.M114.568691
50. Schneider CA, Rasband WS, Eliceiri KW. **NIH Image to ImageJ: 25 years of image analysis**. *Nat. Methods* (2012.0) **9** 671-675. DOI: 10.1038/nmeth.2089
51. Dobin A. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013.0) **29** 15-21. DOI: 10.1093/bioinformatics/bts635
52. Liao Y, Smyth GK, Shi W. **featureCounts: an efficient general purpose program for assigning sequence reads to genomic features**. *Bioinformatics* (2014.0) **30** 923-930. DOI: 10.1093/bioinformatics/btt656
53. McCarthy DJ, Chen Y, Smyth GK. **Differential expression analysis of multifactor RNA-Seq experiments with respect to biological variation**. *Nucleic Acids Res.* (2012.0) **40** 4288-4297. DOI: 10.1093/nar/gks042
54. Robinson MD, McCarthy DJ, Smyth GK. **edgeR: a Bioconductor package for differential expression analysis of digital gene expression data**. *Bioinformatics* (2010.0) **26** 139-140. DOI: 10.1093/bioinformatics/btp616
55. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: a practical and powerful approach to multiple testing**. *J. R. Stat. Soc. Ser. B Methodol.* (1995.0) **57** 289-300
56. Thomas PD. **PANTHER: a library of protein families and subfamilies indexed by function**. *Genome Res.* (2003.0) **13** 2129-2141. DOI: 10.1101/gr.772403
57. Krämer A, Green J, Pollard J, Tugendreich S. **Causal analysis approaches in Ingenuity Pathway Analysis**. *Bioinformatics* (2014.0) **30** 523-530. DOI: 10.1093/bioinformatics/btt703
|
---
title: Ecological shifts of salivary microbiota associated with metabolic-associated
fatty liver disease
authors:
- Min Wang
- Li-Ya Yan
- Cai-Yun Qiao
- Chu-Chu Zheng
- Chen-Guang Niu
- Zheng-Wei Huang
- Yi-Huai Pan
journal: Frontiers in Cellular and Infection Microbiology
year: 2023
pmcid: PMC9971218
doi: 10.3389/fcimb.2023.1131255
license: CC BY 4.0
---
# Ecological shifts of salivary microbiota associated with metabolic-associated fatty liver disease
## Abstract
### Introduction
Metabolic-associated fatty liver disease (MAFLD) is the most common chronic liver disease related to metabolic syndrome. However, ecological shifts in the saliva microbiome in patients with MAFLD remain unknown. This study aimed to investigate the changes to the salivary microbial community in patients with MAFLD and explore the potential function of microbiota.
### Methods
Salivary microbiomes from ten MAFLD patients and ten healthy participants were analyzed by 16S rRNA amplicon sequencing and bioinformatics analysis. Body composition, plasma enzymes, hormones, and blood lipid profiles were assessed with physical examinations and laboratory tests.
### Results
The salivary microbiome of MAFLD patients was characterized by increased α-diversity and distinct β-diversity clustering compared with control subjects. Linear discriminant analysis effect size analysis showed a total of 44 taxa significantly differed between the two groups. Genera Neisseria, Filifactor, and Capnocytophaga were identified as differentially enriched genera for comparison of the two groups. Co-occurrence networks suggested that the salivary microbiota from MAFLD patients exhibited more intricate and robust interrelationships. The diagnostic model based on the salivary microbiome achieved a good diagnostic power with an area under the curve of 0.82($95\%$ CI: 0.61–1). Redundancy analysis and spearman correlation analysis revealed that clinical variables related to insulin resistance and obesity were strongly associated with the microbial community. Metagenomic predictions based on Phylogenetic Investigation of Communities by Reconstruction of Unobserved States revealed that pathways related to metabolism were more prevalent in the two groups.
### Conclusions
Patients with MAFLD manifested ecological shifts in the salivary microbiome, and the saliva microbiome-based diagnostic model provides a promising approach for auxiliary MAFLD diagnosis.
## Introduction
Metabolic-associated fatty liver disease (MAFLD), a terminology updated from non-alcoholic fatty liver disease (NAFLD) in 2020, is the hepatic manifestation of the metabolic syndrome with a broad spectrum of liver conditions ranging from simple hepatic steatosis to steatohepatitis to stage 4 fibrosis (Demirtas and Yilmaz, 2020; Eslam et al., 2020). It usually manifests clinically silent and has no approved pharmacotherapy; with time, it can gradually progress to end-stage liver diseases, such as cirrhosis and hepatocellular carcinoma(Eslam et al., 2020). Accumulating evidence has indicated that MAFLD is strongly associated with increased risks of incident diabetes, chronic kidney disease, and cardiovascular disease, with the magnitude of risk seeming to parallel the severity of MAFLD(Liang et al., 2021; Targher et al., 2021). Globally, MAFLD affects approximately a quarter of the adult population, placing an enormous burden on healthcare systems and society(Eslam et al., 2020). Thus, convenient diagnostic screening and timely strategic intervention for MAFLD are of great significance.
The oral microbiota is the second most diverse microbial ecosystem after that of the gut in the human body, consisting of over 700 bacterial species. It plays an essential role in physiological, metabolic, and immunological functions, which include nutrient digestion, metabolic regulation, immune response, and antibacterial activity (Kilian et al., 2016). In addition to oral diseases, oral microbiota dysbiosis has also been closely linked with metabolic disorders, including MAFLD (Zhao et al., 2020; Peng et al., 2022). An animal study demonstrated that the endotoxemia from *Porphyromonas gingivalis* was a remarkable risk factor for NAFLD pathogenesis, and the altered glucose/lipid metabolism may facilitate disease progression(Sasaki et al., 2018). Our previous work has also confirmed the community structure alterations and microbial dysbiosis changes of supragingival microbiota in patients with MAFLD(Zhao et al., 2020). Given the oral cavity being a complex microbial environment, further insights into the alterations of the oral microbial community under the pathological state of MAFLD are hence warranted.
Saliva is a heterogeneous biofluid containing various proteins, metabolites, microbes, and their genes, which are essential for maintaining oral homeostasis(Belstrøm, 2020). Spreading throughout the entire oral cavity, saliva theoretically acts as a reservoir pool of microorganisms detached from various oral niches and appears representative of the overall oral microbiome(Yoshizawa et al., 2013; Belstrøm, 2020). Besides, saliva-based microbial, immunologic, and molecular biomarkers have been progressively investigated as diagnostic tools for several diseases due to the remarkable advantages of a non-invasive, stress-free, and cost-effective sampling manner(Yoshizawa et al., 2013; Zhang et al., 2016). With the rapid development of next-generation sequencing technology, a growing number of studies have evaluated the alterations of salivary microbial communities and the diagnostic utility in various diseases, such as type 2 diabetes mellitus, schizophrenia, and rheumatic heart disease, among others(Zhang et al., 2016; Liu et al., 2021; Qing et al., 2021; Shi et al., 2021). However, there is a paucity of information about salivary microbiota’s potential role in MAFLD.
This study aimed to investigate the ecological shifts in the salivary microbiome of MAFLD patients and explore the potential function of salivary microbiota. Firstly, microbiota profiles of saliva samples were characterized by 16S rRNA gene sequencing to identify changes between MAFLD patients and control subjects. Then, the potential value of the saliva-based microbiota diagnostic model was evaluated to discriminate MAFLD patients from control subjects. Finally, the relationships between clinical variables and specific microbial genera were assessed. These findings could provide more significant insights into the salivary ecological dysbiosis associated with MAFLD.
## Study population
A total of ten MAFLD patients and ten healthy control individuals were recruited from a health census. Both groups were statistically comparable in age and gender. For each subject, upper abdomen ultrasonography was performed and interpreted by experienced sonographers. MAFLD was diagnosed based on ultrasonographic findings and the exclusion of known etiologic factors of chronic liver disease as well as excessive alcohol consumption. The exclusion criteria were similar to a previous study as follows(Zhao et al., 2020):[1] excessive alcohol consumption: >30g/d for males and >20g/d for females; [2] other liver diseases, including viral hepatitis, autoimmune hepatitis, and hepatolenticular degeneration; [3] drug-induced steatohepatitis (e.g., tamoxifen, amiodarone, valproate, methotrexate, and glucocorticoids); [4] Other factors that may result in hepatic steatosis, including total parenteral nutrition, inflammatory bowel disease, celiac disease, hypothyroidism, Cushing’s syndrome, lipoprotein deficiency, lipid-atrophic diabetes, etc.; [ 5] type I or type II diabetes; [6] the use of lipid-lowering medication within the past six months; [7] other conditions, including pregnant or lactating women, prolonged heavy smoking, use of antibiotics for more than five days within six months, etc.; [ 8] untreated oral abscess, oral precancerous lesions and oral cancer, oral fungal infections; more than eight teeth missing. This study followed the Declaration of Helsinki on medical protocols and ethics, and was approved by the Ethics Committee of School and Stomatology Wenzhou Medical University (approval no. WYKQ2021006). Written informed consent was obtained from all participants.
## Anthropometric and clinical variables measurement
All participants underwent physical examinations, oral examinations, and anthropometric measurements, together with fasting blood sample collection for biochemical tests. The anthropometric parameters included body weight, height, body mass index (BMI), waist and hip circumferences, waist-hip ratio (WHR), blood pressure, and heart rate. The blood biochemical indicators included total cholesterol (TC), total triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma glutamyl transpeptidase (GGT), fasting plasma glucose (FPG), fasting serum insulin (FSI), glycosylated hemoglobin (HbA1c) and C-reactive protein (CRP). In addition, the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) was used to ascertain IR. The formula is as follow: HOMA-IR = FPG (mmol/L) × FSI (mU/L)/22.5(Matthews et al., 1985). A thorough dental exam was conducted to assess dental caries, periodontal status, and other dental conditions listed above by the same dentist. Interdental clinical attachment loss presents at ≥ 2 non-adjacent teeth, or buccal or oral clinical attachment loss ≥ 3 mm with pocketing >3 mm presents at ≥ 2 teeth are diagnosed as periodontitis(Tonetti et al., 2018). The number of decayed, missing, and filled teeth (DMFT) for each subject were recorded. The unpaired Student’s t-test was carried out to analyze anthropometric and biochemical indicators, except for sex and periodontitis, for which the chi-square test was used.
## Specimen collection
Unstimulated saliva samples were collected from each subject according to the Human Microbiome Project Core Sampling Protocol A (https://www.hmpdacc.org/hmp/doc/HMP_MOP_Version12_0_072910.pdf) with minor adjustments. Prior to salivary microbiota sampling, all participants were requested to avoid oral hygiene procedures for 24 h, abstain from drinking, eating at least 2 h in advance. At least 2 ml of unstimulated saliva samples were collected in sterile tubes between 9 am and 11 am. All samples were transported at 4°C to the laboratory as soon as possible and then stored in liquid nitrogen until DNA extraction.
## DNA extraction, amplification, and high-throughput sequencing
The total microbial community DNA was extracted by QIAamp DNA Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions, and the quality was determined using a NanoDrop 2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and $1\%$ agarose gel electrophoresis. The V3-V4 hypervariable region of bacterial 16S rRNA was amplified by a thermocycler PCR system (GeneAmp 9700; Applied Biosystems, Carlsbad, CA, USA) with the forward primer 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and the reverse primer 806R (5′-GGACTACH VGGGTWTCTAAT-3′). The PCR products were purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) and quantified using the QuantiFluor ® Single-Tube Fluorometer (Promega Corporation, Madison, WI, USA), following the manufacturer’s instructions. Finally, the purified amplicons were paired-end sequenced on an Illumina Miseq PE300 platform (Illumina, San Diego, CA, USA). The sequencing work was completed by Majorbio (Shanghai, China).
## Data processing and bioinformatics analysis
The raw sequence reads were quality-filtered using Fastp (version 0.19.6, https://github.com/OpenGene/fastp) and merged by FLASH software (version 1.2.11; https://ccb.jhu.edu/software/FLASH/index.shtml). The specific criteria are consistent with the previous study(Zhao et al., 2020). UPARSE (version 7.1, http://drive5.com/uparse/) was used to cluster the sequences into operational taxonomic units (OTUs) with a threshold of $97\%$ similarity. The taxonomy assignment of each OTU was carried out by RDP Classifier (version 2.11, https://github.com/OpenGene/fastp) against the SILVA 16S rRNA database (Release 132, https://www.arb-silva.de/) using a confidence threshold of 0.7. Prior to further analysis, the OTUs tables were subsampled to equal depths according to the fewest sample sequence.
Alpha diversity indexes (Shannon, Simpson, Ace, and Chao1) based on the OTUs profiles were applied to analyze microbial diversity, which were calculated with Mothur software (version 1.30.2, http://www.mothur.org/). The principal coordinates analysis (PCoA) and partial least squares discriminant analysis (PLS-DA) plots based on Euclidean distance at the genus level were performed to visualize the differences in species composition between the MAFLD and control groups. The species relative abundance was visualized by bar plots at the phylum and genus levels. Besides, the differences in relative abundance were analyzed at the genus level by Analysis of Composition of Microbiomes (ANCOM) via ANCOM 2.0 package in R platform. Statistical significance was defined as W>0.7.
The linear discriminant analysis (LDA) combined effect size (LEfSe; http://huttenhower.sph.harvard.edu/galaxy) was employed to identify the significantly different taxa between the groups at the phylum to genus levels. The threshold on the logarithmic LDA score for discriminative features was set to 2.0. Co-occurrence networks at the genus level were built by pairwise correlation spearman analysis (Spearman’s coefficient > 0.5 and P-value < 0.05). For each network, two properties named “average shortest path length” and “transitivity” were computed using NetworkX86 (version 2.4) according to a previous study(Loftus et al., 2021). A random forest classifier was trained to discriminate the MAFLD patients from healthy control subjects based on the genus abundance profile, and its prediction accuracy performance was accessed by the area under the ROC curve (AUC). ( randomForest package in R V.4.3.2 and plotROC package in R V.3.4.4). The potential relationships between microbial population distribution and clinical variables were evaluated through redundancy analysis (RDA). Moreover, Spearman correlation coefficients were carried out to analyze correlations between clinical variables and the top 20 abundant genera, and the results were visualized by heat maps using the R platform. Furthermore, the functional profiles of microbial communities were predicted using PICRUSt2(version 2.2.0; http://picrust.github.io/picrust/) with reference to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Statistical differences between the two groups were determined by Wilcoxon rank-sum test with Benjamini-Hochberg false discovery rate (FDR) correction. Statistical significance was set to $P \leq 0.05$
## Subject characteristics and clinical data
The demographic data, clinical characteristics, and laboratory data of the participants were summarized in Table 1. No statistical difference was observed in age, gender, blood pressure, heart rate, periodontitis, and DMFT between the two groups, whereas BMI and WHR were higher in the MAFLD group compared to the control group. With regard to biochemical indicators, FPG, FSI, HOMA-IR, and HbA1C were significantly elevated in the MAFLD group, suggesting the presence of IR in MAFLD patients. In addition, the MAFLD group had substantially higher TG levels and lower HDL-C levels, resulting in a higher TG/HDL-C ratio than the control group. Although there was no statistical difference between ALT and AST levels, the AST/ALT ratio was significantly reduced in the MAFLD group.
**Table 1**
| Parameter | Healthy (n=10) | MAFLD (n=10) | P |
| --- | --- | --- | --- |
| Age (y) | 33.30 ± 5.12 | 35.80 ± 9.11 | 0.459 |
| Gender (male/female) | 6/4 | 9/1 | 0.303 |
| BMI (kg/cm2) | 22.07 ± 1.24 | 26.33 ± 2.68 | <0.001* |
| WHR | 0.84 ± 0.07 | 0.91 ± 0.05 | 0.019* |
| SBP (mm HG) | 118.30 ± 14.08 | 123.30 ± 13.40 | 0.427 |
| DBP (mm HG) | 75.10 ± 9.59 | 79.50 ± 7.28 | 0.263 |
| HR | 74.80 ± 5.87 | 79.80 ± 7.13 | 0.104 |
| TC (mmol/L) | 4.81 ± 0.84 | 4.66 ± 1.09 | 0.740 |
| TG (mmol/L) | 0.89 ± 0.33 | 1.84 ± 1.23 | 0.039* |
| LDL-C (mmol/L) | 3.00 ± 0.95 | 2.94 ± 0.98 | 0.892 |
| HDL-C (mmol/L) | 1.44 ± 0.39 | 1.02 ± 0.16 | 0.005* |
| ALT (U/L) | 32.60 ± 39.83 | 55.40 ± 30.35 | 0.167 |
| AST (U/L) | 26.50 ± 14.37 | 29.50 ± 11.30 | 0.610 |
| AST/ALT | 1.21 ± 0.55 | 0.63 ± 0.22 | <0.001* |
| GGT (U/L) | 24.00 ± 19.55 | 43.30 ± 24.32 | 0.066 |
| FPG (mmol/L) | 4.43 ± 0.46 | 5.26 ± 0.40 | <0.001* |
| FSI (mU/L) | 7.71 ± 3.02 | 14.80 ± 4.89 | 0.001* |
| HOMA-IR | 1.49 ± 0.52 | 3.44 ± 1.11 | <0.001* |
| HbA1C | 5.00 ± 0.23 | 5.46 ± 0.29 | 0.001* |
| CRP | 0.76 ± 1.01 | 1.15 ± 1.11 | 0.421 |
| Periodontitis (%) | 10.00 | 20.00 | 0.531 |
| DMFT | 4.50 ± 1.35 | 5.40 ± 2.59 | 0.343 |
## Community structure and microbiota composition of salivary microbiota
Twenty saliva samples contained 16 phyla, 24 classes, 62 orders, 100 families, 205 genera, 383 species, and 453 OTUs. In terms of α-diversity, as shown in Table 2, the Ace and Chao1 richness index of the MAFLD group significantly increased compared with those in the control group, indicating that MAFLD considerably altered the diversity and richness of the salivary microbiota. Principal coordinate analysis (PCoA) based on Euclidean distance at the genus level was employed to evaluate the β-diversity of the salivary microbiota community, despite partial overlapping, which showed a significant difference between the two groups (AMOSIM, $$P \leq 0.04$$; Figure 1A). The visualization of sample clustering using a supervised PLS-DA plot further confirmed the remarkable separation between groups (Figure 1B).
The most abundant phyla in both groups were Firmicutes, Proteobacteria, Bacteroidota, Fusobacteriota, Actinobacteriota, Patescibacteria, and Spirochaetota, accounting for over $99\%$ of community abundance. Notably, the relative abundance of Firmicutes increased in comparison to the controls, while the relative abundance of Proteobacteria declined in the MAFLD patients, thereby resulting in a lower Firmicutes/Proteobacteria ratio in the MAFLD group (Figure 1C). At the genus level, the predominant bacteria included *Streptococcus with* a relative abundance of $22.09\%$ and $18.70\%$, and Neisseria with a relative abundance of $7.89\%$ and $17.44\%$ in the control and MAFLD groups, respectively. ( Figure 1D).
## Alterations of the salivary microbial taxa
The genus abundance was compared based on community abundance data to assess the specific alterations in salivary microbiota. ANCOM analysis showed that the genera Filifactor, Neisseria, and Capnocytophaga had significantly different abundances between the two groups (Table 3). LEfSe analysis revealed that the MAFLD group had a higher abundance of genera Howardella, Treponema, Desulfobulbus, Bulleidia, Propionibacterium, Filifactor, Eggerthia, Fretibacterium, Shuttleworthia, and Roseburia. By contrast, the control subjects had a higher abundance of genera Neisseria and Capnocytophaga. ( Table 3). Figure 2 depicted the differentially abundant microbial taxa from phylum to genus level with LDA scores higher than 2.0 based on the LEfSe analysis. Overall, 4 phyla, 5 classes, 9 orders, 11 families, and 15 genera were identified to be significantly discriminant, with thirty-six taxa enriched in the MAFLD group and eight taxa enriched in the control group. These results suggested the presence of specific bacteria in the saliva of MAFLD patients and healthy subjects.
Co-occurrence networks among the top 50 abundant genus-level taxa were conducted further to investigate the biological interactions within the microbial community. As shown in Figure 3, the microbial networks slightly differ between the two groups, and the taxa within the MAFLD groups exhibited more intricate and robust interrelationships. The network transitivity for the MAFLD and health groups were 0.522 and 0.438, respectively. In contrast, the average shortest path length was 2.240 for the MAFLD group and 3.207 for the health group.
**Figure 3:** *Microbial co-occurrence networks of the top 50 abundant genera constructed on the MAFLD (A) and control (B) groups. Each node represents a genus, and the size of the node is proportional to the mean relative abundance. The same color represents the genera belonging to the same phylum. The thickness of each connection is proportional to the coefficient values. The red and green lines indicate positive and negative interactions, respectively.*
## Diagnostic model for MAFLD based on salivary microbiota
Random forest analysis was conducted to identify the diagnostic potential of salivary microbiota for MAFLD. It is generally accepted that when the value of AUC is greater, the biomarkers have higher diagnostic accuracy. Among salivary microbiota, a panel of seven genera, including Fretibacterium, Neisseria, Treponema, Delftia, Capnocytophaga, Dialister, and Erysipelotrichaceae_UCG-003 were identified as optimal biomarkers with an AUC of 0.82($95\%$ CI: 0.61–1) (Figure 4). These outcomes indicated that salivary microbial markers achieved a good diagnostic potential for discriminating MAFLD patients from the healthy cohort.
**Figure 4:** *Diagnostic model based on the salivary microbiome for MAFLD. (A) Barplot of the genera ranked by importance in the random forest model. (B) The receiver operating characteristic curve achieved an AUC value of 0.82 on the genus level. And the green shade was a confidence interval.*
## Correlation between clinical variables and microbial communities
To explore the potential associations between microbial community composition and multiple clinical variables, the following ten main clinical variables were chosen for the RDA test: BMI, WHR, TC, TG, LDL-C, HDL-C, AST/ALT, GGT, HOMA-IR, and HbA1c. As depicted in Figure 5A, HOMA-IR (r2 = 0.4896, $$p \leq 0.007$$), BMI (r2 = 0.4819, $$p \leq 0.0.003$$), WHR (r2 = 0.4595, $$p \leq 0.006$$) and HbA1c (r2 = 0.3874, $$p \leq 0.018$$) played significant roles in the salivary community composition. Furthermore, the Spearman correlation heatmap portrayed the association between the top 20 abundant genera and clinical variables (Figure 5B).
**Figure 5:** *Correlation analysis between microbial community and clinical variables. (A) RDA analysis on microbial population distribution is explained by the clinical variables at the genus level. Arrows represent clinical variables. The long arrow indicates a high correlation with the distribution of the salivary microbiome. The acute angle of the two arrow lines indicates a positive correlation between the clinical variables, and the obtuse angle is a negative correlation. (B) Spearman correlation heatmap based on the top 20 abundant salivary microbiota and clinical variables. The color range on the right shows the color partitioning of the different R values. A clustering tree for each species is on the left side of the heat map. *P < 0.05; ** P < 0.01; *** P < 0.001.*
## Alterations of functional pathway
To investigate the functional implications of salivary microbiota, PICRUSt2 was performed to predict the metagenome functional content from 16S rRNA data. Notably, among the top 15 significantly abundant KEGG pathways, seven pathways belonging to the metabolism category were observed (Figure 6). MAFLD group exhibited a significant increase in pathways associated with pyrimidine metabolism and fructose and mannose metabolism, whereas a decrease in pathways related to 2-Oxocarboxylic acid metabolism, sulfur metabolism, glutathione metabolism, tyrosine metabolism, and ascorbate and aldarate metabolism, etc.
**Figure 6:** *Predictive metagenome functional profiling of the top 15 significantly abundant KEGG pathways using PICRUSt2 analysis. Bar plots on the left side show the mean proportion of each KEGG pathway; on the right display the differences between proportions.*
## Discussion
The salivary microbiota, which serves as a reservoir of microorganisms from different sites within the oral cavity, could represent the whole oral microbiota and reflect the oral and general health status(Belstrøm, 2020). An increasing number of metabolic disorders have been reported in potential relation to the shifts in salivary microbial ecology, such as atherosclerosis, type 2 diabetes mellitus, polycystic ovary syndrome, and hepatic encephalopathy (Yoshizawa et al., 2013; Zhang et al., 2016; Liu et al., 2021). However, to the best of our knowledge, few studies have explored the microbiome-level association between salivary microbiota and MAFLD. This study discovered that abnormal metabolic levels in MAFLD patients had significantly altered the composition and structure of salivary microbiota, which had good diagnostic power in discriminating MAFLD patients from healthy controls.
Oral microbial diversity is essential for maintaining health and varies with the physiological state of the host. Generally, elevated microbiome richness and diversity have been recognized as hallmarks of a healthy ecosystem, especially for the gut microbial ecosystem(Falony et al., 2018). However, our present study showed that the salivary microbiota in MAFLD patients exhibited increased α-diversity (ACE index and Chao1 index). Although no significant differences were observed in the Shannon and Simpson indices, both showed marginally elevated diversity among patients with MAFLD. These findings are consistent with our previous study that focused on the shifts in supragingival microbiota associated with MAFLD(Zhao et al., 2020). Following the concept proposed by Takeshita et al. ( Takeshita et al., 2016), poor oral health could lead to an increased taxonomic richness in saliva. Because prolonged plaque accumulation may cause the multiplication of attached bacteria, and the gingival bleeding could provide rich nutrients for bacteria growth. These ecological shifts may promote bacterial assemblage in saliva. This phenomenon has been previously observed in patients with other metabolic diseases(Si et al., 2017). Furthermore, the PCoA and PLS-DA results revealed a significant clustering of microbial communities between the two groups. Thus, the altered metabolic level in patients with MAFLD would provoke salivary microbial shifts and increase the risk of oral diseases.
Phyla Firmicutes and Proteobacteria showed divergent abundance trends, with Firmicutes enrichment and Proteobacteria depletion from the control group to the MAFLD group, resulting in an increase in Firmicutes/Proteobacteria ratio. The increased Firmicutes/Proteobacteria ratio was also detected in the salivary microbiome of patients with primary Sjögren’s syndrome(Van der Meulen et al., 2018) and schizophrenia(Qing et al., 2021). Similar to MAFLD, these two diseases are characterized by chronic low-grade inflammation. Lau et al. also revealed a shift in gut microbiome toward decreased Firmicutes/Proteobacteria ratio in type 2 diabetic adults with mild obesity following metabolic surgery (Lau et al., 2021). A probable reason is that Proteobacteria are a major phylum of Gram-negative bacteria, mainly associated with glucose homeostasis improvement, metabolic amelioration, and inflammatory response reduction(Carvalho et al., 2012; Lau et al., 2021). In contrast, the phylum Firmicutes are mostly Gram-positive bacteria playing a pivotal role in the fermentation and metabolism of carbohydrates and lipids through chain-breaking fatty acid synthesis, which usually facilitates obesity development(Stojanov et al., 2020). Therefore, it is indicative that *Firmicutes bacteria* could have a competitive advantage over *Proteobacteria bacteria* in niche occupancy during the MAFLD state. However, whether the Firmicutes/Proteobacteria ratio could be used as a marker of ecosystem health in microbiome research still requires more studies to confirm.
LEfSe and ANCOM analyses revealed notable discrepancies between the two groups. Genus Filifactor was enriched in MAFLD patients, whereas genera Neisseria and Capnocytophaga were more abundant in healthy controls. Of the remaining genera, Treponema, Fretibacterium, Propionibacterium, Shuttleworthia, Eggerthia, Bulleidia, Howardella, and Desulfobulbus, presented a higher proportion in the MAFLD group. Filifactor spp. are well-known periodontal pathogens (Takeshita et al., 2016) and have been reported to exist in high abundance in the salivary microbiota of patients with hepatocellular carcinoma (Park et al., 2021). The genus Treponema consists of a diverse group of pathogenic or commensal microorganisms that not only cause syphilis, yaws, and pinta infections(Shukla et al., 2022) but are also associated with chronic liver disease(Ling et al., 2015), infective endocarditis(Hijikata et al., 2019), and Alzheimer’s disease (Riviere et al., 2002). Besides, oral Treponema denticola, together with *Porphyromonas gingivalis* and Tannerella forsythia, are known as the “red complex” which have been widely considered essential pathogens in periodontal disease etiology and pathogenesis. A series of animal and human studies have confirmed that the genera Fretibacterium (Yamamoto et al., 2021), Propionibacterium (Del Chierico et al., 2017), Shuttleworthia (Xie et al., 2016), Eggerthia (Yamamoto et al., 2021), and Bulleidia (Bashiardes et al., 2016)are prevalent in the oral or gut microbiota with liver disease including nonalcoholic fatty liver disease, hepatocellular carcinoma. These findings could be interpreted by the emerging oral-gut-liver axis that oral microbes could translocate into the gastrointestinal tract, where they would spread to the liver and induce hepatic diseases(Imai et al., 2021). On the other hand, Neisseria spp. ( Liu et al., 2015) are generally Gram-negative microorganisms comprising mainly non-pathogenic species in the oral cavity. Previously, we also demonstrated a higher abundance of Neisseria in the supragingival plaque among healthy individuals compared to those with MAFLD (Zhao et al., 2020). Capnocytophaga is a genus of Gram-negative anaerobic bacteria reportedly linked with improved oral health and reduced caries experience (Schoilew et al., 2019). Interestingly, the genus Neisseria and Capnocytophaga have been identified as part of human oral health’s “core microbiome” (Zaura et al., 2009).
Microorganisms coexist in intricate networks of interactions, which in turn affect the species involved and may lead to disease progression. As shown in the co-occurrence networks diagram, the inter-genera interactions in MAFLD saliva exhibit higher network transitivity and lower average shortest path lengths. Transitivity measures the average connectedness of a network, with higher values indicating the presence of more tightly connected clusters (more inter-genus interactions) (Sisk-Hackworth et al., 2021). By contrast, the lower average shortest path lengths within microbial networks suggest species are interconnected through shorter paths (Loftus et al., 2021). The underlying reason may be that short path lengths and intimate connections within the salivary networks in patients with MAFLD could transmit signals rapidly between bacterial species, thereby potentially promoting shifts in community metabolism.
A timely and accurate diagnosis of MAFLD is a prerequisite for expeditious therapeutic interventions, which could inhibit disease progression. The random forest analysis showed that the salivary microbiota signature has good diagnostic power (bacterial taxa achieved an AUC of 0.82) to discriminate MAFLD patients from healthy controls. These findings reinforce the notion that saliva microbes potentially rich in diagnostic biomarkers could serve as indicators of both oral and systemic diseases(Yoshizawa et al., 2013; Zhang et al., 2016). Using 16S rRNA sequencing of fecal samples from children, Schwimmer et al. reported that prediction models combining serum ALT levels and relative abundance of encoding genes could discriminate participants in NAFLD, NASH, and severe fibrosis with AUCs of 0.95, 0,92, and 0.87, respectively(Schwimmer et al., 2019). Although gut microbiota prediction seems superior for MAFLD, given concerns about noninvasiveness and accessibility, the biochemical blood indices were not incorporated into the saliva-based biomarker model in this study. The key next step is to generalize and transform these critical microbial biomarkers into available tools for clinical practice(Mouzaki and Loomba, 2020).
With numerous studies reporting the complex etiology and pathogenesis of MAFLD, it has become clear that underlying risk factors for MAFLD development encompass IR, obesity, and dyslipidemia(Duseja and Chalasani, 2013), among which IR takes center stage (Kuraji et al., 2021). IR means the cells in the body muscles, fat, and liver cannot respond well to the normal concentration of insulin hormone that promotes glucose uptake and fat storage. Therefore, IR caused by the imbalance between energy intake and expenditure could lead to fat accumulation in the liver and promote dyslipidemia through increased circulating free fatty acids in blood(Kuraji et al., 2021). As first described by Matthews et al. in 1985, HOMA-IR has been commonly used for IR estimation in clinical studies, with a HOMA-IR value ≥2.5 indicating IR(Matthews et al., 1985).*In this* research, HOMA-IR in the MAFLD group exceeded the threshold value and was significantly higher than in the control group, suggesting that IR is highly prevalent among patients with MAFLD. Moreover, the elevated HbA1c and TG/HDL-C ratio, known as a simple and reliable marker of IR, further demonstrated the propensity for IR in the MAFLD group(Borai et al., 2011; Fan et al., 2019). In addition, elevated serum transaminase levels and a decreased AST/ALT ratio are markers of ongoing hepatocellular injury. They are commonly deranged in patients with progressive MAFLD(Sheka et al., 2020), implying that enrolled MAFLD patients had different degrees of hepatocellular damage.
The RDA analyses further verified the significant influences of IR ($$p \leq 0.007$$ for HOMA-IR; $$p \leq 0.018$$ for HbA1c) and obesity ($$p \leq 0.003$$ for BMI; $$p \leq 0.006$$ for WHR) over the salivary microbiota distribution. As shown in the spearman correlation heatmap (between the top 20 abundant genera and clinical variables), although some correlations did not reach statistical significance, Neisseria, Capnocytophaga, Haemophilus, and Rothia tended to be negatively associated with IR, obesity, and dyslipidemia index. *These* genera are members of nitrate-reducing bacteria residing in the oral cavity and exerting beneficial effects on hosts through the nitrate–nitrite–nitric oxide pathway, such as regulating glucose metabolism, lowering lipid levels, and reducing inflammation(Rosier et al., 2022). However, the specific role and related mechanisms of these oral nitrate-reducing microbes in MAFLD remain to be further delineated. The negative association results support the presumption that metabolic abnormalities (e.g., dyslipidemia, IR, and obesity) may play a role in modulating oral microbial composition and cause oral dysbiosis(Negrini et al., 2021). Genus Fusobacterium was positively correlated with TC and LDL-C levels, consistent with the concept that the presence of Fusobacterium spp. is implicated in the pathogenesis of atherosclerosis via aberrant lipid metabolism(Zhou et al., 2022).
The analysis with PICRUSt2 demonstrated that the main functional alterations between the two groups were metabolic-related pathways. The upregulation of the pathways for pyrimidine metabolism and fructose and mannose metabolism is consistent with our previous supragingival plaque study on MAFLD patients(Zhao et al., 2020), further confirming the metabolic-related pathways of oral bacteria are associated with the occurrence of MAFLD. Pathways related to amino acid metabolism (including glutathione metabolism and tyrosine metabolism) were more prevalent in healthy subjects, possibly because the MAFLD patients have disturbances in amino acid metabolism caused by hepatocellular injury. However, further metabolome studies are needed to investigate the metabolism in MAFLD and its relationship to the salivary microbiome.
A limitation of this study was the limited sample size. Although the inclusion criteria were rigorous, future investigations with a larger sample size were needed to control for potential confounding factors. Moreover, due to the limitation of 16s rRNA sequencing, metagenomics sequencing may be required to genetically annotate and validate functional information related to salivary microbiome changes in patients with MAFLD.
In conclusion, we comprehensively described salivary microbiome alterations between MAFLD patients and healthy individuals, evaluated the potential value of salivary microbiota as an auxiliary diagnostic tool to predict MAFLD, and demonstrated the role of biochemical variables, especially the IR, in microbial ecological shifts. This study provides an in-depth view of the association between the salivary microbiome and MAFLD, which may contribute to developing strategies for the prevention, diagnosis, and treatment of MAFLD.
## Data availability statement
The data presented in the study are deposited in the NCBI Sequence Read Archive (SRA) repository, accession number PRJNA929590 (https://www.ncbi.nlm.nih.gov/bioproject/929590, PRJNA929590).
## Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of School and Stomatology Wenzhou Medical University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
MW, C-GN and C-YQ performed the study design, data analysis, drafting, and revising of the work. L-YY and C-CZ performed the data analysis and acquisition. Y-HP and Z-HW contributed to the study design, clinical sample collection, data analysis, drafting, revising, and final approval. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Bashiardes S., Shapiro H., Rozin S., Shibolet O., Elinav E.. **Non-alcoholic fatty liver and the gut microbiota**. *Mol. Metab.* (2016) **5** 782-794. DOI: 10.1016/j.molmet.2016.06.003
2. Belstrøm D.. **The salivary microbiota in health and disease**. *J. Oral. Microbiol.* (2020) **12**. DOI: 10.1080/20002297.2020.1723975
3. Borai A., Livingstone C., Abdelaal F., Bawazeer A., Keti V., Ferns G.. **The relationship between glycosylated haemoglobin (HbA1c) and measures of insulin resistance across a range of glucose tolerance**. *Scand. J. Clin. Lab. Invest.* (2011) **71** 168-172. DOI: 10.3109/00365513.2010.547947
4. Carvalho B., Guadagnini D., Tsukumo D., Schenka A., Latuf-Filho P., Vassallo J.. **Modulation of gut microbiota by antibiotics improves insulin signalling in high-fat fed mice**. *Diabetologia* (2012) **55** 2823-2834. DOI: 10.1007/s00125-012-2648-4
5. Del Chierico F., Nobili V., Vernocchi P., Russo A., De Stefanis C., Gnani D.. **Gut microbiota profiling of pediatric nonalcoholic fatty liver disease and obese patients unveiled by an integrated meta-omics-based approach**. *Hepatology* (2017) **65** 451-464. DOI: 10.1002/hep.28572
6. Demirtas C. O., Yilmaz Y.. **Metabolic-associated fatty liver disease: Time to integrate ground-breaking new terminology to our clinical practice**. *Hepatol. Forum* (2020) **1** 79-81. DOI: 10.14744/hf.2020.2020.0024
7. Duseja A., Chalasani N.. **Epidemiology and risk factors of nonalcoholic fatty liver disease (NAFLD)**. *Hepatol. Int.* (2013) **7** 755-764. DOI: 10.1007/s12072-013-9480-x
8. Eslam M., Newsome P. N., Sarin S. K., Anstee Q. M., Targher G., Romero-Gomez M.. **A new definition for metabolic dysfunction-associated fatty liver disease: An international expert consensus statement**. *J. Hepatol.* (2020) **73** 202-209. DOI: 10.1016/j.jhep.2020.03.039
9. Falony G., Vieira-Silva S., Raes J.. **Richness and ecosystem development across faecal snapshots of the gut microbiota**. *Nat. Microbiol.* (2018) **3** 526-528. DOI: 10.1038/s41564-018-0143-5
10. Fan N., Peng L., Xia Z., Zhang L., Song Z., Wang Y.. **Triglycerides to high-density lipoprotein cholesterol ratio as a surrogate for nonalcoholic fatty liver disease: A cross-sectional study**. *Lipids Health Dis.* (2019) **18** 39-44. DOI: 10.1186/s12944-019-0986-7
11. Hijikata S., Hongo I., Yamaguchi T., Sekikawa Y., Nozato T.. **Infective endocarditis due to treponema pallidum: A case diagnosed using polymerase chain reaction analysis of aortic valve**. *Can. J. Cardiol.* (2019) **35** 104.e109-104.e111. DOI: 10.1016/j.cjca.2018.11.009
12. Hijikata S., Hongo I., Nakayama S. I., Yamaguchi T., Sekikawa Y., Nozato T.. **Infective endocarditis due to treponema pallidum: A case diagnosed using polymerase chain reaction analysis of aortic valve**. *Can. J. Cardiol.* (2019) 104.e109-104.e111. DOI: 10.1016/j.cjca.2018.11.009
13. Imai J., Kitamoto S., Kamada N.. **The pathogenic oral–gut–liver axis: New understandings and clinical implications**. *Expert Rev. Clin. Immunol.* (2021) **17** 727-736. DOI: 10.1080/1744666X.2021.1935877
14. Kilian M., Chapple I., Hannig M., Marsh P., Meuric V., Pedersen A.. **The oral microbiome–an update for oral healthcare professionals**. *Br. Dent. J.* (2016) **221** 657-666. DOI: 10.1038/sj.bdj.2016.865
15. Kuraji R., Sekino S., Kapila Y., Numabe Y.. **Periodontal disease–related nonalcoholic fatty liver disease and nonalcoholic steatohepatitis: An emerging concept of oral-liver axis**. *Periodontol. 2000* (2021) **87** 204-240. DOI: 10.1111/prd.12387
16. Lau E., Belda E., Picq P., Carvalho D., Ferreira-Magalhães M., Silva M. M.. **Gut microbiota changes after metabolic surgery in adult diabetic patients with mild obesity: A randomised controlled trial**. *Diabetol. Metab. Syndr.* (2021) **13** 56. DOI: 10.1186/s13098-021-00672-1
17. Liang Y., Chen H., Liu Y., Hou X., Wei L., Bao Y.. **Association of MAFLD with diabetes, chronic kidney disease, and cardiovascular disease: A 4.6-year cohort study in China**. *J. Clin. Endocrinol. Metab.* (2021) **107** 88-97. DOI: 10.1210/clinem/dgab641
18. Ling Z., Liu X., Cheng Y., Jiang X., Jiang H., Wang Y.. **Decreased diversity of the oral microbiota of patients with hepatitis b virus-induced chronic liver disease: A pilot project**. *Sci. Rep.* (2015) **5**. DOI: 10.1038/srep17098
19. Liu Y.-k., Chen V., He J.-z., Zheng X., Xu X., Zhou X.-d.. **A salivary microbiome-based auxiliary diagnostic model for type 2 diabetes mellitus**. *Arch. Oral. Biol.* (2021) **126**. DOI: 10.1016/j.archoralbio.2021.105118
20. Liu G., Tang C. M., Exley R. M.. **Non-pathogenic neisseria: Members of an abundant, multi-habitat, diverse genus**. *Microbiology* (2015) **161** 1297-1312. DOI: 10.1099/mic.0.000086
21. Loftus M., Hassouneh S. A., Yooseph S.. **Bacterial associations in the healthy human gut microbiome across populations**. *Sci. Rep.* (2021) **11** 2828. DOI: 10.1038/s41598-021-82449-0
22. Matthews D. R., Hosker J., Rudenski A., Naylor B., Treacher D., Turner R.. **Homeostasis model assessment: Insulin resistance and β-cell function from fasting plasma glucose and insulin concentrations in man**. *Diabetologia* (1985) **28** 412-419. DOI: 10.1007/BF00280883
23. Mouzaki M., Loomba R.. **An update on the role of the microbiome in non-alcoholic fatty liver disease pathogenesis, diagnosis, and treatment**. *Curr. Treat Options Gastroenterol.* (2020) **18** 270-280. DOI: 10.1007/s11938-020-00290-2
24. Negrini T., Carlos I. Z., Duque C., Caiaffa K. S., Arthur R. A.. **Interplay among the oral microbiome, oral cavity conditions, the host immune response, diabetes mellitus, and its associated-risk factors–an overview**. *Front. Oral. Health* (2021) **2**. DOI: 10.3389/froh.2021.697428
25. Park S.-Y., Hwang B.-O., Lim M., Ok S.-H., Lee S.-K., Chun K.-S.. **Oral–gut microbiome axis in gastrointestinal disease and cancer**. *Cancers* (2021) **13** 2124. DOI: 10.3390/cancers13092124
26. Peng X., Cheng L., You Y., Tang C., Ren B., Li Y.. **Oral microbiota in human systematic diseases**. *Int. J. Oral. Sci.* (2022) **14** 14. DOI: 10.1038/s41368-022-00163-7
27. Qing Y., Xu L., Cui G., Sun L., Hu X., Yang X.. **Salivary microbiome profiling reveals a dysbiotic schizophrenia-associated microbiota**. *NPJ Schizophr.* (2021) **7** 1-10. DOI: 10.1038/s41537-021-00180-1
28. Riviere G. R., Riviere K., Smith K.. **Molecular and immunological evidence of oral treponema in the human brain and their association with alzheimer's disease**. *Oral. Microbiol. Immunol.* (2002) **17** 113-118. DOI: 10.1046/j.0902-0055.2001.00100.x
29. Rosier B. T., Takahashi N., Zaura E., Krom B. P., MartÍnez-Espinosa R. M., van Breda S. G.. **The importance of nitrate reduction for oral health**. *J. Dent. Res.* (2022) **101** 887-897. DOI: 10.1177/00220345221080982
30. Sasaki N., Katagiri S., Komazaki R., Watanabe K., Maekawa S., Shiba T.. **Endotoxemia by porphyromonas gingivalis injection aggravates non-alcoholic fatty liver disease, disrupts Glucose/Lipid metabolism, and alters gut microbiota in mice**. *Front. Microbiol.* (2018) **9**. DOI: 10.3389/fmicb.2018.02470
31. Schoilew K., Ueffing H., Dalpke A., Wolff B., Frese C., Wolff D.. **Bacterial biofilm composition in healthy subjects with and without caries experience**. *J. Oral. Microbiol.* (2019) **11**. DOI: 10.1080/20002297.2019.1633194
32. Schwimmer J. B., Johnson J. S., Angeles J. E., Behling C., Belt P. H., Borecki I.. **Microbiome signatures associated with steatohepatitis and moderate to severe fibrosis in children with nonalcoholic fatty liver disease**. *Gastroenterology* (2019) **157** 1109-1122. DOI: 10.1053/j.gastro.2019.06.028
33. Sheka A. C., Adeyi O., Thompson J., Hameed B., Crawford P. A., Ikramuddin S.. **Nonalcoholic steatohepatitis: A review**. *Jama* (2020) **323** 1175-1183. DOI: 10.1001/jama.2020.2298
34. Shi X. R., Chen B. Y., Lin W. Z., Li Y. L., Wang Y. L., Liu Y.. **Microbiota in gut, oral cavity, and mitral valves are associated with rheumatic heart disease**. *Front. Cell Infect. Microbiol.* (2021) **11**. DOI: 10.3389/fcimb.2021.643092
35. Shukla M., Pereira L., Pillay A.. **Treponema**. *Molecular typing in bacterial infections* (2022) **Volume I** 191-213. DOI: 10.1007/978-3-030-74018-4
36. Si J., Lee C., Ko G.. **Oral microbiota: Microbial biomarkers of metabolic syndrome independent of host genetic factors**. *Front. Cell Infect. Microbiol.* (2017) **7**. DOI: 10.3389/fcimb.2017.00516
37. Sisk-Hackworth L., Ortiz-Velez A., Reed M. B., Kelley S. T.. **Compositional data analysis of periodontal disease microbial communities**. *Front. Microbiol.* (2021) **12**. DOI: 10.3389/fmicb.2021.617949
38. Stojanov S., Berlec A., Štrukelj B.. **The influence of probiotics on the Firmicutes/Bacteroidetes ratio in the treatment of obesity and inflammatory bowel disease**. *Microorganisms* (2020) **8**. DOI: 10.3390/microorganisms8111715
39. Takeshita T., Kageyama S., Furuta M., Tsuboi H., Takeuchi K., Shibata Y.. **Bacterial diversity in saliva and oral health-related conditions: The hisayama study**. *Sci. Rep.* (2016) **6**. DOI: 10.1038/srep22164
40. Targher G., Corey K. E., Byrne C. D., Roden M.. **The complex link between NAFLD and type 2 diabetes mellitus - mechanisms and treatments**. *Nat. Rev. Gastroenterol. Hepatol.* (2021) **18** 599-612. DOI: 10.1038/s41575-021-00448-y
41. Tonetti M. S., Greenwell H., Kornman K. S.. **Staging and grading of periodontitis: Framework and proposal of a new classification and case definition**. *J. Periodontol.* (2018) **89** S159-S172. DOI: 10.1002/JPER.18-0006
42. Van der Meulen T. A., Harmsen H. J., Bootsma H., Liefers S. C., Vich Vila A., Zhernakova A.. **Dysbiosis of the buccal mucosa microbiome in primary sjögren’s syndrome patients**. *Rheumatology* (2018) **57** 2225-2234. DOI: 10.1093/rheumatology/key215
43. Xie G., Wang X., Liu P., Wei R., Chen W., Rajani C.. **Distinctly altered gut microbiota in the progression of liver disease**. *Oncotarget* (2016) **7** 19355-19366. DOI: 10.18632/oncotarget.8466
44. Yamamoto K., Honda T., Ito T., Ishizu Y., Kuzuya T., Nakamura M.. **The relationship between oral-origin bacteria in the fecal microbiome and albumin–bilirubin grade in patients with hepatitis c**. *J. Gastroenterol. Hepatol.* (2021) **36** 790-799. DOI: 10.1111/jgh.15206
45. Yoshizawa J. M., Schafer C. A., Schafer J. J., Farrell J. J., Paster B. J., Wong D. T. W.. **Salivary biomarkers: Toward future clinical and diagnostic utilities**. *Clin. Microbiol. Rev.* (2013) **26** 781-791. DOI: 10.1128/CMR.00021-13
46. Zaura E., Keijser B. J., Huse S. M., Crielaard W.. **Defining the healthy" core microbiome" of oral microbial communities**. *BMC Microbiol.* (2009) **9**. DOI: 10.1186/1471-2180-9-259
47. Zhang C.-Z., Cheng X.-Q., Li J.-Y., Zhang P., Yi P., Xu X.. **Saliva in the diagnosis of diseases**. *Int. J. Oral. Sci.* (2016) **8** 133-137. DOI: 10.1038/ijos.2016.38
48. Zhao F., Dong T., Yuan K.-Y., Wang N.-J., Xia F.-Z., Liu D.. **Shifts in the bacterial community of supragingival plaque associated with metabolic-associated fatty liver disease**. *Front. Cell Infect. Microbiol.* (2020) **10**. DOI: 10.3389/fcimb.2020.581888
49. Zhou J., Liu L., Wu P., Zhao L., Wu Y.. **Fusobacterium nucleatum accelerates atherosclerosis**. *Front. Microbiol.* (2022) **13**. DOI: 10.3389/fmicb.2022.798685
|
---
title: Association of thyroid function with abnormal lipid metabolism in young patients
with first-episode and drug naïve major depressive disorder
authors:
- Jieqiong Hu
- Yunxin Ji
- Xiaoe Lang
- Xiang-Yang Zhang
journal: Frontiers in Psychiatry
year: 2023
pmcid: PMC9971224
doi: 10.3389/fpsyt.2023.1085105
license: CC BY 4.0
---
# Association of thyroid function with abnormal lipid metabolism in young patients with first-episode and drug naïve major depressive disorder
## Abstract
### Introduction
Abnormal lipid metabolism in patients with major depressive disorder (MDD) has received increasing attention. The coexistence of MDD and abnormal thyroid function has been intensively studied. Moreover, thyroid function is closely related to lipid metabolism. The aim of this study was to investigate the relationship between thyroid function and abnormal lipid metabolism in young patients with first-episode and drug naïve (FEDN) MDD.
### Methods
A total of 1,251 outpatients aged 18–44 years with FEDN MDD were enrolled. Demographic data were collected, and lipid and thyroid function levels were measured, including total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), thyroid stimulating hormone (TSH), free triiodothyronine (FT3), free tetraiodothyronine (FT4), anti-thyroglobulin antibody (TG-Ab), and anti-thyroid peroxidase antibody (TPO-Ab). The Hamilton Rating Scale for Depression (HAMD), Hamilton Anxiety Rating Scale (HAMA), and Positive and Negative Syndrome Scale (PANSS) positive subscale were also assessed for each patient.
### Results
Compared with young MDD patients without comorbid lipid metabolism abnormalities, patients with comorbid lipid metabolism abnormalities had higher body mass index (BMI) values, HAMD score, HAMA score, PANSS positive subscale score, TSH levels, TG-Ab levels, and TPO-Ab levels. Binary logistic regression analysis showed that TSH level, HAMD score and BMI were risk factors for abnormal lipid metabolism. TSH levels were an independent risk factor for abnormal lipid metabolism in young MDD patients. Stepwise multiple linear regression showed that both TC and LDL-C levels were positively correlated with TSH levels, HAMD and PANSS positive subscale scores, respectively. HDL-C levels were negatively correlated with TSH levels. TG levels were positively correlated with TSH and TG-Ab levels and HAMD score.
### Discussion
Our results show that thyroid function parameters, especially TSH levels, are implicated in abnormal lipid metabolism in young patients with FEDN MDD.
## 1. Introduction
Mental disorders are increasingly becoming a major cause of the global burden of disease, with depression at the top of the list [1]. As a major contributor to this burden, major depressive disorder (MDD) accounts for the largest proportion of disability-adjusted years (DALYs) for mental disorders [2], with relatively higher prevalence estimates and disability weights than many other disorders. In recent years, the prevalence and burden of MDD has increased dramatically due to the COVID-19 pandemic [3].
Epidemiological investigations have found that depression is associated with increased morbidity and mortality from cardiovascular disease [4, 5]. It is well established that dyslipidemia increases the risk of cardiovascular disease. In particular, elevated levels of total cholesterol (TC), especially low-density lipoprotein cholesterol (LDL-C) [6] and triglycerides (TG) [7], and decreased levels of high-density lipoprotein cholesterol (HDL-C) [8] are strongly associated with the development of cardiovascular atherosclerosis [9]. Abnormalities in lipid levels are also quite common in patients with MDD, but the findings are inconsistent. For example, Wei et al. found that patients with first-episode MDD had higher TG and lower HDL-C levels in the Chinese population, while no significant differences were found in LDL-C and TC levels [10]. Bharti et al. found that, compared to controls, MDD patients had higher TG levels and lower TC levels, while HDL-C levels were lower in patients older than 40 years [11]. Bot et al. found that depressed patients had higher TG levels and lower HDL-C levels, and this association was consistent across age groups [12]. These inconsistent findings may be related to factors such as race, age, and medication use. In addition, previous studies have reported that antidepressant medication can improve lipid profiles and thus reduce the risk of comorbid hyperlipidemia in MDD patients [13]. Therefore, the current status of abnormal lipid metabolism in MDD patients and related factors need further study.
Abnormal thyroid function is another common comorbidity in patients with MDD. Either hypothyroidism or hyperthyroidism was positively associated with the risk of clinical depression [14, 15]. Subclinical hypothyroidism is associated with suicide risk and psychotic symptoms in patients with MDD [16]. Thyroid function indicators are associated with suicidal ideation in patients with MDD [17]. Previous studies have found an interaction between thyroid function and lipid metabolism. Thyroid dysfunction, particularly hypothyroidism, can lead to the development of hypercholesterolemia. This is mainly due to reduced activity of the low-density lipoprotein (LDL) receptor. This is accompanied by reduced control of the sterol regulatory element binding protein 2 (SREBP-2) by triiodothyronine (T3), which regulates cholesterol biosynthesis by limiting the activity of the degrading enzyme 3-hydroxy-3-methylglutaryl coenzyme a reductase (HMG-CoA) [18]. Meanwhile, hypercholesterolemia may be lipotoxic to the pituitary-thyroid axis. Electron microscopy of mice fed a high-cholesterol diet revealed a significant accumulation of lipid droplets, cytoplasmic loss, and mitochondrial degeneration in thyroid follicular cells [19]. It was also found that compared with a mice fed a normal diet, mice fed a high cholesterol diet had elevated serum cholesterol level and TSH levels, significantly increased cholesterol levels in the pituitary gland, and changes in the cytoarchitecture of TSH in the anterior pituitary gland [20]. However, the relationship between abnormal lipid metabolism and thyroid function in patients with MDD is unclear.
In this study, we purposefully investigated the incidence of abnormal lipid metabolism and its clinical correlates, especially with thyroid function, in Chinese patients with first-episode drug naïve (FEND) MDD. We hypothesized that the prevalence of abnormal lipid metabolism would be increased in young patients with first-episode and drug naïve (FEDN) MDD and correlate with the clinical presentation of MDD. Moreover, thyroid function plays a role in the development of lipid metabolism abnormalities in young patients with FEDN MDD.
## 2.1. Subjects
This was a cross-sectional study, approved by the Ethics Committee of the First Hospital, Shanxi Medical University (No. 2016-Y27). All participants signed an informed consent form prior to enrollment.
Participants were recruited consecutively from September 2015 to December 2017 at the psychiatric outpatient department of the First Affiliated Hospital of Shanxi Medical University. Young adults were defined as 18–44 years old [21]. All participants met the following inclusion criteria: [1] age 18–44 years, Han ethnicity; [2] elementary school education or above; [3] meeting the diagnostic criteria for MDD in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV), which was validated by two experienced psychiatrists according to the Chinese version of the Structured Clinical Interview for DSM-IV (SCID); and [4] first-episode without any prior psychotropic medication such as antidepressants, antipsychotics, or anxiolytics. Exclusion criteria included: [1] other psychiatric disorders diagnosed on Axis I; [2] serious medical abnormalities, such as infection, organ dysfunction, cancer, and trauma; [3] drug/alcohol abuse or dependence, except for nicotine; and [4] pregnant or breastfeeding women.
The sample size was determined by using the following formula: n = Z2p(1−p)/d. n, sample size; Z, $95\%$ confidence interval, equal to 1.96; d, marginal error, equal to 0.05 ($5\%$); and p, expected prevalence, equal to 0.5. The estimated sample size was 384 cases. A total of 1,310 outpatients were screened and 59 patients were excluded for the following reasons: refusal to participate in the study ($$n = 15$$), severe personality disorder ($$n = 11$$), pregnant or breastfeeding women ($$n = 10$$), having a substance use disorder ($$n = 6$$), severe physical illness ($$n = 7$$), being unable to be interviewed due to an acute clinical condition ($$n = 4$$), and having other unspecified reasons ($$n = 6$$). Finally, a total of 1,251 individuals were included in this study. Thus, the sample size in this study was able to provide sufficient power for statistical analysis.
## 2.2. Clinical measures
A self-designed questionnaire was used, which included age, age of onset, sex, marital status, education, duration of illness (months), and body mass index (BMI). BMI was calculated as weight (kg) divided by the square of height (m) (kg/m2).
We used the 17-item Hamilton Rating Scale for Depression (HAMD) [22] to assess depressive symptoms. Eight of its items were scored on a 5-point Likert scale from 0 (none) to 4 (severe), and the other 9 items were scored from 0 (none) to 2 (specific description of symptoms or severity).
The 14-item Hamilton Anxiety Rating Scale (HAMA) was used to assess anxiety symptoms, each scored on a 5-point Likert scale from 0 (none) to 4 (severe) [23]. Patients with a HAMA total score of 29 or more were considered to have severe anxiety symptoms [24].
The positive subscale of the Positive and Negative Syndrome Scale (PANSS) was used to assess participants’ psychotic symptoms [25]. Each item was scored on a 7-point Likert scale from 1 (none) to 7 (extremely severe). Patients were considered to have psychotic symptoms when they scored 15 or more on the PANSS positive subscale [25, 26].
Prior to the study, two raters were simultaneously trained in the use of the aforementioned scales to ensure consistency and reliability of assessment throughout the study. Their inter-rater correlation coefficient was greater than 0.8. Both raters were blinded to the clinical status of each patient.
## 2.3. Biochemical measurements
Fasting blood was collected from each participant between 7:00 AM and 9:00 AM and immediately sent to the clinical laboratory at the hospital. Thyroid function and blood lipid levels were measured on the same morning. Thyroid function included thyroid stimulating hormone (TSH), free triiodothyronine (FT3), free tetraiodothyronine (FT4), anti-thyroglobulin antibody (TG-Ab), and anti-thyroid peroxidase antibody (TPO-Ab). Lipids included TG, TC, HDL-C, and LDL-C.
Participants in this study with TC ≥5.2 mmol/L or TG ≥1.7 mmol/L or LDL-C ≥3.4 mmol/L or HDL-C <1.0 mmol/L were considered to have abnormal lipid metabolism [27, 28].
## 2.4. Statistical analysis
The Kolmogorov–Smirnov one-sample test was used to determine the normality of all continuous variables. Normally distributed continuous variables were expressed as mean ± standard deviation (M ± SD), and non-normally distributed variables were expressed as median (quartiles) [M(Q1, Q3)]. Categorical variables were expressed as frequencies and percentages. Demographic and clinical variables were compared between groups of young MDD patients with and without comorbid lipid metabolism abnormalities. Analysis of variance (ANOVA) was used for continuous variables and chi-square test for categorical variables. Bonferroni correction was used to adjust for multiple testing. Subsequently, binary logistic regression (Forward: Wald) was performed to analyze the independent factors affecting the comorbid abnormal lipid metabolism in young MDD patients using the presence or absence of comorbid abnormal lipid metabolism as the dependent variable and the statistically significant variables from the univariate analysis as covariates. Finally, Pearson correlation analysis and further stepwise multiple regression analysis were used to examine the correlation between the levels of each lipid and thyroid function parameter and clinical variables, respectively. Bonferroni correction was used to adjust for multiple testing.
Statistical analyses were performed using SPSS (version 25.0) with a two-tailed p-value and significance level set at 0.05.
## 3.1. Prevalence of abnormal lipid metabolism in young FEDN MDD patients
A total of 1,251 individuals were included in this study. The prevalence of abnormal lipid metabolism in young MDD patients was $74.58\%$ ($\frac{933}{1}$,251). The rates of high TC, high TG, high LDL-C, and low HDL-C were $48.84\%$ ($\frac{611}{1}$,251), $61.31\%$ ($\frac{767}{1}$,251), $27.81\%$ ($\frac{348}{1}$,251), and $21.98\%$ ($\frac{275}{1}$,251), respectively.
## 3.2. Comparison of demographic and clinical characteristics between young FEDN MDD patients with and without abnormal lipid metabolism
As shown in Table 1, young MDD patients with lipid metabolism abnormalities had higher BMI, HAMD score, HAMA score, PANSS positive subscale score, TSH levels, TG-Ab levels, and TPO-Ab levels compared with patients without lipid metabolism abnormalities. However, the significance of TG-Ab and TPO-Ab failed to pass the Bonferroni correction (Bonferroni-corrected $p \leq 0.05$/14 = 0.0036). Further binary logistic regression analysis showed that TSH levels (odd ratio = 1.271, $95\%$ CI = 1.183–1.365, Wald = 43.177, $p \leq 0.0001$), HAMD score (odd ratio = 1.248, $95\%$ CI = 1.177–1.323, Wald = 55.253, $p \leq 0.0001$), and BMI (odd ratio = 1.088, $95\%$ CI = 1.008–1.174, Wald = 4.716, $p \leq 0.0001$) were risk factors for abnormal lipid metabolism.
**TABLE 1**
| Variable | MDD without abnormal lipid metabolism | MDD with abnormal lipid metabolism | F/X2 | p-Value |
| --- | --- | --- | --- | --- |
| Sample size | 318 (25.4%) | 933 (74.6%) | | |
| Sex (male, %) | 127 (39.9%) | 331 (35.5%) | 2.033 | 0.154 |
| Age (years) | 28.63 ± 8.17 | 28.87 ± 8.43 | 0.191 | 0.662 |
| Education (%) | | | 0.169 | 0.982 |
| Primary school | 41 (12.9%) | 122 (13.1%) | | |
| High school | 153 (48.1%) | 446 (47.8%) | | |
| University | 101 (31.8%) | 303 (32.5%) | | |
| Master’s | 23 (7.2%) | 62 (6.6%) | | |
| Marital status (single, %) | 128 (40.3%) | 359 (38.5%) | 0.314 | 0.575 |
| Age of onset (years) | 28.47 ± 8.09 | 28.72 ± 8.33 | 0.22 | 0.639 |
| Duration of illness (months) | 5.58 ± 4.31 | 5.65 ± 4.0 | 0.057 | 0.812 |
| BMI (kg/m2) | 23.95 ± 1.87 | 24.48 ± 1.96 | 17.289 | 0.0 |
| HAMD | 28.50 ± 2.63 | 30.77 ± 2.79 | 161.406 | 0.0 |
| HAMA | 19.58 ± 3.23 | 21.01 ± 3.36 | 43.595 | 0.0 |
| PANSS (P sub-scale) | 7.70 ± 3.07 | 8.93 ± 4.26 | 22.655 | 0.0 |
| TSH (mIU/L) | 3.51 ± 1.92 | 5.42 ± 2.53 | 150.239 | 0.0 |
| A-TG (IU/ml) | 59.48 ± 120.68 | 93.77 ± 239.61 | 5.995 | 0.014 |
| A-TPO (IU/ml) | 46.57 ± 92.00 | 75.79 ± 173.37 | 8.238 | 0.004 |
| FT3 (pmol/L) | 4.86 ± 0.72 | 4.95 ± 0.737 | 3.587 | 0.058 |
| FT4 (pmol/L) | 16.64 ± 3.04 | 16.77 ± 3.10 | 0.472 | 0.492 |
## 3.3. Correlation of TC, LDL-C, HDL-C, and TG levels and clinical variables
The results of Pearson correlation analysis between the levels of each lipid component and clinical variables are shown in Table 2. TC levels were significantly correlated with age ($r = 0.070$, $$p \leq 0.013$$), age at onset ($r = 0.070$, $$p \leq 0.013$$), duration of disease ($r = 0.064$, $$p \leq 0.024$$), BMI ($r = 0.060$, $$p \leq 0.033$$), HAMD score ($r = 0.569$, $p \leq 0.001$), HAMA score ($r = 0.291$, $p \leq 0.001$), PANSS positive subscale score ($r = 0.211$, $p \leq 0.001$), TSH level ($r = 0.559$, $p \leq 0.001$), TG-Ab level ($r = 0.098$, $p \leq 0.001$), and TPO-Ab level ($r = 0.135$, $p \leq 0.001$). However, the significance of age, age at onset, duration of disease, and BMI failed to pass the Bonferroni correction (Bonferroni-corrected $p \leq 0.05$/15 = 0.0033). Further stepwise multiple linear regression showed that HAMD scores (β = 0.472, $t = 17.693$, $p \leq 0.001$), TSH levels (β = 0.392, $t = 16.284$, $p \leq 0.001$), and PANSS positive subscale scores (β = −0.172, t = −6.92, $p \leq 0.001$) were independently associated with TC levels.
**TABLE 2**
| Variables | Age | Age of onset | Duration of illness | HAMD | HAMA | P-PANSS | TSH | A-TG | A-TPO | FT3 | FT4 | BMI | TC | LDL-C | HDL-C | TG |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Age | 1.0 | 0.999** | 0.313** | 0.034 | 0.009 | -0.042 | 0.034 | 0.008 | 0.063 | -0.034 | -0.015 | 0.048 | 0.070* | 0.055 | 0.008 | 0.032 |
| Age of onset | 0.0 | 1 | 0.278** | 0.035 | 0.010 | -0.040 | 0.031 | 0.009 | 0.034 | -0.033 | -0.013 | 0.045 | 0.070* | 0.053 | 0.008 | 0.034 |
| Duration of illness | 0.0 | 0.000 | 1 | 0.054 | 0.020 | -0.018 | 0.141** | -0.018 | 0.031 | -0.069* | -0.009 | 0.074** | 0.064* | 0.059* | -0.008 | -0.006 |
| HAMD | 0.227 | 0.222 | 0.056 | 1.0 | 0.597** | 0.528** | 0.479** | 0.092** | 0.140** | 0.034 | 0.008 | 0.061* | 0.569** | 0.395** | -0.144** | 0.151** |
| HAMA | 0.738 | 0.717 | 0.490 | 0.0 | 1 | 0.582** | 0.321** | 0.121** | 0.131** | -0.001 | 0.029 | 0.025 | 0.291** | 0.205** | -0.089** | 0.091** |
| P-PANSS | 0.141 | 0.156 | 0.528 | 0.0 | 0.000 | 1 | 0.341** | 0.092** | 0.123** | -0.010 | 0.024 | 0.027 | 0.211** | 0.134** | -0.104** | 0.118** |
| TSH | 0.225 | 0.267 | 0.000 | 0.0 | 0.000 | 0.000 | 1 | 0.213** | 0.274** | 0.033 | 0.033 | 0.156** | 0.559** | 0.362** | -0.343** | 0.169** |
| A-TG | 0.777 | 0.746 | 0.527 | 0.001 | 0.000 | 0.001 | 0.000 | 1 | 0.433** | 0.026 | -0.022 | -0.037 | 0.098** | 0.027 | -0.097** | 0.095** |
| A-TPO | 0.217 | 0.233 | 0.278 | 0.0 | 0.000 | 0.000 | 0.000 | 0.000 | 1 | 0.000 | 0.021 | -0.012 | 0.135** | 0.069* | -0.101** | 0.059* |
| FT3 | 0.223 | 0.246 | 0.014 | 0.236 | 0.981 | 0.723 | 0.243 | 0.357 | 0.997 | 1 | 0.234** | 0.019 | 0.007 | -0.024 | -0.065* | 0.025 |
| FT4 | 0.599 | 0.651 | 0.743 | 0.79 | 0.309 | 0.403 | 0.241 | 0.444 | 0.462 | 0.000 | 1 | 0.037 | -0.016 | -0.007 | -0.032 | -0.022 |
| BMI | 0.091 | 0.009 | 0.091 | 0.032 | 0.385 | 0.338 | 0.000 | 0.190 | 0.679 | 0.508 | 0.192 | 1 | 0.060* | 0.061* | -0.038 | 0.048 |
| TC | 0.013 | 0.013 | 0.024 | 0.0 | 0.000 | 0.000 | 0.000 | 0.001 | 0.000 | 0.818 | 0.577 | 0.033 | 1 | 0.589** | -0.234** | 0.224** |
| LDL-C | 0.053 | 0.059 | 0.038 | 0.0 | 0.000 | 0.000 | 0.000 | 0.348 | 0.015 | 0.387 | 0.792 | 0.030 | 0.000 | 1 | -0.180** | 0.000 |
| HDL-C | 0.791 | 0.780 | 0.780 | 0.0 | 0.002 | 0.000 | 0.000 | 0.001 | 0.000 | 0.022 | 0.251 | 0.182 | 0.000 | 0.000 | 1 | -0.137** |
| TG | 0.252 | 0.232 | 0.844 | 0.0 | 0.001 | 0.000 | 0.000 | 0.001 | 0.036 | 0.380 | 0.433 | 0.087 | 0.000 | 0.986 | 0.000 | 1 |
Low-density lipoprotein cholesterol levels were significantly associated with duration of disease ($r = 0.059$, $$p \leq 0.038$$), BMI ($r = 0.061$, $$p \leq 0.030$$), HAMD score ($r = 0.395$, $p \leq 0.001$), HAMA score ($r = 0.205$, $p \leq 0.001$), PANSS positive subscale score ($r = 0.134$, $p \leq 0.001$), TSH level ($r = 0.362$, $p \leq 0.001$), and TPO-Ab level ($r = 0.069$, $$p \leq 0.015$$). However, the significance of duration of disease, BMI and TPO-Ab level failed to pass the Bonferroni correction (Bonferroni-corrected $p \leq 0.05$/15 = 0.0033). Further stepwise multiple linear regression showed that HAMD score (β = 0.351, $t = 10.971$, $p \leq 0.001$), TSH levels (β = 0.239, $t = 8.257$, $p \leq 0.001$), and PANSS positive subscale score (β = −0.134, t = −4.466, $p \leq 0.001$) were independently associated with LDL-C levels.
High-density lipoprotein cholesterol levels were significantly associated with HAMD score (r = −0.144, $p \leq 0.001$), HAMA score (r = −0.089, $$p \leq 0.002$$), PANSS positive subscale score (r = −0.104, $p \leq 0.001$), TSH levels (r = −0.343, $p \leq 0.001$), TG-Ab levels (r = −0.097, $$p \leq 0.001$$), TPO-Ab level (r = −0.101, $p \leq 0.001$), and FT3 level (r = −0.065, $$p \leq 0.022$$). However, the significance of FT3 level failed to pass the Bonferroni correction (Bonferroni-corrected $p \leq 0.05$/15 = 0.0033). Stepwise multiple linear regression showed that only TSH levels (β = −0.343, t = −12.891, $p \leq 0.001$) were independently associated with HDL-C levels.
Triglycerides levels were correlated with HAMD score ($r = 0.151$, $p \leq 0.001$), HAMA score ($r = 0.091$, $$p \leq 0.001$$), PANSS positive subscale score ($r = 0.118$, $p \leq 0.001$), TSH level ($r = 0.169$, $p \leq 0.001$), TG-Ab level ($r = 0.095$, $$p \leq 0.001$$), and TPO-Ab levels ($r = 0.059$, $$p \leq 0.036$$). However, the significance of TPO-Ab level failed to pass the Bonferroni correction (Bonferroni-corrected $p \leq 0.05$/15 = 0.0033). Stepwise multiple linear regression showed that TSH levels (β = 0.112, $t = 3.478$, $$p \leq 0.001$$), HAMD scores (β = 0.091, $t = 2.881$, $$p \leq 0.004$$), and TG-Ab levels (β = 0.063, $t = 2.207$, $$p \leq 0.028$$) were independently associated with TG levels.
## 4. Discussion
To our knowledge, this is the first study in China on the correlation between abnormal lipid metabolism and thyroid function in young patients with FEND MDD. The main findings of this study were that patients with abnormal lipid metabolism had higher HAMD, HAMA, and PANSS positive subscale scores, higher TSH, TG-Ab, TPO-Ab, and BMI levels. In addition, TSH levels were an independent risk factor for comorbid lipid metabolism abnormalities in young patients with FEND MDD.
Our study found that depressive symptoms were more severe in young FEND MDD patients with than without lipid metabolism abnormalities. In addition, HAMD score was significantly correlated with all four lipid components, specifically, positively with TC, LDL-C, and TG levels and negatively with HDL-C levels. Previous studies have shown that the severity of depressive symptoms is associated with lipid metabolism, but with inconsistent results. For example, some studies have found that the severity of depressive symptoms is positively correlated with TG, TC, and LDL-C levels and negatively correlated with HDL-C levels, consistent with the results of our present study [29, 30]. Several other studies have found low HDL-C levels to be a predictor of the severity of depressive symptoms in patients with FEDN MDD [31]. A positive correlation between TC and LDL-C levels and the severity of depressive symptoms has also been reported [32]. In the general population, low HDL-C levels are found to be associated with the development of depressive symptoms [33], while in the general middle-aged population, high HDL-C levels or high TC levels are found to be risk factors for the development of depressive symptoms (34–36). The results of studies on the causal relationship between depression and lipid composition have also been inconsistent. In a Mendelian randomization study of a European population, one result showed a possible causal relationship between TC and depression [37], while another did not find a causal relationship between the two [38]. We also found that patients with abnormal lipid metabolism also had higher HAMA score compared to patients without abnormal lipid metabolism. Moreover, HAMA score was positively correlated with TC, LDL-C, and TG levels and negatively correlated with HDL-C levels, which is consistent with the findings of Yan et al. [ 39], who observed a negative correlation between HAMA score and HDL levels in depressed patients [40]. In addition to depressive symptoms and anxiety symptoms, we also found higher PANSS positive subscale score in MDD patients with comorbid lipid metabolism abnormalities. Similarly, psychotic symptoms were positively correlated with TC, LDL-C, and TG levels and negatively correlated with HDL-C, which is consistent with the findings of Wang et al. [ 41]. However, another cross-sectional study showed that MDD patients with comorbid severe anxiety or psychotic symptoms had higher TC, TG, and LDL-C levels and lower HDL-C levels compared to those without comorbidities [42]. Taken together, despite the differences in the results of these aforementioned studies, they largely support our findings that clinical features including depressive symptoms, anxiety symptoms, and psychotic symptoms are more pronounced in MDD patients with abnormal lipid metabolism and are significantly associated with lipid levels.
This study also found that young MDD patients with lipid metabolism abnormalities had higher levels of TSH, TPO-Ab, and TG-Ab than patients without lipid metabolism abnormalities. Furthermore, TSH levels were an independent risk factor for abnormal lipid metabolism in young MDD patients. Previous studies have shown that thyroid hormone (TH) and TSH play an important role in dyslipidemia (43–45). TH has contradictory effects on cholesterol absorption and production. TH increases cholesterol synthesis by directly inducing hepatic HMG-COA reductase (HMGCR) expression [46], but TH decreases cholesterol absorption by affecting Niemann-Pick C1-like protein (NPC1L1) in the intestine [47] and increases catabolism by enhancing free fatty acid β-oxidation to increase catabolism [48]. TSH can directly affect cholesterol synthesis by upregulating HMGCR expression and activity through the cAMP/PKA/CREB signaling pathway [49], increasing HMGCR mRNA levels [50], and increasing phosphorylated hormone-sensitive lipase (HSL) to increase lipolysis [51]. On the other hand, TSH plays an important role in the clearance of LDL and induces PI3K/AKT/SREBP2 and SREBP2/HNF4/cholesterol 7α-hydroxylase (CYP7A1) signaling pathways to inhibit hepatic bile acid synthesis [52]. Higher TSH levels are associated with a greater risk of dyslipidemia [53]. Furthermore, TSH levels are positively correlated with TC, LDL-C, and TG levels, while the correlation between TSH levels and HDL-C levels is uncertain (54–56). Recent studies have shown that lipids can also adversely affect thyroid function. The thyroid may be one of the target organs for lipotoxicity, and high circulating TG are a potential risk factor for subclinical hypothyroidism [57]. Moreover, excessive accumulation of cholesterol may induce thyroid dysfunction [58, 59]. Previous studies in depressed patients have found a positive correlation between TSH levels and HDL-C levels [60]. In depressed patients with long duration of symptoms, TSH levels were positively correlated with TC and LDL-C levels [61]. In this study, TSH levels were positively correlated with TC, TG, and LDL-C levels and negatively correlated with HDL-C levels in FEDN patients with MDD.
In addition, previous studies have reported an association between thyroid autoimmune function and depression, but the results have been inconsistent. Siegmann et al. [ 62] has predicted that more than $20\%$ of patients with autoimmune hypothyroidism will develop depression annually, suggesting a strong association between thyroid autoimmunity and depression. In contrast, a study by Bode et al. [ 63] revealed no statistically significant association between autoimmunity (mainly TPO antibody status) and depression. In this study, we found that both TPO-Ab and TG-Ab levels were positively correlated with HAMD score in young FEDN MDD patients. Previous studies have shown an association between thyroid autoimmune antibody status and lipid metabolism, with varying results. Li et al. [ 64] found that TG-Ab was positively correlated with LDL-C and HDL-C, but not with TC and TG, while TPO-Ab was positively correlated with LDL-C only. Cengiz et al. [ 65] found that TG-Ab and TPO-Ab were positively correlated with TC, TG, and LDL-C, respectively, and not with HDL-C. Studies also found that TG-Ab was negatively correlated with TG, whereas TPO-Ab was negatively correlated with HDL-C [66, 67]. In this study, TG-Ab was found to be positively correlated with TC and TG and negatively correlated with HDL-C in young FEDN MDD patients. TPO-Ab was positively correlated with TC, TG, and LDL-C and negatively correlated with HDL-C. Taken together, these findings suggested a correlation between thyroid autoimmune antibodies and lipid components in MDD patients, while only TG-Ab was independently associated with TG.
This study had several limitations. First, this is a cross-sectional study and cannot demonstrate a causal relationship between abnormal lipid metabolism and thyroid function in young patients with FEDN MDD. Second, although the potential confounders had been adjusted for, we did not exclude factors influencing lipid metabolism, such as dietary structure, exercise status, smoking, and alcohol consumption, which may had partially biased the results. Third, a comprehensive assessment of thyroid status may not be possible due to the lack of imaging data of the thyroid gland. Fourth, we excluded pregnant and lactating women, taking into account different endocrine levels. However, abnormalities of thyroid function and lipid metabolism in MDD patients during pregnancy and postpartum need to be investigated in future studies. Fifth, as this study was conducted in only one general hospital in China, the results of this study may not be generalizable to other settings. Finally, there was no healthy control group in this study. Therefore, this is only a preliminary result.
## 5. Conclusion
In summary, in this study, we found that the comorbid lipid metabolism abnormalities in young patients with FEDN MDD were associated with thyroid function parameters, especially TSH levels, which was an independent risk factor. Moreover, TSH levels were associated with TC, LDL-C, HDL-C, and TG levels in young patients with FEDN MDD. Simultaneously, TG-Ab levels were independently associated with TG levels. In the future, a longitudinal study needs to be designed that includes healthy controls and excludes confounding factors that may affect abnormal lipid metabolism, such as the presence of a high-fat diet, alcohol consumption, and sedentary lifestyle. We also need to refine thyroid imaging for a more comprehensive assessment in order to perform an exhaustive analysis according to different thyroid states. This allows us to explore in greater depth the relationship between thyroid function and abnormal lipid metabolism in MDD patients to further validate the results of this study. Abnormal lipid metabolism in young MDD patients is significantly associated with TSH levels and there is potential for future interventions in TSH to prevent dyslipidemia and thereby reduce the incidence of cardiovascular events.
## Data availability statement
The original contributions presented in this study are included in this article/supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
This study was reviewed and approved by the Ethics Committee of the First Hospital, Shanxi Medical University (No. 2016-Y27). All participants signed an informed consent form prior to enrollment.
## Author contributions
JH: conceptualization, methodology, analysis and interpretation of data for the work, and drafting the work. YJ: funding acquisition, resources, and supervision. XL: investigation and data curation. X-YZ: substantial contributions to the conception or design of the work and revising the work critically for important intellectual content. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Patel V, Saxena S, Lund C, Thornicroft G, Baingana F, Bolton P. **The Lancet Commission on global mental health and sustainable development.**. (2018) **392** 1553-98. DOI: 10.1016/S0140-6736(18)31612-X
2. **Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019.**. (2022) **9** 137-50. DOI: 10.1016/S2215-0366(21)00395-3
3. **Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic.**. (2021) **398** 1700-12. DOI: 10.1016/S0140-6736(21)02143-7
4. Meng R, Yu C, Liu N, He M, Lv J, Guo Y. **Association of depression with all-cause and cardiovascular disease mortality among adults in China.**. (2020) **3**. DOI: 10.1001/jamanetworkopen.2019.21043
5. Melo A, Dippenaar I, Johnson S, Weaver N, de Assis Acurcio F, Malta D. **All-cause and cause-specific mortality among people with severe mental illness in Brazil’s public health system, 2000-15: a retrospective study.**. (2022) **9** 771-81. DOI: 10.1016/S2215-0366(22)00237-1
6. Ference B, Ginsberg H, Graham I, Ray K, Packard C, Bruckert E. **Low-density lipoproteins cause atherosclerotic cardiovascular disease. 1. Evidence from genetic, epidemiologic, and clinical studies. A consensus statement from the European atherosclerosis society consensus panel.**. (2017) **38** 2459-72. DOI: 10.1093/eurheartj/ehx144
7. Ginsberg H, Packard C, Chapman M, Borén J, Aguilar-Salinas C, Averna M. **Triglyceride-rich lipoproteins and their remnants: metabolic insights, role in atherosclerotic cardiovascular disease, and emerging therapeutic strategies-a consensus statement from the European atherosclerosis society.**. (2021) **42** 4791-806. DOI: 10.1093/eurheartj/ehab551
8. Magnoni M, Andreini D, Pirillo A, Uboldi P, Latini R, Catapano A. **Predictive value of HDL function in patients with coronary artery disease: relationship with coronary plaque characteristics and clinical events.**. (2022) **54** 1036-46. DOI: 10.1080/07853890.2022.2063374
9. Patel J, Pallazola V, Dudum R, Greenland P, McEvoy J, Blumenthal R. **Assessment of coronary artery calcium scoring to guide statin therapy allocation according to risk-enhancing factors: the multi-ethnic study of atherosclerosis.**. (2021) **6** 1161-70. DOI: 10.1001/jamacardio.2021.2321
10. Wei Y, Cai D, Liu J, Liu R, Wang S, Tang Y. **Cholesterol and triglyceride levels in first-episode patients with major depressive disorder: a meta-analysis of case-control studies.**. (2020) **266** 465-72. DOI: 10.1016/j.jad.2020.01.114
11. Bharti V, Bhardwaj A, Hood K, Elias D, Metcalfe A, Kim JS. **A systematic review and meta-analysis of lipid metabolomic signatures of major depressive disorder.**. (2021) **139** 197-205. DOI: 10.1016/j.jpsychires.2021.05.036
12. Bot M, Milaneschi Y, Al-Shehri T, Amin N, Garmaeva S, Onderwater G. **Metabolomics profile in depression: a pooled analysis of 230 metabolic markers in 5283 cases with depression and 10,145 controls.**. (2020) **87** 409-18. DOI: 10.1016/j.biopsych.2019.08.016
13. Zhang Z, Du Y, Chen L, Liu Y, Du B. **Effects of the selective serotonin reuptake inhibitor fluoxetine on glucose metabolism: a systematic review.**. (2022) **73**. DOI: 10.1016/j.ajp.2022.103092
14. Tang R, Wang J, Yang L, Ding X, Zhong Y, Pan J. **Subclinical hypothyroidism and depression: a systematic review and meta-analysis.**. (2019) **10**. DOI: 10.3389/fendo.2019.00340
15. Bode H, Ivens B, Bschor T, Schwarzer G, Henssler J, Baethge C. **Hyperthyroidism and clinical depression: a systematic review and meta-analysis.**. (2022) **12**. DOI: 10.1038/s41398-022-02121-7
16. Lang X, Hou X, Shangguan F, Zhang X. **Prevalence and clinical correlates of subclinical hypothyroidism in first-episode drug-naive patients with major depressive disorder in a large sample of Chinese.**. (2020) **263** 507-15. DOI: 10.1016/j.jad.2019.11.004
17. Shen Y, Wu F, Zhou Y, Ma Y, Huang X, Ning Y. **Association of thyroid dysfunction with suicide attempts in first-episode and drug naïve patients with major depressive disorder.**. (2019) **259** 180-5. DOI: 10.1016/j.jad.2019.08.067
18. Duntas L, Brenta GA. **Renewed focus on the association between thyroid hormones and lipid metabolism.**. (2018) **9**. DOI: 10.3389/fendo.2018.00511
19. Ayuob N, El-Hawwary A, Huwait E, Mubarak W, Balgoon M. **Red grape juice protects the rat thyroid gland against hypercholesterolemic changes. Ultrastructural and biochemical evidences.**. (2019) **60** 921-9. PMID: 31912105
20. Yang J, Zhang X, Liu Z, Yuan Z, Song Y, Shao S. **High-cholesterol diet disrupts the levels of hormones derived from anterior pituitary basophilic cells.**. (2016) **28**. DOI: 10.1111/jne.12369
21. Jain V, Minhas A, Morris A, Greene S, Pandey A, Khan S. **Demographic and regional trends of heart failure-related mortality in young adults in the US, 1999-2019.**. (2022) **7** 900-4. DOI: 10.1001/jamacardio.2022.2213
22. Hamilton M. **A rating scale for depression.**. (1960) **23** 56-62. DOI: 10.1136/jnnp.23.1.56
23. Hamilton M. **The assessment of anxiety states by rating.**. (1959) **32** 50-5. DOI: 10.1111/j.2044-8341.1959.tb00467.x
24. Eidan A, Al-Harmoosh R, Al-Amarei H. **Estimation of IL-6, INFγ, and lipid profile in suicidal and nonsuicidal adults with major depressive disorder.**. (2019) **39** 181-9. DOI: 10.1089/jir.2018.0134
25. Kay S, Fiszbein A, Opler L. **The positive and negative syndrome scale (PANSS) for schizophrenia.**. (1987) **13** 261-76. DOI: 10.1093/schbul/13.2.261
26. Purnine D, Carey K, Maisto S, Carey M. **Assessing positive and negative symptoms in outpatients with schizophrenia and mood disorders.**. (2000) **188** 653-61. DOI: 10.1097/00005053-200010000-00003
27. **[2016 Chinese guideline for the management of dyslipidemia in adults].**. (2016) **44** 833-53. DOI: 10.3760/cma.j.issn.0253-3758.2016.10.005
28. Liu Y, Kong X, Wang W, Fan F, Zhang Y, Zhao M. **Association of peripheral differential leukocyte counts with dyslipidemia risk in Chinese patients with hypertension: insight from the China stroke primary prevention trial.**. (2017) **58** 256-66. DOI: 10.1194/jlr.P067686
29. Enko D, Brandmayr W, Halwachs-Baumann G, Schnedl W, Meinitzer A, Kriegshäuser G. **Prospective plasma lipid profiling in individuals with and without depression.**. (2018) **17**. DOI: 10.1186/s12944-018-0796-3
30. Wagner C, Musenbichler C, Böhm L, Färber K, Fischer A, von Nippold F. **LDL cholesterol relates to depression, its severity, and the prospective course.**. (2019) **92** 405-11. DOI: 10.1016/j.pnpbp.2019.01.010
31. Kuwano N, Kato T, Setoyama D, Sato-Kasai M, Shimokawa N, Hayakawa K. **Tryptophan-kynurenine and lipid related metabolites as blood biomarkers for first-episode drug-naïve patients with major depressive disorder: an exploratory pilot case-control study.**. (2018) **231** 74-82. DOI: 10.1016/j.jad.2018.01.014
32. Kuperberg M, Köhler-Forsberg O, Shannon A, George N, Greenebaum S, Bowden C. **Cardiometabolic risk markers during mood-stabilizing treatment: correlation with drug-specific effects, depressive symptoms and treatment response.**. (2022) **300** 41-9. DOI: 10.1016/j.jad.2021.12.047
33. Gowey M, Khodneva Y, Tison S, Carson A, Cherrington A, Howard V. **Depressive symptoms, perceived stress, and metabolic health: the REGARDS study.**. (2019) **43** 615-32. DOI: 10.1038/s41366-018-0270-3
34. Shin H, Kang G, Kang H, Kim S, Shin I, Yoon J. **Relationships between high-density lipoprotein cholesterol and depressive symptoms: findings of the Korean national health and nutrition examination survey (KNHANES).**. (2016) **241** 172-4. DOI: 10.1016/j.psychres.2016.05.003
35. Zainal N, Newman M. **Prospective network analysis of proinflammatory proteins, lipid markers, and depression components in midlife community women.**. (2022) **4** 1-12. DOI: 10.1017/S003329172200232X
36. Cepeda M, Kern D, Blacketer C, Drevets W. **Low levels of cholesterol and the cholesterol type are not associated with depression: results of a cross-sectional NHANES study.**. (2020) **14** 515-21. DOI: 10.1016/j.jacl.2020.06.001
37. Khandaker G, Zuber V, Rees J, Carvalho L, Mason A, Foley C. **Shared mechanisms between coronary heart disease and depression: findings from a large UK general population-based cohort.**. (2020) **25** 1477-86. DOI: 10.1038/s41380-019-0395-3
38. Zhang M, Chen J, Yin Z, Wang L, Peng L. **The association between depression and metabolic syndrome and its components: a bidirectional two-sample Mendelian randomization study.**. (2021) **11**. DOI: 10.1038/s41398-021-01759-z
39. Yan L, Kang C, Wang X, Yang L, Zhao N, Zhang X. **Association of serum lipid levels with psychotic symptoms in young, first-episode and drug naïve outpatients with major depressive disorder: a large-scale cross-sectional study.**. (2022) **317**. DOI: 10.1016/j.psychres.2022.114864
40. Wang J, Jiang C, Chen L, Wu S, Lin J, Gao L. **A cross-sectional study to investigate the correlation between depression comorbid with anxiety and serum lipid levels.**. (2016) **69** 163-8. DOI: 10.1016/j.comppsych.2016.05.005
41. Wang X, Yang L, Liu J, Kang C, Zheng Y, Qiu S. **Association of serum lipid levels with psychotic symptoms in first-episode and drug naïve outpatients with major depressive disorder: a large-scale cross-sectional study.**. (2022) **297** 321-6. DOI: 10.1016/j.jad.2021.10.053
42. Liu W, Wu Z, Sun M, Zhang S, Yuan J, Zhu D. **Association between fasting blood glucose and thyroid stimulating hormones and suicidal tendency and disease severity in patients with major depressive disorder.**. (2022) **22** 635-42. PMID: 35238287
43. Liu F, Hwang J, Kuo C, Ko Y, Chen S, Lin J. **Subclinical hypothyroidism and metabolic risk factors association: a health examination-based study in northern Taiwan.**. (2018) **41** 52-8. DOI: 10.1016/j.bj.2018.02.002
44. Martin S, Daya N, Lutsey P, Matsushita K, Fretz A, McEvoy J. **Thyroid function, cardiovascular risk factors, and incident atherosclerotic cardiovascular disease: the atherosclerosis risk in communities (ARIC) study.**. (2017) **102** 3306-15. DOI: 10.1210/jc.2017-00986
45. Zhao M, Liu L, Wang F, Yuan Z, Zhang X, Xu C. **A worthy finding: decrease in total cholesterol and low-density lipoprotein cholesterol in treated mild subclinical hypothyroidism.**. (2016) **26** 1019-29. DOI: 10.1089/thy.2016.0010
46. Choi J, Choi H. **The regulatory effects of thyroid hormone on the activity of 3-hydroxy-3-methylglutaryl coenzyme A reductase.**. (2000) **26** 1-21. DOI: 10.1080/07435800009040142
47. Gälman C, Bonde Y, Matasconi M, Angelin B, Rudling M. **Dramatically increased intestinal absorption of cholesterol following hypophysectomy is normalized by thyroid hormone.**. (2008) **134** 1127-36. DOI: 10.1053/j.gastro.2008.01.032
48. Sinha R, You S, Zhou J, Siddique M, Bay B, Zhu X. **Thyroid hormone stimulates hepatic lipid catabolism via activation of autophagy.**. (2012) **122** 2428-38. DOI: 10.1172/JCI60580
49. Tian L, Song Y, Xing M, Zhang W, Ning G, Li X. **A novel role for thyroid-stimulating hormone: up-regulation of hepatic 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase expression through the cyclic adenosine monophosphate/protein kinase A/cyclic adenosine monophosphate-responsive element binding prote.**. (2010) **52** 1401-9. DOI: 10.1002/hep.23800
50. Moreno-Navarrete J, Moreno M, Ortega F, Xifra G, Hong S, Asara J. **TSHB mRNA is linked to cholesterol metabolism in adipose tissue.**. (2017) **31** 4482-91. DOI: 10.1096/fj.201700161R
51. Gagnon A, Antunes T, Ly T, Pongsuwan P, Gavin C, Lochnan H. **Thyroid-stimulating hormone stimulates lipolysis in adipocytes in culture and raises serum free fatty acid levels in vivo.**. (2010) **59** 547-53. DOI: 10.1016/j.metabol.2009.08.018
52. Song Y, Xu C, Shao S, Liu J, Xing W, Xu J. **Thyroid-stimulating hormone regulates hepatic bile acid homeostasis via SREBP-2/HNF-4α/CYP7A1 axis.**. (2015) **62** 1171-9. DOI: 10.1016/j.jhep.2014.12.006
53. Chang Y, Hua S, Chang C, Kao W, Lee H, Chuang L. **High TSH level within normal range is associated with obesity, dyslipidemia, hypertension, inflammation, hypercoagulability, and the metabolic syndrome: a novel cardiometabolic marker.**. (2019) **8**. DOI: 10.3390/jcm8060817
54. Jain R. **Associations between the levels of thyroid hormones and lipid/lipoprotein levels: data from national health and nutrition examination survey 2007-2012.**. (2017) **53** 133-44. DOI: 10.1016/j.etap.2017.05.002
55. Luxia L, Jingfang L, Songbo F, Xulei T, Lihua M, Weiming S. **Correlation between serum TSH levels within normal range and serum lipid profile.**. (2021) **53** 32-40. DOI: 10.1055/a-1191-7953
56. Kuś A, Marouli E, Del Greco MF, Chaker L, Bednarczuk T, Peeters R. **Variation in normal range thyroid function affects serum cholesterol levels, blood pressure, and type 2 diabetes risk: a mendelian randomization study.**. (2021) **31** 721-31. DOI: 10.1089/thy.2020.0393
57. Zhao M, Tang X, Yang T, Zhang B, Guan Q, Shao S. **Lipotoxicity, a potential risk factor for the increasing prevalence of subclinical hypothyroidism?**. (2015) **100** 1887-94. DOI: 10.1210/jc.2014-3987
58. Song Y, Liu J, Zhao K, Gao L, Zhao J. **Cholesterol-induced toxicity: an integrated view of the role of cholesterol in multiple diseases.**. (2021) **33** 1911-25. DOI: 10.1016/j.cmet.2021.09.001
59. Shao S, Zhao Y, Song Y, Xu C, Yang J, Xuan S. **Dietary high-fat lard intake induces thyroid dysfunction and abnormal morphology in rats.**. (2014) **35** 1411-20. DOI: 10.1038/aps.2014.82
60. Peng R, Dai W, Li Y. **Low serum free thyroxine level is correlated with lipid profile in depressive patients with suicide attempt.**. (2018) **266** 111-5. DOI: 10.1016/j.psychres.2018.05.059
61. Peng R, Li Y. **Low serum thyroid-stimulating hormone levels are associated with lipid profile in depressive patients with long symptom duration.**. (2017) **217** 99-104. DOI: 10.1016/j.jad.2017.03.055
62. Siegmann E, Müller H, Luecke C, Philipsen A, Kornhuber J, Grömer T. **Association of depression and anxiety disorders with autoimmune thyroiditis: a systematic review and meta-analysis.**. (2018) **75** 577-84. DOI: 10.1001/jamapsychiatry.2018.0190
63. Bode H, Ivens B, Bschor T, Schwarzer G, Henssler J, Baethge C. **Association of hypothyroidism and clinical depression: a systematic review and meta-analysis.**. (2021) **78** 1375-83. DOI: 10.1001/jamapsychiatry.2021.2506
64. Li J, Wang Z, Liu H, Fu J, Qin F, Guan H. **Serum lipids are novel predictors for thyroid autoimmunity in the general population with normal TSH levels from a cross-sectional study.**. (2021) **73** 331-8. DOI: 10.1007/s12020-021-02731-z
65. Cengiz H, Demirci T, Varim C, Tamer A. **The effect of thyroid autoimmunity on dyslipidemia in patients with euthyroid hashimoto thyroiditis.**. (2021) **37** 1365-70. DOI: 10.12669/pjms.37.5.3883
66. Tamer G, Mert M, Tamer I, Mesci B, Kilic D, Arik S. **Effects of thyroid autoimmunity on abdominal obesity and hyperlipidaemia.**. (2011) **62** 421-8. PMID: 22069103
67. Zhang J, Gao Y, Li Y, Teng D, Xue Y, Yan L. **The presence of serum tgab suggests lower risks for glucose and lipid metabolic disorders in euthyroid general population from a national survey.**. (2020) **11**. DOI: 10.3389/fendo.2020.00139
|
---
title: Advanced glycation end products regulate the receptor of AGEs epigenetically
authors:
- Xiaoqing Wu
- Xuanren Shi
- Xiaoyong Chen
- Zhanhai Yin
journal: Frontiers in Cell and Developmental Biology
year: 2023
pmcid: PMC9971228
doi: 10.3389/fcell.2023.1062229
license: CC BY 4.0
---
# Advanced glycation end products regulate the receptor of AGEs epigenetically
## Abstract
Advanced glycation end-products (AGEs) can boost their receptor of AGE (RAGE) expression through the downstream signaling pathway to facilitate AGE–RAGE interaction. In this regulation process, the primary signaling pathways are NF-κB and STAT3. However, the inhibition of these transcription factors cannot completely block the upregulation of RAGE, which indicates AGEs may also impact RAGE expression via other pathways. In this study, we revealed that AGEs can exhibit epigenetic impacts on RAGE expression. Here, we used carboxymethyl-lysine (CML) and carboxyethyl-lysine (CEL) to treat liver cells and discovered that AGEs can promote the demethylation of the RAGE promoter region. To verify this epigenetic modification, we employed dCAS9-DNMT3a with sgRNA to specifically modify the RAGE promoter region against the effect of carboxymethyl-lysine and carboxyethyl-lysine. The elevated RAGE expressions were partially repressed after AGE-induced hypomethylation statuses were reversed. Additionally, TET1 were also upregulated in AGE-treated cells, indicating AGEs may epigenetically modulate RAGE through the elevating TET1 level.
## 1 Introduction
Advanced glycation end-products (AGEs) are one kind of heterogeneous compounds that are generated from the Maillard reaction. The Maillard reaction is a series of non-enzymatic glycation that occurs between reducing sugars and free amino groups, lipids, or proteins. AGEs are generated either from dietary food processing or metabolized biological activities in normal physiological processes. Long-term intake of AGEs leads to the accumulation of AGEs in our body, which can trigger the outbreak of multiple chronic abnormalities and toxic pathogenesis (Kang and Yang, 2020; Perrone et al., 2020), e.g., diabetes and various diabetic complications. Some AGEs have been chemically well characterized in the body, such as carboxyethyl-lysine (CEL), and carboxymethyl-lysine (CML).
AGE mediated the damage via the interaction with its cell surface receptor—RAGE. The binding between AGEs and RAGE can promote the activation of inflammatory response through the NF-κB signaling pathway and cellular apoptosis and meanwhile, also elevate the VEGF level to increase vascular endothelial permeability (Ma et al., 2007; Kang and Yang, 2020). Most importantly, AGEs can also facilitate RAGE upregulation via the downstream cellular signal pathways, including NF-κB and STAT3 signaling pathways (Zhou et al., 2020). The AGE–RAGE axis not only brings about the upregulated inflammatory genes expression, but also promotes the positive feedback loops, in which the sustained activation of transcription factors facilitates RAGE expression (Dariya and Nagaraju, 2020). NF-κB is a critical transcription factor for RAGE upregulation in cells (Ott et al., 2014). NF-κB activation is considered an important proinflammatory and proapoptotic signaling to induce AGE-mediated inflammation and apoptosis (Kang and Yang, 2020). RAGE overexpression is also closely associated with the activation of the STAT3 pathway (Abo El-Nasr et al., 2020; Chiappalupi et al., 2020). The interaction of AGEs and RAGE triggers the activation of the STAT3/Pim1/NFAT axis, which can maintain the RAGE high transcription activity by functioning as a positive feedback loop to augment RAGE expression (Meloche et al., 2011).
Epigenetic modification is also a significant regulation for gene expression, especially gene promoter methylation (Dawson and Kouzarides, 2012; Henikoff and Greally, 2016). AGE-mediated epigenetic impacts have been demonstrated to be associated with diabetes and other chronic diseases (Kang and Yang, 2020; Perrone et al., 2020), which are conducted by their interaction with RAGE (Khalid et al., 2022). AGEs can also lead to the downregulation of sirtuin 1 (SIRT1), which provokes high acetylation of the factors, such as STAT3, NF-κB (p65), and FOXO4 (Perrone et al., 2020). Diabetes augments the expression of matrix metalloproteinase (MMP)-9 in the skin and its keratinocytes, and a high level of MMP-9 leads to impairments to skin wound healing, in which AGE-induced demethylation of MMP-9 promoter plays a key role (Perrone et al., 2020). On the other hand, the methylation status in the RAGE promoter region exhibits key effects on multiple diseases (Wang et al., 2019). Previous studies also revealed that RAGE promoter methylation status can function as an indicator for the diabetic retinal inflammation and RAGE gene promoters in DR patients, showing a lower methylation rate than that of healthy adults (Kan et al., 2018). However, few studies have addressed the correlation between AGEs and the status of RAGE promoter methylation. Here, we reveal that AGEs, such as carboxymethyl-lysine and carboxyethyl-lysine, can impact the methylation status of the RAGE promoter and consequently modulate RAGE expression in cells, in which TET1 may play an important role. These results further developed our comprehension of the interaction between AGEs and RAGE.
## 2.1 Cell culture and reagents
Human liver cell lines, LO2, were immortal non-tumor cell lines isolated from normal liver tissues. The LO2 cells were cultured in DMEM (Gibco), with the supplementation of $10\%$ FBS and $1\%$ antibiotics (streptomycin-penicillin) at 37°C, in the cell-culture incubator with $5\%$ CO2. Cell passaging was performed by trypsin-EDTA solution ($0.25\%$) to digest the cells from the culture dish.
CML and CEL were purchased from Toronto Research Chemicals (Canada). NF-κB and STAT3 inhibitors, niclosamide, and QNZ were purchased from Selleck Chemicals and prepared in DMSO for subsequent dilution to the specific dose in the study.
## 2.2 Cell viability assay
The cell viability was measured by the CCK-8 Reagent (Thermo Scientific) according to the manufacturer’s instructions. Cells were treated for 24 h, washed with PBS, and subsequently added to CCK-8 solution. The absorbance was measured at 450 nm. All experiments were repeated at least three times.
## 2.3 qPCR
Total RNA was extracted using RNAzol (Molecular Research Center, OH, United States) according to the instructions. qPCR was performed with iTaq universal SYBR green supermix (Bio-Rad, Hercules, United States) in a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). Briefly, the incubation process includes two main steps. Step one is 94°C for 3 min, and step two is 94°C for 10 s and 60°C for 30 s with 40 cycles. The expression of RAGE was calculated by using the 2−ΔΔCT formula. The primers for qPCR were listed in Table 1, and all samples were performed in triplicate.
**TABLE 1**
| Gene name (RT-qPCR) | Sense (5'-3') | Sense (5'-3').1 | Anti-sense (5'-3') |
| --- | --- | --- | --- |
| Tet1 | CAGGACCAAGTGTTGCTGCTGT | CAGGACCAAGTGTTGCTGCTGT | GACACCCATGAGAGCTTTTCCC |
| RAGE | CACCTTCTCCTGTAGCTTCAGC | CACCTTCTCCTGTAGCTTCAGC | AGGAGCTACTGCTCCACCTTCT |
| GAPDH | AGGTCGGTG TGAACGGATTTG | AGGTCGGTG TGAACGGATTTG | GGGGTCGTTGATGGC AACA |
| BSP primers | ATTTTTGGATAGAGGATATGGG | ATTTTTGGATAGAGGATATGGG | ATTCTATTAATTTAAAATAAACT |
| | | Sequence | Sequence |
| sgRNA-targeting RAGE promoter | sgRNA-targeting RAGE promoter | TCTTTCACGAAGTTCCAAAC | TCTTTCACGAAGTTCCAAAC |
| Scrambled sgRNA | Scrambled sgRNA | CCCCCGGGGGAAAAATTTTT | CCCCCGGGGGAAAAATTTTT |
## 2.4 Luciferase reporter assay
NF-κB and STAT3 luciferase reporter plasmids (YESEN, Shanghai, China) were used to test the activity of NF-κB and STAT3. Transient transfection of liver cells was performed using lipofectamine 3000 (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. The cells were transfected with the luciferase plasmid. pRL-TK renilla luciferase was used as the internal control. The transfected cells were harvested and lysed according to the protocol of dual-luciferase reporter assay kit (Promega, Madison, WI) to analyze the luciferase activity via a luminometer fluorescence reader.
## 2.5 Immunofluorescence assay
The attached cells were fixed by $4\%$ formaldehyde in PBS. After the fix, cells were blocked by $1\%$ BSA. Then, the cells were incubated in the solution of the anti-RAGE primary antibody conjugated to FITC (2 μg/ml) (Santa Cruz Biotechnology), followed by counterstaining with Hochest solution (Invitrogen). Finally, these cells were analyzed under fluorescent microscopy. For flow cytometry analysis, the cells were suspended by trypin and stained as given in the aforementioned protocol.
## 2.6 DNA demethylation blocking, bisulfite conversion, and sequencing
Lentivirus consisting of the sg-RNA-targeting RAGE promoter and dCAS9-DNMT3A (Addgene #84476, #52963) were transducted into cells to specifically prevent DNA demethylation (Kang et al., 2021). Bisulfite modification of DNA was conducted by using EpiTect bisulfite kit (QIAGEN), following the manufacturer’s instructions. Briefly, the genomic DNA were extracted from cells and then incubated with bisulfite mix for 85 min at 60°C. Bisulfated DNA was purified and used as the template for the following PCR and sequencing. The CpG island in the promoter region (Upstream of exon 1 by 1481–1584bp) of modified DNA was amplified by PCR and then purified to sub-clone into the TA cloning vector plasmid. Seven clones in each indicated group were chosen for sequencing randomly. The sequencing results were analyzed in the bisulfite result analysis website: http://quma.cdb.riken.jp/. The information of sgRNA and BSP primers were listed in Table 1.
## 2.7 Statistical analysis
All experiments were conducted in at least three independent trials. Student’s t-test was used for the analysis. p-value less than 0.05, 0.01 and 0.001 were presented as “*,” “**,” and “***,” respectively. p-values less than 0.05 were regarded as being significant statistically.
## 3.1 CML and CEL augment the expression of RAGE in LO2 cells
To evaluate the modulation of CML and CEL on RAGE expression, we incubated LO2 cells with different concentrations of CML and CEL to detect the expression of RAGE, respectively. It was found that 10–50 μM CML or CEL dose-dependently increased cellular RAGE expression, which further increased up to 30 μM and did not bring about any stronger expression of RAGE (Figures 1A, B). The RAGE expression was concentration dependent associated with AGEs—CML and CEL. To preclude the interference from the potential effects of CML and CEL on the viability of LO2 cells, we conducted CCK-8 assay with a range of above mentioned concentrations. CML and CEL showed weak influence on the LO2 cell lines (Figures 1A, B). Hence, 30 μM (non-cytotoxic concentration) of CML or CEL was adopted in subsequent experiments. The positive modulation of RAGE by CML and CEL in LO2 cells was also confirmed by fluorescence microscopy and flow cytometry analysis. The RAGE on the cell membrane was stained by antibody conjugated with FITC. Compared with the control, CML/CEL-treated cells exhibit stronger florescence intensity (Figures 2A, B), directly manifesting the upregulation of RAGE, which also confirmed by the flow cytometer (Figures 2C, D).
**FIGURE 1:** *CML and CEL can upregulate RAGE expression in LO2 cells. (A,B) RAGE expression in cells with different treatment doses of CML and CEL for 24 h. (C,D) Effects of different doses of CML and CEL on cell viability. All dates are shown as the mean ± SD.* **FIGURE 2:** *Expression of RAGE were detected by immunofluorescence assay. Control and CML/CEL cells were stained with an antibody of anti-RAGE-FITC and Hoechst 33342, and examined by fluorescence microscopy (A,B) and flow cytometry (C,D).*
## 3.2 RAGE upregulation by CML and CEL is not mediated solely through NF-κB and STAT3
Previous studies have revealed that the AGE–RAGE axis in the intracellular and extracellular signal transduction eventually give rise to the activation of NF-κB and STAT3. The activation of NF-κB and STAT3 triggers the successive transcription of RAGE and further boosts various RAGE-reliant signaling pathways. These two factors play a critical role in the vicious cycle. To clarify whether the activation of NF-κB and STAT3 is the sole modulation for CML/CEL-mediated RAGE upregulation, niclosamide and QNZ, widely used STAT3 and NF-κB specific inhibitors, were used to see if it could completely block the effect produced by these AGEs.
STAT3 and NF-κB luciferase reporter plasmids were used to detect the inhibitory effect of the drug. These two transcription factors’ activity was repressed dose-dependently (Figures 3A, B). Here, we chose 2 μM and 20 nM for STAT3 and NF-κB inhibition in the following study.
**FIGURE 3:** *Inhibition of NF-κB and STAT3 cannot completely block RAGE upregulation in cells. (A,B) Inhibition of NF-κB and STAT3 by their inhibitors. (C,D) RAGE expression of the indicated treatment in cells. (D,E) Flow cytometry analysis of RAGE expression. All dates are shown as the mean ± SD. *p < 0.05, ***p < 0.001.*
qPCR analysis found STAT3 and NF-κB inhibitors significantly repressed RAGE levels in CML and CEL treated cells, while these inhibitors did not completely block the CML/CEL-induced upregulation of RAGE (Figure 3C, D), and the flow cytometry of the cell stain with the anti-RAGE antibody conjugated with FITC also showed the similar results (Figure 3E, F), suggesting the existence of an NF-κB/STAT3-independent mechanism underlying CML/CEL’s effect.
## 3.3 AGEs can regulate RAGE expression epigenetically
The RAGE gene promoter methylation status is closely associated with AGE-induced chronic disease, such as diabetic complications (Kan et al., 2018), and promoter hypo-/hypermethylation acts as an important epigenetic modulation for gene expression. Having established AGE-induced concentration-dependent upregulation of RAGE expression in cells, we sought to determine their impact on the methylation status in the RAGE promoter region. Here, we used CML and CEL for the following methylation study. To conduct the epigenetic modification at a specific site, we use the dCAS9-DNMT3a to mediate methylation in the RAGE promoter region (Figure 4A), and BSP were preformed to detect the change of the methylation status. The BSP results revealed that CML and CEL prompt the hypomethylation status in the RAGE promoter region, and the average methylation rate of individual CpGs decreased by $23\%$ and $25\%$, respectively (Figures 4B, C).
**FIGURE 4:** *Change of the RAGE promoter methylation status. (A) Representative images of dCas9-DNMT3a-mediated specific modification. (B,C) Changes in the methylation status of the RAGE promoter region. CML and CEL enhanced the demethylation level of the RAGE promoter, and the dCas9-DNMT3a-targeting RAGE promoter recovered the hypermethylation status. (D,E) Relative expression of RAGE in the indicated groups. All dates are shown as the mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.*
To verify this phenomenon, dCAS9-DNMT3a- and sgRNA-targeting RAGE promoter regions were transfected into the cells to counteract the effect of CML/CEL. As expected, CML/CEL-mediated demethylation was abrogated by the dCAS9-DNMT3a-mediated methylation, and the increased fold change of the average DNA methylation rate of individual CpGs between control and dCas9-DNMT3a-treated cells is about $30\%$ (Figures 4B, C). Correspondingly, the dCAS9-DNMT3a-targeting RAGE promoter region partially blocked the CML/CEL-mediated augment of RAGE expression (Figures 4D, E), while the scrambled sgRNA with dCAS9-DNMT3a exhibited no effect on RAGE expression (Supplementary Figure S1).
## 3.4 AGEs upregulate TET1 expression
DNA demethylation is mainly conducted by TET1 that is a maintenance DNA demethylase to prevent methylation spreading in cells. Here, we performed qPCR to analyze the impact of CML and CEL on TET1 expression. Our results revealed that the treatment of CML and CEL upregulates TET1 expression (Figures 5A, B), which indicates AGE and RAGE interactions may facilitate the elevation of TET1 level to mediate epigenetic modulation of RAGE. As Figure 5C shows, the interaction between AGEs and RAGE not only activates its downstream NF-κB/STAT3 signaling pathway to augment RAGE expression as a feedback, but also facilitate the demethylation of the RAGE promoter region to boost its transcription. However, how AGEs upregulate TET1 expression still remains unknown.
**FIGURE 5:** *Change of the RAGE promoter methylation status. Tet1 expression affected by CML (A) and CEL (B). (C) Representative images of the regulatory process. All dates are shown as the mean ± SD. *p < 0.05.*
## 4 Discussion
It is well-known that AGEs can promote RAGE expression through the downstream pathway—NF-κB and STAT3 signaling pathways, to augment the interaction between AGEs and RAGE. In this study, we further discovered AGEs can also epigenetically modulate RAGE expression, which deepens our understanding of this feedback loop. AGEs and their receptor RAGE play critical roles in the progression of chronic disease, such as diabetes and cancer (Sanajou et al., 2018). Notably, the AGE–RAGE axis can activate its downstream signaling to facilitate RAGE expression, which forms a vicious cycle to escalate its impact. In this cycle, the key factor is NF-κB and STAT3 (Fuentes et al., 2007; Oh et al., 2019). However, their inhibitors cannot completely block RAGE upregulation, indicating that other pathways exist in this regulatory phenomenon.
Epigenetic modification is an important modulation of gene expression, which mediates the dynamic processes of transcriptional activation or suppression without the alteration of the DNA sequence (Afanas’ev, 2014; Hogg et al., 2020). The AGE-mediated epigenetic effects have been revealed to involve the occurrence of diabetes and other chronic diseases. Recent studies conducted in human podocytes showed that AGEs enhanced the acetylation of key transcription factors via the downregulation of SIRT1, leading to podocyte apoptosis and consequently kidney diseases (Perrone et al., 2020). AGEs induced the demethylation of the MMP-9 promoter through the downregulation of GADD45a, which is involved in diabetic foot ulcers (Zhong and Kowluru, 2013; Zhou et al., 2018).
DNA methylation has been recently demonstrated to be involved in glycemic memory of diabetic complication (Tewari et al., 2012; Dunn et al., 2014), the epigenetic changes were sustained even if blood glucose return to normal levels (Park et al., 2014; Perrone et al., 2020). DNA methylation also plays an important role in the progression of diabetic foot ulcers. Global hypomethylation is evident in diabetic foot ulcer fibroblasts. Functional enrichment analysis in the previous study emphasized differential methylation of gene clusters that are associated with the myofibril function, angiogenesis, and extracellular matrix for the wound healing process. ( Park et al., 2014). Notably, the methylation status of the RAGE promoter can also perform as indicator for diabetic complication (Rajasekar et al., 2015). Kan and colleagues have proved the association between RAGE promoter methylation and diabetic retinal inflammation, and the hypomethylation of RAGE promoter can raise IL-1β, IL-6, and TNF-α levels in serum of patients (Kan et al., 2018), and the augmentation of RAGE promoter methylation may contribute to reducing the inflammation of DR patients. Moreover, the promoters of RAGE glomeruli from diabetic db/db mice showed enhanced RNA polymerase II recruitment, increased levels of activated marks and decreased levels of repressive marks (Reddy et al., 2014). Thus, the regulation of methylation in the RAGE promoter region provides an optional way to control these chronic diseases.
It is well-known that AGEs can promote RAGE expression to form the feedback cycle and amplify the influence of their interaction, while the epigenetic relationship between AGEs and RAGE remains elusive. AGEs have an environmental influence, and their epigenetic effects on the downstream genes mainly rely on their interaction with RAGE. Based on the AGE-mediated epigenetic action and RAGE promoter methylation changes, we accordingly proposed the hypothesis that whether AGEs can exhibit epigenetic effects on RAGE transcription to facilitate RAGE expression in this vicious cycle. Here, we identified the epigenetic relationship between AGEs and RAGE promoter methylation and demonstrated this association via the dCAS9-DNMT3a system. Meanwhile, we also found out that AGEs can elevate the TET1 levels in the cells, which may contribute to hypomethylation of the RAGE promoter. However, the exact mechanism for this process still remains unknown, and it is also worth exploring the roles of this epigenetic regulation in chronic disease, such as diabetes and cancers, for better therapy.
The damages mediated by the AGE and RAGE axis occur in multiple diseases (Kang and Yang, 2020; Corica et al., 2019; Cai et al., 2004). The binding of AGEs and RAGE induces the activation of multiple downstream signaling pathways, which are associated with numerous cellular processes of inflammation, vasculopathy, apoptosis, nephropathy, and angiogenesis (Ott et al., 2014). Although NF-κB pathway and STAT3 pathway play primary roles in AGE-induced RAGE upregulation, there may still be other unclear downstream signaling pathways in this modulation process. Moreover, with regard to promoter methylation, there may be other factors such as environmental influences to modulate the methylation status. For example, disturbed blood flow has been identified to alter DNA methylation patterns in murine arterial endothelial cells in a DNMT-dependent manner (Dunn et al., 2014).
In conclusion, the present study demonstrates that AGEs can modulate RAGE expression not only through its downstream signaling pathway, but also through the epigenetic modifications. The interaction of AGEs and RAGE induces the demethylation of the RAGE promoter possibly via upregulating the TET1 level. However, the exact mechanism underlying these epigenetic regulatory processes still needs to be explored.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material; further inquiries can be directed to the corresponding author.
## Author contributions
All authors were involved in drafting the manuscript or revising it critically for intellectual content, and all authors approved the final version to be published. XW and ZY designed the study. XW, XS, and XC conducted the experiments. ZY revised the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors, and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcell.2023.1062229/full#supplementary-material
## References
1. Abo El-Nasr N. M. E., Saleh D. O., Mahmoud S. S., Nofal S. M., Abdelsalam R. M., Safar M. M.. **Olmesartan attenuates type 2 diabetes-associated liver injury: Cross-talk of AGE/RAGE/JNK, STAT3/SCOS3 and RAS signaling pathways**. *Eur. J. Pharmacol.* (2020) **874** 173010. DOI: 10.1016/j.ejphar.2020.173010
2. Afanas'ev I.. **New nucleophilic mechanisms of ros-dependent epigenetic modifications: Comparison of aging and cancer**. *Aging Dis.* (2014) **5** 52-62. DOI: 10.14336/AD.2014.050052
3. Cai W., He J. C., Zhu L., Peppa M., Lu C., Uribarri J.. **High levels of dietary advanced glycation end products transform low-density lipoprotein into a potent redox-sensitive mitogen-activated protein kinase stimulant in diabetic patients**. *Circulation* (2004) **110** 285-291. DOI: 10.1161/01.CIR.0000135587.92455.0D
4. Chiappalupi S., Sorci G., Vukasinovic A., Salvadori L., Sagheddu R., Coletti D.. **Targeting RAGE prevents muscle wasting and prolongs survival in cancer cachexia**. *J. Cachexia Sarcopenia Muscle* (2020) **11** 929-946. DOI: 10.1002/jcsm.12561
5. Corica D., Aversa T., Ruggeri R. M., Cristani M., Alibrandi A., Pepe G.. **Could AGE/RAGE-Related oxidative homeostasis dysregulation enhance susceptibility to pathogenesis of cardio-metabolic complications in childhood obesity?**. *Front. Endocrinol. (Lausanne)* (2019) **10** 426. DOI: 10.3389/fendo.2019.00426
6. Dariya B., Nagaraju G. P.. **Advanced glycation end products in diabetes, cancer and phytochemical therapy**. *Drug Discov. Today* (2020) **25** 1614-1623. DOI: 10.1016/j.drudis.2020.07.003
7. Dawson M. A., Kouzarides T.. **Cancer epigenetics: From mechanism to therapy**. *Cell* (2012) **150** 12-27. DOI: 10.1016/j.cell.2012.06.013
8. Dunn J., Qiu H., Kim S., Jjingo D., Hoffman R., Kim C. W.. **Flow-dependent epigenetic DNA methylation regulates endothelial gene expression and atherosclerosis**. *J. Clin. Invest.* (2014) **124** 3187-3199. DOI: 10.1172/JCI74792
9. Fuentes M. K., Nigavekar S. S., Arumugam T., Logsdon C. D., Schmidt A. M., Park J. C.. **RAGE activation by S100P in colon cancer stimulates growth, migration, and cell signaling pathways**. *Dis. Colon Rectum* (2007) **50** 1230-1240. DOI: 10.1007/s10350-006-0850-5
10. Henikoff S., Greally J. M.. **Epigenetics, cellular memory and gene regulation**. *Curr. Biol.* (2016) **26** R644-R648. DOI: 10.1016/j.cub.2016.06.011
11. Hogg S. J., Beavis P. A., Dawson M. A., Johnstone R. W.. **Targeting the epigenetic regulation of antitumour immunity**. *Nat. Rev. Drug Discov.* (2020) **19** 776-800. DOI: 10.1038/s41573-020-0077-5
12. Kan S., Wu J., Sun C., Hao J., Wu Z.. **Correlation between RAGE gene promoter methylation and diabetic retinal inflammation**. *Exp. Ther. Med.* (2018) **15** 242-246. DOI: 10.3892/etm.2017.5378
13. Kang Q., Tong Y., Gowd V., Wang M., Chen F., Cheng K. W.. **Oral administration of EGCG solution equivalent to daily achievable dosages of regular tea drinkers effectively suppresses miR483-3p induced metastasis of hepatocellular carcinoma cells in mice**. *Food Funct.* (2021) **12** 3381-3392. DOI: 10.1039/d1fo00664a
14. Kang Q., Yang C.. **Oxidative stress and diabetic retinopathy: Molecular mechanisms, pathogenetic role and therapeutic implications**. *Redox Biol.* (2020) **37** 101799. DOI: 10.1016/j.redox.2020.101799
15. Khalid M., Petroianu G., Adem A.. **Advanced glycation end products and diabetes mellitus: Mechanisms and perspectives**. *Biomolecules* (2022) **12** 542. DOI: 10.3390/biom12040542
16. Ma W., Lee S. E., Guo J., Qu W., Hudson B. I., Schmidt A. M.. **RAGE ligand upregulation of VEGF secretion in ARPE-19 cells**. *Invest. Ophthalmol. Vis. Sci.* (2007) **48** 1355-1361. DOI: 10.1167/iovs.06-0738
17. Meloche J., Paulin R., Courboulin A., Lambert C., Barrier M., Bonnet P.. **RAGE-dependent activation of the oncoprotein Pim1 plays a critical role in systemic vascular remodeling processes**. *Arterioscler. Thromb. Vasc. Biol.* (2011) **31** 2114-2124. DOI: 10.1161/ATVBAHA.111.230573
18. Oh H., Park S. H., Kang M. K., Kim Y. H., Lee E. J., Kim D. Y.. **Asaronic acid attenuates macrophage activation toward M1 phenotype through inhibition of NF-κB pathway and JAK-STAT signaling in glucose-loaded murine macrophages**. *J. Agric. Food Chem.* (2019) **67** 10069-10078. DOI: 10.1021/acs.jafc.9b03926
19. Ott C., Jacobs K., Haucke E., Navarrete Santos A., Grune T., Simm A.. **Role of advanced glycation end products in cellular signaling**. *Redox Biol.* (2014) **2** 411-429. DOI: 10.1016/j.redox.2013.12.016
20. Park L. K., Maione A. G., Smith A., Gerami-Naini B., Iyer L. K., Mooney D. J.. **Genome-wide DNA methylation analysis identifies a metabolic memory profile in patient-derived diabetic foot ulcer fibroblasts**. *Epigenetics* (2014) **9** 1339-1349. DOI: 10.4161/15592294.2014.967584
21. Perrone A., Giovino A., Benny J., Martinelli F.. **Advanced glycation end products (AGEs): Biochemistry, signaling, analytical methods, and epigenetic effects**. *Oxid. Med. Cell Longev.* (2020) **2020** 3818196. DOI: 10.1155/2020/3818196
22. Rajasekar P., O'Neill C. L., Eeles L., Stitt A. W., Medina R. J.. **Epigenetic changes in endothelial progenitors as a possible cellular basis for glycemic memory in diabetic vascular complications**. *J. Diabetes Res.* (2015) **2015** 436879. DOI: 10.1155/2015/436879
23. Reddy M. A., Sumanth P., Lanting L., Yuan H., Wang M., Mar D.. **Losartan reverses permissive epigenetic changes in renal glomeruli of diabetic db/db mice**. *Kidney Int.* (2014) **85** 362-373. DOI: 10.1038/ki.2013.387
24. Sanajou D., Ghorbani Haghjo A., Argani H., Aslani S.. **AGE-RAGE axis blockade in diabetic nephropathy: Current status and future directions**. *Eur. J. Pharmacol.* (2018) **833** 158-164. DOI: 10.1016/j.ejphar.2018.06.001
25. Tewari S., Zhong Q., Santos J. M., Kowluru R. A.. **Mitochondria DNA replication and DNA methylation in the metabolic memory associated with continued progression of diabetic retinopathy**. *Invest. Ophthalmol. Vis. Sci.* (2012) **53** 4881-4888. DOI: 10.1167/iovs.12-9732
26. Wang J., Zhen Y., Zhou Y., Yan S., Jiang L., Zhang L.. **Promoter methylation cooperates with SNPs to modulate RAGE transcription and alter UC risk**. *Biochem. Biophys. Rep.* (2019) **17** 17-22. DOI: 10.1016/j.bbrep.2018.11.001
27. Zhong Q., Kowluru R. A.. **Regulation of matrix metalloproteinase-9 by epigenetic modifications and the development of diabetic retinopathy**. *Diabetes* (2013) **62** 2559-2568. DOI: 10.2337/db12-1141
28. Zhou L., Wang W., Yang C., Zeng T., Hu M., Wang X.. **GADD45a promotes active DNA demethylation of the MMP-9 promoter via base excision repair pathway in AGEs-treated keratinocytes and in diabetic male rat skin**. *Endocrinology* (2018) **159** 1172-1186. DOI: 10.1210/en.2017-00686
29. Zhou Q., Cheng K. W., Xiao J., Wang M.. **The multifunctional roles of flavonoids against the formation of advanced glycation end products (AGEs) and AGEs-induced harmful effects**. *Trends Food Sci. Technol.* (2020) **103** 333-347. DOI: 10.1016/j.tifs.2020.06.002
|
---
title: Antidiabetic effects of Andrographis paniculata supplementation on biochemical
parameters, inflammatory responses, and oxidative stress in canine diabetes
authors:
- Namphung Suemanotham
- Sataporn Phochantachinda
- Duangthip Chatchaisak
- Walasinee Sakcamduang
- Anchana Chansawhang
- Pornsiri Pitchakarn
- Boonrat Chantong
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9971231
doi: 10.3389/fphar.2023.1077228
license: CC BY 4.0
---
# Antidiabetic effects of Andrographis paniculata supplementation on biochemical parameters, inflammatory responses, and oxidative stress in canine diabetes
## Abstract
Introduction: *Diabetes mellitus* is a common endocrine disorder that causes hyperglycemia in dogs. Persistent hyperglycemia can induce inflammation and oxidative stress. This study aimed to investigate the effects of A. paniculata (Burm.f.) Nees (Acanthaceae) (A. paniculata) on blood glucose, inflammation, and oxidative stress in canine diabetes. A total of 41 client-owned dogs (23 diabetic and 18 clinically healthy) were included in this double-blind, placebo-controlled trial.
Methods: The diabetic dogs were further divided into two treatments protocols: group 1 received A. paniculata extract capsules (50 mg/kg/day; $$n = 6$$) or received placebo for 90 days ($$n = 7$$); and group 2 received A. paniculata extract capsules (100 mg/kg/day; $$n = 6$$) or received a placebo for 180 days ($$n = 4$$). Blood and urine samples were collected every month. No significant differences in fasting blood glucose, fructosamine, interleukin-6, tumor necrosis factor-alpha, superoxide dismutase, and malondialdehyde levels were observed between the treatment and placebo groups ($p \leq 0.05$).
Results and Discussion: The levels of alanine aminotransferase, alkaline phosphatase, blood urea nitrogen, and creatinine were stable in the treatment groups. The blood glucose levels and concentrations of inflammatory and oxidative stress markers in the client-owned diabetic dogs were not altered by A. paniculata supplementation. Furthermore, treatment with this extract did not have any adverse effects on the animals. Non-etheless, the effects of A. paniculata on canine diabetes must be appropriately evaluated using a proteomic approach and involving a wider variety of protein markers.
## 1 Introduction
Diabetes mellitus (DM) is a common endocrine disorder caused by absolute or relative insulin deficiency, which impairs glucose uptake into the cell (Nelson and Reusch, 2014). Insulin-dependent DM, a disease that resembles type 1 diabetes in humans, has been commonly recognized in dogs (Behrend et al., 2018). The diagnosis of DM in dogs is based on the presence of hyperglycemia and glucosuria, along with signs of weight loss, decreased appetite, polyuria, and polydipsia (Nelson and Reusch, 2014). Uncontrolled DM results in persistent hyperglycemia that can cause complications, such as diabetic cardiomyopathy, diabetic neuropathy, diabetic retinopathy, diabetic nephropathy, and atherosclerosis (Papatheodorou et al., 2016). The overproduction of superoxide has been proposed as a unifying mechanism that mediates the tissue-damaging effects of prolonged hyperglycemia (Brownlee, 2005). Thus, the management of DM is aimed at controlling the blood glucose level, which can be accomplished through insulin therapy, dietary modification, and control of concurrent disorders (Papachristoforou et al., 2020).
In combination with standard treatment, botanical drugs are utilized as an adjunct therapy to prevent long-term complications and improve the overall wellbeing of diabetic dogs. A. paniculata (Burm.f.) Nees (Acanthaceae) (A. paniculate), one of the most popular medicinal plants, consists of many active phytochemicals, such as andrographolide, neoandrographolide, andrographiside 14-deoxyandrographolide, 14-deoxy-11,12-didehydroandrographolide, 14-deoxy-11-oxoandrographolide, and β-sitosterol (Dai et al., 2019). The aerial part of this plant was commonly used due to the presence of andrographolide, the primary active component of A. paniculata (Patil and Andrographolide, 2020). The biological activities of andrographolide include anti-inflammatory, antioxidant, antiangiogenic, antidiabetic, antifertility, antiviral, antibacterial, cardioprotective, nephroprotective, and hepatoprotective effects (Okhuarobo et al., 2014; Kishore et al., 2017; Zhang et al., 2021). A. paniculata is known to exert anti-inflammatory properties by down-regulating the levels of cyclooxygenase and proinflammatory cytokines such as tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), and interleukin-10 (IL-10) in human, rats, and mice (Abu-Ghefreh et al., 2009; Zhang et al., 2013; Jaiyesimi et al., 2020; Zeng et al., 2022).
Evidence suggests that A. paniculata might be considered as a promising candidate for the management of DM; for instance, oral A. paniculata supplementation was used as a hypoglycemic agent in streptozotocin-induced diabetic rats (Yu et al., 2003) and high-fat-fructose-fed rats (Nugroho et al., 2012). Andrographolide supplementation ameliorated renal mesangial cell proliferation and inflammation by inhibiting protein kinase B (Akt) and nuclear factor-κB (NF-κB) signaling (Ji et al., 2016). Treatment of streptozotocin-induced diabetic rats with A. paniculata reduced oxidative stress by increasing the superoxide dismutase (SOD) and catalase activities (Zhang and Tan, 2000). Furthermore, decreased levels of malondialdehyde (MDA) and increased levels of glutathione were observed in the kidneys of diabetic rats supplemented with A. paniculata (Hidayat and Wulandari, 2021).
Studies in humans and rodents demonstrated the potential of A. paniculata as a supplemental treatment for DM in tandem with standard medicine. However, to the best of our knowledge, there is no published evidence of the impact and safety of A. paniculata supplementation in client-own diabetic dogs. Therefore, the aim of the present study was to evaluate the outcomes of A. paniculata supplementation in canine DM-associated inflammation and oxidative stress. Additionally, the adverse effects of A. paniculata supplementation were determined in the animals.
## 2.1 Animals
This randomized, double-blind clinical trial was approved by the Committee for the Care and Use of Laboratory Animals, at the Faculty of Veterinary Science, Mahidol University (approval number: MUVS-2020-04-10). A total of 41 client-own dogs (23 diabetic and 18 clinically healthy) from Prasu Arthorn Animal Hospital, Faculty of Veterinary Science, Mahidol University were included in this study. The clinically healthy dogs were used for cross-sectional normal baseline evaluations. All dogs were used after obtaining the signed consent forms from the owners. The design used for the experiments in this study is shown in Figure 1. For the determination of optimum dose and duration of A. paniculata treatment, the diabetic dogs were divided into two treatment protocols. In the first protocol, dogs were given either A. paniculata extracted capsules (50 mg/kg/day; $$n = 6$$) or a placebo ($$n = 7$$) for 90 days. In another protocol, dogs were given A. paniculata extracted capsules (100 mg/kg/day; $$n = 6$$), while others were given placebo ($$n = 4$$) for 180 days. Routine treatment was provided to all the dogs. The clinical parameters and the levels of the inflammatory and oxidative stress biomarkers were evaluated in each group. The characteristics of the dogs in each group are shown in Supplementary Table S1A.
**FIGURE 1:** *Design of experiments. Twenty-three diabetic and 18 healthy dogs were included in the study. The healthy dogs were used for cross-sectional baseline assessments. The diabetic dogs were divided into two treatment protocols to determine the appropriate dose and duration of A. paniculata administration. Dogs were given either A. paniculata extracted capsules (50 mg/kg/day; n = 6) or a placebo (n = 7) for 90 days. In another study, dogs were given A. paniculata extract capsules (100 mg/kg/day; n = 6), whereas others were given a placebo (n = 4) for 180 days.*
The inclusion criteria were diabetic dogs of any breed, age, or sex, with stable blood glucose levels for the previous 3 months. The diabetic dogs were diagnosed with a history of polyuria, polydipsia, polyphagia, weight loss with normal or increased appetite, fasting hyperglycemia, and glucosuria. The exclusion criteria for the study were as follows: dogs with unstable diabetes or diabetic ketoacidosis, those that received corticosteroids, and those with diseases that affect the blood glucose levels, such as hyperadrenocorticism, exocrine pancreatic insufficiency, neoplasia, and acromegaly. To reduce the confounding factors, all the diabetic dogs enrolled in this study were fed a commercial diabetic diet and allowed to live indoors or within their compounds, close to their owners, without any changes in their environment during the study period.
## 2.2 A. paniculata and placebo
A commercial A. paniculata capsule (Abhaibhubejhr ® FAH-TALAI-JONE capsule, lot: MB00916006) was used as a supplement in this study. Each capsule contained 400 mg of the aerial part of A. paniculata, which was standardized to a content of $1\%$ w/w andrographolide. The procedures and findings of the analysis of andrographolide in the capsule of A. paniculata are detailed in Supplementary material B. Lactose powder, which was similar in appearance to the A. paniculata capsule, was used as a placebo in this study. The dosage of the drug was referenced from the dosage given in experimental rats in a previous study (Panossian et al., 2000) and calculated to that used for dogs using the conversion factors proposed by Reagan-Shaw (Reagan-Shaw et al., 2008).
## 2.3 Sample collection
Blood and urine samples were collected from clinically healthy and diabetic dogs on day 0 for the baseline data. The blood and urine samples from the diabetic dogs were reassessed every 30 days for 90 or 180 days, depending on the treatment protocol. Blood samples (3–5 mL) were collected from the cephalic or saphenous vein to evaluate the clinical parameters and the biomarkers for inflammation and oxidative stress. The clinical parameters, which included complete blood count (CBC) and the levels of glucose, fructosamine, alanine aminotransferase (ALT), alkaline phosphatase (ALP), blood urea nitrogen (BUN), and creatinine, were analyzed every 30 days during routine health check-up. IL-6 and TNF-α were selected as the inflammatory biomarkers, whereas SOD and MDA were used as the oxidative stress biomarkers. The biomarkers were evaluated on days 0 and 90 or 180, depending on the treatment protocol.
One drop of the blood sample was used to test the glucose level using the AlphaTRAK glucometer (Zoetis, Parsippany, NJ, United State). The blood samples were then divided into three parts and used for further experiments. The ethylenediaminetetraacetic acid (EDTA) blood samples were stored at 4°C, and the CBCs were measured using an animal blood counter (Horiba Medical, Montpellier, France) within 4 h. The second part was centrifuged at 3500 rpm for 5 min within an hour after blood collection; the resultant plasma was used to measure the ALT, ALP, BUN, and creatinine levels using an automatic analyzer (Chema diagnostica, Monsano AN, Italy), and the serum was collected in a sterile microcentrifuge tube and stored at −80°C to measure the IL-6, TNF-α, SOD, and MDA levels. The third part was collected in plain tubes and sent to a commercial laboratory within 24 h to measure the fructosamine level.
The urine sample was collected by cystocentesis and centrifuged at 2000 rpm for 3 min (Hettich Lab Technology, Tuttlingen, Germany). The physical and chemical properties (color, clarity, specific gravity, pH, protein, glucose, ketone, and bilirubin levels, and erythrocyte counts) of the urine supernatants were tested using the dipstick test (Roche Diagnostics, Indianapolis, IN, United State). The urine sediments were evaluated under a light microscope (ZEISS, Jena, Germany). The white blood cell counts (high-power field: HPF), red blood cell counts (HPF), and presence of amorphous crystals, mucous, bacteria, epithelium cells (HPF), cast (low-power field: LPF), and crystals were determined.
## 2.4 Interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α) determination
The levels of IL-6 and TNF-α in the plasma were measured by an enzyme-linked immunosorbent assay (ELISA) assay using kits obtained from Abcam (Cambridge, United Kingdom) and Thermo Scientific (Waltham, MA, United State), respectively. The procedure was carried out in accordance with the manufacturer’s instructions. An ELISA plate reader (BioTek, Santa Clara, CA, United State) was used to measure the absorbance at 450 nm, and the IL-6 and TNF-α levels were compared with those of the standards.
## 2.5 Superoxide dismutase (SOD) determination
The levels of SOD in the serum were measured by ELISA using commercial kits (Abcam, Cambridge, United Kingdom) according to the manufacturer’s protocols. The microplate reader (BioTek, Santa Clara, CA, United State) was used to read the optical density at 450 nm, and the SOD activity was calculated in percentage.
## 2.6 Malondialdehyde (MDA) determination
The levels of MDA in plasma were measured using the thiobarbituric acid reactive substance assay, which was based on the reaction between MDA and thiobarbituric acid; the reaction resulted in a pink-colored product, which demonstrated absorption at a wavelength of 535 nm (Gheita et al., 2014). Briefly, plasma samples (0.1 mL) were mixed with 1 mL of 6.7 mg/mL of orthophosphoric acid (Sigma, St. Louis, MO, United State) and 1 mL of $0.13\%$ thiobarbituric acid (Sigma, St. Louis, MO, United State). The solution was thoroughly mixed and heated in a boiling water bath for 45 min; after cooling down, 0.8 mL of n-butanol was added to the solution and mixed vigorously. The n-butanol layer was separated by centrifugation at 3,000 rpm for 15 min (Beckman Coulter, Brea, CA, United State) and retained for spectrometric analysis. The absorbance of the pink-colored product within the butanol layer was measured with a microplate reader (BioTek, Santa Clara, CA, United State) at 535 nm and 520 nm for interference subtraction. The blank in this protocol contained distilled water. The concentrations of MDA were calculated from the standard MDA solution (range, 0.1–20 nmol/mL).
## 2.7 Statistical analysis
All data from the experiments were assessed for normal distribution using the Shapiro–Wilk test and for variance using Levene’s test. Data from the clinically healthy dogs were compared with those from the diabetic dogs prior to A. paniculata supplementation (day 0) using the independent t-test. The clinical parameters of DM were compared between day 0 and days 30, 60, 90, 120, 150, and 180 (depending on the treatment protocol) using repeated measures analysis of variance. The inflammatory and oxidative stress parameters were compared between day 0 and day 90 (treatment protocol 1) or day 0 and day 180 (treatment protocol 2) using the paired t-test. Statistical calculations were performed using SPSS version 21 (IBM, Armonk, NY, United State), and the significance level was set at p ≤ 0.05.
## 3.1 Descriptive characteristics of the dogs at inclusion
The baseline parameters of the clinically healthy ($$n = 18$$) and the diabetic ($$n = 23$$) dogs prior to A. paniculata supplementation were evaluated to minimize bias between groups. The average fasting blood glucose level in the diabetic dogs (377.33 ± 91.83 mg/dL) was significantly ($p \leq 0.01$) higher than that in the clinically healthy dogs (64.67 ± 8.36 mg/dL; Figure 2A). Similarly, the average blood fructosamine level in the diabetic dogs (459 ± 159.42 umol/L) was significantly ($p \leq 0.01$) higher than that in the clinically healthy dogs (175 ± 23.22 umol/L; Figure 2B).
**FIGURE 2:** *Bar graphs showing the blood glucose (A), blood fructosamine (B), ALT (C), ALP (D), BUN (E), and creatinine (F) levels in the healthy and diabetic group of dogs **** = p < 0.0001, *** = p < 0.001.*
The blood biochemistry results showed that the ALT levels in the clinically healthy and the diabetic groups were 35 ± 8.99 U/L and 107.08 ± 72.07 U/L, respectively (Figure 2C). The measured ALP was 56.5 ± 13.3 U/L in the clinically healthy group and 213.86 ± 138.23 U/L in the diabetic group (Figure 2D). Likewise, the BUN levels in the clinically healthy and diabetic groups were 14.33 ± 5.50 U/L and 25.42 ± 11.54 U/L, respectively (Figure 2E), while those of creatinine were 1.5 ± 0.2 mg/dL and 1.6 ± 0.18 mg/dL, respectively (Figure 2F). The diabetic dogs had significantly higher levels of ALT, ALP, and BUN compared to the clinically healthy dogs ($p \leq 0.05$), whereas the creatinine levels did not differ significantly between the two groups.
The urine dipstick results showed that all the diabetic dogs, but none of the healthy dogs, had glucosuria. The urine samples from the diabetic and clinically healthy dogs were negative for ketone with inactive urine sediments (<5 white blood cells per HPF, <5 red blood cells per HPF, no bacteria, no epithelium cell, no cast, and no crystal).
The IL-6 levels in the healthy dogs ($130.89\%$ ± $54.06\%$) were not significantly different from those in the diabetic dogs ($206.17\%$ ± $149.32\%$; Figure 3A). The level of TNF-α in the healthy dogs ($102.45\%$ ± $19.64\%$) was slightly higher but not significantly different from that in the diabetic dogs ($93.38\%$ ± $13.05\%$; Figure 3B). Similarly, no significant differences in the levels of SOD (Figure 3C) and MDA (Figure 3D) were observed between the two groups.
**FIGURE 3:** *Bar graphs showing the baseline levels of the inflammatory and oxidative stress parameters in the clinically healthy and diabetic dogs (A) IL-6 (B) TNF-α (C) SOD, and (D) MDA.*
## 3.2 Effect of A. paniculata supplementation on fasting blood glucose and fructosamine levels in the diabetic dogs
The effects of A. paniculata (dose 50 mg/kg/day for 90 days and 100 mg/kg/day for 180 days) on fasting blood glucose and blood fructosamine levels were investigated. The fasting blood glucose (Figure 4A) and fructosamine (Figure 4C) levels in diabetic dogs supplemented with 50 mg/kg/day of A. paniculata for 90 days were not significantly lower than those in the placebo groups. Similarly, supplementation with 100 mg/kg/day of A. paniculata for 180 days did not alter the fasting blood glucose (Figure 4B) and fructosamine (Figure 4D) levels in the diabetic group compared to those in the placebo groups.
**FIGURE 4:** *The effects of A. paniculata (dose 50 mg/kg/day for 90 days and 100 mg/kg/day for 180 days) on fasting blood glucose and blood fructosamine levels. Bar graphs showing blood glucose levels in groups treated with 50 mg/kg of A. paniculata and placebo for 90 days (A), blood glucose levels in groups treated with 100 mg/kg of A. paniculata and placebo for 180 days (B), blood fructosamine levels in groups treated with 50 mg/kg of A. paniculata and placebo for 90 days (C), and blood fructosamine levels in groups treated with 100 mg/kg of A. paniculata and placebo for 180 days (D).*
## 3.3 Effect of A. paniculata supplementation on inflammation in diabetic dogs
The outcomes of A. paniculata (dose 50 mg/kg/day for 90 days and 100 mg/kg/day for 180 days) on inflammation were investigated. Supplementation with 50 and 100 mg/kg/day for 90 and 180 days, respectively, did not significantly alter the levels of IL-6 and TNF-α in the diabetic dogs when compared to those in the placebo groups (Figures 5A–D) levels.
**FIGURE 5:** *The effects of A. paniculata supplementation (doses 50 and 100 mg/kg) on inflammation in the diabetic dogs. Bar graphs showing no significant differences in the (A) IL-6 levels of dogs treated with 50 mg/kg of A. paniculata for 90 days, (B) IL-6 levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days, (C) TNF-α levels of diabetic dogs treated with 50 mg/kg of A. paniculata for 90 days, and (D) TNF-α levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days compared to those in the corresponding placebo groups.*
## 3.4 Effect of A. paniculata supplementation on oxidative stress in diabetic dogs
A. paniculata supplementation with 50 and 100 mg/kg/day for 90 and 180 days, respectively, did not significantly alter the SOD and MDA levels in the diabetic dogs compared to those in the placebo groups (Figures 6A–D).
**FIGURE 6:** *The effects of A. paniculata supplementation (doses 50 and 100 mg/kg for 90 and 180 days, respectively) on the oxidative stress parameters in the diabetic dogs. Bar graphs showing no significant differences in the (A) SOD levels of dogs treated with 50 mg/kg of A. paniculata for 90 days, (B) SOD levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days, (C) MDA levels of diabetic dogs treated with 50 mg/kg of A. paniculata for 90 days, and (D) MDA levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days compared to those in the corresponding placebo groups.*
## 3.5 Effect of A. paniculata supplementation on the biochemical parameters of liver injury and kidney function
Blood ALT and ALP were selected as the liver injury markers, and BUN and creatinine levels were used to examine the kidney function. No statistically significant changes in the levels of the liver injury markers (Figure 7) and kidney function markers (Figure 8) were noted after supplementation with A. paniculata (dose 50 mg/kg/day for 90 days and 100 mg/kg/day for 180 days) compared to those in the placebo groups.
**FIGURE 7:** *The effects of A. paniculata supplementation (doses 50 and 100 mg/kg for 90 and 180 days, respectively) on the biochemical parameters of liver injury in the diabetic dogs compared to those in the placebo groups. Bar graphs showing no significant differences in the (A) ALT levels of dogs treated with 50 mg/kg of A. paniculata for 90 days, (B) ALT levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days, (C) ALP levels of diabetic dogs treated with 50 mg/kg of A. paniculata for 90 days, and (D) ALP levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days compared to those in the corresponding placebo groups.* **FIGURE 8:** *The effects of A. paniculata supplementation (doses 50 and 100 mg/kg for 90 and 180 days, respectively) on the biochemical parameters of renal injury in the diabetic dogs. Bar graphs showing no significant differences in the (A) BUN levels of dogs treated with 50 mg/kg of A. paniculata for 90 days, (B) BUN levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days, (C) creatinine levels of diabetic dogs treated with 50 mg/kg of A. paniculata for 90 days, and (D) creatinine levels of diabetic dogs treated with 100 mg/kg of A. paniculata for 180 days compared to those in the corresponding placebo groups.*
The urinalysis results from day 0 to day 180 showed negative ketone bodies and inactive urine sediments (<5 white blood cells per HPF, <5 red blood cells per HPF, and no bacteria, epithelial cells, casts, and crystals), indicating the absence of diabetic ketosis and bacterial cystitis during the study period. Additionally, the owners reported no adverse effects of A. paniculata supplementation throughout the study.
## 4 Discussion
Canine DM is a prevalent problem in veterinary medicine (Gilor et al., 2016). The application of plant-derived metabolites for the management of canine diabetes has been studied extensively (Russell et al., 2008). Bixa orellana extract was reported to decrease the blood glucose level by increasing peripheral glucose utilization in streptozotocin-induced diabetic dogs (Russell et al., 2008; Ogbu et al., 2013). Gongronema latifolium maintained the plasma glucose levels in alloxan-induced diabetic dogs by delaying stomach emptying (Ogbu et al., 2013). Additionally, numerous medicinal plants are reported to have the potential to improve the management of diabetes in humans (Bindu and Narendhirakannan, 2018). A. paniculata, a medicinal plant typically used as an anti-inflammatory and antimicrobial agent, has been studied for its possible impact on the management of DM in humans (Nugroho et al., 2012; Dai et al., 2019; Adiguna et al., 2021; Hossain et al., 2021). Andrographolide is a potential bioactive phytochemical in A. paniculata, which possesses antidiabetic properties (Nugroho et al., 2012; Ji et al., 2016; Islam, 2017; Naik et al., 2017; Mehta et al., 2021; Syukri et al., 2021). Plasma andrographolide was detected in dogs orally treated with A. paniculata tablets, indicating that andrographolide can be absorbed via the GI tract (Xu et al., 2015). However, it is important to be aware of the adverse effects of these natural products. Knowledge about the effectiveness and safety of A. paniculata in canine DM is limited. Therefore, the aim of this study was to investigate the effects of A. paniculata supplementation on the levels of glucose, fructosamine, inflammatory cytokines, and oxidative stress markers, and to determine the presence of any adverse effects in canine diabetes.
Initially, the animals were treated with 50 mg/kg of A. paniculata for 90 days, based on previous studies on rats, to determine the long-term efficacy and safety of the medicinal plant; however, no improvements were observed in the treatment group. Therefore, the dose and duration of A. paniculata supplementation were increased to 100 mg/kg/day for 180 days; however, no significant differences in blood glucose and fructosamine levels were observed in the diabetic dogs. The dose and frequency of insulin administration to the dogs were not adjusted during A. paniculata supplementation. However, the findings of this study were not in concordance with those reported in mice and rats (Yu et al., 2003; Akhtar et al., 2016; Chen et al., 2020; Jaiyesimi et al., 2020; Wediasari et al., 2020). Oral andrographolide and A. paniculata lowered the blood glucose levels of streptozotocin-induced diabetic rats in a dose-dependent manner (Yu et al., 2003; Wediasari et al., 2020). Furthermore, A. paniculata extract was reported to reduce hyperglycemia by inhibiting β-cell dysfunction in alloxan-induced diabetic rats (Jaiyesimi et al., 2020). An ethanolic extract of A. paniculata and andrographolide lowered the plasma glucose levels by enhancing the translocation of glucose-transporter-4 in insulin-resistant obese mice (Akhtar et al., 2016; Chen et al., 2020).
Hyperglycemia elevates the levels of inflammatory markers and promotes reactive oxygen species (ROS) formation, leading to several DM complications (Siddiqui et al., 2019). Furthermore, increased oxidative stress and inflammation could result in insulin resistance and reduced β cell function. These vicious cycles are linked to the pathogenesis of DM. In the current study, no changes in the plasma levels of IL-6 and TNF-α were observed in the A. paniculata-treated diabetic dogs. However, the extract has been reported to significantly reduce the levels of these markers in diabetic rats (Jaiyesimi et al., 2020). In another study, treatment with andrographolide suppressed cardiac inflammation via nuclear factor-κB (NF-κB) and reduced both cardiac fibrosis and cardiac hypertrophy in streptozotocin-induced diabetic mice (Liang et al., 2018). Similarly, andrographolide reportedly blocked the NF-κB signaling pathway induced by TNF-α in adipocytes, suggesting that it might reduce insulin resistance by modulating the insulin signaling pathway and improving glucose uptake (Jin et al., 2011; Chen et al., 2020). In the current study, no significant differences in the levels of the oxidative stress markers, including SOD and MDA were observed between the healthy and diabetic groups of dogs. This is contrary to previous reports, which indicated the beneficial effect of A. paniculata and andrographolide on DM complications via reductions in oxidative stress. For instance, andrographolide reduced the characteristics of diabetic nephropathy in murine glomerular mesangial cell lines by reducing the intracellular oxidative states. As considered, andrographolide renal hypertrophy and extracellular matrix accumulation in diabetic mice, as well as NADPH oxidase-1 (NOX-1) expression, ROS production, and proinflammatory cytokines (Lee et al., 2010; Ji et al., 2016). In addition, the cardioprotective effects of andrographolides in diabetic cardiomyopathy in mice have been associated with a decrease in ROS produced by NOX activation (Liang et al., 2018).
The etiology and type of diabetes differ across species, which might explain the variability in the responses to A. paniculata treatment. Insulin deficiency, the most common type of canine diabetes, is characterized by the autoimmune destruction of pancreatic β-cells, which leads to insufficient insulin production and glucotoxicity comparable to that seen in human type 1 diabetes (T1DM) (Nelson and Reusch, 2014; Gilor et al., 2016; O'Kell et al., 2017). Assuming dogs with clinical diabetes have end-stage T1DM may result in a lost potential to treat and reverse glucotoxicity (Gilor et al., 2016). This might explain why A. paniculata treatment did not improve the characteristic features of canine DM in the present study. Furthermore, the findings of the current study were inconsistent with those of other rodent studies, which reported lowered glucose levels and decreased DM complications after treatment with A. paniculata and andrographolide (Yu et al., 2003; Lee et al., 2010; Nugroho et al., 2012; Ji et al., 2016; Islam, 2017; Naik et al., 2017; Wediasari et al., 2020; Mehta et al., 2021; Syukri et al., 2021). However, the DM in rodents was chemically induced with streptozotocin or alloxan, which promotes cell death and decreases insulin production; this is different from the canine DM produced by chronic autoantibody destruction over a long period (Al-awar et al., 2016; Furman, 2021; O’Kell et al., 2022). A. paniculata treatment was reported to lead to beneficial outcomes in insulin-resistant obese mice, a model for type 2 diabetes (T2DM), which is not frequent in dogs (Catchpole et al., 2005; Akhtar et al., 2016; Heydemann, 2016; Chen et al., 2020).
Interspecies variations in pharmacokinetics, particularly the absorption process, may impact the response to A. paniculata treatment. Although oral A. paniculata tablets were absorbed via the GI tract in dogs, the information about the pharmacokinetics involved is currently lacking (Xu et al., 2015). Furthermore, variations in bioavailability across studies beyond species variations may be related to the use of various forms of A. paniculata (Loureiro Damasceno et al., 2022; Songvut et al., 2022). Andrographolide metabolism and clearance variations have been observed among species (Panossian et al., 2000; Zhao et al., 2013). DM involves complicated metabolic processes; hence, the parameters utilized to assess the impact of A. paniculata in this study, such as the blood glucose, fructosamine, cytokines (IL-6 and TNF-α), and oxidative stress indicators (SOD and MDA), may not sensitive enough. The pathogenesis and complications of DM, along with the treatment response, have been analyzed using proteomic patterns (Riaz, 2015; Pena et al., 2016; Kim et al., 2019). Candidate proteins which including alpha-1-antichymotrypsin, alpha-1-antitrypsin, apolipoprotein A-I (apoA-I), haptoglobin, retinol-binding protein 4, transthyretin, and zinc-alpha2-glycoprotein were shown to vary significantly between normal and prediabetes/diabetes in human patients (Riaz, 2015; Kim et al., 2019). Plasma proteomic analysis in canine DM showed a differential expression of alpha-2-HS glycoprotein, transthyretin, apolipoprotein A-I, and apolipoprotein A-IV compared with healthy dogs (Suemanotham et al., 2022). Plasma potential biomarkers such as apolipoprotein C-I, apoA-I, transthyretin, and cystatin C were purposed to predict the progression and response to the treatment in diabetic kidney disease (Riaz, 2015; Pena et al., 2016). Consequently, the effects of A. paniculata on canine diabetes should be evaluated using a proteomic approach and involving a wider variety of protein markers.
Confounding factors such as diet, activities, and environment were common limitations in this clinical study. Therefore, dogs that were fed a commercial diabetic diet, allowed to live indoors or within compounds close to their owners, and did not experience any changes in their environment before and during the study period, were included in the present study. The randomized, double-blind placebo control trial, considered as the gold standard method for medical research, was used in this study to mask the bias. Thus, we believe that, based on the strict inclusion and exclusion criteria and the robust study design, the results of this study should adequately represent the effect of A. paniculata, despite the clinical confounding factors.
Physical examinations were conducted, and the biochemical parameters of liver injury and kidney function were evaluated during the treatment periods. Additionally, adverse reactions such as weakness, vomiting, diarrhea, and allergy were monitored. No changes in physical and biochemical parameters, including ALT, ALP, BUN, and creatinine levels, were noted in all dogs supplemented with A. paniculata (dosages, 50 and 100 mg/kg/day for 90 and 180 days, respectively). Andrographolide was reported to exert beneficial effects by lowering the levels of urea, BUN, and creatinine in streptozotocin-induced diabetic rats (Xu et al., 2012). In another study, A. paniculata demonstrated protective effects against hepatic injury induced by paracetamol and ethanol in rodents (Nagalekshmi et al., 2011; Sivaraj et al., 2011; Mondal et al., 2022). Adverse effects, including nausea, vomiting, abdominal discomfort, dizziness, drowsiness, and urticaria, of A. paniculata extracts or pure metabolites, were observed in some human clinical studies (Thamlikitkul et al., 1991; Saxena et al., 2010; Jayakumar et al., 2013). Non-etheless, no adverse effects were observed in any of the treated dogs in the present study. The worsening of the clinical signs of DM, such as weight loss, polyphagia, polyuria, and polydipsia, was not encountered. These findings suggest that the dosages of A. paniculata administered were harmless within the trial durations specified in this investigation.
The supplementation of A. paniculata did not statistically improve the clinical signs and did not affect the DM parameters or the inflammatory and oxidative stress markers in the current study, which might be attributed to three main reasons. First, the doses and treatment period of A. paniculata supplementation were insufficient. Secondly, the pharmacokinetic variations among species and dosage forms of A. paniculata may impact the plasma concentrations of andrographolide, leading to variations in the pharmacological response (Zhao et al., 2013; Loureiro Damasceno et al., 2022; Songvut et al., 2022). Lastly, the biomarkers we used were not sensitive enough to identify the changes in the etiology and complications of DM. Therefore, protein profiling has been suggested to better understand the prognosis, development of complications, and treatment outcome prediction of DM (Chen and Gerszten, 2020; Araumi et al., 2021; Gummesson et al., 2021). However, unless the appropriate biomarkers can be evaluated, there is still a possibility that A. paniculata can contribute to delay the development and consequences of canine diabetes. Taken together, the findings of this study suggest that treatment with A. paniculata for 180 days did not affect the clinical parameters and levels of inflammatory cytokines and oxidative stress markers in canine DM. Additionally, no adverse effects were observed in the diabetic animals. These positive attributes demonstrate the tolerability of A. paniculata for long-term treatment in canine diabetes. Non-etheless, the relevance of other biomarkers such as apolipoprotein A-I (apoA-I), transthyretin, and alpha2-glycoprotein for observing the changes in the pathogenesis of DM must be explored further.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by The Committee on the Care and Use of Laboratory Animals in the Faculty of Veterinary Science, Mahidol University, Thailand (approval number: MUVS-2020-04-10). Written informed consent was obtained from the owners for the participation of their animals in this study.
## Author contributions
NS, SP, and BC enabled with conception; NS and SP contributed to data processing; DC and BC provided funding acquisition; NS, SP, DC, and WS performed investigation; NS and BC designed methodology; WS provided resources; NS, DC, PP, and BC supervised, data collection, and conducted data analysis; NS, SP, WS, and AC conducted validation and visualization; PP performed phytochemical analysis; NS, SP, and BC prepared the original draft; NS, SP, DC, WS, AC, PP, and BC reviewed and edited the manuscript. All authors have read and agreed to the published version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2023.1077228/full#supplementary-material
## References
1. Abu-Ghefreh AaA., Canatan H., Ezeamuzie C. I.. *Int. Immunopharmacol.* (2009) **9** 313-318. DOI: 10.1016/j.intimp.2008.12.002
2. Adiguna SbP., Panggabean J. A., Atikana A., Untari F., Izzati F., Bayu A.. **Antiviral activities of andrographolide and its derivatives: Mechanism of action and delivery system**. *Pharmaceuticals* (2021) **14** 1102. DOI: 10.3390/ph14111102
3. Akhtar M. T., Bin Mohd Sarib M. S., Ismail I. S., Abas F., Ismail A., Lajis N. H.. **Anti-diabetic activity and metabolic changes induced by Andrographis paniculata plant extract in obese diabetic rats**. *Molecules* (2016) **21** 1026. DOI: 10.3390/molecules21081026
4. Al-awar A., Kupai K., Veszelka M., Szűcs G., Attieh Z., Murlasits Z.. **Experimental diabetes mellitus in different animal models**. *J. Diabetes Res.* (2016) **2016** 9051426. DOI: 10.1155/2016/9051426
5. Araumi A., Osaki T., Ichikawa K., Kudo K., Suzuki N., Watanabe S.. **Urinary and plasma proteomics to discover biomarkers for diagnosing between diabetic nephropathy and minimal change nephrotic syndrome or membranous nephropathy**. *Biochem. Biophys. Rep.* (2021) **27** 101102. DOI: 10.1016/j.bbrep.2021.101102
6. Behrend E., Holford A., Lathan P., Rucinsky R., Schulman R.. **2018 AAHA diabetes management guidelines for dogs and cats**. *J. Am. Anim. Hosp. Assoc.* (2018) **54** 1-21. DOI: 10.5326/JAAHA-MS-6822
7. Bindu J., Narendhirakannan R. T.. **Role of medicinal plants in the management of diabetes mellitus: A review**. *Biotech* (2018) **9** 4. DOI: 10.1007/s13205-018-1528-0
8. Brownlee M.. **The pathobiology of diabetic complications: A unifying mechanism**. *Diabetes* (2005) **54** 1615-1625. DOI: 10.2337/diabetes.54.6.1615
9. Catchpole B., Ristic J. M., Fleeman L. M., Davison L. J.. **Canine diabetes mellitus: Can old dogs teach us new tricks?**. *Diabetologia* (2005) **48** 1948-1956. DOI: 10.1007/s00125-005-1921-1
10. Chen C. C., Lii C. K., Lin Y. H., Shie P. H., Yang Y. C., Huang C. S.. **Andrographis paniculata improves insulin resistance in high-fat diet-induced obese mice and tnfα-treated 3T3-L1 adipocytes**. *Am. J. Chin. Med.* (2020) **48** 1073-1090. DOI: 10.1142/S0192415X20500524
11. Chen Z. Z., Gerszten R. E.. **Metabolomics and proteomics in type 2 diabetes**. *Circ. Res.* (2020) **126** 1613-1627. DOI: 10.1161/CIRCRESAHA.120.315898
12. Dai Y., Chen S-R., Chai L., Zhao J., Wang Y., Wang Y.. **Overview of pharmacological activities of Andrographis paniculata and its major compound andrographolide**. *Crit. Rev. Food Sci. Nutr.* (2019) **59** S17-S29. DOI: 10.1080/10408398.2018.1501657
13. Furman B. L.. **Streptozotocin-induced diabetic models in mice and rats**. *Curr. Protoc.* (2021) **1** e78. DOI: 10.1002/cpz1.78
14. Gheita T. A., Kenawy S. A., Eggleton P., Ward F. J.. **Measurement of malondialdehyde, glutathione, and glutathione peroxidase in SLE patients**. *Systemic lupus erythematosus: Methods and protocols* (2014) 193-199
15. Gilor C., Niessen S. J. M., Furrow E., DiBartola S. P.. **What's in a name? Classification of diabetes mellitus in veterinary medicine and why it matters**. *J. Veterinary Intern. Med.* (2016) **30** 927-940. DOI: 10.1111/jvim.14357
16. Gummesson A., Björnson E., Fagerberg L., Zhong W., Tebani A., Edfors F.. **Longitudinal plasma protein profiling of newly diagnosed type 2 diabetes**. *EBioMedicine* (2021) **63** 103147. DOI: 10.1016/j.ebiom.2020.103147
17. Heydemann A.. **An overview of murine high fat diet as a model for type 2 diabetes mellitus**. *J. Diabetes Res.* (2016) **2016** 2902351. DOI: 10.1155/2016/2902351
18. Hidayat R., Wulandari P.. **Effects of Andrographis paniculata (burm. F.) extract on diabetic nephropathy in rats**. *Rep. Biochem. Mol. Biol.* (2021) **10** 445-454. DOI: 10.52547/rbmb.10.3.445
19. Hossain S., Urbi Z., Karuniawati H., Mohiuddin R. B., Moh Qrimida A., Allzrag A. M. M.. **Andrographis paniculata (burm. F.) wall. Ex nees: An updated review of phytochemistry, antimicrobial pharmacology, and clinical safety and efficacy**. *Life (Basel)* (2021) **11** 348. DOI: 10.3390/life11040348
20. Islam M. T.. **Andrographolide, a new hope in the prevention and treatment of metabolic syndrome**. *Front. Pharmacol.* (2017) **8** 571. DOI: 10.3389/fphar.2017.00571
21. Jaiyesimi K. F., Agunbiade O. S., Ajiboye B. O., Afolabi O. B.. **Polyphenolic-rich extracts of Andrographis paniculata mitigate hyperglycemia via attenuating β-cell dysfunction, pro-inflammatory cytokines and oxidative stress in alloxan-induced diabetic Wistar albino rat**. *J. Diabetes Metab. Disord.* (2020) **19** 1543-1556. DOI: 10.1007/s40200-020-00690-2
22. Jayakumar T., Hsieh C. Y., Lee J. J., Sheu J. R.. **Experimental and clinical pharmacology of Andrographis paniculata and its major bioactive phytoconstituent andrographolide**. *Evid. Based Complement. Altern. Med.* (2013) **2013** 846740. DOI: 10.1155/2013/846740
23. Ji X., Li C., Ou Y., Li N., Yuan K., Yang G.. **Andrographolide ameliorates diabetic nephropathy by attenuating hyperglycemia-mediated renal oxidative stress and inflammation via Akt/NF-κB pathway**. *Mol. Cell Endocrinol.* (2016) **437** 268-279. DOI: 10.1016/j.mce.2016.06.029
24. Jin L., Shi G., Ning G., Li X., Zhang Z.. **Andrographolide attenuates tumor necrosis factor-alpha-induced insulin resistance in 3T3-L1 adipocytes**. *Mol. Cell Endocrinol.* (2011) **332** 134-139. DOI: 10.1016/j.mce.2010.10.005
25. Kim S. W., Choi J. W., Yun J. W., Chung I. S., Cho H. C., Song S. E.. **Proteomics approach to identify serum biomarkers associated with the progression of diabetes in Korean patients with abdominal obesity**. *PLoS One* (2019) **14** e0222032. DOI: 10.1371/journal.pone.0222032
26. Kishore V., Yarla N. S., Bishayee A., Putta S., Malla R., Neelapu N. R.. **Multi-targeting andrographolide and its natural analogs as potential therapeutic agents**. *Curr. Top. Med. Chem.* (2017) **17** 845-857. DOI: 10.2174/1568026616666160927150452
27. Lee M. J., Rao Y. K., Chen K., Lee Y. C., Chung Y. S., Tzeng Y. M.. **Andrographolide and 14-deoxy-11,12-didehydroandrographolide from Andrographis paniculata attenuate high glucose-induced fibrosis and apoptosis in murine renal mesangeal cell lines**. *J. Ethnopharmacol.* (2010) **132** 497-505. DOI: 10.1016/j.jep.2010.07.057
28. Liang E., Liu X., Du Z., Yang R., Zhao Y.. **Andrographolide ameliorates diabetic cardiomyopathy in mice by blockage of oxidative damage and NF-κB-Mediated inflammation**. *Oxid. Med. Cell Longev.* (2018) **2018** 9086747. DOI: 10.1155/2018/9086747
29. Loureiro Damasceno J. P., Silva da Rosa H., Silva de Araújo L., Jacometti Cardoso Furtado N. A.. **Andrographis paniculata formulations: Impact on diterpene lactone oral bioavailability**. *Eur. J. Drug Metabolism Pharmacokinet.* (2022) **47** 19-30. DOI: 10.1007/s13318-021-00736-7
30. Mehta S., Sharma A. K., Singh R. K.. **Pharmacological activities and molecular mechanisms of pure and crude extract of Andrographis paniculata: An update**. *Phytomedicine Plus* (2021) **1** 100085. DOI: 10.1016/j.phyplu.2021.100085
31. Mondal M., Sarkar C., Saha S., Hossain M. N., Norouzi R., Mubarak M. S.. **Hepatoprotective activity of andrographolide possibly through antioxidative defense mechanism in Sprague-Dawley rats**. *Toxicol. Rep.* (2022) **9** 1013-1022. DOI: 10.1016/j.toxrep.2022.04.007
32. Nagalekshmi R., Menon A., Chandrasekharan D. K., Nair C. K.. **Hepatoprotective activity of Andrographis paniculata and Swertia chirayita**. *Food Chem. Toxicol.* (2011) **49** 3367-3373. DOI: 10.1016/j.fct.2011.09.026
33. Naik R. R., Munipally P. K., Nagaraju T.. **Andrographolide reorganise hyperglycaemia and distorted antioxidant profile in streptozotocin-induced diabetic rats**. *Cardiovasc Hematol. Agents Med. Chem.* (2017) **15** 121-127. DOI: 10.2174/1871525715666171026115248
34. Nelson R. W., Reusch C. E.. **Animal models of disease: Classification and etiology of diabetes in dogs and cats**. *J. Endocrinol.* (2014) **222** T1-T9. DOI: 10.1530/JOE-14-0202
35. Nugroho A. E., Andrie M., Warditiani N. K., Siswanto E., Pramono S., Lukitaningsih E.. **Antidiabetic and antihiperlipidemic effect of Andrographis paniculata (Burm. f.) Nees and andrographolide in high-fructose-fat-fed rats**. *Indian J. Pharmacol.* (2012) **44** 377-381. DOI: 10.4103/0253-7613.96343
36. O'Kell A. L., Wasserfall C., Catchpole B., Davison L. J., Hess R. S., Kushner J. A.. **Comparative pathogenesis of autoimmune diabetes in humans, NOD mice, and canines: Has a valuable animal model of type 1 diabetes been overlooked?**. *Diabetes* (2017) **66** 1443-1452. DOI: 10.2337/db16-1551
37. Ogbu S. O., Agwu K. K., Asuzu I. U.. **Gongronema latifolium delays gastric emptying of semi-solid meals in diabetic dogs**. *Afr. J. Tradit. Complement. Altern. Med.* (2013) **10** 325-331. DOI: 10.4314/ajtcam.v10i5.17
38. O’Kell A. L., Shome M., Qiu J., Williams S., Chung Y., LaBaer J.. **Exploration of autoantibody responses in canine diabetes using protein arrays**. *Sci. Rep.* (2022) **12** 2490. DOI: 10.1038/s41598-022-06599-5
39. Okhuarobo A., Ehizogie Falodun J., Erharuyi O., Imieje V., Falodun A., Langer P.. **Harnessing the medicinal properties of Andrographis paniculata for diseases and beyond: A review of its phytochemistry and pharmacology**. *Asian Pac. J. Trop. Dis.* (2014) **4** 213-222. DOI: 10.1016/s2222-1808(14)60509-0
40. Panossian A., Hovhannisyan A., Mamikonyan G., Abrahamian H., Hambardzumyan E., Gabrielian E.. **Pharmacokinetic and oral bioavailability of andrographolide from Andrographis paniculata fixed combination Kan Jang in rats and human**. *Phytomedicine* (2000) **7** 351-364. DOI: 10.1016/S0944-7113(00)80054-9
41. Papachristoforou E., Lambadiari V., Maratou E., Makrilakis K.. **Association of glycemic indices (hyperglycemia, glucose variability, and hypoglycemia) with oxidative stress and diabetic complications**. *J. Diabetes Res.* (2020) **2020** 7489795. DOI: 10.1155/2020/7489795
42. Papatheodorou K., Papanas N., Banach M., Papazoglou D., Edmonds M.. **Complications of diabetes 2016**. *J. Diabetes Res.* (2016) **2016** 6989453. DOI: 10.1155/2016/6989453
43. Patil R., Andrographolide Jain V.. **Andrographolide: A review of analytical methods**. *J. Chromatogr. Sci.* (2020) **59** 191-203. DOI: 10.1093/chromsci/bmaa091
44. Pena M. J., Mischak H., Heerspink H. J.. **Proteomics for prediction of disease progression and response to therapy in diabetic kidney disease**. *Diabetologia* (2016) **59** 1819-1831. DOI: 10.1007/s00125-016-4001-9
45. Reagan-Shaw S., Nihal M., Ahmad N.. **Dose translation from animal to human studies revisited**. *Fed. Am. Soc. Exp. Biol.* (2008) **22** 659-661. DOI: 10.1096/fj.07-9574LSF
46. Riaz S.. **Study of protein biomarkers of diabetes mellitus type 2 and therapy with vitamin B1**. *J. Diabetes Res.* (2015) **2015** 150176. DOI: 10.1155/2015/150176
47. Russell K. R., Omoruyi F. O., Pascoe K. O., Morrison E. Y.. **Hypoglycaemic activity of Bixa orellana extract in the dog**. *Methods Find. Exp. Clin. Pharmacol.* (2008) **30** 301-305. DOI: 10.1358/mf.2008.30.4.1186073
48. Saxena R. C., Singh R., Kumar P., Yadav S. C., Negi M. P., Saxena V. S.. **A randomized double blind placebo controlled clinical evaluation of extract of Andrographis paniculata (KalmCold) in patients with uncomplicated upper respiratory tract infection**. *Phytomedicine* (2010) **17** 178-185. DOI: 10.1016/j.phymed.2009.12.001
49. Siddiqui A., Desai N. G., Sharma S. B., Aslam M., Sinha U. K., Madhu S. V.. **Association of oxidative stress and inflammatory markers with chronic stress in patients with newly diagnosed type 2 diabetes**. *Diabetes/metabolism Res. Rev.* (2019) **35** e3147. DOI: 10.1002/dmrr.3147
50. Sivaraj A., Vinothkumar P., Sathiyaraj K., Sundaresan S., Devi K., Senthilkumar B.. **Hepatoprotective potential of Andrographis paniculata aqueous leaf extract on ethanol induced liver toxicity in albino rats**. *J. Appl. Pharm. Sci.* (2011) **2011** 204-208
51. Songvut P., Suriyo T., Panomvana D., Rangkadilok N., Satayavivad J.. **A comprehensive review on disposition kinetics and dosage of oral administration of Andrographis paniculata, an alternative herbal medicine, in co-treatment of coronavirus disease**. *Front. Pharmacol.* (2022) **13** 952660. DOI: 10.3389/fphar.2022.952660
52. Suemanotham N., Photcharatinnakorn P., Chantong B., Buranasinsup S., Phochantachinda S., Sakcamduang W.. **Curcuminoid supplementation in canine diabetic mellitus and its complications using proteomic analysis**. *Front. Veterinary Sci.* (2022) **9** 1057972. DOI: 10.3389/fvets.2022.1057972
53. Syukri Y., Taher M., Martien R., Lukitaningsih E., Nugroho A. E., Zakaria Z. A.. **Self-nanoemulsifying delivery of andrographolide: Ameliorating islet beta cells and inhibiting adipocyte differentiation**. *Adv. Pharm. Bull.* (2021) **11** 171-180. DOI: 10.34172/apb.2021.018
54. Thamlikitkul V., Dechatiwongse T., Theerapong S., Chantrakul C., Boonroj P., Punkrut W.. **Efficacy of Andrographis paniculata, Nees for pharyngotonsillitis in adults**. *J. Med. Assoc. Thai* (1991) **74** 437-442. PMID: 1797953
55. Wediasari F., Nugroho G. A., Fadhilah Z., Elya B., Setiawan H., Mozef T.. **Hypoglycemic effect of a combined**. *Adv. Pharmacol. Pharm. Sci.* (2020) **2020** 8856129. DOI: 10.1155/2020/8856129
56. Xu F., Fu S., Gu S., Wang Z., Wang Z., He X.. **Simultaneous determination of andrographolide, dehydroandrographolide and neoandrographolide in dog plasma by LC–MS/MS and its application to a dog pharmacokinetic study of Andrographis paniculata tablet**. *J. Chromatogr. B* (2015) **990** 125-131. DOI: 10.1016/j.jchromb.2015.03.014
57. Xu J., Li Z., Cao M., Zhang H., Sun J., Zhao J.. **Synergetic effect of Andrographis paniculata polysaccharide on diabetic nephropathy with andrographolide**. *Int. J. Biol. Macromol.* (2012) **51** 738-742. DOI: 10.1016/j.ijbiomac.2012.06.035
58. Yu B. C., Hung C. R., Chen W. C., Cheng J. T.. **Antihyperglycemic effect of andrographolide in streptozotocin-induced diabetic rats**. *Planta Med.* (2003) **69** 1075-1079. DOI: 10.1055/s-2003-45185
59. Zeng B., Wei A., Zhou Q., Yuan M., Lei K., Liu Y.. **Andrographolide: A review of its pharmacology, pharmacokinetics, toxicity and clinical trials and pharmaceutical researches**. *Phytotherapy Res.* (2022) **36** 336-364. DOI: 10.1002/ptr.7324
60. Zhang C., Gui L., Xu Y., Wu T., Liu D.. **Preventive effects of andrographolide on the development of diabetes in autoimmune diabetic NOD mice by inducing immune tolerance**. *Int. Immunopharmacol.* (2013) **16** 451-456. DOI: 10.1016/j.intimp.2013.05.002
61. Zhang H., Li S., Si Y., Xu H.. **Andrographolide and its derivatives: Current achievements and future perspectives**. *Eur. J. Med. Chem.* (2021) **224** 113710. DOI: 10.1016/j.ejmech.2021.113710
62. Zhang X-F., Tan B. K-H.. **Antihyperglycaemic and anti-oxidant properties of Andrographis paniculata in normal and diabetic rats**. *Clin. Exp. Pharmacol. Physiology* (2000) **27** 358-363. DOI: 10.1046/j.1440-1681.2000.03253.x
63. Zhao H-Y., Hu H., Wang Y-T.. **Comparative metabolism and stability of andrographolide in liver microsomes from humans, dogs and rats using ultra-performance liquid chromatography coupled with triple-quadrupole and Fourier transform ion cyclotron resonance mass spectrometry**. *Rapid Commun. Mass Spectrom.* (2013) **27** 1385-1392. DOI: 10.1002/rcm.6585
|
---
title: 'Intraoperative mean arterial pressure and acute kidney injury after robot-assisted
laparoscopic prostatectomy: a retrospective study'
authors:
- Tae Lim Kim
- Namo Kim
- Hye Jung Shin
- Matthew R. Cho
- Hae Ri Park
- So Yeon Kim
journal: Scientific Reports
year: 2023
pmcid: PMC9971240
doi: 10.1038/s41598-023-30506-1
license: CC BY 4.0
---
# Intraoperative mean arterial pressure and acute kidney injury after robot-assisted laparoscopic prostatectomy: a retrospective study
## Abstract
Intraoperative hemodynamics can affect postoperative kidney function. We aimed to investigate the effect of intraoperative mean arterial pressure (MAP) as well as other risk factors on the occurrence of acute kidney injury (AKI) after robot-assisted laparoscopic prostatectomy (RALP). We retrospectively evaluated the medical records of 750 patients who underwent RALP. The average real variability (ARV)-MAP, standard deviation (SD)-MAP, time-weighted average (TWA)-MAP, area under threshold (AUT)-65 mmHg, and area above threshold (AAT)-120 mmHg were calculated using MAPs collected within a 10-s interval. Eighteen ($2.4\%$) patients developed postoperative AKI. There were some univariable associations between TWA-MAP, AUT-65 mmHg, and AKI occurrence; however, multivariable analysis found no association. Alternatively, American Society of Anesthesiologists physical status ≥ III and the low intraoperative urine output were independently associated with AKI occurrence. Moreover, none of the five MAP parameters could predict postoperative AKI, with the area under the receiver operating characteristic curve values for ARV-MAP, SD-MAP, TWA-MAP, AUT-65 mmHg, and AAT-120 mmHg being 0.561 ($95\%$ confidence interval [CI], 0.424–0.697), 0.561 ($95\%$ CI, 0.417–0.704), 0.584 ($95\%$ CI, 0.458–0.709), 0.590 ($95\%$ CI, 0.462–0.718), and 0.626 ($95\%$ CI, 0.499–0.753), respectively. Therefore, intraoperative MAP changes may not be a determining factor for AKI after RALP.
## Introduction
Robot-assisted laparoscopic prostatectomy (RALP) is preferred to open prostatectomy since it allows minimal invasion, better short-term outcomes, and improved functional results1,2. However, pneumoperitoneum during RALP can induce direct compression of the renal vasculature, ureter, and kidney, which causes a reduction in the renal blood flow and glomerular filtration rate as well as oliguria3,4. The incidence of acute kidney injury (AKI) after RALP has been reported to be significantly lower than that after open radical prostatectomy5; however, AKI still occurs in approximately $5\%$ of patients undergoing RALP5,6. Moreover, an acute increase in serum creatinine (SCr) levels after RALP usually occurs on the operation day7,8. Patients who develop AKI on the operation day have been reported to present reduced renal function at 12 months after RALP8. Therefore, careful perioperative management may be crucial for preventing AKI.
The steep Trendelenburg position with carbon dioxide (CO2) insufflation is necessary for optimizing surgical exposure during RALP. It results in remarkable hemodynamic alterations, including a > $30\%$ increase in the mean arterial pressure (MAP)9–11. Moreover, there is a considerable abrupt decrease in MAP after resuming a supine position with CO2 desufflation11,12. Intraoperative blood pressure (BP) variability has an independent positive correlation with the risk of AKI after non-cardiac surgery13. Additionally, intraoperative low MAP is an established risk factor for AKI after non-cardiac surgery13–17. Therefore, abrupt changes in MAP during RALP can affect postoperative kidney function. Moreover, patients undergoing RALP are mostly older adults with various comorbidities, including hypertension and diabetes mellitus5,6,8,12, which may further increase BP variability and the intraoperative risk of hypotension13,18. However, the relationship between intraoperative MAP changes and postoperative AKI in patients undergoing RALP remains unclear. Therefore, this retrospective study aimed to investigate the association of intraoperative MAP (MAP variability, hypotension, and hypertension) with AKI occurrence after RALP. Additionally, we aimed to investigate the association between AKI following RALP and other factors.
## Patient characteristics
Among 833 screened patients, we excluded two patients with a preoperative estimated glomerular filtration rate (eGFR) < 30 mL/min/1.73 m2 and 81 patients with missing intraoperative BP for > 10 min. Finally, we included 750 patients (Fig. 1). Out of the 750 patients, 18 ($2.4\%$) patients developed postoperative AKI. Table 1 summarizes the patient characteristics. Patients who developed AKI had a significantly higher American Society of Anesthesiologists (ASA) physical status, and were more likely to have diabetes mellitus, coronary artery disease, and atrial fibrillation than those who did not develop AKI. Additionally, a significantly greater proportion of patients with AKI was taking angiotensin-converting enzyme inhibitor (ACEI)/angiotensin receptor blocker (ARB) or diuretics than those without AKI. Table 2 summarizes the intraoperative characteristics. Patients with AKI had significantly longer Trendelenburg position time, lesser amount of urine output, and greater number of transfusions compared with those without AKI.Figure 1Flow chart of the study population. Table 1Patient characteristics. VariablesAll ($$n = 750$$)No AKI ($$n = 732$$)AKI ($$n = 18$$)P valueAge, y68 [63–73]68 [63–73]71 [63–75]0.438Body mass index, kg/m225.0 [23.1–26.7]25.0 [23.1–26.7]26.4 [24.4–28.1]0.084ASA *Physical status* < 0.001 I117 ($15.6\%$)116 ($15.9\%$)1 ($5.6\%$) II402 ($53.6\%$)398 ($54.4\%$)4 ($22.2\%$) III231 ($30.8\%$)218 ($29.8\%$)13 ($72.2\%$)Comorbidities Hypertension428 ($57.1\%$)414 ($56.6\%$)14 ($77.8\%$)0.072 Diabetes mellitus133 ($17.7\%$)126 ($17.2\%$)7 ($38.9\%$)0.027 Coronary artery disease50 ($6.7\%$)46 ($6.3\%$)4 ($22.2\%$)0.027 COPD67 ($8.9\%$)65 ($8.9\%$)2 ($11.1\%$)0.671 Atrial fibrillation27 ($3.6\%$)24 ($3.3\%$)3 ($16.7\%$)0.024 Chronic kidney disease21 ($2.8\%$)19 ($2.6\%$)2 ($11.1\%$)0.087Preoperative medication Calcium channel blocker249 ($33.2\%$)243 ($33.2\%$)6 ($33.3\%$)0.990 β-blocker67 ($8.9\%$)65 ($8.9\%$)2 ($11.1\%$)0.671 ACEI or ARB321 ($42.8\%$)308 ($42.1\%$)13 ($72.2\%$)0.011 Diuretics80 ($10.7\%$)75 ($10.3\%$)5 ($27.8\%$)0.034Preoperative laboratory values Hematocrit, %42.1 [39.6–44.3]42.1 [39.6–44.3]43.2 [39.6–44.6]0.427 Serum albumin, g/dL4.6 [4.4–4.8]4.6 [4.4–4.8]4.6 [4.4–4.7]0.481 Serum sodium, mmol/L140 [139–142]140 [139–142]141 [139–141]0.849 eGFR, mL/min/1.73 m286 [77–93]86 [77–93]78 [56–91]0.081Preoperative MAP, mmHg97 [91–103]97 [91–103]97 [92–104]0.888Values are median (interquartile rage) or number of patients (percentage).AKI acute kidney injury, ASA American Society of Anesthesiology, ACEI angiotensin-converting enzyme inhibitor, ARB angiotensin receptor blocker, COPD chronic obstructive pulmonary disease, eGFR estimated glomerular filtration rate, MAP mean arterial pressure. Table 2Intraoperative characteristics. VariablesAll ($$n = 750$$)No AKI ($$n = 732$$)AKI ($$n = 18$$)P valueAnesthesia time, min140 [125–160]140 [125–160]153 [120–185]0.497Trendelenburg position time, min57 [50–74]57 [49–73]76 [56–96]0.013Fluid intake, mL1100 [910–1300]1100 [905–1300]1200 [1000–1600]0.066Urine output, mL200 [100–250]200 [100–250]115 [100–200]0.028Blood loss, mL300 [200–500]300 [200–500]400 [200–550]0.222Transfusion of red blood cells, yes5 ($0.7\%$)3 ($0.4\%$)2 ($11.1\%$)0.005Administered dose of ephedrine, mg4 (0–8)4 (0–8)4 (0–8)0.391Use of NE or Phenyl, yes49 ($6.5\%$)48 ($6.6\%$)1 ($5.6\%$) >.999Administered dose of remifentanil, µg386 [304–500]385 [304–500]471 [330–612]0.077Values are median (interquartile rage) or number of patients (percentage).AKI acute kidney injury, NE norepinephrine, Phenyl phenylephrine.
## Mean arterial pressure (MAP) parameters
Among the 750 included patients, 642,336 measurements of continuous MAP were obtained. The median [IQR] number of measured MAPs per surgery was 817 [728–930]. Figure 2 shows the distributions of MAP parameters. The median [IQR] values of the average real variability (ARV)-MAP, standard deviation (SD)-MAP, time-weighted average (TWA)-MAP, area under threshold (AUT)-65 mmHg, and area above threshold (AAT)-120 mmHg were 7 [6–8] mmHg/min, 13 [11–15] mmHg, 81 [77–86] mmHg, 32 [5–89] mmHg × min, and 2 [0–28] mmHg × min, respectively. There were significant correlations among MAP parameters (Fig. 3). ARV-MAP was positively correlated with all other MAP parameters except AUT-65 mmHg (Fig. 3). Body mass index (BMI) and preoperative MAP showed a significant positive correlation with ARV-MAP; whereas, patients with diabetes mellitus had significantly lower ARV-MAP values (Table 3).Figure 2Scatterplots of the mean arterial pressure parameters. The distribution of the mean arterial pressure parameters: (A) average real variability of mean arterial pressure (ARV-MAP), (B) standard deviation of mean arterial pressure (SD-MAP), (C) time-weighted average of mean arterial pressure (TWA-MAP), (D) area under threshold of mean arterial pressure of 65 mmHg (AUT-65 mmHg), and (E) area above threshold of mean arterial pressure of 120 mmHg (AAT-120 mmHg) within the study population. Q1, first quartile; Q3, third quartile. Figure 3Heat map showing the correlations among mean arterial pressure parameters. The squares show the correlation coefficients (− 1 to 1), with P values, between the variables according to row and column. The red and blue show positive and negative correlations, respectively. ARV, average real variability; SD, standard deviation; TWA, time-weighted average; MAP, mean arterial pressure; AUT, area under threshold; AAT, area above threshold. Table 3Correlation between average real variability of mean arterial pressure and clinical factors. VariablesCoefficientP valueAge− 0.0690.060Body mass index0.0830.022ASA physical status− 0.0090.809Hypertension− 0.0300.408Diabetes mellitus− 0.0810.026ACEI or ARB− 0.0260.479Preoperative MAP0.158< 0.001ASA American Society of Anesthesiology, ACEI angiotensin-converting enzyme inhibitor, ARB angiotensin receptor blocker, MAP mean arterial pressure.
## Primary and secondary outcomes
Among the 18 patients who developed AKI, 12 ($66.7\%$) patients, 3 ($16.7\%$) patients, and 3 ($16.7\%$) patients were stage 1, 2, and 3, respectively. In the univariable analysis of risk factors for AKI, 19 factors including TWA-MAP and AUT-65 mmHg revealed $P \leq 0.2$ (Table 4). Out of these 19 factors, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, and chronic kidney disease were excluded in the multivariable analysis because they are strongly related to ASA physical status. Moreover, exclusion of chronic kidney disease can be rationalized due to the selection of preoperative eGFR. The ‘Trendelenburg position time’ and the ‘anesthesia time’ were highly correlated (correlation coefficient = 0.84), thus we chose the ‘Trendelenburg position time’ due to its lower P value compared to the ‘anesthesia time’ ($$P \leq 0.004$$ vs $$P \leq 0.107$$). Finally, 13 factors were included in the multivariable analysis. ASA physical status ≥ III (odd ratio [OR] = 4.97, $95\%$ confidence interval [CI] 1.51–16.34) and the intraoperative urine output (OR = 0.61, $95\%$ CI 0.42–0.89 per 50 mL increase) were independently associated with the occurrence of AKI (Table 4).Table 4Univariable and multivariable analyses of risk factors for postoperative acute kidney injury. UnivariableMultivariableOR ($95\%$ CI)P valueOR ($95\%$ CI)P valueARV-MAP, 10 mmHg/min increase2.94 (0.32–27.1)0.342SD-MAP, 10 mmHg increase1.58 (0.38–6.63)0.534TWA-MAP, 10 mmHg increase1.55 (0.81–2.99)0.1891.71 (0.61–4.77)0.310AUT-65 mmHg, 50 mmHg × min increase0.70 (0.44–1.13)0.1460.62 (0.35–1.09)0.095AAT-120 mmHg, 50 mmHg × min increase1.11 (0.76–1.61)0.600Age, 10 y increase1.29 (0.65–2.55)0.465Body mass index, 5 kg/m2 increase2.05 (0.94–4.49)0.0731.87 (0.68–5.09)0.224ASA physical status ≥ III (referent I & II)6.13 (2.16–17.4)0.0014.97 (1.51–16.3)0.008Comorbidities Hypertension2.69 (0.88–8.25)0.084 Diabetes mellitus3.06 (1.16–8.05)0.023 Coronary artery disease4.26 (1.35–13.5)0.014 COPD1.28 (0.29–5.70)0.743 Atrial fibrillation5.90 (1.60–21.8)0.008 Chronic kidney disease4.69 (1.01–21.9)0.049Preoperative medication β-blocker1.28 (0.29–5.70)0.743 Calcium channel blocker1.01 (0.37–2.71)0.990 ACEI or ARB3.58 (1.26–10.14)0.0162.79 (0.78–9.96)0.115 Diuretics3.37 (1.17–9.71)0.0251.67 (0.47–5.93)0.432Preoperative MAP, 10 mmHg increase0.92 (0.53–1.62)0.777Preoperative laboratory values Hematocrit, $3\%$ increase1.18 (0.79–1.77)0.427 Serum albumin, 1 g/dL increase0.62 (0.12–3.30)0.571 eGFR, 10 mL/min/1.73 m2 increase0.67 (0.50–0.89)0.0070.75 (0.54–1.05)0.094 Serum sodium, 5 mmol/L increase1.15 (0.37–3.59)0.814Trendelenburg position time, 10 min increase1.17 (1.05–1.29)0.0041.01 (0.80–1.28)0.936Anesthesia time, 30 min increase1.28 (0.95–1.73)0.107Fluid intake, 100 mL increase1.14 (1.04–1.26)0.0071.27 (0.99–1.62)0.058Urine output, 50 mL increase0.74 (0.56–0.99)0.0390.61 (0.42–0.89)0.009Blood loss, 100 mL increase1.13 (1.00–1.28)0.0590.77 (0.58–1.02)0.065Transfusion of red blood cells, yes30.4 (4.75–194) <.00122.4 (0.51–988)0.107Administered dose of remifentanil, 100 µg increase1.29 (1.05–1.60)0.0161.13 (0.77–1.66)0.522Administered dose of ephedrine, 4 mg increase0.91 (0.67–1.23)0.526Use of NE or phenyl, yes0.84 (0.11–6.43)0.865OR odds ratio, CI confidence interval, ARV average real variability, SD standard deviation, TWA time-weighted average, MAP mean arterial pressure, AUT area under threshold, AAT area above threshold, ASA American Society of Anesthesiologists, COPD chronic obstructive pulmonary diseas, ACEI angiotensin-converting enzyme inhibitor, ARB angiotensin receptor blocker, eGFR estimated glomerular filtration rate, NE norepinephrine, Phenyl phenylephrine.
Table 5 presents the area under the receiver operating characteristic curve (AUROC) values of each MAP parameter for predicting AKI after RALP. No MAP parameter could predict postoperative AKI. Specifically, the AUROC values for ARV-MAP, SD-MAP, TWA-MAP, AUT-65 mmHg, and AAT-120 mmHg were 0.561 ($95\%$ confidence interval [CI], 0.424–0.697), 0.561 ($95\%$ CI, 0.417–0.704), 0.584 ($95\%$ CI, 0.458–0.709), 0.590 ($95\%$ CI, 0.462–0.718), and 0.626 ($95\%$ CI, 0.499–0.753), respectively. Table 5Area under the receiver operating characteristic curve of each mean arterial pressure parameter for predicting postoperative acute kidney injury. ParametersArea under the curve$95\%$ CIP valueARV-MAP, mmHg/min0.5610.424–0.6970.342SD-MAP, mmHg0.5610.417–0.7040.534TWA-MAP, mmHg0.5840.458–0.7090.189AUT-65 mmHg, mmHg × min0.5900.462–0.7180.146AAT-120 mmHg, mmHg × min0.6260.499–0.7530.325CI confidence interval, ARV average real variability, SD standard deviation, TWA time-weighted average, MAP mean arterial pressure, AUT area under threshold, AAT area above threshold.
## Discussion
This is the first retrospective study to explore the relationship between intraoperative MAP and AKI occurrence after RALP. Pneumoperitoneum and the steep Trendelenburg position during RALP cause diverse intraoperative MAP changes. However, we observed no association between these changes (MAP variability, hypotension, and hypertension) and AKI after RALP. In contrast, ASA physical status ≥ III and the low intraoperative urine output were independently associated with the occurrence of AKI.
Pneumoperitoneum, which is dependent on the amount of intra-abdominal pressure, can induce direct renal vascular and parenchymal compression as well as the release of antidiuretic hormone, renin, and aldosterone, which results in decreased renal blood flow, GFR, and renal excretory function3,4. However, it remains unclear whether compared with open radical prostatectomy, RALP increases the risk of AKI. A study reported that compared with open radical prostatectomy, RALP involves a significantly lower incidence of AKI5. However, in another study, 25 ($13.4\%$) out of 187 patients who underwent RALP showed an acute increase in SCr levels on the operation day, which met the Kidney Disease Improving Global Outcomes (KDIGO) criteria for AKI; however, none of the patients who underwent open radical prostatectomy met this criteria7. Therefore, AKI after RALP remains a concern; accordingly, there have been clinical trials for mitigating AKI in patients undergoing RALP. Intraoperative infusion of low-dose (0.5 µg/kg/min) nicardipine, which is a calcium channel blocker, has been found to improve renal function on postoperative day 119. Contrastingly, intraoperative infusion of mannitol (0.5 g/kg) did not facilitate the prevention of AKI after RALP6.
Numerous factors, including CO2 gas insufflation and desufflation as well as performing a steep Trendelenburg position and resuming a supine position, induce abrupt changes in BP during RALP9–12. A large-scale study on patients undergoing non-cardiac surgery reported a positive correlation of intraoperative MAP variability with the risk of postoperative AKI, regardless of intraoperative hypotension13. Similarly, systolic BP variability is negatively correlated with renal function in patients with hypertension20,21. Renal perfusion is maintained by neurohormonal responses over time19,22; therefore, abrupt BP fluctuations may exceed the capacity of such adaptations, which may result in kidney damage. However, we found that ARV-MAP was not related to and a poor predictor of AKI after RALP (Tables 4 and 5). The median [IQR] ARV-MAP was 7 [6–8] mmHg/min, with the lowest and highest values being 3 and 19 mmHg/min, respectively (Fig. 2). Therefore, the BP variability during RALP may be tolerable with respect to renal function.
There remains no recommended standard measurement for BP variability. We assessed MAP variability using the SD-MAP and ARV-MAP. ARV-MAP may be more appropriate than SD-MAP since ARV represent consecutive changes in MAP, while SD does not consider the timing of measurements13,18. Since we preferred ARV-MAP over SD-MAP, we identified preoperative factors correlated with ARV-MAP (Table 3). BMI was positively correlated with intraoperative ARV-MAP, which is consistent with findings from previous reports that showed that compared with normal-weight patients, overweight and obese patients present higher BP variability during their daily lives23,24. We found that patients with diabetes mellitus showed low ARV-MAP, which is inconsistent with the results of a previous report of high intraoperative BP variability in patients with diabetes mellitus13. Although hypertension and treatment with ACEI or ARB did not affect ARV-MAP, intraoperative ARV-MAP was positively correlated with preoperative MAP. Additionally, ARV-MAP showed a positive correlation with TWA-MAP and AUT-120 mmHg as well as a negative correlation with AUT-65 mmHg (Fig. 3). These findings suggest a positive correlation of BP with BP variability, which is consistent with a previous finding of high BP variability in patients with uncontrolled hypertension25.
Intraoperative hypotension is known to be strongly correlated with AKI after non-cardiac surgery13–17. Absolute and relative (reduction from baseline) MAP thresholds have been used to define hypotension14,15,26. However, the association of relative and absolute hypotension thresholds with AKI have similar strengths15. Absolute thresholds are easier to use in decision-making without requiring preoperative BP data, with an absolute MAP threshold of < 65 mmHg being the most commonly used14,26. Therefore, we used a MAP threshold of < 65 mmHg and calculated AUT-65 mmHg, which characterizes the hypotension duration and severity (amount of hypotension). However, AUT-65 mmHg was not associated with and a poor predictor of AKI after RALP (Tables 4 and 5), which could be attributed to our low incidence of AKI. A previous study on 138,021 non-cardiac surgeries demonstrated that the relationship of intraoperative hypotension with AKI varied according to the underlying patient and procedural risks. Specifically, intraoperative hypotension was associated with AKI in patients with medium and high but not low risk27. The AKI incidence was $1.7\%$ and ≥ $4.6\%$ in patients with low and medium risk, respectively27. In our study, the AKI incidence was $2.4\%$, which suggests that patients undergoing RALP had a low risk of AKI and that intraoperative hypotension is not an important determinant of AKI. However, further studies are warranted to elucidate the impact of intraoperative hypotension on AKI after RALP in high-risk patients. Alternatively, TWA-MAP, which represent the overall BP and AAT-120 mmHg, indicating the hypertension duration and severity, were also not associated with and not good predictors of AKI after RALP (Tables 4 and 5). Therefore, the high BP during RALP might be tolerable and not cause renal damage. Although the association of high BP with increased postoperative morbidity remains unclear, compared with hypotension, elevated intraoperative BP may not be as strongly associated with postoperative morbidity26.
In our study, ASA physical status ≥ III and the low intraoperative urine output were revealed to be independent predictors of postoperative AKI. After adjusting for confounding factors, patients with an ASA physical status ≥ III had an approximately fivefold increased risk of AKI compared to the ones with an ASA physical status I or II. Consistent with our result, ASA physical status was proven as a determinant for postoperative AKI in non-cardiac surgery28,29 and an ASA physical status ≥ III showed an approximately twofold increased risk of AKI28. The predictive value of intraoperative urine output for postoperative AKI remains controversial. There are some studies stating that there is no association between intraoperative urine output and the occurrence of AKI30,31, while others are advocating for the association between oliguria and the occurrence of AKI after major non-cardiac surgery28,32,33. Although some previous studies investigated the risk factors for AKI after RALP, none of those included the urine output as a candidate variable6–8. Therefore, our study is the first to identify the low intraoperative urine output as a risk factor for AKI in RALP. However, future studies investigating whether increasing the urine output with the use of diuretics would prevent the development of AKI after RALP are needed to confirm the causative effect of intraoperative urine output on AKI after RALP.
Our findings have strength in terms of the elaborate data quality. Specifically, the MAP values were obtained through invasive arterial monitoring and collected in 10-s intervals; accordingly, 642,336 continuous MAP measurements were included in the analysis. Nevertheless, this study has several limitations. First, this was a single-center retrospective study, which increases the risk of bias and the influence of confounding factors. Although we adjusted for confounding factors based on a previous report34, there could have been unknown and unadjusted confounding factors. Second, we did not consider the use of vasopressors or vasodilators given their dose complexity and inaccuracies. Further studies that consider BP-modifying drugs are warranted to validate our findings. Finally, the lower incidence of AKI in our study, may have led to the conclusion of no association of any MAP parameters with AKI. The AKI incidence in our study was $2.4\%$, which is half of the previously reported incidence ($5\%$) in RALP5,6. Some studies have demonstrated that a longer operation time represents a risk factor for AKI after RALP6–8. Therefore, our low AKI incidence could be attributed to shorter operation times than those observed in previous studies in which the mean operation time was around 170–180 min6. Therefore, multi-center large-scale trials are needed to confirm the association between MAP parameters and AKI after RALP.
In conclusion, ASA physical status ≥ III and the low intraoperative urine output were independent risk factors for AKI after RALP. At the levels observed in the present study, we were unable to demonstrate an association between intraoperative MAP changes, including MAP variability, hypotension, and hypertension, and AKI after RALP. Therefore, dynamic BP changes during RALP may be within the acceptable range for kidney perfusion; accordingly, intraoperative MAP changes may not be an important determinant of postoperative AKI.
## Selection of participants
This single-center retrospective study was approved by the Institutional Review Board and Hospital Research Ethics Committee of Severance Hospital, Yonsei University Health System, Seoul, Korea (number: 4–2021-0434, approved on 25 May 2021) and was performed in accordance with relevant guidelines and regulations. The requirement for informed consent was waived by the Severance hospital ethical committee's institutional review board given the retrospective nature of the study. We screened adult patients who underwent RALP between January 2020 and April 2021; among them, we excluded patients with preoperative eGFR < 30 mL/min/1.73 m2 and those who had missing intraoperative BP for > 10 min.
## Data collection
We retrospectively collected the following demographic characteristics from the patients’ electronic medical records: age, BMI, ASA physical status, and comorbidities such as hypertension (previously diagnosed and currently taking antihypertensive medications), diabetes mellitus (previously diagnosed and taking antidiabetic medications), coronary artery disease (history of coronary angioplasty and stent insertion), chronic obstructive pulmonary disease (emphysema or chronic bronchitis under use of bronchodilators), atrial fibrillation (confirmed by preoperative electrocardiogram), and chronic kidney disease (eGFR < 60 ml/min/1.73 m2 for > 3 months).
Moreover, we collected preoperative medication data, including the use of calcium channel blocker, β-blocker, ACEI or ARB, and diuretics, as well as preoperative laboratory data, including hematocrit, serum albumin, serum sodium levels, and eGFR, which are established predictors for AKI34. The eGFR was calculated based on SCr levels using the Chronic Kidney Disease Epidemiology Collaboration Eq. 35. Regarding preoperative MAP, we calculated the average of two MAPs calculated as follows: MAP = diastolic BP + $\frac{1}{3}$ [systolic BP–diastolic BP] based on BP measurements obtained twice preoperatively. We collected the following intraoperative information: anesthesia duration, duration of Trendelenburg position, fluid intake volume, urine output, blood loss, transfusion of red blood cells (yes/no), administered dose of remifentanil and ephedrine, and use of norepinephrine/phenylephrine (yes/no). Postoperative SCr levels were measured within the first 7 postoperative days.
## Intraoperative blood pressure
Noninvasive BP measurements were obtained using BP cuffs at 1- to 5-min intervals during the anesthesia-induction periods. After anesthesia induction, all patients underwent radial arterial cannulation with a 20-gauge catheter. Recordings of noninvasive and invasive BPs were automatically saved in the VitalDB database via the VitalRecorder, which is a recent software for querying large databases containing high-resolution time-synchronized physiological data obtained using multiple anesthesia devices (https://vitaldb.net)36,37. We downloaded raw data obtained at 10-s intervals and removed artifacts based on the following respective criteria: [1] only systolic or diastolic BP was measured, [2] the systolic BP was > 10 times the diastolic BP, [3] systolic BP > 300 mmHg or diastolic BP > 200 mmHg, [4] systolic BP < 30 mmHg or diastolic BP < 20 mmHg, and [5] diastolic BP was > $95\%$ of the systolic BP13.
## Mean arterial pressure parameters
We calculated five MAP parameters using the measured MAP values. ARV-MAP (mmHg/min) and SD-MAP (mmHg) were used as measures of MAP variability18. The ARV-MAP and SD-MAP were calculated using the following equations:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{ARV}} - {\text{MAP}} = \frac{1}{{\text{T}}}\mathop \sum \limits_{{{\text{i}} = 1}}^{{{\text{N}} - 1}} \left| {{\text{X}}_{{{\text{i}} + 1}} - {\text{X}}_{{\text{i}}} } \right|,{\text{SD}} - {\text{MAP}} = \sqrt {\frac{{\mathop \sum \nolimits_{{{\text{i}} = 1}}^{{\text{N}}} \left({{\text{x}}_{{\text{i}}} - {\overline{\text{x}}}} \right)^{2} }}{{\left({{\text{N}} - 1} \right)}}}$$\end{document}ARV-MAP=1T∑$i = 1$N-1Xi+1-Xi,SD-MAP=∑$i = 1$Nxi-x¯2N-1 N, T, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\overline{\text{x}}}$$\end{document}x¯ represent the number of BP readings, total time (min) from first to last BP reading, and mean MAP value, respectively.
TWA-MAP (mmHg) was calculated as the area under the curve value of the MAP measurements divided by the total measurement time for the mean MAP. AUT-65 mmHg was calculated as the area under a MAP of 65 mmHg while AAT-120 mmHg was calculated as the area above a MAP of 120 mmHg. AUT-65 mmHg (mmHg × min) and AAT-120 mmHg (mmHg × min) present the duration and severity of hypotension and hypertension, respectively26.
## Study outcome
The primary endpoint was the association of intraoperative MAP (MAP variability, hypotension, and hypertension) with AKI after RALP. The secondary endpoint was the association of other factors with AKI after RALP. AKI occurrence was based on the KDIGO criteria, i.e., an increase in the SCr level by ≥ 0.3 mg/dL within 48 h or an increase in the SCr level to ≥ 1.5 times the baseline value within 7 postoperative days38. AKI was further categorized into the following three stages: stage 1, an increase in SCr level by 1.5–1.9 times baseline or by ≥ 0.3 mg/dL; stage 2, an increase in SCr level by 2.0–2.9 times baseline; stage 3, an increase in SCr level by at least 3.0 times baseline or to ≥ 4.0 mg/dL, or the initiation of renal replacement therapy38.
## Statistical analysis
Continuous and categorical variables are presented as median [IQR] and number (percentage), respectively. Spearman’s correlation coefficients were used to determine the correlations among MAP parameters and the correlations of ARV-MAP with clinical factors. A binary logistic regression analysis was performed to find risk factors for the occurrence of AKI. Variables with $P \leq 0.2$ in the univariable analysis were included in the multivariable logistic regression model. However, in order to avoid multicollinearity, only one variable was chosen from a group of highly correlated continuous variables, and the same applies to a group of highly related categorical variables. Receiver operating characteristic curve analysis was performed and the AUROC value was calculated to evaluate the discrimination ability of each MAP parameter to predict AKI. Statistical significance was set at $P \leq 0.05$; moreover, all tests were two-tailed. Statistical analyses were performed using SAS version 9.4 (SAS Institute) and R version 4.0.4 (R Project for Statistical Computing).
## References
1. Novara G. **Systematic review and meta-analysis of perioperative outcomes and complications after robot-assisted radical prostatectomy**. *Eur. Urol.* (2012) **62** 431-452. DOI: 10.1016/j.eururo.2012.05.044
2. Leow JJ. **Robot-assisted versus open radical prostatectomy: A contemporary analysis of an all-payer discharge database**. *Eur. Urol.* (2016) **70** 837-845. DOI: 10.1016/j.eururo.2016.01.044
3. Bishara B. **Impact of pneumoperitoneum on renal perfusion and excretory function: Beneficial effects of nitroglycerine**. *Surg. Endosc.* (2009) **23** 568-576. DOI: 10.1007/s00464-008-9881-4
4. Demyttenaere S, Feldman LS, Fried GM. **Effect of pneumoperitoneum on renal perfusion and function: A systematic review**. *Surg. Endosc.* (2007) **21** 152-160. DOI: 10.1007/s00464-006-0250-x
5. Joo EY. **Comparison of acute kidney injury after robot-assisted laparoscopic radical prostatectomy versus retropubic radical prostatectomy: A propensity score matching analysis**. *Medicine (Baltimore)* (2016) **95** e2650. DOI: 10.1097/MD.0000000000002650
6. Kong Y-G. **Effect of intraoperative mannitol administration on acute kidney injury after robot-assisted laparoscopic radical prostatectomy: A propensity score matching analysis**. *Medicine* (2018) **97** 55. DOI: 10.1097/MD.0000000000011338
7. Naito A. **Transient acute kidney injury observed immediately after robot-assisted radical prostatectomy but not after open radical prostatectomy**. *Mol. Clin. Oncol.* (2020) **13** 18. DOI: 10.3892/mco.2020.2087
8. Sato H. **Acute kidney injury and its impact on renal prognosis after robot-assisted laparoscopic radical prostatectomy**. *Int. J. Med. Robot.* (2020) **16** 1-7. DOI: 10.1002/rcs.2117
9. Lestar M, Gunnarsson L, Lagerstrand L, Wiklund P, Odeberg-Wernerman S. **Hemodynamic perturbations during robot-assisted laparoscopic radical prostatectomy in 45° Trendelenburg position**. *Anesth. Analg.* (2011) **113** 1069-1075. DOI: 10.1213/ANE.0b013e3182075d1f
10. Pawlik MT. **Pronounced haemodynamic changes during and after robotic-assisted laparoscopic prostatectomy: A prospective observational study**. *BMJ Open* (2020) **10** e038045-e038045. DOI: 10.1136/bmjopen-2020-038045
11. Rosendal C, Markin S, Hien MD, Motsch J, Roggenbach J. **Cardiac and hemodynamic consequences during capnoperitoneum and steep Trendelenburg positioning: Lessons learned from robot-assisted laparoscopic prostatectomy**. *J. Clin. Anesth.* (2014) **26** 383-389. DOI: 10.1016/j.jclinane.2014.01.014
12. Kim NY. **Prediction of hypotension after postural change in robot-assisted laparoscopic prostatectomy using esophageal Doppler monitoring: A prospective observational trial**. *Sci. Rep.* (2021) **11** 14589. DOI: 10.1038/s41598-021-93990-3
13. Park S. **Intraoperative arterial pressure variability and postoperative acute kidney injury**. *Clin. J. Am. Soc. Nephrol.* (2020) **15** 35-46. DOI: 10.2215/cjn.06620619
14. Wesselink EM, Kappen TH, Torn HM, Slooter AJC, van Klei WA. **Intraoperative hypotension and the risk of postoperative adverse outcomes: A systematic review**. *Br. J. Anaesth.* (2018) **121** 706-721. DOI: 10.1016/j.bja.2018.04.036
15. Salmasi V. **Relationship between intraoperative hypotension, defined by either reduction from baseline or absolute thresholds, and acute kidney and myocardial injury after noncardiac surgery: A retrospective cohort analysis**. *Anesthesiology* (2017) **126** 47-65. DOI: 10.1097/aln.0000000000001432
16. Ahuja S. **Associations of intraoperative radial arterial systolic, diastolic, mean, and pulse pressures with myocardial and acute kidney injury after noncardiac surgery: A retrospective cohort analysis**. *Anesthesiology* (2020) **132** 291-306. DOI: 10.1097/aln.0000000000003048
17. Gregory A. **Intraoperative hypotension is associated with adverse clinical outcomes after noncardiac surgery**. *Anesth. Analg.* (2021) **132** 1654-1665. DOI: 10.1213/ANE.0000000000005250
18. Mascha EJ, Yang D, Weiss S, Sessler DI. **Intraoperative mean arterial pressure variability and 30-day mortality in patients having noncardiac surgery**. *Anesthesiology* (2015) **123** 79-91. DOI: 10.1097/aln.0000000000000686
19. Bonventre JV, Weinberg JM. **Recent advances in the pathophysiology of ischemic acute renal failure**. *J. Am. Soc. Nephrol.* (2003) **14** 2199-2210. DOI: 10.1097/01.asn.0000079785.13922.f6
20. Wang X. **Twenty-four-hour systolic blood pressure variability and renal function decline in elderly male hypertensive patients with well-controlled blood pressure**. *Clin. Interv. Aging* (2018) **13** 533-540. DOI: 10.2147/CIA.S161752
21. Farrag HMA, Amin AS, Abdel-Rheim AR. **Relation of short-term blood pressure variability to early renal effects in hypertensive patients with controlled blood pressure**. *Blood Press Monit.* (2019) **24** 221-224. DOI: 10.1097/MBP.0000000000000383
22. Munshi R, Hsu C, Himmelfarb J. **Advances in understanding ischemic acute kidney injury**. *BMC Med.* (2011) **9** 11. DOI: 10.1186/1741-7015-9-11
23. Chen H. **Impact of body mass index on long-term blood pressure variability: A cross-sectional study in a cohort of Chinese adults**. *BMC Public Health* (2018) **18** 1193. DOI: 10.1186/s12889-018-6083-4
24. Tadic M. **The association between obesity, blood pressure variability, and right ventricular function and mechanics in hypertensive patients**. *J. Am. Soc. Echocardiogr.* (2016) **29** 802-811. DOI: 10.1016/j.echo.2016.04.001
25. Pengo MF. **Systolic and diastolic short-term blood pressure variability and its determinants in patients with controlled and uncontrolled hypertension: A retrospective cohort study**. *Blood Press* (2015) **24** 124-129. DOI: 10.3109/08037051.2014.992187
26. Sessler DI. **Perioperative Quality Initiative consensus statement on intraoperative blood pressure, risk and outcomes for elective surgery**. *Br. J. Anaesth.* (2019) **122** 563-574. DOI: 10.1016/j.bja.2019.01.013
27. Mathis MR. **Preoperative risk and the association between hypotension and postoperative acute kidney injury**. *Anesthesiology* (2020) **132** 461-475. DOI: 10.1097/ALN.0000000000003063
28. Shiba A, Uchino S, Fujii T, Takinami M, Uezono S. **Association between intraoperative oliguria and acute kidney injury after major noncardiac surgery**. *Anesth. Analg.* (2018) **127** 1229-1235. DOI: 10.1213/ane.0000000000003576
29. Abelha FJ, Botelho M, Fernandes V, Barros H. **Determinants of postoperative acute kidney injury**. *Crit Care* (2009) **13** R79. DOI: 10.1186/cc7894
30. Kheterpal S. **Predictors of postoperative acute renal failure after noncardiac surgery in patients with previously normal renal function**. *Anesthesiology* (2007) **107** 892-902. DOI: 10.1097/01.anes.0000290588.29668.38
31. Goren O, Matot I. **Perioperative acute kidney injury**. *Br. J. Anaesth.* (2015) **115** 3-14. DOI: 10.1093/bja/aev380
32. Mizota T. **Intraoperative oliguria predicts acute kidney injury after major abdominal surgery**. *Br. J. Anaesth.* (2017) **119** 1127-1134. DOI: 10.1093/bja/aex255
33. Zhao BC. **Assessment of prognostic value of intraoperative oliguria for postoperative acute kidney injury: A retrospective cohort study**. *Br. J. Anaesth.* (2021) **126** 799-807. DOI: 10.1016/j.bja.2020.11.018
34. Park S. **Simple postoperative AKI risk (SPARK) classification before noncardiac surgery: A prediction index development study with external validation**. *J. Am. Soc. Nephrol.* (2019) **30** 170-181. DOI: 10.1681/ASN.2018070757
35. Rule AD. **The CKD-EPI equation for estimating GFR from serum creatinine: Real improvement or more of the same?**. *Clin. J. Am. Soc. Nephrol.* (2010) **5** 951. DOI: 10.2215/CJN.03110410
36. Lee HC, Jung CW. **Vital Recorder-a free research tool for automatic recording of high-resolution time-synchronised physiological data from multiple anaesthesia devices**. *Sci. Rep.* (2018) **8** 1527. DOI: 10.1038/s41598-018-20062-4
37. Vistisen ST, Pollard TJ, Enevoldsen J, Scheeren TWL. **VitalDB: Fostering collaboration in anaesthesia research**. *Br. J. Anaesth.* (2021) **127** 184-187. DOI: 10.1016/j.bja.2021.03.011
38. Khwaja A. **KDIGO clinical practice guidelines for acute kidney injury**. *Nephron. Clin. Pract.* (2012) **120** c179-184. DOI: 10.1159/000339789
|
---
title: Sodium benzoate attenuates 2,8-dihydroxyadenine nephropathy by inhibiting monocyte/macrophage
TNF-α expression
authors:
- Yoichi Oshima
- Shu Wakino
- Takeshi Kanda
- Takaya Tajima
- Tomoaki Itoh
- Kiyotaka Uchiyama
- Keiko Yoshimoto
- Jumpei Sasabe
- Masato Yasui
- Hiroshi Itoh
journal: Scientific Reports
year: 2023
pmcid: PMC9971245
doi: 10.1038/s41598-023-30056-6
license: CC BY 4.0
---
# Sodium benzoate attenuates 2,8-dihydroxyadenine nephropathy by inhibiting monocyte/macrophage TNF-α expression
## Abstract
Sodium benzoate (SB), a known D-amino acid oxidase (DAO) enzyme inhibitor, has an anti-inflammatory effect, although its role in renal damage has not been explored. 2,8-dihydroxyadenine crystal induced chronic kidney disease, in which TNF-α is involved in the pathogenesis, was established by oral adenine administration in C57BL/6JJcl mice (AdCKD) with or without SB to investigate its renal protective effects. SB significantly attenuated AdCKD by decreasing serum creatinine and urea nitrogen levels, and kidney interstitial fibrosis and tubular atrophy scores. The survival of AdCKD mice improved 2.6-fold by SB administration. SB significantly decreased the number of infiltrating macrophages observed in the positive F$\frac{4}{80}$ immunohistochemistry area and reduced the expression of macrophage markers and inflammatory genes, including TNF-α, in the kidneys of AdCKD. Human THP-1 cells stimulated with either lipopolysaccharide or TNF-α showed increased expression of inflammatory genes, although this was significantly reduced by SB, confirming the anti-inflammatory effects of SB. SB exhibited renal protective effects in AdCKD in DAO enzyme deficient mice, suggesting that anti-inflammatory effect of SB was independent of DAO enzyme activity. Moreover, binding to motif DNA sequence, protein level, and mRNA level of NF-κB RelB were significantly inhibited by SB in AdCKD kidneys and lipopolysaccharide treated THP-1 cells, respectively. We report that anti-inflammatory property of SB is independent of DAO enzymatic activity and is associated with down regulated NF-κB RelB as well as its downstream inflammatory genes such as TNF-α in AdCKD.
## Introduction
Chronic kidney disease (CKD) is a health problem affecting over $10\%$ of the global population1. CKD results in end-stage renal disease that requires dialysis. Patients with CKD who are on dialysis are susceptible to multiple comorbidities, including life-threatening cardiovascular diseases2,3 and cancer4. Therefore, effective treatment of CKD is essential.
Crystal nephropathy is a form of CKD characterized by the crystallization of uric acid, calcium oxalate, calcium phosphate, and adenine5. These crystals induce renal inflammation in an auto-amplification manner5. Adenine-induced 2,8-dihydroxyadenine (2,8-DHA) crystal nephropathy is an animal CKD model of a human autosomal recessive genetic disorder caused by adenine phosphoribosyl transferase (APRT) deficiency6. Adenine-containing chow, 2,8-DHA nephropathy, or adenine-induced CKD (AdCKD) induced the deposition of crystals in renal tubules, causing inflammation and tubular injury that was followed by interstitial fibrosis and collagen deposition in mice7. The AdCKD model mimics human CKD features, including kidney atrophy and fibrosis, elevated urea nitrogen and creatinine levels, anemia, cardiovascular calcifications, cardiac hypertrophy, and elevated blood pressure8. Several studies have shown that inflammation and macrophage activation are involved in the pathogenesis of AdCKD. Deletion of Tnfr1, the gene encoding a TNF-α receptor, ameliorated disease progression in AdCKD mice7. Inhibition of NF-κB, a molecule involved in the TNFα/Tnfr1 cascade, by pyrrolidine dithiocarbamate also attenuated AdCKD progression9. Ozone therapy attenuated AdCKD by decreasing the expression of toll-like receptor 4 (TLR4)10, a lipopolysaccharide (LPS) receptor. Endoplasmic reticulum (ER) stress was also elevated in the kidneys of AdCKD mice, although this was reduced by the fatty acid receptor GPR40 agonist and aggravated in GPR40 knockout mice11. Since GPR40 signaling reduces ER stress12 and ER stress and NF-κB have an integrated crosstalk13, the inflammatory pathway induced by NF-κB signaling is also important as a downstream inducer of ER stress. AdCKD was also attenuated in CCL3 and CCR5 knockout mice by the monocyte/macrophage depleting agent, clodronate liposomes, suggesting that monocyte/macrophage chemotaxis contributes to the pathogenesis14.
Sodium benzoate (SB) is a salt of benzoic acid, a cinnamon-derived metabolite used as a food and cosmetic additive15. SB is also a competitive inhibitor of d-amino acid oxidase (DAO), a molecule found in the brain, liver, and kidneys in humans and oxidizes d-amino acids to α-keto acids and hydrogen peroxide16. Additionally, SB has been reported to possess anti-inflammatory property17,18, although whether the effect is associated with DAO or not have not been determined.
Therefore, we have investigated whether SB could exert anti-inflammatory effects in AdCKD or lipopolysaccharide treated THP-1 cells. We also describe for the first time the underlying mechanism of anti-inflammation leading to alleviated AdCKD.
## Adenine-induced nephropathy is attenuated by oral administration of sodium benzoate
We divided mice into four groups; control group, adenine-induced CKD (AdCKD) group, sodium benzoate group, and sodium benzoate treated AdCKD group (Fig. 1a). Body weight decreased in AdCKD mice compared to control mice and was improved in the AdCKD + SB mice compared to AdCKD mice (Fig. 1b). Body weights were similar between the control mice and SB mice. Plasma urea nitrogen and creatinine, which is the hallmark of kidney function, was increased in the AdCKD compared to control, whereas theses were significantly lower in the AdCKD + SB mice (Fig. 1c,d). Proteinuria was similar among the four groups (Fig. 1e), whereas renal tubule damage was aggravated in AdCKD compared to control mice, which was significantly attenuated in the AdCKD + SB mice as shown in tubule damage marker lipocalin-2 (LCN-2), also known as the neutrophil gelatinase-associated lipocalin, mRNA levels (Fig. 1f) and interstitial fibrosis tubular atrophy score (Fig. 1g,h). The elevation of LCN-2 mRNA level in AdCKD was compatible with previous report7. AdCKD mice did not have increased proteinuria compared to controls, as shown in Fig. 1e, which was compatible with previous report19. This is likely explained by the C57BL/6 strain's known resistance towards development of proteinuria in combination with the tubulointerstitial nature of the renal damage20, as reviewed previously19. Aquaporin-1 immunofluorescence was lower in AdCKD kidneys, indicating loss of proximal tubular cells in the kidneys of AdCKD mice which was improved in AdCKD + SB mice (Fig. 1g). The average survival period in the AdCKD group was significantly shortened to 113 days compared to control group. However, the survival was significantly improved more than 2.6-fold in the AdCKD + SB group of 296 days compared to AdCKD group (Fig. 1i).Figure 1SB attenuates adenine induced CKD. ( a) Seven-week-old mice were divided into four groups: control, adenine CKD (AdCKD), SB, and AdCKD + SB. The relevant groups were treated with SB in the last two weeks during eight weeks of adenine intake and the mice were sacrificed six weeks later. Changes in body weight (b), plasma urea nitrogen (c), plasma creatinine (d), and urine total protein creatinine ratios (e) were examined ($$n = 10$$ each). ( f) Lcn2 mRNA levels in kidney cortex homogenates were measured using RT-PCR ($$n = 4$$–10). ( g) Representative images of masson-trichrome staining and aquaporin-1 immunofluorescent staining. ( h) Violin plot showing IFTA scores of randomly acquired images ($$n = 50$$ each). ( i) Survival curves of all groups of mice were examined ($$n = 10$$ each). ( j) *Plasma urea* nitrogen and creatinine levels in AdCKD mice treated with different concentrations of SB ($$n = 4$$–5). For urea nitrogen, p-value for respective pair was as follows. Control vs AdCKD, $p \leq 0.0001$; AdCKD versus AdCKD 0.1 mM SB, $$p \leq 0.0318$$; AdCKD 1 mM SB vs AdCKD 5 mM SB, $p \leq 0.0001$; AdCKD 5 mM SB versus AdCKD 20 mM SB, $$p \leq 0.0459.$$ For creatinine, p-value for respective pair was as follows. Control versus AdCKD, $p \leq 0.0001$; AdCKD 1 mM SB versus AdCKD 5 mM SB, $p \leq 0.0001.$ ( k) The effectiveness of late-stage SB administration was confirmed in the AdCKD model by measuring plasma urea nitrogen and creatinine levels ($$n = 5$$ each). * $p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. AdCKD, adenine induced chronic kidney disease; SB, sodium benzoate; TP/Cre, total protein creatinine ratio; LCN2, lipocalin-2; IFTA, interstitial fibrosis and tubular atrophy; CKD, chronic kidney disease.
Treatment with 1 mM of SB had no therapeutic effect on kidney function, based on evaluation of creatinine concentrations in plasma, whereas 5 mM of SB exerted a therapeutic effect, showing a dose-dependent response (Fig. 1j). For plasma urea nitrogen levels, 0.1 mM of SB in AdCKD exerted a significant decrease compared to AdCKD alone (Fig. 1j). Urea nitrogen level improved also showing a dose responsiveness as depicted in Fig. 1j. Therefore, a dose dependent improvement of kidney function was confirmed by SB in AdCKD.
As shown in Fig. 1k, we established AdCKD by administering adenine for six weeks prior to SB addition to seek the SB renal protection is seen in already damaged kidneys. The therapeutic effect was still observed, as confirmed by reduced plasma urea nitrogen and creatinine levels (Fig. 1k).
## SB attenuates AdCKD regardless of DAO enzyme activity
Since SB is a well-known inhibitor of DAO enzyme activity, we examined whether DAO activity in mice plays a role in AdCKD using the DAO-deficient mouse with C57BL6 background (DAO − / −)21. We obtained littermates of wild-type DAO (DAO + / +), heterozygous deficient DAO(+ / −), and homozygous deficient DAO(− / −) mice. The littermates were fed chow containing $0.2\%$ adenine. However, no significant differences were observed in the plasma concentrations of urea nitrogen and creatinine between the study groups (Fig. 2a). To examine whether SB exerts a therapeutic effect in AdCKD DAO(− / −) mice, the mice were divided into three groups: DAO(− / −) control, AdCKD-induced DAO(− / −), and AdCKD-induced DAO(− / −) mice treated with SB. We observed a significant reduction in serum urea nitrogen and creatinine levels in AdCKD-induced DAO(− / −) mice treated with SB compared with that of AdCKD-induced DAO(− / −) mice that were not treated with SB (Fig. 2b). We also confirmed the direct inhibition of normal kidney DAO activity by SB (Fig. 2c). Collectively, these results suggest that SB exerts a therapeutic effect regardless of DAO enzymatic activity in AdCKD mice. Figure 2AdCKD is attenuated by SB regardless of DAO enzymatic activity. ( a) AdCKD was induced in DAO + / +, DAO + / −, and DAO-/- littermates and plasma urea nitrogen and creatinine concentrations compared ($$n = 6$$–17). ( b) DAO − / − mice were divided into control, AdCKD, and AdCKD + SB groups. SB significantly reduced plasma urea nitrogen and creatinine concentrations ($$n = 4$$–11). ( c) DAO activity assay of normal kidney homogenate shows SB significantly reduced DAO activity ($$n = 5$$ each). *** $p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. AdCKD, adenine induced chronic kidney disease; DAO, D-amino acid oxidase.
## SB decreases macrophage infiltration and the expression of inflammatory genes in AdCKD mice
Macrophage infiltration was elevated in the kidneys of AdCKD mice, consistent with previous data14. This increase in macrophage infiltration was significantly reduced by SB, based on F$\frac{4}{80}$ immunohistochemistry results (Fig. 3a). RT-PCR of kidney cortex samples showed that the expression of macrophage markers such as F$\frac{4}{80}$, Iba-1, CD 80, CD163, and CD206 was upregulated in AdCKD mice compared with that in the control mice, and this expression was ameliorated by SB (Fig. 3b–f). F$\frac{4}{80}$ and Iba-1 are pan-macrophage markers, CD 80 is a type 1 (M1) macrophage marker, and CD163 and CD206 are type 2 (M2) macrophage markers. Additionally, the expression of inflammatory M1 macrophage cytokines, including TNF-α, IL-1β, MCP-1, and IL-6, and the M2 cytokine, TGF-β, was upregulated in the kidneys of AdCKD mice compared with that in the control mice. However, the levels of these markers were significantly decreased following SB treatment (Fig. 3g–k). Inducible NOS (iNOS or NOS2) is a hallmark M1 macrophage molecule that produces nitric oxide (NO)22. Expression levels of iNOS protein (Fig. 3l) and tissue concentrations of NO (Fig. 3m) were significantly elevated in AdCKD-induced kidneys compared with those in the control kidneys, although this was significantly reduced following SB treatment. These data indicate that both M1 and M2 macrophage infiltration in the kidneys of AdCKD mice was reduced by SB treatment. Finally, the expression of ICAM-1, which is important for monocyte or macrophage tissue adhesion, increased in the kidneys of AdCKD mice, but was attenuated by SB treatment (Fig. 3n).Figure 3SB attenuated macrophage infiltration and various macrophage markers. ( a) Plot of F$\frac{4}{80}$ immunohistochemistry staining areas of randomly acquired high-power field images. The bar graph in the right panel shows the quantifications of the stained area ($$n = 25$$ each). SB reduced relative expression of pan-macrophage markers (b) F$\frac{4}{80}$ and (c) Iba-1, M1 macrophage markers (d) CD80, (g) TNF-α, (h) IL-1β, (i) MCP-1, and (j) IL-6, M2 macrophage markers (e) CD163, (f) CD206, and (k) TGF-β, and contact molecule (n) ICAM-1 in the kidney ($$n = 3$$–10). HPRT was used as an internal control. Western blots and nitric oxide assays, respectively, showed that the abundance of the M1 macrophage marker (l) iNOS and (m) nitric oxide level was decreased following SB treatment ($$n = 6$$–10). * $p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. Iba-1, ionized calcium binding adaptor molecule 1; TNF-α, tumor necrosis factor α; IL-1β, interleukin-1 beta; MCP-1, monocyte chemoattractant protein-1; iNOS, inducible nitric oxide synthase; ICAM-1, intercellular adhesion molecule-1; HPRT, hypoxanthine guanine phosphoribosyl transferase; SB, sodium benzoate.
## SB reduces MAPK and NF-κB p65 signals in the kidneys of AdCKD mice
To investigate the involvement of the signal transduction pathway in the kidneys of SB-treated AdCKD mice, we examined the activation of the MAP kinase (MAPK), PI3 kinase/Akt, and NF-κB p65 pathways, since these pathways are closely associated with proinflammatory cytokine signal transduction23–25 (Fig. 4a). The expression of phospho-JNK, phospho-ERK, phospho-Akt, and phospho-NF-κB p65 was upregulated in the kidneys of AdCKD mice, whereas phospho-p38 levels remained constant. The upregulated expression of phospho-JNK, phospho-ERK, and phospho-NF-κB p65 was significantly attenuated following SB treatment, whereas that of phospho-Akt was unaltered (Fig. 4b–f). These changes were consistent with the reduced expression of inflammatory molecules such as TNF-α, IL-1β, and MCP-1, since these molecules have downstream mediators such as MAPK and NF-κB signals25,26.Figure 4SB reduces the phosphorylation of JNK, ERK, and NF-κB p65 in the kidneys of AdCKD mice. ( a) Immunoblots of the respective groups are shown. Band intensities of (b) phopho-p38 MAPK (p-p38 MAPK), (c) phospho-JNK (p-JNK), (d) phospho-ERK (p-ERK), (e) phospho-Akt (p-Akt), and (f) phospho-NF-κB p65 (p-NF-κB p65) were quantified and adjusted to the intensities of the internal control molecule β-actin ($$n = 4$$–10). Mice in the AdCKD + SB group showed decreased levels of p-JNK, p-ERK, and p-NF-κB p65 compared with mice in the AdCKD group. * $p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. SB, sodium benzoate.
## SB suppresses expression of LPS or TNF-α-induced inflammatory genes in human THP-1 monocytes
To investigate the direct effects of SB on the expression of inflammatory genes, we stimulated the human monocytic cell line, THP-1, with LPS (Fig. 5a) or TNF-α (Fig. 5e), because their respective receptors, TLR47 and TNFR19, have been implicated in the pathogenesis of adenine-induced nephropathy. LPS increased the mRNA levels of inflammatory genes such as TNF-α and IL-1β, as well as that of chemokine MCP-1, with the levels being significantly decreased by SB (Fig. 5b–d). TNF-α induced the expression of TNF-α, IL-1β, and MCP-1 significantly, but this was significantly inhibited by SB (Fig. 5f–h). These data suggest that SB inhibits the expression of inflammatory genes in monocytic human THP-1 cells. Figure 5SB reduced the expression of inflammatory genes in stimulated THP-1 cells. ( a) Protocol for LPS stimulation in THP-1 cells with or without SB preincubation. RT-PCR analysis of (b) TNF-α, (c) IL-1β, and (d) MCP-1. mRNA levels were adjusted to that of GAPDH ($$n = 5$$–10 each). ( e) Protocol for TNF-α stimulation in THP-1 cells with or without SB preincubation. RT-PCR analysis of (f) TNF-α, (g) IL-1β, and (h) MCP-1. mRNA levels were adjusted to that of GAPDH ($$n = 6$$ each). ( i) Immunoblotting (left panel) and quantification (right panel) of phosphorylated signal transduction molecules ($$n = 5$$–6). Band intensities were quantified using ImageJ software and adjusted to the intensity of the internal control molecule, β-actin. LPS (lane 2) significantly increased phospho-JNK and phospho-NF-κB p65 compared with the control (lane 1). SB did not decrease the phosphorylation of either molecule (lane 4). For comparison, the MEK inhibitor U0126 significantly decreased phospho-ERK levels (lane 5), while the JNK inhibitor SP600125 significantly decreased phospho-JNK (lane 6) and phospho-Akt (lane 6) levels. * $p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. LPS, lipopolysaccharide; SB, sodium benzoate.
To closely examine the SB inhibitory mechanism, THP-1 cells were stimulated with LPS for 15 min and the phosphorylation levels of MAPK, Akt, and NF-κB p65 were measured (Fig. 5i). Phospho-ERK and phospho-Akt were abundantly expressed in baseline THP-1 cells, consistent with previous immunoblot experiments27. LPS upregulated phospho-JNK and phospho-NF-κB p65 levels compared with those in the control (lane 1 vs. 2). SB treatment did not decrease phosphorylation (lane 4), whereas the MEK inhibitor U0126 (lane 5) and JNK inhibitor SP600125 (lane 6) significantly inhibited phospho-ERK and phospho-JNK expression, respectively. These results are discrepant to those obtained in AdCKD mice experiments where both MAPK and NF-κB p65 signals are inhibited (Fig. 3). Since MAPK and NF-κB p65 signals are not only required for proinflammatory gene transcription but also downstream signaling elicited by these genes25,26, SB may inhibit primarily the expressions of proinflammatory cytokine expression in AdCKD mice thereafter block downstream signaling of these cytokines.
## SB reduces THP-1 cell motility
Since the ability of monocytes to mobilize and traffic themselves is one of the central functions driving inflammatory diseases28, we confirmed the inhibitory effect of SB on THP-1 cell motility using Transwell assays. When the cells were incubated with SB, the increase in THP-1 cell migration induced by FBS was significantly attenuated in a dose-dependent manner (Fig. 6a,b), confirming the inhibitory effect of SB on cell motility. These results were consistent with in vivo data that showed that the expression of MCP-1 and ICAM-1, which are crucial adhesion molecules in monocytes29,30, was downregulated by SB treatment of AdCKD mice (Fig. 3n).Figure 6Transwell assay shows that SB reduces THP-1 migration. ( a) The total number of cells in each well in the lower chamber of the 5 μm-pore transwell were counted and plotted after 24 h of incubation ($$n = 6$$ each). Cell counts per well were significantly lower in the FBS + SB group compared with the FBS group. ( b) SB dose dependency was confirmed by applying different concentrations of SB to the upper chamber of the transwell ($$n = 4$$–6). Higher SB concentrations significantly reduced the number of migrating cells. * $p \leq 0.05$; ***$p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. FBS, fetal bovine serum; SB, sodium benzoate.
## SB reduces NF-κB RelB level of protein, mRNA, and binding to motif DNA sequence
To investigate the underlying mechanism for SB reducing the inflammatory gene expression, we focused on NF-κB RelB, as an upstream molecule associated with inflammatory gene regulation31,32. Protein level of NF-κB RelB was increased significantly in AdCKD kidneys compared to controls, which was significantly decreased in AdCKD + SB compared to AdCKD (Fig. 7a). Protein level of NF-κB RelB showed similar results in LPS stimulated THP-1 cells (Fig. 7b). mRNA levels of NF-κB RelB also showed similar results both in vivo and in vitro (Fig. 7c–e). We also confirmed that the mRNA level of NF-κB RelB was decreased in THP-1 cells treated with TNF-α + SB compared to TNF-α alone (Fig. 7e). The binding capacity of NF-κB RelB to its motif DNA sequence was increased in AdCKD kidneys compared to controls whereas it was significantly decreased in AdCKD + SB (Fig. 7f). The binding also was decreased in LPS stimulated THP-1 cells (Fig. 7g). We also confirmed that SB inhibition of motif DNA sequence binding was specific to NF-κB RelB because NF-κB p65, which is also an upstream inflammatory regulatory molecule, was unaltered by SB treatment in LPS stimulated THP-1 cells (Fig. 7h). Figure 7d and g presents that the inhibiting effect of NF-κB RelB by SB showed dose dependency. Figure 7SB reduces NF-κB RelB levels of protein, mRNA, and binding to motif DNA sequence. ( a) Band intensities of NF-κB RelB relative to beta-actin were plotted. In AdCKD kidney, NF-κB RelB protein expression was increased compared to control kidney, which was significantly decreased in AdCKD + SB kidney. ( b) NF-κB RelB protein expression was increased in LPS stimulated THP-1 cells compared to control cells, which was significantly decreased in SB treated LPS stimulated THP-1 cells. ( c) mRNA levels of NF-κB RelB in AdCKD kidney were increased compared to control, which was significantly decreased in AdCKD + SB. ( d) mRNA levels of NF-κB RelB in LPS stimulated THP-1 cells were increased compared to control, which was significantly decreased in SB-treated LPS stimulated cells in a dose dependent manner. ( e) mRNA levels of NF-κB RelB in TNF-α stimulated THP-1 cells were increased compared to control, which was significantly decreased in SB-treated TNF-α stimulated cells. ( f) NF-κB RelB binding to motif DNA sequence was increased in AdCKD kidney compared to control, which was significantly decreased in AdCKD + SB kidney. ( g) NF-κB RelB binding to motif DNA sequence was increased in LPS stimulated THP-1 cells compared to control, which was significantly decreased in SB-treated LPS stimulated cells in a dose dependent manner. ( h) NF-κB p65 binding to motif DNA sequence was increased in LPS stimulated THP-1 cells compared to control, which was similar levels compared to SB-treated LPS stimulated cells. ** $p \leq 0.01$; ***$p \leq 0.001$; ns, not significant. Each bar represents mean ± SEM. LPS, lipopolysaccharide; TNF-α, tumor necrosis factor α; SB, sodium benzoate.
## Discussion
We have presented data showing that AdCKD can be improved by orally administering SB to mice, independent of their DAO enzymatic activity. TNF-α (Fig. 3g), IL-1β (Fig. 3h), and MCP-1 (Fig. 3i) expression, and MAPK (Fig. 4c,d) and NF-κB p65 (Fig. 4f) phosphorylation in the kidneys of SB-treated AdCKD mice were significantly downregulated. As a result, macrophage infiltration and kidney fibrosis were attenuated (Fig. 3a), resulting in improved kidney function (Fig. 1c,d). Studies have shown that when the expression of various inflammatory and chemotactic genes is induced and inflammation is amplified once the input signal crosses a given threshold33,34, forming a positive feedback loop. In the kidneys of AdCKD mice, the expression of these inflammatory genes influenced the infiltrated monocytic cells to develop into M1 macrophages, which released cytokines that inhibit the proliferation of surrounding cells and damaged contiguous tissue35. M2 macrophages released cytokines that promote tissue repair in response to M1 macrophages35. These cytokines include TGF-β, a critical regulator of kidney fibrosis36. Our study showed that both the M1 macrophage marker CD80 (Fig. 3d) and the M2 macrophage markers CD163 and CD206 (Fig. 3e and f, respectively) were upregulated in AdCKD mice, suggesting that both the inflammatory and repair processes caused by these macrophages are activated. TNF-α signaling directly contributes to the development of inflammation in the kidney of AdCKD by increasing the crystal deposition area7, whereas IL-1β signaling promotes kidney fibrosis37, both of which were significantly suppressed by SB. Additionally, MCP-1, which upregulates TNF-α, IL-1β, and TGF-β38, was suppressed by SB. MCP-1 is associated with inflammation and CKD progression in various human and experimental kidney diseases39,40. The suppression of these pro-inflammatory molecules by SB contributed to the blockade of the positive feedback loop of inflammation in AdCKD mice, preventing the progression of kidney injury and renal failure (Fig. 8).Figure 8Graphical abstract showing the beneficial effects of SB in the kidneys of AdCKD mice. Left panel: In AdCKD mice, 2,8-DHA crystal deposits accumulated in renal tubules leads to monocyte/macrophage infiltration into the kidney tissue following upregulated expression of inflammatory genes such as TNF-α, IL-1β, and MCP-1, as well as TNFR1 and TLR4. In this process, NF-κB RelB is also increased. The upregulated inflammatory genes positively regulate additional crystal deposition7, monocyte/macrophage infiltration, and the expression of inflammatory genes. This positive feedback loop resulted in renal failure. Right Panel: SB treatment reduces the expression of NF-κB RelB and inflammatory genes, thereby down regulating monocyte/macrophage infiltration and the expression of the inflammatory genes, improving renal function. AdCKD, adenine induced chronic kidney disease; SB, sodium benzoate; CKD, chronic kidney disease; TNFR1, tumor necrosis factor receptor 1; TNF-α, tumor necrosis factor α; IL-1β, Interleukin 1 beta; MCP-1, Monocyte chemoattractant protein-1; TLR4, Toll-like receptor 4.
SB reduced the expression of inflammatory molecules such as iNOS and IL-1β, and increased the expression of contact molecules such as ICAM-1 and E-selectin in a mouse model of experimental allergic encephalomyelitis17. SB also reduced TNF-α levels following LPS stimulation in mouse BV-2 microglial cells18. In this study, we determined that the inhibitory effect of SB on AdCKD was due to the suppression of activated monocytic cells (Fig. 5). The expression of inflammatory genes, including TNF-α, was upregulated by LPS or TNF-α stimulation in THP-1 cells, which was significantly downregulated by SB treatment (Figure 5b–d, f–h). TNF-α was inhibited under every experimental condition, thus leading to significantly improved AdCKD in SB treated mice compared to AdCKD mice because TNF-α plays a central role in adenine-induced nephropathy pathophysiology as reported previously7.
A mechanism for SB inhibition of inflammation has been proposed. Based on previous reports18,41, benzoate is converted into benzoyl-CoA, which inhibits the mevalonate pathway, thereby inhibiting phenyl pyrophosphate production, resulting in inhibition of the Ras-ERK/MAPK cascade and downstream NF-κB-targeted gene transcription. However, these studies focused on the NF-κB reporter or iNOS expression and did not measure the phosphorylation of ERK or NF-κB. Our in vivo analysis showed that phosphorylation of ERK, NF-κB p65, and JNK was attenuated in the kidneys of SB-treated AdCKD mice (Fig. 4). In contrast, in vitro analysis of THP-1 cells showed that SB treatment did not inhibit either MAPK or NF-κB p65 phosphorylation (Fig. 5i). Generally, inflammatory molecules such as TNF-α, IL-1β, MCP-1, and TLR4 have downstream mediators such as MAPK and NF-κB signal25,26, and activation of MAPK and NF-κB signals are, in turn, necessary to increase the mRNA transcription of inflammatory molecules such as TNF-α, IL-1β, and MCP-125,26. Taken together, the inhibition of ERK, JNK, and NF-κB p65 phosphorylation in SB-treated AdCKD mice can be interpreted as a secondary inhibitory effect of SB on the mRNA expression of TNF-α, IL-1β, and MCP-1 rather than direct effect. Finally, we have successfully demonstrated that SB inhibits NF-κB RelB, which is an upstream regulatory molecule for broad inflammatory molecules including TNF-α, IL-1β, and MCP-131,32. Taken all together, SB inhibition of proinflammatory cytokines is due to inhibited NF-κB RelB expression, ultimately leading to alleviated kidney injury in AdCKD mice model.
In addition to its anti-inflammatory action, SB inhibited the motility of THP-1 monocytic cells (Fig. 6). Cell migration and trafficking are multistep processes that involve local recruitment of inflammatory cells28. Recruited monocytes can participate in the initial inflammatory response by releasing TNF-α and IL-1β, as well as pattern recognition receptors such as TLRs42. MCP-1 is a chemoattractant that induces THP-1 cell migration43. Thus, cell motility is significant in initiating inflammation. AdCKD pathophysiology is associated with inflammatory molecules such as TNF-α7 and NF-κB9 and chemotactic molecules such as CCL3 and CCR514. The beneficial effect of SB in AdCKD seems to be associated with a reduction in the chemotactic molecule, MCP-1 (Fig. 3i) and the trafficking molecule, ICAM-1 (Fig. 3n). Additionally, in vitro analysis showed that SB inhibited the transcription of inflammatory genes which was activated upon LPS or TNF-α stimulation (Fig. 5). SB also inhibited THP-1 cell motility (Fig. 6). Since cell motility and inflammation are interrelated28,42, the presented results show that SB inhibited mRNA of several proinflammatory molecules through inhibition of NF-κB RelB, leading to down regulated cell motility.
The potential limitation of this study is that, first, SB could exhibit hematological abnormality including anemia in mice44, therefore SB usage should be handled with caution to avoid overdose. However, the average dosage of SB in our study was 500 mg/kg/day per mouse which human equivalent dose is 40 mg/kg/day, and this is only 16 percent of the admitted human dose of SB in human urea cycle disorder of 250 mg/kg/day45. The side effect may not be significant, although, long-term administration side effects should carefully be determined if applying to diseases including AdCKD and human CKD related to APRT deficiency in the future. Secondly, we have not presented the blood concentration of SB in mice, so this could be a limitation in this study. Besides, as presented in Fig. 1j, we have confirmed the dose dependency of SB on kidney function. This could support that blood concentration of SB is adequately elevated triggering beneficial effects to kidney function in AdCKD. Also, according to Fig. 1j, the minimal effective dose of SB in AdCKD could be considered as 5 mM, which triggered reduction of both plasma urea nitrogen and creatinine. Thirdly, since this study focused on treating AdCKD mouse model, we have not conducted experiments for human participants. Future study should aim to treat human CKD patients. Lastly, we have employed modified experimental protocol for obtaining MCP-1 mRNA level in THP-1 cells stimulated by LPS as shown in the method section, in which we used medium without FBS. Using the FBS containing medium during LPS stimulation made MCP-1 increase less detectable compared to control46, because FBS itself induces MCP-147. This is not the case for TNF-α or IL-1β, because FBS reduces the level of these molecules48,49. The modified protocol successfully increased the level of MCP-1 mRNA by LPS, which was significantly reduced by SB (Fig. 5d).
This study is novel because DAO-deficient mice were used to confirm the therapeutic effect of SB on AdCKD. Although SB is a well-known DAO inhibitor16, previous studies have not examined the relevance of DAO. We have clearly shown that the beneficial effect of SB on AdCKD is independent of DAO. Additionally, we have presented that the underlying mechanism of inhibiting inflammation by SB involves inhibited NF-κB RelB, an upstream inflammation regulatory molecule. SB exerts beneficial effects by reducing inflammation and protecting the kidneys during diseases such as AdCKD and opens future investigations in human CKD related to APRT deficiency.
## Animal experiments
All animal experiment protocols were approved by the Institutional Animal Care and Use Committee of Keio University (Tokyo, Japan) (Approval No. 21011-[0]), and all experiments were performed following relevant guidelines and regulations and in compliance with the ARRIVE guidelines. C57BL/6JJcl-specific pathogen-free (SPF) male mice (CLEA Japan, Tokyo, Japan) were fed a standard CE-2 diet (CLEA Japan) and provided ad libitum access to tap water. Seven-week-old mice were randomly assigned into four groups with or without adenine in chow and with or without sodium benzoate (SB) (20 mM) in drinking water: control group, 2,8-DHA nephropathy group (AdCKD group), SB-treated control group (SB group), and AdCKD with SB treatment group (AdCKD + SB group) (Fig. 1a). Each group consisted of ten mice each. Five mice were kept in each cage. Mice cages were made of transparent plastic supplemented with chip bedding and the cages were randomly placed in a rack to minimize potential confounders such as location of a cage. Mice weighted between 18 and 24 g were used. Mice in the control group were fed a normal diet whereas those in the AdCKD group were given CE-2 supplemented with $0.2\%$ adenine (Wako, Japan) for six or eight weeks. SB was dissolved in distilled water to give final concentrations of 0.1 mM, 1 mM, 5 mM, and 20 mM. These were used for dose-dependent analysis, and each group consisted of four to five mice each. Survival analysis was performed using ten mice from each group who were monitored over a 336-day period. Mice were monitored if they were alive every day and the date of the demise of a mouse was recorded as the endpoint of the analysis. Demise of a mouse was defined as confirmation of postmortem rigidity or whole-body tremor which was considered a humane endpoint. Finally, the mice were weighed, euthanized using intraperitoneal injection of 0.3 mg/kg medetomidine, 4.0 mg/kg midazolam, and 5.0 mg/kg butorphanol, and all efforts were made to minimize animal suffering. Blood samples were collected from the inferior vena cava and sacrificed. Kidney tissues were extracted and snap frozen in liquid nitrogen and stored at − 80 °C for further use. Serum creatinine and urea nitrogen (UN), urine creatinine, total protein, and N-acetyl-β-D-glucosaminidase (NAG) were measured as described previously50. Kidneys which developed hydronephrosis were not included in the study. Blinding was not conducted in the experiments. No adverse events were observed.
## DAO mouse line carrying a G181R mutation
A ddY/DAO − mouse line lacking DAO activity due to a G181R point-mutation in the DAO gene (DAO − / −) was backcrossed with C57BL/6JJcl mice 15 times, as described previously21. The female mice were crossed with C57BL/6JJcl male mice to obtain a heterozygous DAO + / − mouse strain. The G181R mutation was genotyped in littermates of DAO + / − crosses. The littermates were then divided into wild-type (DAO + / +), heterozygous DAO G181R (DAO + / −), or homozygous DAO G181R (DAO-/-) groups ($$n = 6$$–17 per group). The mice were fed with CE-2 chow containing $0.2\%$ adenine to induce AdCKD. DAO − / − mice were divided into three groups ($$n = 4$$–11 per group): control, $0.2\%$ adenine-induced CKD (AdCKD group), and $0.2\%$ adenine-induced CKD treated with 20 mM sodium benzoate (AdCKD + SB group). Blood was collected similarly. For genotyping, the tails of the mice were cut and vortexed in 50 mM NaOH (180 μL) and then incubated at 95 °C for 10 min. The reaction was stopped by adding 1 M Tris–HCl (pH 8.0, 20 μL). Two microliters of the sample lysate were used as a PCR template, which was performed using Quick Taq® HS DyeMix DTM-101 (Toyobo, Japan). Nested PCR was performed using the following primers: 1st PCR forward 5ʹ-GAAGAGGGAGAGGAGGAGAAT-3ʹ and reverse 5ʹ-TTTGGTTAAGATGGTGATGTG-3ʹ; 2nd PCR forward 5ʹ-GGGAGAGGG CACAGCACAGTC-3ʹ, reverse 5ʹ-ACACCAGGGCAGGAGTAGGC-3ʹ. PCR products were electrophoresed on a $3\%$ agarose gel containing $0.1\%$ ethylene bromide. Bands were detected under UV light. The product of the 1st PCR was diluted 200 times and used as a template in the 2nd PCR.
## Histological staining and assessment
The kidney tissues were fixed in $10\%$ formalin neutral buffer solution (Fujifilm, Japan) and embedded in paraffin. 4-μm slices were stained with masson trichrome (MT) according to a standard protocol. For F$\frac{4}{80}$ immunohistochemistry (IHC) staining, 4-μm slices of the paraffin-embedded tissues were processed as follows: deparaffinized slices were treated with proteinase K and $3\%$ hydroxy peroxide for antigen retrieval. The slices were incubated in anti-mouse F$\frac{4}{80}$ rat monoclonal antibody (Bio-Rad, 1:200) at 20 to 25 °C for 50 min, washed in PBS, enhanced by histofine simple stain max-po (Nichirei Bioscience, Japan), and stained with 3,3'-diaminobenzidine and then with Mayer's hematoxylin nucleus stain. Digital images were obtained using a NanoZoomer-XR C12000 virtual slide scanner (Hamamatsu Photonics K.K., Japan). Ten high power field pictures of every MT slide were captured for the assessment. Each picture was given an interstitial fibrosis and tubular atrophy (IFTA) score based on a previously described criteria51. Briefly, IFTA scores were assigned as follows: 0 for images with < $25\%$ renal tissue affected, 1 for images with 25–$50\%$ of the renal tissue affected, 2 for images with 50–$75\%$ of the renal tissue affected, and 3 for images with $75\%$ or more of the renal tissue affected. Five images of each stained kidney were captured for IHC analysis. Color deconvolution was applied to each image using ImageJ software (National Institutes of Health, USA) using a color deconvolution plug, after the 3,3″-diaminobenzidine tetrahydrochloride area was calculated. Antibody details are provided in Table S2.
## Cell culturing and stimulation
Cells from the human monocytic leukemia cell line THP-1 were grown in RPMI-1640 medium (ATCC, USA) supplemented with $10\%$ fetal bovine serum (FBS) at 37 °C in the presence of $5\%$ CO2. THP-1 cells (1 × 106 cells/mL) were stimulated with 1 µg/mL LPS or 50 ng/mL TNF-α for 2 h. For Fig. 5d, THP-1 cells were incubated in medium without FBS for 10 h and stimulated with 1 µg/mL LPS for 6 h. SB was pre-treated for 1 h before cell stimulation. RNA was subsequently extracted from the cells. For immunoblotting, THP-1 cells were stimulated with 1 µg/mL LPS for 15 min. For Fig. 7, THP-1 cells were stimulated with 1 µg/mL LPS for 2 h.
## THP-1 cell migration assay
Migration assays were performed in a 5-µm diameter Transwell (Corning, Corning, NY, USA). THP-1 monocytes (2 × 105 cells) were suspended in serum-free RPMI-1640 medium (200 µL) in the upper chamber before and after preincubating with 10 mM SB for 1 h. RPMI-1640 medium (600 µL) with or without $10\%$ FBS was added to the lower transwell chamber. The medium in the lower chamber was collected and centrifuged (4 °C, 5000 g, 5 min) after 24 h. The supernatant was removed, and $0.4\%$ trypan blue stain (Gibco, USA) used to stain the cells. The total number of cells was counted using a hemocytometer and the total cell count per well was plotted.
## RNA extraction and real-time polymerase chain reaction (RT-PCR)
RNA extraction and quantitative real-time PCR were performed as follows: briefly, RNA was purified from mouse kidney homogenates using the RNeasy Mini Kit (Qiagen, Germantown, MD, USA) and from cell lysates using TRIZOL (Invitrogen, Waltham, MA, USA). mRNA concentration was measured using a Nanodrop One C (Thermo Fisher Scientific, Waltham, MA, USA). RNA (500 μg) was reverse transcribed into cDNA using PrimeScript RT Master Mix (Takara, Japan). PCR was performed on a StepOnePlus Real-Time PCR system (Applied Biosystems, Waltham, MA, USA) using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific). mRNA levels were normalized to those of the housekeeping gene HPRT (mouse kidney assays) or GAPDH (in vitro analysis), and the levels relative to those in the control group were plotted. Expression levels were calculated using the ΔΔCT method. The sequences of the primers used are listed in Table S1.
## Immunoblotting
Immunoblotting was performed as follows: briefly, kidney lysate was obtained by homogenizing kidney cortex tissue in modified RIPA buffer (50 mM Tris–HCl [pH 8.0], 150 mM NaCl, $1\%$ NP-40, 5 mM EDTA, 5 mM MgCl2) supplemented with protease inhibitor cocktail (Roche, Basel, Switzerland), 20 mM NaF, 0.5 mM PMSF, 10 mM nicotinamide, and 330 nM Trichostatin A. Homogenates were centrifuged at 4 °C and 12,000 g for 15 min. Protein concentration in the supernatant was measured using the Bradford method. Sample supernatants were incubated at 95 °C for 5 min with Laemmli sample buffer (Bio-Rad, Hercules, CA, USA) mixed with 2-mercaptoethanol to denature proteins. Protein extracts were separated via electrophoresis on 8–$15\%$ sodium dodecyl sulfate–polyacrylamide gels. Proteins were transferred from the gels to PVDF membranes (Bio-Rad) using a transblot turbo system (Bio-Rad). The membranes were blocked with $5\%$ skim milk or EzBlock Chemi (Atto, Japan) for 1 h at 20–25 °C and then incubated overnight at 4 °C with primary antibodies diluted (1:1000) with either $5\%$ skim milk or Can Get Signal Immunoreaction Enhancer Solution (Toyobo, Japan). The membranes were thereafter washed three times in TBST ($0.1\%$ tween-20 in TBS), incubated for 1 h at room temperature with diluted (1:5000) secondary antibody, and then washed three times with TBST. Chemiluminescence was detected using ECL prime (Cytiva, Malborough, MA, USA) and imaged using the LAS-4000 mini (GE healthcare, Chicago, IL, USA). Band intensities were quantified using ImageJ software and normalized to those of internal proteins. Antibody specifications are listed in Table S2.
## Immunofluorescent of kidney tissue, nitrite production, DAO activity assay, and assay of NF-κB binding to motif DNA sequence
These protocols are described in the Supplementary Methods.
## Statistical analysis
Data are expressed as mean ± SEM and analyzed by the student’s t test for two group comparison and by the Tukey–Kramer test for multiple group comparisons. Survival was analyzed using log-rank test. The sample size was determined empirically and confirmed by samplesize calculator (http://powerandsamplesize.com/) using power of 0.8, α of 0.05, standard deviation of 0.1 to 1. Statistical significance was set at $p \leq 0.05.$ All analyses were performed, and graphs were generated using either EZR version 1.35 (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) or GraphPad Prism, version 9 (https://www.graphpad.com/).
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30056-6.
## References
1. Mills KT. **A systematic analysis of worldwide population-based data on the global burden of chronic kidney disease in 2010**. *Kidney Int.* (2015) **88** 950-957. DOI: 10.1038/ki.2015.230
2. Ortiz A. **Epidemiology, contributors to, and clinical trials of mortality risk in chronic kidney failure**. *Lancet (London, England)* (2014) **383** 1831-1843. DOI: 10.1016/s0140-6736(14)60384-6
3. Foley RN, Parfrey PS, Sarnak MJ. **Clinical epidemiology of cardiovascular disease in chronic renal disease**. *Am. J. Kidney Dis.* (1998) **32** S112-119. DOI: 10.1053/ajkd.1998.v32.pm9820470
4. Wong G. **Association of CKD and cancer risk in older people**. *J. Am. Soc. Nephrol.* (2009) **20** 1341-1350. DOI: 10.1681/asn.2008090998
5. Mulay SR, Anders HJ. **Crystal nephropathies: Mechanisms of crystal-induced kidney injury**. *Nat. Rev. Nephrol.* (2017) **13** 226-240. DOI: 10.1038/nrneph.2017.10
6. Bollée G. **Phenotype and genotype characterization of adenine phosphoribosyltransferase deficiency**. *J. Am. Soc. Nephrol.* (2010) **21** 679-688. DOI: 10.1681/asn.2009080808
7. Klinkhammer BM. **Cellular and molecular mechanisms of kidney injury in 2,8-dihydroxyadenine nephropathy**. *J. Am. Soc. Nephrol.* (2020) **31** 799-816. DOI: 10.1681/asn.2019080827
8. Diwan V, Brown L, Gobe GC. **Adenine-induced chronic kidney disease in rats**. *Nephrology (Carlton, Vic.)* (2018) **23** 5-11. DOI: 10.1111/nep.13180
9. Okabe C. **NF-κB activation mediates crystal translocation and interstitial inflammation in adenine overload nephropathy**. *Am. J. Physiol. Renal Physiol.* (2013) **305** F155-163. DOI: 10.1152/ajprenal.00491.2012
10. Chen Z. **Ozone therapy ameliorates tubulointerstitial inflammation by regulating TLR4 in adenine-induced CKD rats**. *Ren. Fail.* (2016) **38** 822-830. DOI: 10.3109/0886022x.2016.1143757
11. Thibodeau JF. **PBI-4050 via GPR40 activation improves adenine-induced kidney injury in mice**. *Clin. Sci. (London, England 1979)* (2019) **133** 1587-1602. DOI: 10.1042/cs20190479
12. Verma MK. **Activation of GPR40 attenuates chronic inflammation induced impact on pancreatic β-cells health and function**. *BMC Cell Biol.* (2014) **15** 24. DOI: 10.1186/1471-2121-15-24
13. Schmitz ML, Shaban MS, Albert BV, Gökçen A, Kracht M. **The crosstalk of endoplasmic reticulum (ER) stress pathways with NF-κB: Complex mechanisms relevant for cancer, inflammation and infection**. *Biomedicines* (2018). DOI: 10.3390/biomedicines6020058
14. Correa-Costa M. **Macrophage trafficking as key mediator of adenine-induced kidney injury**. *Med. Inflamm.* (2014) **2014** 291024. DOI: 10.1155/2014/291024
15. Nair B. **Final report on the safety assessment of Benzyl Alcohol, Benzoic Acid, and Sodium Benzoate**. *Int. J. Toxicol.* (2001) **20** 23-50. DOI: 10.1080/10915810152630729
16. Sacchi S, Cappelletti P, Murtas G. **Biochemical properties of human D-amino acid oxidase variants and their potential significance in pathologies**. *Front. Mol. Biosci.* (2018) **5** 55. DOI: 10.3389/fmolb.2018.00055
17. Brahmachari S, Pahan K. **Sodium benzoate, a food additive and a metabolite of cinnamon, modifies T cells at multiple steps and inhibits adoptive transfer of experimental allergic encephalomyelitis**. *J. Immunol. (Baltimore, Md: 1950)* (2007) **179** 275-283. DOI: 10.4049/jimmunol.179.1.275
18. Brahmachari S, Jana A, Pahan K. **Sodium benzoate, a metabolite of cinnamon and a food additive, reduces microglial and astroglial inflammatory responses**. *J. Immunol.* (2009) **183** 5917-5927. DOI: 10.4049/jimmunol.0803336
19. Jia T. **A novel model of adenine-induced tubulointerstitial nephropathy in mice**. *BMC Nephrol.* (2013) **14** 116. DOI: 10.1186/1471-2369-14-116
20. Gurley SB. **Influence of genetic background on albuminuria and kidney injury in Ins2(+/C96Y) (Akita) mice**. *Am. J. Physiol. Renal Physiol.* (2010) **298** F788-795. DOI: 10.1152/ajprenal.90515.2008
21. Sasabe J. **D-amino acid oxidase controls motoneuron degeneration through D-serine**. *Proc. Natl. Acad. Sci. U.S.A.* (2012) **109** 627-632. DOI: 10.1073/pnas.1114639109
22. Xue Q, Yan Y, Zhang R, Xiong H. **Regulation of iNOS on immune cells and its role in diseases**. *Int. J. Mol. Sci.* (2018). DOI: 10.3390/ijms19123805
23. Li MD, Yang X. **A retrospective on nuclear receptor regulation of inflammation: Lessons from GR and PPARs**. *PPAR Res.* (2011) **2011** 742785. DOI: 10.1155/2011/742785
24. Laban H. **VASP regulates leukocyte infiltration, polarization, and vascular repair after ischemia**. *J. Cell Biol.* (2018) **217** 1503-1519. DOI: 10.1083/jcb.201702048
25. Beck IM. **Crosstalk in inflammation: The interplay of glucocorticoid receptor-based mechanisms and kinases and phosphatases**. *Endocr. Rev.* (2009) **30** 830-882. DOI: 10.1210/er.2009-0013
26. Garcia-Bonilla L, Benakis C, Moore J, Iadecola C, Anrather J. **Immune mechanisms in cerebral ischemic tolerance**. *Front. Neurosci.* (2014) **8** 44. DOI: 10.3389/fnins.2014.00044
27. Wang Y. **Cordycepin induces apoptosis of human acute monocytic leukemia cells via downregulation of the ERK/Akt signaling pathway**. *Exp. Ther. Med.* (2017) **14** 3067-3073. DOI: 10.3892/etm.2017.4855
28. Shi C, Pamer EG. **Monocyte recruitment during infection and inflammation**. *Nat. Rev. Immunol.* (2011) **11** 762-774. DOI: 10.1038/nri3070
29. Edwards EE, Thomas SN. **P-Selectin and ICAM-1 synergy in mediating THP-1 monocyte adhesion in hemodynamic flow is length dependent**. *Integr. Biol.* (2017) **9** 313-327. DOI: 10.1039/c7ib00020k
30. Das P. **A PKCβ-LYN-PYK2 signaling axis is critical for MCP-1-dependent migration and adhesion of monocytes**. *J. Immunol.* (2021) **206** 181-192. DOI: 10.4049/jimmunol.1900706
31. Khoyratty TE. **Distinct transcription factor networks control neutrophil-driven inflammation**. *Nat. Immunol.* (2021) **22** 1093-1106. DOI: 10.1038/s41590-021-00968-4
32. Hsu KH. **Upregulation of RelB in the miR-122 knockout mice contributes to increased levels of proinflammatory chemokines/cytokines in the liver and macrophages**. *Immunol. Lett.* (2020) **226** 22-30. DOI: 10.1016/j.imlet.2020.06.015
33. Dorrington MG, Fraser IDC. **NF-κB signaling in macrophages: Dynamics, crosstalk, and signal integration**. *Front. Immunol.* (2019) **10** 705. DOI: 10.3389/fimmu.2019.00705
34. Anders HJ. **Of inflammasomes and alarmins: IL-1β and IL-1α in kidney disease**. *J. Am. Soc. Nephrol.* (2016) **27** 2564-2575. DOI: 10.1681/asn.2016020177
35. Wang N, Liang H, Zen K. **Molecular mechanisms that influence the macrophage m1–m2 polarization balance**. *Front. Immunol.* (2014) **5** 614. DOI: 10.3389/fimmu.2014.00614
36. Meng XM, Nikolic-Paterson DJ, Lan HY. **TGF-β: The master regulator of fibrosis**. *Nat. Rev. Nephrol.* (2016) **12** 325-338. DOI: 10.1038/nrneph.2016.48
37. Lemos DR. **Interleukin-1β activates a MYC-dependent metabolic switch in kidney stromal cells necessary for progressive tubulointerstitial fibrosis**. *J. Am. Soc. Nephrol.* (2018) **29** 1690-1705. DOI: 10.1681/asn.2017121283
38. Xu L, Sharkey D, Cantley LG. **Tubular GM-CSF promotes late MCP-1/CCR2-mediated fibrosis and inflammation after ischemia/reperfusion injury**. *J. Am. Soc. Nephrol.* (2019) **30** 1825-1840. DOI: 10.1681/asn.2019010068
39. Tam FWK, Ong ACM. **Renal monocyte chemoattractant protein-1: An emerging universal biomarker and therapeutic target for kidney diseases?**. *Nephrol. Dial. Transpl.* (2020) **35** 198-203. DOI: 10.1093/ndt/gfz082
40. Zheng Z. **Hippo-YAP/MCP-1 mediated tubular maladaptive repair promote inflammation in renal failed recovery after ischemic AKI**. *Cell Death Dis.* (2021) **12** 754. DOI: 10.1038/s41419-021-04041-8
41. Pahan K. **Immunomodulation of experimental allergic encephalomyelitis by cinnamon metabolite sodium benzoate**. *Immunopharmacol. Immunotoxicol.* (2011) **33** 586-593. DOI: 10.3109/08923973.2011.561861
42. Auffray C. **Monitoring of blood vessels and tissues by a population of monocytes with patrolling behavior**. *Science* (2007) **317** 666-670. DOI: 10.1126/science.1142883
43. Zhou J, Bai W, Liu Q, Cui J, Zhang W. **Intermittent hypoxia enhances THP-1 monocyte adhesion and chemotaxis and promotes M1 macrophage polarization via RAGE**. *Biomed. Res. Int.* (2018) **2018** 1650456. DOI: 10.1155/2018/1650456
44. Olofinnade AT, Onaolapo AY, Onaolapo OJ, Olowe OA. **The potential toxicity of food-added sodium benzoate in mice is concentration-dependent**. *Toxicol. Res.* (2021) **10** 561-569. DOI: 10.1093/toxres/tfab024
45. Raina R. **Consensus guidelines for management of hyperammonaemia in paediatric patients receiving continuous kidney replacement therapy**. *Nat. Rev. Nephrol.* (2020) **16** 471-482. DOI: 10.1038/s41581-020-0267-8
46. Lim SG, Kim JK, Suk K, Lee WH. **Crosstalk between signals initiated from TLR4 and cell surface BAFF results in synergistic induction of proinflammatory mediators in THP-1 cells**. *Sci. Rep.* (2017) **7** 45826. DOI: 10.1038/srep45826
47. Li S. **MCP-1 overexpressed in tuberous sclerosis lesions acts as a paracrine factor for tumor development**. *J. Exp. Med.* (2005) **202** 617-624. DOI: 10.1084/jem.20042469
48. Bas S, James RW, Gabay C. **Serum lipoproteins attenuate macrophage activation and Toll-Like Receptor stimulation by bacterial lipoproteins**. *BMC Immunol.* (2010) **11** 46. DOI: 10.1186/1471-2172-11-46
49. McKenzie RC, Harley CB, Matic S, Sauder DN. **Fetal bovine serum contains an inhibitor of interleukin-1**. *J. Immunol. Methods* (1990) **133** 99-105. DOI: 10.1016/0022-1759(90)90323-n
50. Minakuchi H. **The role of adipose tissue asymmetric dimethylarginine/dimethylarginine dimethylaminohydrolase pathway in adipose tissue phenotype and metabolic abnormalities in subtotally nephrectomized rats**. *Nephrol. Dial. Transpl.* (2016) **31** 413-423. DOI: 10.1093/ndt/gfv367
51. Tervaert TW. **Pathologic classification of diabetic nephropathy**. *J. Am. Soc. Nephrol.* (2010) **21** 556-563. DOI: 10.1681/asn.2010010010
|
---
title: The activation mechanism and antibody binding mode for orphan GPR20
authors:
- Xi Lin
- Shan Jiang
- Yiran Wu
- Xiaohu Wei
- Gye-Won Han
- Lijie Wu
- Junlin Liu
- Bo Chen
- Zhibin Zhang
- Suwen Zhao
- Vadim Cherezov
- Fei Xu
journal: Cell Discovery
year: 2023
pmcid: PMC9971246
doi: 10.1038/s41421-023-00520-8
license: CC BY 4.0
---
# The activation mechanism and antibody binding mode for orphan GPR20
## Abstract
GPR20 is a class-A orphan G protein-coupled receptor (GPCR) and a potential therapeutic target for gastrointestinal stromal tumors (GIST) owing to its differentially high expression. An antibody-drug conjugate (ADC) containing a GPR20-binding antibody (Ab046) was recently developed in clinical trials for GIST treatment. GPR20 constitutively activates Gi proteins in the absence of any known ligand, but it remains obscure how this high basal activity is achieved. Here we report three cryo-EM structures of human GPR20 complexes including Gi-coupled GPR20 in the absence or presence of the Fab fragment of Ab046 and Gi-free GPR20. Remarkably, the structures demonstrate a uniquely folded N-terminal helix capping onto the transmembrane domain and our mutagenesis study suggests a key role of this cap region in stimulating the basal activity of GPR20. We also uncover the molecular interactions between GPR20 and Ab046, which may enable the design of tool antibodies with enhanced affinity or new functionality for GPR20. Furthermore, we report the orthosteric pocket occupied by an unassigned density which might be essential for exploring opportunities for deorphanization.
## Introduction
Orphan G protein-coupled receptors (oGPCRs) are pathologically related to many human diseases such as schizophrenia, type 2 diabetes, hyperactivity, cognitive impairment, brain malformation, Alzheimer’s disease and others1–7. Many oGPCRs including the adhesion family (aGPCRs) are constitutively active receptors, presenting additional technical hurdles for deorphanization8. Since the report of the first ligand-free orphan GPR52 structure by Lin et al.2, several structures of oGPCRs in complex with G proteins in the absence of exogenous ligand stimulation have been determined—such as GPR17-Gi9, GPR88-Gi10, GPR119-Gs11 and aGPCR-G protein complexes12–15. All these oGPCRs were activated by either their own motif or an endogenous lipid from the cell membrane: the receptor’s extracellular loop 2 (ECL2) folds into the orthosteric pocket and functions as a built-in agonist for GPR522 and GPR17;9 the “stalk” region functions as a tethered agonist for aGPCRs;12–15 an unassigned density in the orthosteric pocket may represent a putative endogenous ligand for GPR88;10 a lipid molecule occupies the orthosteric pocket responsible for the high basal activity of GPR11911.
GPR20 is an orphan receptor with its endogenous ligand remaining unknown16. It is expressed in many tissues with notably high expression in the intestine. Previous studies identified GPR20 as a novel non-tyrosine kinase target in gastrointestinal stromal tumors (GIST) given its differentially high expression17,18. To date, the only approved treatments for GIST are tyrosine kinase inhibitors (TKI), but patients ultimately experience disease progression most often due to the development of heterogeneous secondary resistance mutations in tyrosine-protein kinase KIT17. Recently, Daiichi Sankyo developed a DXd-ADC drug (DS-6157a) derived from an anti-GPR20 antibody (Ab046) to inhibit tumor growth in GIST17. However, the drug failed in Phase I clinical trial due to a lack of response. Nevertheless, the potential of developing new antibody or derivatives with higher potency and/or new function on GPR20 would deserve further investigation for the treatment of GIST and other related diseases. A previous study suggested that the intrinsic expression of GPR20 is involved in the regulation of cell proliferation by controlling cellular cAMP levels19. Indeed, GPR20 shows high level of constitutive activity in the absence of ligand, leading to continuous activation of its coupled Gi proteins20. It is unclear how the high basal activity of GPR20 is achieved, and whether a stimulator is required for G-protein coupling and signal transduction.
Here we report the atomic-resolution structures of the ligand-free human orphan GPR20 in the Gi-coupled states in the absence or presence of the Fab fragment of Ab046 (hereafter named as GPR20-Gi and GPR20-Gi-Fab046 respectively) as well as in the Gi-free state (GPR20-Fab046) using single particle cryo-electron microscopy technique (cryo-EM). These structures reveal a uniquely folded N-terminal α-helix (cap) of GPR20 that might be essential in conferring the receptor’s constitutive activity. Our results provide an integrated understanding of the structure and function of GPR20, which paves the way for uncovering the landscape of structural basis for orphan GPCR’s constitutive activity. Structural analysis of the antibody-binding interface as well as the observation of an unassigned density may present opportunity for developing structure-based tool ligands and antibodies for GPR20 deorphanization.
## Structure of the ligand-free GPR20-Gi complex
To test whether GPR20 can signal through Gi and understand its high level of constitutive activity20, we first performed bioluminescence resonance energy transfer (BRET) assays21 to measure G-protein heterotrimer dissociation. The results confirmed the Gi activity by ligand-free GPR20 as shown by the reduced BRET signal compared to the negative controls including the empty vector (mock control) and the adenosine A2A receptor (A2AR, which was known to not couple to Gi22). It was observed that ligand-free GPR20 showed comparable BRET signal to the apelin receptor (APJ, which was known to couple to Gi23) in the presence of an agonist (Fig. 1a). In order to obtain stable GPR20-Gi complex sample amenable for structural investigation by cryo-EM, we tried different strategies (see “Materials and methods” and Supplementary Fig. S1). The final complex was composed of the N-terminal BRIL-fused24 wild-type (WT) GPR20, dominant-negative mutant of Gαi1 (containing three mutations: S47N, G203A and A326S25), WT Gβ1Gγ2, and a single-chain stabilizing antibody fragment scFv1626, which all could be clearly identified by two-dimensional (2D) classification (Supplementary Fig. S1c). Finally, the structure of the ligand-free GPR20-Gi complex was determined at a global resolution of 3.14 Å (Fig. 1b; Supplementary Fig. S1 and Table S1).Fig. 1Cryo-EM structure of the ligand-free GPR20-Gi complex.a Left, schematic diagram of the bioluminescence resonance energy transfer (BRET2) assay to measure Gi heterotrimer dissociation upon activation. Right, normalized BRET values of HEK293T cells transiently co-transfected with the Gi BRET sensor along with either empty vector (mock control), GPR20, the adenosine A2A receptor (A2AR, negative control) or the apelin receptor (APJ, positive control) in the absence and presence of 10 μM agonist cpd644. Data are normalized to the mock control and represented as the mean ± s.e.m. for $$n = 3$$ biologically independent experiments. Significance was determined by two-way analysis of variance (ANOVA) without repeated measures, followed by Dunnett’s post hoc test (***$P \leq 0.001$, *$P \leq 0.05$; n.s., not significant). b Left, cryo-EM density map of the GPR20-GαiGβGγ-scFv16 complex. Right, two orthogonal views of the cartoon representation of the atomic model of the complex. Color coding is annotated for each protein component. c Left, cryo-EM density map of the GPR20-GαiGβGγ-scFv16-Fab046 complex. Right, two orthogonal views of the cartoon representation of the atomic model of the complex. Color coding is annotated for each protein component.
To date, there are no ligands available for GPR20; the only reported tool molecule that binds to GPR20 is an anti-GPR20 antibody (Ab046) derived from an ADC drug17 for potential GIST treatment. To understand how Ab046 binds and whether it affects the structure and function of GPR20, we also solved the structure of Gi-coupled GPR20 complexed with Fab046. This structure was determined to a global resolution of 3.22 Å with the density maps sufficiently clear to place part of Fab046 (VL and VH of Fab046) into the GPR20-Gi-Fab046 complex (Fig. 1c; Supplementary Fig. S2 and Table S1). The overall structures of the GPR20-Gi-Fab046 complex and GPR20-Gi complex are nearly identical with root mean square deviation (RMSD) values of 0.71 Å for the whole GPR20-Gi complex and 0.70 Å for the receptor alone, which is consistent with our BRET results that Ab046 is a non-functional antibody for GPR20 (Supplementary Fig. S2a).
The overall structures of the GPR20-Gi-Fab046 complex and GPR20-Gi complex show canonical GPCR-G protein complex features; no extra density was observed in the transmembrane helical bundle (7TM) of GPR20 confirming they are both in ligand-free states. We used the structure with higher resolution in transmembrane region (GPR20-Gi-Fab046) for structural illustration in the following analysis unless otherwise noted.
## The N-terminal cap
Overall, the orphan GPR20 adopts a canonical 7TM architecture resembling other class-A GPCRs (Fig. 2a). Remarkably, the N terminal residues 33–49 fold over the top of the receptor forming a unique α-helical cap (Fig. 2a, b). Compared to Sphingosine 1-phosphate receptor subtype 1 (S1P1R, PDB: 7TD3)27 and Lysophosphatidic acid receptor 1 (LPA1R, PDB: 7TD0)27, both of which contain a N-terminal helix essential for lipid ligand entry, the N-terminal cap of GPR20 locates deeper in the 7TM core by 9.5 Å than LPA1R and S1P1R (the Cα atoms of the residues embedded deepest in the pocket were used: F38 for GPR20, Y34 for LPA1R and Y29 for S1P1R) (Fig. 2c). Structural superposition showed that the positions of ligands in S1P1R (sphingosine 1-phosphate, S1P) and LPA1R (lysophosphatidic acid, LPA) would clash with the N-terminal cap of GPR20 (Supplementary Fig. S3). The structural characterization hints that the N-terminal cap of GPR20 may behave as a ligand or play important roles in ligand entry and function. Fig. 2Unique N-terminal region of GPR20.a Two orthogonal views of the cartoon representation of GPR20 (cyan) to show the α-helical N-terminal cap (blue). b The electron density (gray mesh) of the N-terminal cap region is shown by overlying on the model. c Superposition of GPR20 (cyan), S1P1R (PDB 7TD3, orange) and LPA1R (PDB 7TD0, green) structures to show the respective position of the N-terminal cap in each structure (all aligned to β2AR structure (PDB: 3SN6), the Cα atoms of the residues embedded deepest in the pocket were used: F38 for GPR20, Y34 for LPA1R and Y29 for S1P1R). d Magnified view of the N-terminal cap of GPR20. The N-terminal cap and 7TM domain are in transparent blue and cyan cartoon representations, respectively. The side chains of key interacting residues are shown as sticks and overlaid with electron densities (gray mesh). e Mutations that may interfere with the conformation of the N-terminal cap impaired GPR20’s basal activity. ΔBRET: the change of bioluminescence resonance energy transfer value (Materials and methods). The ΔBRET was compared between the wild-type (WT) GPR20 and various mutants. Significance was determined by two-way analysis of variance (ANOVA) without repeated measures, followed by Dunnett’s post hoc test (***$P \leq 0.001$, **$P \leq 0.01$). Data are mean ± s.e.m. ( $$n = 3$$ independent biological experiments).
Different to S1P1R and LPA1R, whose N-terminal helix does not interact with the 7TM region directly, the intramolecular interactions around the N-terminal cap of GPR20 mainly engage TM2, TM3 and TM7 residues: F38N-term is in close contact with Y1303.32 in TM3 (superscripts denote Ballesteros–Weinstein numbering28), L41N-term forms a hydrophobic interaction with L2767.32 in TM7, D42N-term forms a salt bridge with R1092.60 in TM2, and L45N-term is in close contact with H2807.36 in TM7 (Fig. 2d). Mutagenesis and cellular functional assays showed that mutating the single key residues (F38N-termA, L41N-termE, D42N-termR and L45N-termE) on the N-terminal cap all markedly reduced the basal signaling activity of GPR20 (Fig. 2e). This result confirmed our hypothesis that the uniquely folded N-terminal cap may have a key role in stimulating the basal activity of GPR20.
## The activation mechanism of GPR20
A structural comparison of GPR20-Gi-Fab046 complex with all reported GPCR-G protein structures enabled us to examine the conformational features of GPR20 in G protein-coupled state. We found the structure of the ligand-free GPR20-Gi highly resembled that of galanin receptors29: galanin-bound GAL1R-Gi (PDB: 7WQ3) with RMSD value for aligned Cα atoms (RMSDCα) of 1.24 Å, and galanin-bound GAL2R-Gq (PDB: 7WQ4) with RMSDCα of 1.66 Å, although the sequence identities of GPR20 with GAL1R and GAL2R are only $25.1\%$ and $32.2\%$, respectively (Supplementary Fig. S4). Interestingly, structural superposition of GPR20 and GAL1R/GAL2R reveals that the position of the N-terminal cap of GPR20 overlaps with the position of galanin in GAL1R/GAL2R (Fig. 3a). Additionally, we found that the key residue Y9P on the galanin peptide, which mutation to alanine disrupts galanin’s binding29, is very close to F38N-term of GPR20 (Fig. 3a). It was reported that Y9P in galanin engaged extensive hydrophobic network in the receptor to transmit the signal toward the toggle switch29. Such an analogy prompts that F38 may be essential for activating GPR20. Indeed, the alanine mutation of F38 significantly decreased basal activity while the tryptophan mutation of F38 had no effect (Fig. 3b).Fig. 3The activation mechanism of GPR20.a Left, structural superposition of the ligand-free GPR20-Gi, galanin-bound GAL1R-Gi (PDB 7WQ3) and galanin-bound GAL2R-Gq (PDB 7WQ4) structures. Right, magnified view of the N-terminal cap of GPR20 (gray), galanin in GAL1R (green) and galanin in GAL2R (purple). Key residues are shown as sticks. b Constitutive activities of WT GPR20 and F38 mutations are measured by BRET assay. ΔBRET: the change of bioluminescence resonance energy transfer value (Materials and methods). Significance was determined by two-way analysis of variance (ANOVA) without repeated measures, followed by Dunnett’s post hoc test (**$P \leq 0.01$, n.s., not significant). Data are mean ± s.e.m. ( $$n = 3$$ independent biological experiments). c Four key residues involved in signal transmission from the N-terminal cap to the transmembrane region in GPR20. d The effects of key residue (shown in (c)) mutations on GPR20’s constitutive activity are measured by BRET assay. Significance was determined by two-way analysis of variance (ANOVA) without repeated measures, followed by Dunnett’s post hoc test (**$P \leq 0.01$). Data are mean ± s.e.m. ( $$n = 3$$ independent biological experiments).
To understand GPR20’s activation pathway, we focused on the key amino acid F38 in the N-terminal cap and found several hydrophobic residues on TM3 and TM6, including Y1303.32, M1343.36 and F2576.51 that are located below F38 and may transmit the signal to the toggle switch F2546.48 through a hydrophobic network (Fig. 3c). Mutagenesis and cellular functional assays showed that the single mutations of these residues impaired basal activity of GPR20 (Fig. 3d). Moreover, we investigated whether the common activation mechanism of class-A receptors can be observed in GPR20. Structural comparison of GPR20-Gi-Fab046 with active and inactive β2 adrenergic receptor (β2AR) (PDB: 3SN630 and 2RH131, respectively) reveals an outward movement of TM6 in the G protein-coupled GPR20 relative to inactive β2AR (Supplementary Fig. S5a). Several highly conserved micro-switches including the toggle switch (W/F6.48), PIF motif (P5.50, I3.40 and F6.44), DRY motif (D3.49 and R3.50), and NPxxY motif (Y7.53) (Supplementary Fig. S5b) in GPR20 resemble the conformation of the active β2AR structure. Thus, we conclude that a “signal” initiated by the N-terminal cap might be transmitted through a hydrophobic network toward the toggle switch and activation motifs, thus leading to the active conformation of GPR20 in the ligand-free and G protein-coupled state.
## Structure of the Gi-free GPR20-Fab046 complex
To understand the conformational changes induced by G protein coupling, we aimed to determine the structure of Fab046-bound GPR20 without Gi heterotrimer. Therefore, we assembled the purified GPR20 protein with Fab046 (see Materials and methods) for structural investigation. To improve the stability and expression level of WT GPR20 in the absence of co-expressed Gi proteins, two point mutations were introduced on the basis of homology to other class-A receptors: D2937.49N32 and L1393.41W33, which were essential for GPR20’s purification. Size-exclusion chromatography (SEC) and SDS-PAGE analysis revealed that GPR20 protein containing the two stabilizing mutations can form a stable complex with Fab046 amenable for cryo-EM studies (Supplementary Fig. S6b). We then determined the GPR20-Fab046 complex structure at a nominal global resolution of 3.08 Å (Fig. 4a, b; Supplementary Fig. S6 and Table S1). This structure is of sufficiently high resolution to allow assignment of the majority of Fab046 residues consisting of the intact light chain and VH + CH1 region of the heavy chain. Fig. 4Antibody-bound GPR20 structure.a, b Cryo-EM density map and atomic models of the GPR20-Fab046 structure. c Overlay between GPR20-Fab046 and GPR20-Gi-Fab046 structures in three representative views. Transmembrane helices TM1-TM7 and helix 8 (H8) are labeled. The movements of TM5 and TM6 on the intracellular side are highlighted as red arrows. d–f The conformational rearrangement of residues in conserved “micro-switches” upon Gi coupling: CF(W)xP, toggle switch, PIF, DRY and NPxxY motifs. g The effects of mutation in micro-switches are measured by BRET assay. Significance was determined by two-way analysis of variance (ANOVA) without repeated measures, followed by Dunnett’s post hoc test (**$P \leq 0.01$). Data are mean ± s.e.m. ( $$n = 3$$ independent biological experiments).
Structural comparison of GPR20-Fab046 and GPR20-Gi-Fab046 reveals an outward movement of TM6 for about 7.6 Å (based on Cα of G2326.26) and an inward movement of TM5 for about 4.5 Å (based on Cα of L2235.65) at the intracellular side in the Gi-coupled relative to Gi-free GPR20 structures (Fig. 4c; Supplementary Fig. S7). These movements are accompanied by structural transformations including the toggle switch (F2546.48), PIF motif (P2085.50, I1383.40 and F2506.44), DRY motif (D1473.49 and R1483.50), and NPxxY motif (Y2977.53) (Fig. 4d–f). In agreement with the structural findings, mutagenesis and cellular functional assays showed that mutating a single residue of each of these micro-switches reduced the signaling activity of GPR20 (Fig. 4g). Moreover, through structural comparison of GPR20-Fab046 with active and inactive β2AR structures, we found that the GPR20-Fab046 complex was captured in an inactive-like state resembling the conformation of the inactive β2AR structure (Supplementary Figs. S8a–c). Closer examination of inactive β2AR and GPR20-Fab046 structures revealed different conformations of TMs 5, 6 and 7 that may be partially attributed to the low sequence identity ($21.45\%$) between the two receptors. Also, the position of the ligand carazolol in the inactive β2AR forms direct interaction with toggle switch (W2866.48) and induces the TM6’s inward movement (Supplementary Fig. S8d). Another possible reason for this difference on TMs 5–7 is that the presence of the N-terminal cap region may stabilize GPR20 in the intermediate state, so that GPR20 could be quickly transformed into the fully active state in the presence of the G protein. Furthermore, an unidentified density, as will be discussed later, was observed in the transmembrane core of GPR20-Fab046 structure which might also impact the structure of the intracellular side of the GPR20 (Supplementary Fig. S8e).
In contrast to the intracellular side, the extracellular end of GPR20 does not exhibit notable changes between the GPR20-Fab046 and GPR20-Gi-Fab046 except for the flexible extracellular loops (Fig. 4c). Interestingly, ECL2 of GPR20 is very short and flexible which lacks a class-A conserved disulfide bond between ECL2 and TM3. This structural feature of ECL2 may be associated with the accommodation of the unique N-terminal cap which folds over next to the position of ECL2 in all three GPR20 structures (Figs. 2a and 4c).
## Binding interfaces for the antibody and G protein
In both GPR20-Fab046 and GPR20-Gi-Fab046 structures, the antibody fragment stabilizes the complex through an antibody binding interface composed of CDR2 and CDR3 of Fab046 heavy chain and the extracellular side of GPR20 mainly consisting of the N-terminal cap, extracellular side of TM1 and ECL1 of GPR20 (Supplementary Fig. S9), which is consistent with the previous epitope mapping study that Ab046 mainly binds to the N-terminal domain and ECL1 of GPR2017. The interface is maintained by seven pairs of hydrogen bonds between Fab046 and GPR20: side chain of N71CDR2 with E43N-term, S74CDR2 forming two hydrogen bonds with E43N-term, backbone carbonyl of S74CDR2 with R40N-term, backbone carbonyl of G121CDR3 with R117ECL1, the main-chain carbonyl oxygen of F122CDR3 with Y114ECL1, backbone carbonyl of F122CDR3 with H46N-term; and one π-π interaction: F122CDR3 with Y114ECL1 (Fig. 5a-b; Supplementary Table S2). The molecular interactions identified between GPR20 and Fab046 may aid in the rational design/optimization of tool antibodies with different affinities or new functions for therapeutics development. Fig. 5Binding interfaces for the Fab046 antibody and Gi protein on GPR20.a The binding interface between Fab046 (heavy chain in green, light chain in purple, spheres) and GPR20 (yellow, ribbon and transparent surface) is shown in two different views. b Magnified view of the antibody binding interface. CRD2 is shown in dark green, CDR3 is shown in light green, key residues are shown as sticks, and polar interactions are highlighted as red dashed lines. c The Gi binding interface. d Polar interactions (left) and hydrophobic interactions (right) between GPR20 (cyan) and Gi protein (orange). Key residues are shown as sticks, and polar interactions are highlighted as red dashed lines.
The structures of the GPR20-Gi complex with or without Fab046 show almost identical G protein coupling interface (Supplementary Fig. S7), which consists of TM2-3, TM5-6, ICL1-3 and H8 of GPR20, as well as the α5 helix of the Gα subunit: Q2316.25 forms hydrogen bonds with E318h4s6.12 and I319S6.01 in Gαi34, R2356.29 and R2366.30 form salt bridges with E318h4s6.12 and D341H5.13, and other hydrophobic interactions are formed by a series of hydrophobic residues (Fig. 5c, d; Supplementary Fig. S10a). Consistent with the structural findings, mutagenesis and cellular functional assays showed that most of the mutations at the interface reduced the signaling activity of GPR20 (Supplementary Fig. S10b). Compared to several representative GPCR-Gi structures including APJ23, CB2R35, D2R36, A1R37 and GAL1R29, where the αN helices of the Gαi subunits interact with the ICL2 of receptors to stabilize the GPCR-Gi protein complex, such a contact does not exist in GPR20 suggesting that it may utilize a slightly different mechanism for Gi coupling (Supplementary Fig. S10c, d).
## Discussion
Orphan GPCR research is still in its infancy; however new opportunities in this area are emerging given the increased number of receptor structures that have been reported and the diverse mechanisms for their constitutive activity uncovered. Here we report the structures of the orphan GPR20 with different states (GPR20-Gi, GPR20-Gi-Fab046 and GPR20-Fab046). These structures reveal a new mechanism to confer the orphan receptor’s high basal activity which might be attributed to the uniquely folded N-terminal α-helical cap region. Structural findings together with mutagenesis analysis suggest an “agonist-like” role of this N-terminal helix. In particular, the key residue F38N-term is located right above a hydrophobic network in GPR20-Gi and GPR20-Gi-Fab046 complexes (Fig. 3c; Supplementary Fig. S7c, d), through which the activation signal can be transmitted to the toggle switch to trigger conformational changes at the intracellular side. We also report the structures of GPR20 in complex with the selective but non-functional antibody, which uncover the specific binding interface between GPR20 and Fab046, thus providing the accurate template for tool antibody discovery. Moreover, as GPR20 shows therapeutic potential in GIST and other intestinal disorders, our structures may offer opportunity for rational drug discovery targeting GPR20 for related diseases.
As mentioned above, an extra cryo-EM density in the core of the 7TM region was observed in the GPR20-Fab046 map (Fig. 6a, b). Structural superposition of this extra density of GPR20 with corresponding ligand positions from representative ligand-bound GPCRs—such as LPA in the LPA1R27, S1P in the S1P1R27, LSD in the 5-HT2BR38, BI-167107 in the β2AR30, and AM12033 in the CB2R35—shows that it overlays with the orthosteric ligand-binding pocket in these receptors (Supplementary Fig. S11a). Moreover, we identified several residues surrounding this density that may contribute to the formation of the pocket: F38N-term, Y1303.32, M1343.36, F2546.48, F2576.51, H2586.52, Y2797.35 (Supplementary Fig. S11b). Mutagenesis and cellular functional assays showed that mutating any single residue in this putative pocket reduced the basal activity of GPR20 (Supplementary Fig. S11c and Table S3), suggesting that these residues may constitute the orthosteric pocket and the density here may represent an unknown ligand. As it was previously reported that GPR20 might be closely related to lipid receptors20, we tried modeling the endogenous ligands from all known structures of lipid GPCRs into this unassigned density one by one, however, none of them could be fit into the density properly (data not shown). It is also worth mentioning that this density was only observed in the GPR20-Fab046 structure but not the other two Gi-coupled GPR20 complex structures reported in this study. Though we cannot rule out the caveat originated from the two stabilizing mutations (maintained around $50\%$ of the basal activity relative to the WT protein, Supplementary Fig. S6a) in this specific construct, another possibility of assigning this density to an endogenously inverse agonist may merit future investigation. Fig. 6Unassigned density in the orthosteric pocket.a Two orthogonal views of the cartoon representation of GPR20 in yellow. The unassigned electron density observed in the canonical orthosteric pocket of GPR20 in the Gi-free GPR20-Fab046 complex is shown as a blue mesh. b A cross-section view of GPR20 in the GPR20-Fab046 structure (yellow) to show the shape complementarity of the unassigned electron density (blue mesh) with the orthosteric pocket. c Overlay between GPR20-Fab046 (yellow, unassigned density shown as a blue mesh) and GPR20-Gi-Fab046 (cyan) structures. Key residues of the hydrophobic network are shown as sticks. d, e Working model of putative activation mechanism for GPR20.
To conclude, we can reconcile all the structural findings in this study in a following working model for GPR20 (Fig. 6c–e): in the absence of the endogenous ligand, GPR20 can couple with Gi proteins to form a constitutively active complex (referring to the two Gi-coupled GPR20 structures), which is achieved by the N-terminal cap especially the F38N-term residue and the hydrophobic network that it engaged (Fig. 6d); when the ligand binds (referring to the unassigned density in the Gi-free GPR20 structure), the F38 is pushed upward away from the orthosteric pocket, followed by a disruption of the hydrophobic network where the key residue movements in the active-state structure would otherwise clash with the unassigned density (Fig. 6c, e). Thus the G protein is de-coupled and receptor activation is inhibited (Fig. 6e). Though this model requires identity of the unassigned density along with extensive experimental validations which are beyond the scope of current study, our structural findings together with mutagenesis data may provide a starting point for design of inverse agonist for therapeutic opportunity as well as to guide ligand discovery toward deorphanization for GPR20.
## Constructs and expression of GPR20 and Gi heterotrimer for cryo-EM study
The codon-optimized nucleotide sequence of human WT GPR20 (UniProt ID Q99678) was synthesized by GenScript. The human GPR20 gene was subcloned into an expression vector pFastBac1 (Invitrogen) with the addition of a haemagglutinin signal peptide, Flag tag, and a thermostabilized *Escherichia coli* apocytochrome b562RIL (BRIL)24 at the N-terminus of the receptor gene as well as an HRV 3 C protease recognition site followed by a 10x His tag at the C-terminus. The human dominant-negative Gαi1 (DNGαi subunit was generated by introducing three mutations: S47N, G203A, A326S) and the human WT Gβ1γ2 subunits (codon-optimized and synthesized by GenScript) were cloned into a pFastBac1 and pFastBacDual (Invitrogen) vector, respectively. The Gi protein-bound GPR20 complexes were obtained by co-expressing the receptor, DNGαi and Gβ1γ2 in Trichuplusia ni Hi5 insect cells (Invitrogen, B85502) using the Bac-to-Bac Baculovirus Expression System (Invitrogen). Trichuplusia ni Hi5 cells were infected at a cell density of 2–2.5 × 106 cells per mL with three separate virus (MOI = 5) preparations for GPR20, DNGαi and Gβ1γ2 at a ratio of 1:2:2. The infected cells were cultured at 27 °C for 48 h before collection by centrifugation, and the cell pellets were stored at −80 °C for future use.
## Expression and purification of scFv16
The codon-optimized nucleotide sequence of scFv16 was synthesized by GenScript and subcloned into an expression vector pFastBac1 with an 8x His tag at the C-terminus. The scFv16 used in this paper was the same as that used in the structures of the CB1-Gi-scFv1635. In brief, scFv16 was expressed in secreted form from Trichuplusia ni Hi5 insect cells and purified by Ni-NTA chromatography. The supernatant was incubated with Ni-NTA resin (GenScript) at 4 °C for 2 h. The resin was then loaded to a gravity column and washed with 15 column volumes (CV) of wash I buffer containing 20 mM HEPES (pH7.5), 100 mM NaCl and 10 mM imidazole; followed by 15 CV of wash II buffer containing 20 mM HEPES (pH7.5), 100 mM NaCl and 30 mM imidazole. The protein was eluted with elute buffer containing 20 mM HEPES (pH7.5), 100 mM NaCl and 250 mM imidazole. The elute was collected and further purified using a Superdex 200 $\frac{10}{300}$ column (GE Healthcare). Monomeric fractions were pooled, concentrated 10 mg/mL with a 10-kDa cut-off concentrator (Millipore), and flash frozen in liquid nitrogen, then stored at −80 °C for further use.
## Purification and formation of ligand-free GPR20-Gi complex
The cell pellets corresponding to 1 L GPR20-Gi co-expression culture were thawed and lysed in the hypotonic buffer of 10 mM HEPES (pH 7.5), 10 mM MgCl2, 20 mM KCl with EDTA-free complete protease inhibitor cocktail tablets (Roche). The GPR20-Gi complex was formed in membranes by addition of 1 unit of apyrase (NEB). The lysate was incubated overnight at 4 °C, and the supernatant was discarded after centrifugation at 40,000 rpm for 30 min. The complex was solubilized from membranes in the buffer containing 50 mM HEPES (pH 7.5), 100 mM NaCl, $1\%$ (w/v) lauryl maltose neopentyl glycol (LMNG, Anatrace), $0.2\%$ (w/v) cholesteryl hemisuccinate (CHS) (Sigma), 2 units of apyrase at 4 °C for 2 h. The supernatant was isolated by ultracentrifugation at 35,000 rpm for 30 min, and then incubated with TALON IMAC resin (Clontech) and 20 mM imidazole overnight at 4 °C. The resin was washed with 15 CV (column volumes) of washing buffer I containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $5\%$ (v/v) glycerol, $0.1\%$ (w/v) LMNG, $0.02\%$ (w/v) CHS, 30 mM imidazole, and 15 CV of washing buffer II containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $5\%$ (v/v) glycerol, $0.03\%$ (w/v) LMNG, $0.006\%$ (w/v) CHS and 50 mM imidazole. The protein was eluted using 3 column volumes of elution buffer containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $10\%$ (v/v) glycerol, $0.01\%$ (w/v) LMNG, $0.002\%$ (w/v) CHS and 250 mM imidazole. The eluate (GPR20-Gi complex) and scFv16 were mixed in a 1:1.5 ratio for 0.5 h, then concentrated and injected onto a Superdex200 $\frac{10}{300}$ GL column (GE Healthcare) equilibrated in the buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, $0.00075\%$ (w/v) LMNG, $0.00025\%$ (w/v) glyco-diosgenin (GDN, Anatrace), $0.0001\%$ (w/v) CHS, 100 μM TCEP. The complex peak fractions were collected and concentrated to 2.5 mg/mL with a 100-kDa cut-off concentrator (Millipore) for electron microscopy experiments.
## Constructs, expression, and purification of Fab046
We used the previously reported IgG04617 to generate a Fab fragment (Fab046, codon-optimized and synthesized by GenScript, containing intact light chain and part of heavy chain (VH + CH1) of IgG046). The light and heavy chains of Fab046 were cloned into a pFastBacDual vector. The Trichuplusia ni Hi5 insect cells were infected with baculovirus at a density of 2 × 106 cells per mL. Cells were grown at 27 °C and collected 48 h after infection. The cells were centrifuged at 2000 rpm for 30 min, and the 1 L supernatant was loaded onto a 2 mL Ni-NTA resin. The column was washed with 15 CV of wash buffer containing 20 mM Tris-HCI (pH 7.55), 150 mM NaCl, and 20 mM imidazole and the protein was eluted with the same buffer supplemented with 250 mM imidazole, the protein was collected and purified over gel filtration chromatography using a Superdex 200 $\frac{10}{300}$ column equilibrated in the buffer containing 20 mM Tris-HCI (pH 7.55), 100 mM NaCl, and $10\%$ glycerol. Monomeric fractions were pooled, concentrated to 4 mg/mL with a 30-kDa cut-off concentrator (Millipore), and flash frozen in liquid nitrogen, then stored at −80 °C for further use.
## Constructs, expression, and purification of Fab046 bound Gi-free and Gi-coupled complex
For GPR20-Gi-Fab046, the complex protein was purified as described above for the GPR20-Gi complex, except that Fab046 was added during the purification process: the eluate (GPR20-Gi complex) and Fab046 were mixed in a 1:1.5 ratio for 6 h, then concentrated and injected onto a Superdex200 $\frac{10}{300}$ GL column (GE Healthcare) equilibrated in the buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, $0.00075\%$ (w/v) LMNG, $0.00025\%$ (w/v) GDN, $0.0001\%$ (w/v) CHS, 100 μM TCEP. The complex peak fractions were collected and concentrated to 2.5 mg/mL with a 100-kDa cut-off concentrator (Millipore) for electron microscopy experiments.
For GPR20-Fab046, to improve protein stability of GPR20 alone, two stabilizing mutations (L1393.41W and D2937.49N, which were essential for GPR20’s purification) were introduced based on the above construct for GPR20-Gi complex. We used the Bac-to-*Bac baculovirus* system in *Spodoptera frugiperda* (Sf9) cells for expression. These cells were infected with baculovirus (MOI = 5) at a density of 2 × 106 cells per mL. Cells were grown at 27 °C and collected 48 h after infection. The cells were washed once with a low-salt buffer containing 10 mM HEPES (pH 7.5), 20 mM KCl, 10 mM MgCl2 and protease inhibitor cocktail (Roche), and three times with a high-salt buffer containing 10 mM HEPES (pH 7.5), 1 M NaCl, 20 mM KCl, 10 mM MgCl2 and protease inhibitor cocktail. Before solubilization, purified membranes were incubated with 2 mg/mL iodoacetamide (Sigma) at 4 °C for 0.5 h. The protein was extracted from the membrane by 50 mM HEPES (pH 7.5), 500 mM NaCl, $1.0\%$ (w/v) LMNG and $0.2\%$ (w/v) CHS and stirred for 2 h at 4 °C. After centrifugation, the supernatant was incubated with TALON IMAC resin (Clontech) at 4 °C overnight. Then the resin was washed with 15 CV of buffer I containing 50 mM HEPES (pH 7.5), 500 mM NaCl, $5\%$ (v/v) glycerol, $0.05\%$ (w/v) LMNG, $0.01\%$ (w/v) CHS, 10 mM MgCl2 and 20 mM imidazole. Then the resin was washed with 10 CV of buffer II containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $5\%$ (v/v) glycerol, $0.01\%$ (w/v) LMNG, $0.002\%$ (w/v) CHS, 40 mM imidazole. Next, the protein was eluted with 3 CV of buffer III containing 25 mM HEPES (pH 7.5), 100 mM NaCl, $5\%$ (v/v) glycerol, $0.005\%$ (w/v) LMNG, $0.001\%$ (w/v) CHS and 220 mM imidazole. The purified GPR20 protein was concentrated with a 50-kDa cut-off concentrator to around 2 mg/mL, mixed with Fab046 in a 1:1.5 mole ratio and incubated on ice overnight to form the complex. The mixture was loaded on a Superdex 200 $\frac{10}{300}$ column with a running buffer of 20 mM HEPES (pH 7.5), 100 mM NaCl, $0.00075\%$ (w/v) LMNG, $0.00025\%$ (w/v) GDN, $0.0001\%$ (w/v) CHS, 100 μM TCEP. Peak fractions containing the GPR20-Fab046 complex were pooled and concentrated with a 100-kDa cut-off concentrator to 2 mg/mL for cryo-EM studies.
## Preparation of vitrified samples for Cryo-EM
In total, 3 μL of the purified samples (GPR20-Gi, GPR20-Fab046 and GPR20-Gi-Fab046) at a concentration of around 2 mg/mL were applied to glow-discharged 300-mesh Au grids (Quantifoil, R$\frac{1.2}{1.3}$). Excess sample was removed by blotting with filter paper for 3.5 s before plunge-freezing in liquid ethane using a FEI Vitrobot Mark IV at $100\%$ humidity and 8 °C.
## Cryo-EM data collection
All datasets were collected on a Titan Krios 300 kV electron microscope (Thermo Fisher Scientifics, USA) equipped with a GIF Quantum energy filter (20 eV energy slit width, Gatan Inc., USA). All the GPR20 datasets were recorded by a K3 camera (Gatan) at a nominal magnification of 105,000 (calibrated pixel size: 0.832 Å/pixel) and 15 e-/pixel2/s. The movies were recorded using the super resolution counting mode by SerialEM which applied the beam image shift acquisition method with one image near the edge of each hole. A 50 µm C2 aperture was always inserted during the data collection period. The defocus ranged from −0.7 to −2.2 µm. For each movie stack, a total of 40 frames were recorded, yielding a total dose of 60 e−/Å2.
## Cryo-EM image processing
For GPR20-Gi complex, 4347 movies were recorded and processed with cryoSPARC v.3.339. Patch motion correction was used for beam-induced motion correction. Then, contrast transfer function (CTF) parameters for each dose-weighted micrograph were estimated by patch CTF estimation. Only images with the highest resolution of less than 4 Å were selected for further processing. A total of 4129 images were selected for auto blob picking, and particles were extracted to do 2D classification. 2D class averages with diverse orientations and clear secondary features were selected as 2D templates for another round of autopicking process by cryoSPARC. A total of 386,819 particles were selected from good 2D classification to generate the initial models. These particles and initial models were used to do 3D classification in heterogeneous refinement in cryoSPARC. 256,813 particles were selected for the final homogeneous refinement followed by nonuniform refinement and local refinement in cryoSPARC, resulting in density map with nominal resolution of 3.14 Å for the GPR20-Gi complex (determined by gold-standard Fourier shell correlation (FSC), 0.143 criterion). Estimation of local resolution was performed in cryoSPARC.
For GPR20-Gi-Fab046 complex, 4443 movies were recorded and processed with cryoSPARC. Motion correction and CTF were applied and estimated as in the case of the GPR20-Gi complex. Only images with the highest resolution of less than 4 Å were selected for further processing. A total of 4247 images were selected for auto blob picking, and particles were extracted to do 2D classification. 2D class averages with diverse orientations and clear secondary features were selected as the 2D templates for another round of autopicking process by cryoSPARC. A total of 314,744 particles were selected from good 2D classification to generate the initial models. These particles and initial models were used to do 3D classification in heterogeneous refinement in cryoSPARC. 164,932 particles were selected for the final homogeneous refinement followed by nonuniform refinement and local refinement in cryoSPARC, resulting in density map with nominal resolution of 3.22 Å for the GPR20-Gi-Fab046 complex (FSC = 0.143). Estimation of local resolution was performed in cryoSPARC.
For GPR20-Fab046 complex, 4458 movies were recorded and processed with cryoSPARC. Motion correction and CTF were corrected and estimated as GPR20-Gi complex. Only images with the highest resolution of less than 4 Å were selected for further processing. A total of 4,145 images were selected to do auto blob picking and particles were extracted to do 2D classification. 2D class averages with diverse orientations and clear secondary features were selected as the 2D templates for another round of autopicking process by cryoSPARC. A total of 839,525 particles were selected from good 2D classification to generate the initial models. These particles and initial models were used to do 3D classification in heterogeneous refinement in cryoSPARC. 418,288 particles were selected for final homogeneous refinement followed by nonuniform refinement and local refinement in cryoSPARC, resulting in density map with nominal resolution of 3.08 Å for the GPR20-Fab046 complex (FSC = 0.143). Estimation of local resolution was performed with local resolution estimation in cryoSPARC.
## Cryo-EM model building and refinement
The homology models of GPR20 and Fab046 were initially generated by Alphafold40. For Gi trimer and scFv16, the model 6KPF (PDB)35 was chosen. Each part of the target models was docked into the electron microscopy density map using UCSF Chimera41. Then, these models were used for model building and refinement against the electron density map. Subsequently, the generated model was manually adjusted in Coot42 followed by automatic real space refinement in real space in Phenix43 for several iterations. The model statistics were validated using Phenix43. The final refinement statistics are provided in Supplementary Table S1.
## BRET2 assay
To measure the dissociation of Gαβγ heterotrimer directly, we applied the BRET2 assay system as reported before21. In brief, HEK293T cells were plated in a 6-well plate. After 2 h, cells were transiently co-transfected with plasmids encoding WT or mutated GRR20 together with Gi BRET probe (Gαi1-RLuc8, Gβ3, Gγ9-GFP2) using Lipofectamine 2000 reagent (Life Technologies). Adenosine A2A receptor (A2AR) that does not couple to Gi proteins was used as a negative control, Apelin receptor (APJ) that couples to Gi proteins was used as a positive control for the Gi BRET assay. 24 h after transfection, cells were distributed into a 96-well microplate (30,000–50,000 cells per well) and incubated for additional 24 h at 37 °C. For the constitutive activity measurement, white backings (Perkin Elmer) were applied to the plate bottoms, the transfected cells were washed once with HBSS and supplemented with 100 µL of 5 µM coelenterazine 400a (Nanolight Technologies). Plates were read in EnVision plate reader (Perkin Elmer) with 410 nm (RLuc8) and 515 nm (GFP2) emission filters with an integration time of 1 s per well. The GFP2 emission to RLuc8 emission ratio was used to compute the BRET2 ratios. ΔBRET represent the change of bioluminescence resonance energy transfer value. ΔBRET = BRET ratio (GPCR with G protein sensor) - BRET ratio (only G protein sensor).
## Supplementary information
Supplementary Information The online version contains supplementary material available at 10.1038/s41421-023-00520-8.
## References
1. Zhou Y. **Molecular insights into ligand recognition and G protein coupling of the neuromodulatory orphan receptor GPR139**. *Cell Res.* (2022) **32** 210-213. DOI: 10.1038/s41422-021-00591-w
2. Lin X. **Structural basis of ligand recognition and self-activation of orphan GPR52**. *Nature* (2020) **579** 152-157. DOI: 10.1038/s41586-020-2019-0
3. Leonard S, Kinsella GK, Benetti E, Findlay JBC. **Regulating the effects of GPR21, a novel target for type 2 diabetes**. *Sci. Rep.* (2016) **6** 27002. DOI: 10.1038/srep27002
4. Sun H. **First-time disclosure of CVN424, a potent and selective gpr6 inverse agonist for the treatment of Parkinson’s disease: discovery, pharmacological validation, and identification of a clinical candidate**. *J. Med. Chem.* (2021) **64** 9875-9890. DOI: 10.1021/acs.jmedchem.0c02081
5. Azam S, Haque ME, Jakaria M. **G-Protein-coupled receptors in CNS: a potential therapeutic target for intervention in neurodegenerative disorders and associated cognitive deficits**. *Cells* (2020) **9** 506. DOI: 10.3390/cells9020506
6. Ganesh RA, Venkataraman K, Sirdeshmukh R. **GPR56: An adhesion GPCR involved in brain development, neurological disorders and cancer**. *Brain Res.* (2020) **1747** 147055. DOI: 10.1016/j.brainres.2020.147055
7. Su XQ, Wang J, Sinclair AJ. **Plasmalogens and Alzheimer’s disease: a review**. *Lipids Health Dis.* (2019) **18** 100. DOI: 10.1186/s12944-019-1044-1
8. Martin AL, Steurer MA, Aronstam RS. **Constitutive Activity among Orphan Class-A G Protein Coupled Receptors**. *PLoS One* (2015) **10** e0138463. DOI: 10.1371/journal.pone.0138463
9. Ye F. **Cryo-EM structure of G-protein-coupled receptor GPR17 in complex with inhibitory G protein**. *MedComm (2020)* (2022) **3** e159. PMID: 36105372
10. Chen G. **Activation and allosteric regulation of the orphan GPR88-Gi1 signaling complex**. *Nat. Commun.* (2022) **13** 2375. DOI: 10.1038/s41467-022-30081-5
11. Xu P. **Structural identification of lysophosphatidylcholines as activating ligands for orphan receptor GPR119**. *Nat. Struct. Mol. Biol.* (2022) **29** 863-870. DOI: 10.1038/s41594-022-00816-5
12. Ping YQ. **Structural basis for the tethered peptide activation of adhesion GPCRs**. *Nature* (2022) **604** 763-770. DOI: 10.1038/s41586-022-04619-y
13. Qu X. **Structural basis of tethered agonism of the adhesion GPCRs ADGRD1 and ADGRF1**. *Nature* (2022) **604** 779-785. DOI: 10.1038/s41586-022-04580-w
14. Xiao P. **Tethered peptide activation mechanism of the adhesion GPCRs ADGRG2 and ADGRG4**. *Nature* (2022) **604** 771-778. DOI: 10.1038/s41586-022-04590-8
15. Barros-Álvarez X. **The tethered peptide activation mechanism of adhesion GPCRs**. *Nature* (2022) **604** 757-762. DOI: 10.1038/s41586-022-04575-7
16. O’Dowd BF. **Cloning and chromosomal mapping of four putative novel human G-protein-coupled receptor genes**. *Gene* (1997) **187** 75-81. DOI: 10.1016/S0378-1119(96)00722-6
17. Iida K. **Identification and therapeutic targeting of gpr20, selectively expressed in gastrointestinal stromal tumors, with DS-6157a, a first-in-class antibody-drug conjugate**. *Cancer Discov.* (2021) **11** 1508-1523. DOI: 10.1158/2159-8290.CD-20-1434
18. Allander SV. **Gastrointestinal stromal tumors with KIT mutations exhibit a remarkably homogeneous gene expression profile**. *Cancer Res.* (2001) **61** 8624-8628. PMID: 11751374
19. Vélez JI. **APOE*E2 allele delays age of onset in PSEN1 E280A Alzheimer’s disease**. *Mol. Psychiatry* (2016) **21** 916-924. DOI: 10.1038/mp.2015.177
20. Hase M, Yokomizo T, Shimizu T, Nakamura M. **Characterization of an orphan G protein-coupled receptor, GPR20, that constitutively activates Gi proteins**. *J. Biol. Chem.* (2008) **283** 12747-12755. DOI: 10.1074/jbc.M709487200
21. Olsen RHJ. **TRUPATH, an open-source biosensor platform for interrogating the GPCR transducerome**. *Nat. Chem. Biol.* (2020) **16** 841-849. DOI: 10.1038/s41589-020-0535-8
22. Kooistra AJ. **GPCRdb in 2021: integrating GPCR sequence, structure and function**. *Nucleic Acids Res.* (2021) **49** D335-d343. DOI: 10.1093/nar/gkaa1080
23. Yue Y. **Structural insight into apelin receptor-G protein stoichiometry**. *Nat. Struct. Mol. Biol.* (2022) **29** 688-697. DOI: 10.1038/s41594-022-00797-5
24. Chun E. **Fusion partner toolchest for the stabilization and crystallization of G protein-coupled receptors**. *Structure* (2012) **20** 967-976. DOI: 10.1016/j.str.2012.04.010
25. Liang YL. **Dominant negative G proteins enhance formation and purification of Agonist-GPCR-G protein complexes for structure determination**. *ACS Pharm. Transl. Sci.* (2018) **1** 12-20. DOI: 10.1021/acsptsci.8b00017
26. Maeda S. **Development of an antibody fragment that stabilizes GPCR/G-protein complexes**. *Nat. Commun.* (2018) **9** 3712. DOI: 10.1038/s41467-018-06002-w
27. Liu S. **Differential activation mechanisms of lipid GPCRs by lysophosphatidic acid and sphingosine 1-phosphate**. *Nat. Commun.* (2022) **13** 731. DOI: 10.1038/s41467-022-28417-2
28. Ballesteros JA, Weinstein H. **Integrated methods for the construction of three-dimensional models and computational probing of structure-function relations in G protein-coupled receptors**. *Methods Neurosci.* (1995) **25** 366-428. DOI: 10.1016/S1043-9471(05)80049-7
29. Duan J. **Molecular basis for allosteric agonism and G protein subtype selectivity of galanin receptors**. *Nat. Commun.* (2022) **13** 1364. DOI: 10.1038/s41467-022-29072-3
30. Rasmussen SG. **Crystal structure of the β2 adrenergic receptor-Gs protein complex**. *Nature* (2011) **477** 549-555. DOI: 10.1038/nature10361
31. Cherezov V. **High-resolution crystal structure of an engineered human beta2-adrenergic G protein-coupled receptor**. *Science* (2007) **318** 1258-1265. DOI: 10.1126/science.1150577
32. Popov P. **Computational design of thermostabilizing point mutations for G protein-coupled receptors**. *Elife* (2018) **7** e34729. DOI: 10.7554/eLife.34729
33. Roth CB, Hanson MA, Stevens RC. **Stabilization of the human beta2-adrenergic receptor TM4-TM3-TM5 helix interface by mutagenesis of Glu122(3.41), a critical residue in GPCR structure**. *J. Mol. Biol.* (2008) **376** 1305-1319. DOI: 10.1016/j.jmb.2007.12.028
34. Flock T. **Universal allosteric mechanism for Gα activation by GPCRs**. *Nature* (2015) **524** 173-179. DOI: 10.1038/nature14663
35. Hua T. **Activation and signaling mechanism revealed by cannabinoid receptor-G(i) complex structures**. *Cell* (2020) **180** 655-665.e618. DOI: 10.1016/j.cell.2020.01.008
36. Zhuang Y. **Structural insights into the human D1 and D2 dopamine receptor signaling complexes**. *Cell* (2021) **184** 931-942.e918. DOI: 10.1016/j.cell.2021.01.027
37. Draper-Joyce CJ. **Structure of the adenosine-bound human adenosine A(1) receptor-G(i) complex**. *Nature* (2018) **558** 559-563. DOI: 10.1038/s41586-018-0236-6
38. Wacker D. **Crystal structure of an LSD-bound human serotonin receptor**. *Cell* (2017) **168** 377-389.e312. DOI: 10.1016/j.cell.2016.12.033
39. Punjani A, Rubinstein JL, Fleet DJ, Brubaker MA. **cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination**. *Nat. Methods* (2017) **14** 290-296. DOI: 10.1038/nmeth.4169
40. Jumper J. **Highly accurate protein structure prediction with AlphaFold**. *Nature* (2021) **596** 583-589. DOI: 10.1038/s41586-021-03819-2
41. Pettersen EF. **UCSF Chimera-a visualization system for exploratory research and analysis**. *J. Comput. Chem.* (2004) **25** 1605-1612. DOI: 10.1002/jcc.20084
42. Emsley P, Lohkamp B, Scott WG, Cowtan K. **Features and development of Coot**. *Acta Crystallogr. D. Biol. Crystallogr.* (2010) **66** 486-501. DOI: 10.1107/S0907444910007493
43. McCoy AJ. **Phaser crystallographic software**. *J. Appl. Crystallogr.* (2007) **40** 658-674. DOI: 10.1107/S0021889807021206
|
---
title: 'Mind matters: A narrative review on affective state-dependency in non-invasive
brain stimulation'
authors:
- Dennis J.L.G. Schutter
- Fenne Smits
- Jana Klaus
journal: 'International Journal of Clinical and Health Psychology : IJCHP'
year: 2023
pmcid: PMC9971283
doi: 10.1016/j.ijchp.2023.100378
license: CC BY 4.0
---
# Mind matters: A narrative review on affective state-dependency in non-invasive brain stimulation
## Abstract
Variability in findings related to non-invasive brain stimulation (NIBS) have increasingly been described as a result of differences in neurophysiological state. Additionally, there is some evidence suggesting that individual differences in psychological states may correlate with the magnitude and directionality of effects of NIBS on the neural and behavioural level. In this narrative review, it is proposed that the assessment of baseline affective states can quantify non-reductive properties which are not readily accessible to neuroscientific methods. Particularly, affective-related states are theorized to correlate with physiological, behavioural and phenomenological effects of NIBS. While further systematic research is needed, baseline psychological states are suggested to provide a complementary cost-effective source of information for understanding variability in NIBS outcomes. Implementing measures of psychological state may potentially contribute to increasing the sensitivity and specificity of results in experimental and clinical NIBS studies.
## Introduction
Transcranial magnetic (TMS) and electric stimulation (tES) techniques are non-invasive methods to probe and modulate brain physiology, allowing for causal inferences between neural activity and behaviour. Owing to its unique ability to influence nerve tissue, non-invasive brain stimulation (NIBS) provides an imperative means to study the workings of the human brain. In addition to investigating the basic neural properties underlying mental processes and behaviour, the modulatory effects of NIBS on neurophysiology that can outlast the stimulation period make tES and TMS suitable candidates for the treatment of mental and neurological disorders. NIBS has proven invaluable for expanding our knowledge about the neurobiological underpinnings of human behaviour in health and disease, yet fluctuations in, for example, ongoing neural activity can greatly influence the outcomes of NIBS on brain and behaviour (e.g., Feurra et al., 2013; Schutter & Hortensius, 2011). Within this context, the concept of neural state-dependency has been proposed, in which the effects caused by NIBS at least in part depend on ‘types’ of brain activity at the time of stimulation (Kasten & Herrmann, 2022). Acknowledging variability of physiological states within and across individuals may contribute to a better understanding of how NIBS interacts with brain matter. Moreover, it provides information on how accounting for individual differences in neural state may increase reliability and optimise the effects of NIBS (Bradley, Nydam, Dux & Mattingley, 2022).
Variation in neural activity due to external (e.g., task demands, time of day) or internal (e.g., spontaneous electroencephalography [EEG] fluctuations) factors has been demonstrated to predict magnitude and directionality of the effects of NIBS on brain physiology and behaviour (Bradley et al., 2022; Hartwigsen & Silvanto, 2022; Penton, Catmur, Banissy, Bird & Walsh, 2022). Information that can be extracted from the brain with conventional neuroimaging techniques, such as functional magnetic resonance imaging (fMRI) and EEG, is an important source for studying the brain-NIBS coupling in humans. While the biological-centred approach is critical, these techniques are restricted by physical and spatiotemporal constraints and can only focus on a limited number of neural features. Identifying response markers for neuromodulation treatment of psychiatric disorders has become an important topic in the field of NIBS, but the extent to which these markers can reliably predict behavioural outcomes to NIBS-based interventions is still part of ongoing neuroscientific research.
An additional source of information that may be relevant is one's psychological state. The added value of measuring psychological states is proposed to lie in [1] providing unique information, such as the phenomenological (subjective) experience of anxiety that cannot be disclosed with existing neuroimaging techniques, and [2] not having to rely on expensive neuroimaging equipment and expertise. Within the psychological framework of embodiment, psychological states can be understood as subjective phenomenological read-outs determined by sensory signals from the external surroundings (exteroception), (somato-visceral) signals from within the body (interoception), and prior experience (Northoff, 2012; Oosterwijk et al., 2012). One category of psychological states are affective states or the conscious experience of feeling the underlying emotion (Panksepp & Biven, 2012). Affective states can be either focused, short-lived sensations or more diffuse experiences that extend over time. The former would be more closely related to emotions (e.g., anxiety), whereas the latter is more associated with mood (e.g., depression) and personality characteristics (e.g., neurotic). In addition, motivational dispositions based on, for example, reward and punishment sensitivity may reflect the more latent aspects of affective states. Affective states may provide a non-reductive proxy for an individual's overall (bodily) state. It is proposed that such states can be informative for understanding and predicting the effects of NIBS on brain and behaviour both on the population and the individual level.
The aim of this theoretical perspective is to illustrate the potential role of affective states in NIBS by reviewing studies that provide evidence for associations between baseline affective states and the magnitude and directionality of tES- and TMS-related effects on brain and behaviour. It is proposed that heterogeneity of results and null findings of NIBS studies on the group level may at least in part be accounted for by individual differences in affective states (Fig. 1).Fig. 1Illustration of the association between affective state and NIBS-induced effects. When not accounting for individual differences in affective state (left panel), effects from participants showing inhibition from active NIBS relative to a control condition (green shaded points) and from participants showing facilitation (red shaded points) cancel each other out, resulting in a null effect of NIBS on the group level. Individual differences in affective state may account for differences in polarity and magnitude of the NIBS effect (right panel): Individuals scoring low on an affective measure show NIBS-induced inhibition, which flips towards facilitation with increasing affective state scores. Note that the directionality of effects is hypothetical and the association can also be positive or non-linear. Fig 1
## Affective state-dependency in TMS research
Ever since the pioneering research demonstrating that manipulating the state of neural excitability can induce opposite behavioural effects of TMS (for a review see Silvanto & Pascual-Leone, 2008), neuroscientific research has yielded important insights into the role of neural state-dependency in basic and clinical research (e.g., Borgomaneri et al., 2020; Janssens & Sack, 2021; Siebner et al., 2022; Wendt et al., 2022). Next to physiological measures, however, there is an increasing number of studies showing that psychological measures depicting individuals’ affective states prior to or during TMS can substantially contribute to observed outcomes as well (Table 1). Increases in cortical excitability during the processing of threat-related stimuli have, for instance, been repeatedly shown in healthy volunteers (Coombes et al., 2009; Hortensius, de Gelder, & Schutter, 2016; Schutter, Hofman & Honk, 2008). Moreover, anticipatory anxiety was shown to have a facilitating effect on cortical excitability in response to a suprathreshold TMS pulse (Oathes, Bruce & Nitschke, 2008), providing further evidence for relations between affective states, including worrying, nervousness, action preparedness, and cortical excitability levels. Table 1Overview of reviewed studies examining the interaction between TMS effects and affective states. Table 1StudySampleTMS detailsOutcome variable(s)Affective state measurementFindingBaeken et al. [ 2011]$$n = 24$$ (all female); healthy volunteers; age 19–29 years10-Hz offline TMS, 1560 pulses, right DLPFC or sham, $110\%$ rMTSalivary cortisol, Mood (POMS)STAI-stateHigher state anxiety scores associated with higher cortisol increase following active right DLPFC TMS; no significant TMS effects on moodBerlim et al. [ 2013]$$n = 14$$ (7 M/8F); MDD patients; age: 33–61 years10-Hz offline TMS, 3000 pulses, $120\%$ rMT, left DLPFC, 20 sessionsDepressive symptoms (HAM-D21)Big Five Inventory: Extraversion and Neuroticism personality domainsHigher baseline Extraversion scores associated with more depressive symptom improvement; no effect of baseline Neuroticism scoresChen et al. [ 2020]$$n = 97$$ (20F/77 M); MUD patients; age >18 yearsActive or sham iTBS over the left DLPFC; 20 daily sessions over 4 weeks; $100\%$ rMT; 50 Hz, burst frequency 5 Hz, 900 pulses spread over five minutes; train duration: 2 s, intertrain-interval: 6 s; between-participantVAS for spontaneous and cue-induced cravingPHQ-9, GAD-7Milder anxiety and depressive symptoms correlated with reduced craving in response to active iTBS vs. shamDe Witte et al. [ 2020]$$n = 38$$ (all female); Mage = 23.5, SD = 3.0Active or sham iTBS over the left DLPFC; $110\%$ rMT; 50 Hz, burst frequency 5 Hz, 1620 pulses spread over 54 cycles (10 bursts of 3 pulses each); train duration: 2 s, inter-train interval: 6 s; within-participantTrait rumination measured by Ruminative Response Scale (reflective pondering and depressive brooding); momentary rumination measured by Ruminative Self-Focus Scale before TSST and after TBS; cortisol levels 15 and 25 min after TSSTTSSTUnder sham cTBS, rumination increased with increasing brooding levels, but this effect was not observed under active cTBSHigher levels of brooding related to decreased cortisol secretion under active, but not sham cTBSDownar et al. [ 2014]$$n = 47$$ (20 M/27F); patients with unipolar ($$n = 38$$) or bipolar depression ($$n = 9$$)10-Hz offline TMS, 3000 pulses, $120\%$ rMT, DMPFC, 20 sessionsDepressive symptoms (HAM-D17)BDI-II, QIDSHigher baseline anhedonia symptoms (BDI-II items ‘Pessimism’ and ‘Loss of Pleasure’, QIDS item ‘General Interest’) associated with lower symptom improvementGuo et al. [ 2020]$$n = 44$$ (24F/20 M); healthy volunteers; Mage = 21, SEM = 0.39Single-pulse TMS 90, 100, 110, 120, 130, 140 or 150 ms after stimulus onset, or no TMS over early visual cortex at $120\%$ phosphene detection thresholdEmotion detection task (angry vs. happy vs. fearful faces)STAI, BAIHigher TMS-induced disruption of anger recognition for individuals with lower anxiety levels; no modulation of fear or happiness recognitionSagliano et al. [ 2016]$$n = 22$$ (all female); healthy volunteers; age: 19–30 yearsSingle-pulse TMS 100 or 200 ms after stimulus onset to left or right DLPFC or sham vertexAttentional bias for threatening pictures (exogenous cueing task)STAI-traitHigher trait anxiety associated with higher disengagement bias during left DLPFC stimulationVanderhasselt et al. [ 2011]$$n = 28$$ (all female); healthy volunteers; age: 18–29 years10-Hz offline TMS, 1560 pulses, right DLPFC or shamAttentional bias for angry faces (exogenous cueing task)POMS, STAI-stateHigher state anxiety scores associated with increased attentional bias for angry faces following active right DLPFC TMSBDI-II = Beck Depression Inventory-II; cTBS = continuous Theta Burst Stimulation; DLPFC: Dorsolateral Prefrontal Cortex; DMPFC: Dorsomedial Prefrontal Cortex; F = Females; GAD-7 = Generalized Anxiety Disorder-7; HAM-D17/HAM-D21 = $\frac{17}{21}$-item Hamilton Depression Rating Scale; iTBS = intermittent Theta Burst Stimulation; M = Males; M = Mean; MDD = Major Depressive Disorder; MEP = Motor-Evoked Potential; MUD = Methamphetamine Use Disorder; NEO-PI-R = NEO Personality Inventory Revised; PANAS = Positive and Negative Affect Schedule; PEMS = Palatable Eating Motives Scale; PHQ-9: Patient Health Questionnare-9; POMS = Profile of Mood States; QIDS = Quick Inventory of Depressive Symptomatology–Self-rated 16-item scale; rMT = resting Motor Threshold; SD = Standard Deviation; SEM = Standard Error of the Mean; STAI = State-Trait Anxiety Inventory; TMS = Transcranial Magnetic Stimulation; TSST = Trier Social Stress Test; VAS = Visual Analogue Scale.
Additionally, affective state-dependent modulation of NIBS effects have been reported for different outcome measures. A sham-controlled study tested the influence of state anxiety on the endocrinological response to a single session of high-frequency repetitive TMS (rTMS) over the right dorsolateral prefrontal cortex (DLPFC) in healthy female volunteers (Baeken, Vanderhasselt & De Raedt, 2011). Results showed that individuals with higher state anxiety displayed significantly higher levels of the stress hormone cortisol after real TMS as compared to those scoring low on state anxiety. The authors proposed that the inclusion of individual anxiety states in experimental rTMS research could lead to a deeper understanding of the effects of NIBS on the brain's stress system (Baeken et al., 2011). Another study examined the effects of state anxiety prior to a session of high- frequency rTMS to the right DLPFC on attention processing in healthy volunteers (Vanderhasselt, Baeken, Hendricks & De Raedt, 2011). Self-report measures of state anxiety prior to stimulation correlated positively with behavioural performance. More specifically, participants with higher state anxiety showed a more pronounced increase of attentional bias towards negative information after high-frequency rTMS over the right DLPFC. Interference in the prefrontal-amygdala pathway was speculated to provide a possible neural basis for the findings (Vanderhasselt et al., 2011). Indeed, the close PFC-amygdala link to the brain's stress axis offers a neurobiological framework for unifying the cortisol findings by Baeken et al. [ 2011] and the effects on attentional bias reported by Vanderhasselt et al. [ 2011]. In another attentional bias study, the lateralized role of the left and right DLPFC in early threat processing was examined in low- and high-anxious participants using a single-pulse online TMS protocol to interfere with cortical processing (Sagliano, D'Olimpio, Panico, Gagliardi & Trojano, 2016). Here, left DLPFC interference in high-anxious participants increased individual attentional bias to threat cues. By contrast, low-anxious participants showed an attentional bias away from threat cues in response to TMS-induced interference with left-sided cortical processing (Sagliano et al., 2016). Similarly, Guo, Calver, Soornack and Bourke [2020] found that single-pulse TMS to the visual cortex disrupted emotion recognition of angry faces more in low-trait anxiety participants than in high-trait anxiety participants.
Research in clinical settings also provides evidence for affective state-dependent effects. For example, an open-label adjuvant study investigated the antidepressant effects of bilateral high-frequency rTMS of the prefrontal cortex in medication-resistant patients with a major depressive episode (Downar et al., 2014). Higher levels of baseline anhedonia (e.g., loss of pleasure and interest) were observed in patients who failed to respond to rTMS as compared to patients who showed an antidepressant effect. Moreover, higher anhedonia levels were associated with reduced resting-state functional connectivity in the brain's reward circuit, suggesting a potential neurobiological link between anhedonic state and antidepressant response to rTMS. Berlim, McGirr, Beaulieu, Van den Eynde and Turecki [2013] reported a positive association between extraversion and treatment outcome of high-frequency TMS in major depressive disorder (MDD) patients, while neuroticism had no influence. In another study, the role of baseline anxiety, depression and impulsivity on the efficacy of adjuvant intermittent theta burst stimulation (iTBS) treatment to reduce craving was explored in a group of patients with methamphetamine use disorder (Chen et al., 2020). Lower levels of anxiety and depression and higher levels of non-planning impulsivity were associated with a higher probability of a positive treatment outcome to iTBS. Interestingly, this finding fits results showing that the effects of iTBS on the speed of cortisol response recovery after the Trier Social Stress Test depend on the individual tendency to use the maladaptive rumination style ‘depressive brooding’ (De Witte et al., 2020). Altogether, these studies suggest that the interactions between affective state and TMS can be informative about research in healthy volunteers as well as treatment of patients.
## Affective state-dependency in tES research
Following the empirical confirmation that weak electric currents applied to the scalp induce changes in cortical excitability (Nitsche & Paulus, 2000), transcranial direct current stimulation (tDCS) has become a widely used technique to modulate brain physiology and behaviour in healthy volunteers. Despite its significant value to brain research, large individual differences in responsivity have been observed. These can in part be explained by, for instance, variation in scalp-to-cortex distances, neuro-anatomical features (e.g., gyral folding), electrode montages, and physiological states (Bradley et al., 2022; Opitz, Paulus, Will, Antunes & Thielscher, 2015). However, the role of affective state on the effects of tDCS provides another point of entry for accounting for intra- and interindividual variability. Evidence for such an association comes from studies that explored the effects of tDCS on motivation and emotion (Table 2).Table 2Overview of reviewed studies examining the interaction between tES effects and affective states. Table 2StudySampletES detailsOutcome variable(s)Affective state measurementFindingAbend et al. [ 2019]$$n = 16$$ (7F/9 M); healthy volunteers; Mage = 25.6, SD = 2.5tDCS; 15 min, 1.5 mA; online; anode centrally over forehead, cathode below inion; active or sham; within-subjects designResponse to emotional and neutral video clipsBDI-IIStronger tDCS-induced increase in sgACC activity during negative video clips associated with higher baseline depressive symptomsAnkri et al. [ 2020]$$n = 69$$ (all female); healthy volunteers; Mage=24, SD=2tDCS, 20 min, 2 mA, anode placed right frontal (F4-AF4)/cathode placed on Cz, or shamWorking memory performance (n-back task)High stress condition (TSST) vs. low stress condition (control)Reduced accuracy following tDCS in high vs. low stress conditionEsposito et al. [ 2022]$$n = 18$$ (8 M/10F); healthy volunteers; Mage = 24, SD = 4tDCS, 17 min, 1 mA, anode placed left frontal (F3)/cathode placed on right supraorbital area, or shamReaction times (RT) and pupil dilation in auditory oddball taskSTAI-stateHigher state anxiety scores and reduced pupillary response associated with slower RT following tDCSGilam et al. [ 2018]$$n = 25$$ (15F/10 M); healthy volunteers; Mage = 26.16, SD =3.631.5 mA anodal tDCS, 22 min, offline (during provocation); anode over central forehead, cathode on right shoulder; active or sham; within-subjects designAggressive behaviour towards fictional opponent in TAP following anger-infused ultimatum game (aiUG)Emotion regulation strategies: Trait suppression and trait reappraisalHigher tendency for emotion suppression associated with larger tDCS-induced reduction in anger following provocation; no effect of participants who reported a stronger tendency to use suppression as an emotion regulation strategy, showed a greater effect of reappraisalHortensius et al. [ 2012]$$n = 60$$; healthy volunteers; no age and sex distributions reported2 mA tDCS; 15 min offline; bilateral DLPFC (F3/F4 or F4/F3) or sham between participantsAggressive behaviour towards fictional opponent in TAP, following interpersonal insultsSelf-reported anger (following insult minus baseline)Increased aggression in high-anger participants receiving anodal tDCS to the left DLPFCPeña-Gómez et al. [ 2011]$$n = 16$$ (all female); healthy volunteers; Meanage= 22.93, SD = 4.18tDCS, 20 min, 1 mA; anode over left PFC (F3), cathode over right motor cortex (C4), or sham; within-subjects designValence rating of IAPS pictures (positive, negative, neutral)Before and after tDCS: VAS (nervousness, contentment, sadness, hope and annoyance); before tDCS: PANAS, STAI-state, NEO-FFINegative correlation between extraversion and reduced negative ratings of negative (but not positive or neutral) picturesRabipour et al. [ 2018]single-blind: $$n = 52$$ (31F/21 M), Mage = 20.5, SD = 1.9; double-blind: $$n = 38$$ (22F/16 M), Mage = 20.6, SD = 3.4tDCS, 20 min, 2 mA; online; anode over F3, cathode over right supraorbital area; active or sham, between-subjects designN-back working memory taskExpectation priming: High (i.e., tDCS is effective in improving performance) vs. low (i.e., no known benefits of tDCS)Better performance under active tDCS following high vs. low expectation primingRabipour, Vidjen, Remaud, Davidson, & Tremblay, 2019N = 121 (88F/33 M); healthy volunteers; Mage = 21.1, SD = 3.6tDCS, 20 min, 2 mA; anode over left or right motor cortex (corresponding to preferred or non-preferred hand, respectively), cathode over contralateral supraorbital region; active or sham; between-subjects designGrooved Pegboard TestExpectation priming: High (i.e., tDCS is effective in improving performance) vs. low (i.e., no known benefits of tDCS)No modulation of motor performance as a function of tDCS or expectation primingRay et al. [ 2017]$$n = 18$$ (10F/8 M) obese volunteers; Mage = 22.7, SD = 7.9tDCS, 20 min, 2 mA; cathode over F3, anode over F4; active or sham; within-subjects designFood craving pre- and post-tDCS; actual food intake (amount of calories eaten) post-tDCSBIS-11, DEBQ-RLower food craving following active tDCS in females scoring lower on attentional impulsivity; lower preferred-food intake in males not focused on calorie restriction; lower total food intake in males with higher non-planning impulsivityRay et al. [ 2019]$$n = 74$$ (44F/30 M) obese volunteers; Mage = 19.9, SD = 3.4,tDCS, 20 min, 2 mA; cathode over F3, anode over F4; active or sham; within-subjects designFood craving pre- and post-tDCS; actual food intake (amount of calories eaten) post-tDCSManipulation of expectation of behavioural modulation of tDCS; BIS-11, DEBQ-R, PEMS, BESExpecting positive effect of tDCS decreased food craving and caloric intake, regardless of tDCS condition (active/sham); no effect of questionnaire scoresRiddle et al. [ 2022]$$n = 82$$ (66F/16 M); MDD patients ($$n = 41$$) and healthy volunteers ($$n = 41$$); age: 18–65 yearstACS at individual alpha frequency, 40 min, 1 mA zero-to-peak, bifrontal (F3/F4, return electrode: Cz) or shamAlpha power in resting-state and during emotionally salient IAPS picture presentationsBDI-IItACS reduced the left frontal alpha power increase during resting-state and positive picture viewing in the MDD group; higher baseline depression scores were associated with a stronger reduction in alpha power following tACSRiddle et al. [ 2022]$$n = 3$$ (all female); PMDD patients; age not reported10-Hz tACS, 40 min, 1 mA zero-to-peak, bifrontal (return electrode not reported)Alpha power in resting-stateLate luteal phase (PMDD symptoms high) vs. follicular phase (PMDD symptoms low) of menstrual cycleIn the late luteal phase, tACS increased midline frontal alpha power in all three participants, while in the follicular phase, tACS had no consistent effect on alpha powerSagliano et al. [ 2017]$$n = 40$$ (all females); healthy volunteers; Mage = 22.95, SEM = 0.48tDCS; 15 min, 1 mA; offline; bilateral prefrontal montage (F3/F4); left-anodal/right-cathodal, left-cathodal/right-anodal, or sham; within-subjects designExogenous cueing task with threatening and non-threatening IAPS picturesSTAIFollowing right-cathodal/left-anodal tDCS, higher attentional capture by threatening stimuli for individuals scoring lower on trait anxiety, and longer attentional holding by threatening stimuli for individuals scoring higher on trait anxietyWang et al. [ 2022]$$n = 70$$ (all female); healthy volunteers; Mage = 19.6, SD = 1.55tDCS, 20 min, 1.5 mA; anode over right PFC (F4), cathode over left PFC (F3); active or sham; between-subjects designPerformance on AUT and RAT before and after stress inductionSTAI, BDI-II, PANASTSST before tDCSLower tDCS-induced performance impairment on flexibility component of AUT mediated by state anxiety; no effects of other measuresAUT = Alternative Uses Task; BAI = Beck Anxiety Inventory; BES = Binge Eating Scale; BIS-11 = Barratt Impulsiveness Scale 11; DEBQ-R = Dutch Eating Behaviour Questionnaire-Restraint; F = Females; IAPS = International Affective Picture System; M = Males; M = Mean; MDD = Major Depressive Disorder; NEO-FFI = shortened version of NEO Personality Inventory; PANAS = Positive and Negative Affect Schedule; PEMS = Palatable Eating Motives Scale; PMDD = Premenstrual Dysphoric Disorder; PSAP = Point Subtraction Aggression Paradigm; RAT = Remote Associations Task; SD = Standard Deviation; SEM = Standard Error of the Mean; sgACC = subgenual Anterior Cingulate Cortex; STAI = State-Trait Anxiety Inventory; tACS = transcranial Alternating Current Stimulation; TAP = Taylor Aggression Paradigm; tDCS = transcranial Direct Current Stimulation; TSST = Trier Social Stress Test; VAS = Visual Analogue Scale.
In a sham-controlled study, tDCS over the frontal cortex was combined with a revised Taylor aggression paradigm in healthy volunteers (Hortensius, Schutter & Harmon-Jones, 2012). State anger was manipulated by presenting participants with insulting feedback from a fictional person. Results showed no main effect of tDCS, but when levels of induced state anger were taken into account, tDCS was found to increase the administration of noise blasts to the fictional person. Behavioural effects of tDCS on anger in the anger-infused ultimatum game have also been shown to depend on individuals’ emotion regulation style as shown by a cross-over sham-controlled double-blind study (Gilam et al., 2018). Individuals with a tendency to suppress rather than reappraise negative emotions showed decreased anger under active relative to sham tDCS.
The importance of considering individual differences is further underlined by a study showing that individuals scoring high on the introversion personality dimension are more susceptible to the modulatory effects of tDCS on emotional reactivity than individuals scoring high on extraversion (Peña-Gómez, Vidal-Piñeiro, Clemente, Pascual-Leone & Bartrés-Faz, 2011). In a more recent sham-controlled tDCS study on stress-induced creativity in healthy volunteers, mediation analysis showed that state anxiety explained more than $60\%$ of the stress-reduced performance decrement following active as compared to sham bilateral prefrontal tDCS (Wang et al., 2022). The modulation of neural excitability of the DLPFC was speculated to allow for more effective top-down regulation of the subcortical brain regions associated with anxiety. Levels of self-reported depression have also been found to predict the effects of tDCS on behaviour (Abend et al., 2019). Compared to sham tDCS, active fronto-cerebellar tDCS reduced perceived negative emotions in an emotion induction task in healthy volunteers. Importantly, individual depression scores during sham tDCS were negatively correlated with the neural response in the subgenual anterior cingulate cortex, while this pattern was reversed during active tDCS. This suggests that individual differences in depressive mood can moderate the effect of tDCS during emotion regulation, with differences becoming more pronounced in individuals with high levels of depressive mood. Moreover, Sagliano, D'Olimpio, Izzo and Trojano [2017] showed that anodal tDCS to the right DLPFC increased attentional bias towards threatening stimuli in high-anxious women, while it increased attentional capture by threatening stimuli in low-anxious women. Recently, Esposito, Ferrari, Fracassi, Miniussi and Brignani [2022] found that anodal tDCS over the left prefrontal cortex improved reaction times in an auditory oddball task relative to baseline performance in healthy volunteers with low self-reported state anxiety. By contrast, individuals with higher state anxiety performed significantly worse under anodal tDCS relative to baseline performance. These findings were corroborated by concomitant pupil dilation measures, leading the authors to emphasise the crucial role of taking baseline arousal into account when anticipating tDCS-induced performance changes. Moreover, it has been shown that administering tDCS over the right prefrontal cortex can improve working memory performance in no-stress conditions, while under stress (Trier Social Stress task), the application of tDCS can impair performance (Ankri, Braw, Luboshits & Meiron, 2020).
In clinical populations, there is increasing evidence for associations between affective state and NIBS-induced effects as well. A study on alpha hyperactivity in MDD patients (Riddle, Alexander, Schiller, Rubinow & Frohlich, 2022) showed that transcranial alternating current stimulation (tACS) at the individual alpha peak frequency (8–12 Hz) attenuated left frontal alpha hyperactivity to a significantly stronger extent in patients with higher MDD symptom severity at baseline. The state-dependency of this effect was later replicated in a case series with three women with premenstrual dysphoric disorder, who showed a significant increase in frontal alpha hypoactivity when alpha-tACS was applied in the luteal phase compared to the follicular phase of their menstrual cycle, for example, when symptom severity is highest (Riddle, Rubinow and Frohlich, 2022). In another study on obesity and food intake using a food photo craving task, the effects of bipolar tDCS of the DLPFC depended on sex and one's ability to control impulses (Ray et al., 2017). More specifically, tDCS significantly reduced food craving in obese females with lower attentional impulsivity as assessed with the self-report Barratt Impulsiveness Scale. By contrast, for obese males, tDCS significantly reduced food intake in individuals with higher non-planning impulsivity scores. Despite the small sample size (10 female and 8 male participants), these findings suggest an association between sex and trait impulsivity on the magnitude of the tDCS response. In a between-participant design, a follow-up study investigated to what extent expectation priming of the tDCS effect modulated obesity-related eating and craving patterns (Ray et al., 2019). While all volunteers were told at the beginning of the study that active tDCS had been shown to reduce craving and caloric intake, the participant-specific instruction (i.e., whether they would receive active or sham tDCS) as well as the actual stimulation administered (active or sham) were manipulated. Results showed no tDCS-induced effect, but participants being told to receive beneficial tDCS (regardless of the actual stimulation condition) reported craving reduced by about $10\%$ and caloric intake by almost $40\%$. These findings suggest that prior subjective expectations of the efficacy of tDCS are a powerful psychological mechanism potentially obscuring any (weak) stimulation effect. However, Rabipour, Wu, Davidson and Iacoboni [2018] found an interaction of expectation priming (i.e., being told that tDCS improves or impairs performance, respectively) and actual stimulation (active or sham prefrontal tDCS) on performance in an n-back task. That is, individuals receiving active tDCS who were told that tDCS had a negative effect on memory performance indeed performed worse than participants receiving active tDCS who were primed positively. Interestingly, these findings could not be replicated with tDCS over bilateral motor cortex aimed at improving motor performance (Rabipour, Vidjen, Remaud, Davidson, & Tremblay, 2019), suggesting the importance of context in expectation-related effects. Collectively, these findings suggest that aside from individual differences in state and trait emotions, subjective expectations of the effects of stimulation may influence the effect of NIBS on behaviour. Crucially, the mentioned studies targeted different domains and brain regions, and the extent to which positive expectations influence the effects of NIBS on performance warrants further research. Nonetheless, assessing or even actively manipulating individuals’ expectations on the efficacy of the stimulation holds promise of augmenting more subtle NIBS effects.
## Linking affective state to neural excitability
Affective state-dependency of NIBS may at least in part be explained by the link between affective state and cortical excitability, which influences the susceptibility to TMS and tES. Cortical excitability can be indexed by recording the motor evoked potential (MEP) from different finger muscles following a suprathreshold TMS pulse administered over the motor cortex. Associations between resting-state cortical excitability and affective states have been extensively studied. For example, a previous study has shown that asymmetries in left and right cortical excitability are correlated with approach and avoidance-related motivational tendencies (Schutter, De Weijer, Meuwese, Morgan & Van Honk, 2008). Moreover, anticipatory anxiety and exposure to threat-related stimuli were reported to increase cortical excitability levels as indexed by higher MEP amplitudes (Coombes et al., 2009; Hortensius, de Gelder, & Schutter, 2016; Oathes et al., 2008; Schutter et al., 2008). Importantly, as the MEP in response to single-pulse TMS applied to the motor cortex results from signal propagation along the cortico-spinal tract to the motor endplates, it represents only a proxy for cortical excitability. A more direct correlate of neural excitability can be obtained by applying so-called paired-pulse TMS. In this protocol, a conditioning TMS pulse is preceded by a suprathreshold test pulse to the primary motor cortex (for a review see Hallett, 2007). Paired-pulse TMS allows for studying intracortical inhibitory and facilitatory process underlying the MEP. Higher levels of neuroticism have been associated with increased resting-state cortical excitability caused by lower intracortical inhibition in healthy volunteers (Wassermann, Greenberg, Nguyen & Murphy, 2001). Moreover, novelty seeking/approach-motivation was negatively correlated with intracortical inhibition in social anxiety disorder patients (Pallanti et al., 2010), while higher anxiety symptom severity was linked to higher intracortical facilitation in patients with generalised anxiety disorder (Li et al., 2017). Interestingly, in patients with knee osteoarthritis, higher levels of self-reported anxiety and pain perception were correlated with lower intracortical facilitation and higher intracortical inhibition (Simis et al., 2021). In contrast to the relationship with these affective measures, the pain experienced in specific movement situations, such as walking, was associated with higher intracortical facilitation and lower intracortical inhibition. The complex pattern of relations may suggest differential contributions of affective state and situational pain to cortical excitability in these patients (Simis et al., 2021).
In sum, previous research indicates that affective states are linked to cortical excitability levels, thereby providing a psychobiological basis for interactions between affective states and NIBS interventions.
## Limitations and challenges
Different findings have been found for self-reported state and trait anxiety as assessed by Spielberger's State and Trait Anxiety Inventory (STAI). For example, Vanderhasselt et al. [ 2011] found associations with state anxiety, whereas Sagliano et al. [ 2016, 2017] observed a moderating effect of trait, but not state anxiety on NIBS-induced outcomes. This begs the question whether timepoint-specific (i.e., state) or persistent (i.e., trait) measures of individual affective states are more promising in accounting for interindividual variability in NIBS outcomes. While trait and state anxiety of the STAI are positively correlated both during baseline as well as during threatening situations (Leal, Goes, da Silva & Teixeira-Silva, 2017), there is some evidence to suggest that the trait scale of the STAI also measures depression and general negative affect (Bieling, Antony & Swinson, 1998). These findings may point towards the relevance of selecting affective measures tailored to the experimental design and research question.
Self-report questionnaires are not without shortcomings and can include response biases (e.g., social desirability), and a lack of a person's ability for introspection can have a negative impact on the reliability and validity of the measurement. Differences in construct validity of self-report scales and questionnaires as well as contextual factors like subject's expectations and motivation to participate could therefore potentially explain inconsistencies observed in the literature so far. Higher levels of anxiety have, for instance, been associated both with increased and decreased intracortical facilitation, but in the first case the association was based on the Hamilton anxiety rating scale in generalized anxiety patients (Li et al., 2017) and in the latter case on the Hospital Anxiety and Depression Scale in arthritis patients (Simis et al., 2021). The relationship between affective state and NIBS outcome may seem inconsistent when different measurement instruments are used or different populations are tested. By consistently surveying both state and trait levels of affect in the context of NIBS application, future studies could assess this discrepancy and contribute to a better understanding of the driving force behind affective state-dependency.
It is important to emphasise that we do not aim to equate affective state with disease-specific predictors (e.g., symptom severity in depression to foresee treatment success). Instead, we propose that fluctuating experiences of, for example, anxiety or anger, which are specific to a given stimulation session, may be more informative to anticipate the observed NIBS-induced effect. Crucially, based on the reviewed studies, such a pattern has been observed in both clinical and non-clinical samples, suggesting that it could be a generic mechanism not restricted to specific (psycho)pathologies. Still, it needs to be stressed that assessing individual affective states to better explain NIBS effects has predominantly been done in a post-hoc manner so far. Arguably, this increases the risk of false positives. It is therefore suggested to systematically investigate the predictive value of affective states on the direction and size of NIBS effects to scrutinize whether this concept is tenable. In addition, assessing psychological states beyond cross-sectional self-reports may include longitudinal follow-ups to monitor changes in psychological states and adding behavioural correlates by administering psychological tasks may further contribute to the validity of affective states.
Furthermore, the studies reviewed here included measures of either natural fluctuations (e.g., self-reported anxiety across several time points) or experimentally induced (e.g., self-reported anxiety after a stress test) affective states (see also Tables 1 and 2). It is not unreasonable to assume that affective states caused by experimental manipulation show a different relation to the effects of NIBS as compared to affective states acquired during rest prior to NIBS. However, the extent to which these states within and across individuals differentially affect NIBS remains unclear and should be addressed empirically. Also, the vast majority of studies report findings from single experimental sessions. How ‘natural’ or NIBS-induced changes in psychological states relate to long-term effects in the treatment of disorders is an important avenue for further research. To the best of our knowledge, there is no strong evidence yet that changes in psychological states during multiple sessions lead to more predictable outcomes of NIBS.
Lastly, based on the limited existing evidence it is not yet viable to determine whether effects of affective state reflect, for example, arousal and/or are driven by valence (e.g., positive or negative). In light of improving the predictive value of affective state-dependency, it may be worthwhile to try to disentangle these two aspects empirically. For example, anxiety and anger are both affective states associated with increased arousal, but the specific emotional experience of the two states is different. At this point it remains unclear whether the measured affective state needs to be aligned with the study outcome measure, or whether a generic arousal measurement suffices. Gaining more insight into this association may shed more light on the link between affective states and NIBS effects prospectively. Administering validated questionnaires, such as a shortened version of the Profile of Mood States (Shacham, 1983), the Behavioural Inhibition (BIS) and Activations Scales (BAS) (Carver & White, 1994), or the State-Trait Anxiety and Anger Inventory (Spielberger, Sydeman, Owen, & Marsh, 1999) allows to quickly and easily assess the emotional and motivational aspects of affective states prior to NIBS. This way, it becomes feasible to investigate potential interactions between individual differences in affective states and effects of interventions in healthy as well as treatments in clinical populations. Similar associations have been reported in the context of other forms of treatment. For example, positive emotion scores on the center for Epidemiologic Studies Depression scale were found to correlate with higher overall functional status, including motor and cognitive performance at three-month follow-up after adjustment for relevant risk factors in stroke patients who underwent medical rehabilitation (Ostir, Berges, Ottenbacher, Clow, & Ottenbacher, 2008). In another study, baseline BAS scores were inversely correlated with the six-month course of depression (McFarland, Shankman, Tenke, Bruder, & Klein, 2006). The correlation was controlled for clinical depressive symptoms, indicating that the BAS score, as a subjective measure of dispositional tendencies associated with reward and appetitive motivation, explains unique variance and may potentially contribute to the antidepressant efficacy of NIBS. These studies suggest that the use of questionnaires may have added value in predicting responses to NIBS by taking into account individual differences in affective states.
## Conclusion
Unravelling the physiological mechanisms by which NIBS techniques establish their effects in the brain remains critical for addressing basic neuroscientific research questions and developing effective neuromodulation-based treatments. In addition to studying neural states directly, evaluation of affective states could be a cost-effective way to obtain valuable non-reductive readouts which together with neurophysiological markers offer important mediators of the effects of NIBS both in basic research as well as in applied settings. Here, empirical evidence was presented to illustrate this idea and to raise the possibility that affective states can have added value in explaining intra- and interindividual study-outcome variability. Systematic assessment of affective states in experimental and clinical research may contribute to further increasing the sensitivity and specificity of tES- and TMS-based interventions.
## Funding
This work was supported by the Dutch Research Foundation (NWO, VI.C.181.005).
## Declaration of competing interests
The authors have no competing interests to declare.
## References
1. Abend R., Sar-el R., Gonen T., Jalon I., Vaisvaser S., Bar-Haim Y.. **Modulating emotional experience using electrical stimulation of the medial-prefrontal cortex: A preliminary tDCS-fMRI study**. *Neuromodulation: Technology at the Neural Interface* (2019) **22** 884-893. DOI: 10.1111/ner.12787
2. Ankri Y.L.E., Braw Y., Luboshits G., Meiron O.. **The effects of stress and transcranial direct current stimulation (tDCS) on working memory: A randomized controlled trial**. *Cognitive, Affective, & Behavioral Neuroscience* (2020) **20** 103-114. DOI: 10.3758/s13415-019-00755-7
3. Baeken C., Vanderhasselt M.A., De Raedt R.. **Baseline ‘state anxiety’ influences HPA-axis sensitivity to one sham-controlled HF-rTMS session applied to the right dorsolateral prefrontal cortex**. *Psychoneuroendocrinology* (2011) **36** 1. DOI: 10.1016/j.psyneuen.2010.06.006
4. Berlim M.T., McGirr A., Beaulieu M.-.M., Van den Eynde F., Turecki G.. **Are neuroticism and extraversion associated with the antidepressant effects of repetitive transcranial magnetic stimulation (rTMS)? An exploratory 4-week trial**. *Neuroscience Letters* (2013) **534** 306-310. DOI: 10.1016/j.neulet.2012.12.029
5. Bieling P.J., Antony M.M., Swinson R.P.. **The State–trait anxiety inventory, trait version: Structure and content re-examined**. *Behaviour Research and Therapy* (1998) **36** 777-788. DOI: 10.1016/S0005-7967(98)00023-0
6. Borgomaneri S., Battaglia S., Garofalo S., Tortora F., Avenanti A., di Pellegrino G.. **State-dependent TMS over prefrontal cortex disrupts fear-memory reconsolidation and prevents the return of fear**. *Current Biology* (2020) **30** 3672-3679. DOI: 10.1016/j.cub.2020.06.091
7. Bradley C., Nydam A.S., Dux P.E., Mattingley J.B.. **State-dependent effects of neural stimulation on brain function and cognition**. *Nature Reviews. Neuroscience* (2022) **23** 459-475. DOI: 10.1038/s41583-022-00598-1
8. Carver C.S., White T.L.. **Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: the BIS/BAS scales**. *Journal of Personality and Social Psychology* (1994) **67** 319-333. DOI: 10.1037/0022-3514.67.2.319
9. Chen T., Su H., Jiang H., Li X., Zhong N., Du J.. **Cognitive and emotional predictors of real versus sham repetitive transcranial magnetic stimulation treatment response in methamphetamine use disorder**. *Journal of Psychiatric Research* (2020) **126** 73-80. DOI: 10.1016/j.jpsychires.2020.05.007
10. Coombes S.A., Tandonnet C., Fujiyama H., Janelle C.M., Cauraugh J.H., Summers J.J.. **Emotion and motor preparation: A transcranial magnetic stimulation study of corticospinal motor tract excitability**. *Cognitive, Affective, & Behavioral Neuroscience* (2009) **9** 380-388. DOI: 10.3758/CABN.9.4.380
11. De Witte S., Baeken C., Pulopulos M.M., Josephy H., Schiettecatte J., Anckaert E.. **The effect of neurostimulation applied to the left dorsolateral prefrontal cortex on post-stress adaptation as a function of depressive brooding**. *Progress in Neuro-Psychopharmacology and Biological Psychiatry* (2020) **96**. DOI: 10.1016/j.pnpbp.2019.109687
12. Downar J., Geraci J., Salomons T.V., Dunlop K., Wheeler S., McAndrews M.P.. **Anhedonia and reward-circuit connectivity distinguish nonresponders from responders to dorsomedial prefrontal repetitive transcranial magnetic stimulation in major depression**. *Biological Psychiatry* (2014) **76** 176-185. DOI: 10.1016/j.biopsych.2013.10.026
13. Esposito M., Ferrari C., Fracassi C., Miniussi C., Brignani D.. **Responsiveness to left-prefrontal tDCS varies according to arousal levels**. *European Journal of Neuroscience* (2022) **55** 762-777. DOI: 10.1111/ejn.15584
14. Feurra M., Pasqualetti P., Bianco G., Santarnecchi E., Rossi A., Rossi S.. **State-dependent effects of transcranial oscillatory currents on the motor system: what you think matters**. *Journal of Neuroscience* (2013) **33** 17483-17489. DOI: 10.1523/JNEUROSCI.1414-13.2013
15. Gilam G., Abend R., Gurevitch G., Erdman A., Baker H., Ben-Zion Z.. **Attenuating anger and aggression with neuromodulation of the vmPFC: A simultaneous tDCS-fMRI study**. *Cortex; A Journal Devoted to the Study of the Nervous system and Behavior* (2018) **109** 156-170. DOI: 10.1016/j.cortex.2018.09.010
16. Guo K., Calver L., Soornack Y., Bourke P.. **Valence-dependent disruption in processing of facial expressions of emotion in early visual cortex—A transcranial magnetic stimulation study**. *Journal of Cognitive Neuroscience* (2020) **32** 906-916. DOI: 10.1162/jocn_a_01520
17. Hallett M.. **Transcranial magnetic stimulation: A primer**. *Neuron* (2007) **55** 187-199. DOI: 10.1016/j.neuron.2007.06.026
18. Hartwigsen G., Silvanto J.. **Noninvasive brain stimulation: Multiple effects on cognition**. *The Neuroscientist* (2022). DOI: 10.1177/10738584221113806
19. Hortensius R., de Gelder B., Schutter D.J.L.G.. **When anger dominates the mind: Increased motor corticospinal excitability in the face of threat**. *Psychophysiology* (2016) **53** 9. DOI: 10.1111/psyp.12685
20. Hortensius R., Schutter D.J.L.G., Harmon-Jones E.. **When anger leads to aggression: Induction of relative left frontal cortical activity with transcranial direct current stimulation increases the anger–aggression relationship**. *Social Cognitive and Affective Neuroscience* (2012) **7** 3. DOI: 10.1093/scan/nsr012
21. Janssens S.E.W., Sack A.T.. **Spontaneous fluctuations in oscillatory brain state cause differences in transcranial magnetic stimulation effects within and between individuals**. *Frontiers in Human Neuroscience* (2021) **15**. DOI: 10.3389/fnhum.2021.802244
22. Kasten F.H., Herrmann C.S.. **The hidden brain-state dynamics of tACS aftereffects**. *NeuroImage* (2022) **264**. DOI: 10.1016/j.neuroimage.2022.119713
23. Leal P.C., Goes T.C., da Silva L.C.F., Teixeira-Silva F.. **Trait vs. State anxiety in different threatening situations**. *Trends in Psychiatry and Psychotherapy* (2017) **39** 147-157. DOI: 10.1590/2237-6089-2016-0044
24. Li C.-.T., Lu C.-.F., Lin H.-.C., Huang Y.-.Z., Juan C.-.H., Su T.-.P.. **Cortical inhibitory and excitatory function in drug-naive generalized anxiety disorder**. *Brain Stimulation* (2017) **10** 604-608. DOI: 10.1016/j.brs.2016.12.007
25. McFarland B.R., Shankman S.A., Tenke C.E., Bruder G.E., Klein D.N.. **Behavioral activation system deficits predict the six-month course of depression**. *Journal of Affective Disorders* (2006) **91** 229-234. DOI: 10.1016/j.jad.2006.01.012
26. Nitsche M.A., Paulus W.. **Excitability changes induced in the human motor cortex by weak transcranial direct current stimulation**. *The Journal of Physiology* (2000) **527** 3. DOI: 10.1111/j.1469-7793.2000.t01-1-00633.x
27. Northoff G.. **From emotions to consciousness – A neuro-phenomenal and neuro-relational approach**. *Frontiers in Psychology* (2012) **3** 303. DOI: 10.3389/fpsyg.2012.00303
28. Oathes D.J., Bruce J.M., Nitschke J.B.. **Worry facilitates corticospinal motor response to transcranial magnetic stimulation**. *Depression and Anxiety* (2008) **25** 969-976. DOI: 10.1002/da.20445
29. Oosterwijk S., Lindquist K.A., Anderson E., Dautoff R., Moriguchi Y., Barrett L.F.. **States of mind: Emotions, body feelings, and thoughts share distributed neural networks**. *NeuroImage* (2012) **62** 2110-2128. DOI: 10.1016/j.neuroimage.2012.05.079
30. Opitz A., Paulus W., Will S., Antunes A., Thielscher A.. **Determinants of the electric field during transcranial direct current stimulation**. *NeuroImage* (2015) **109** 140-150. DOI: 10.1016/j.neuroimage.2015.01.033
31. Ostir G.V., Berges I.M., Ottenbacher M.E., Clow A., Ottenbacher K.J.. **Associations between positive emotion and recovery of functional status following stroke**. *Psychosomatic Medicine* (2008) **70** 404. DOI: 10.1097/PSY.0b013e31816fd7d0
32. Pallanti S., Borgheresi A., Pampaloni I., Giovannelli F., Bernardi S., Cantisani A.. **Motor cortex excitability correlates with novelty seeking in social anxiety: A transcranial magnetic stimulation investigation**. *Journal of Neurology* (2010) **257** 1362-1368. DOI: 10.1007/s00415-010-5533-4
33. Panksepp J., Biven L.. (2012)
34. Peña-Gómez C., Vidal-Piñeiro D., Clemente I.C., Pascual-Leone Á., Bartrés-Faz D.. **Down-regulation of negative emotional processing by transcranial direct current stimulation: Effects of personality characteristics**. *PloS one* (2011) **6** 7. DOI: 10.1371/journal.pone.0022812
35. Penton T., Catmur C., Banissy M.J., Bird G., Walsh V.. **Non-invasive stimulation of the social brain: The methodological challenges**. *Social Cognitive and Affective Neuroscience* (2022) **17** 1. DOI: 10.1093/scan/nsaa102
36. Rabipour S., Vidjen P.S., Remaud A., Davidson P.S.R., Tremblay F.. **Examining the interactions between expectations and tDCS effects on motor and cognitive performance**. *Frontiers in Neuroscience* (2019) **12** 999. DOI: 10.3389/fnins.2018.00999
37. Rabipour S., Wu A.D., Davidson P.S.R., Iacoboni M.. **Expectations may influence the effects of transcranial direct current stimulation**. *Neuropsychologia* (2018) **119** 524-534. DOI: 10.1016/j.neuropsychologia.2018.09.005
38. Ray M.K., Sylvester M.D., Helton A., Pittman B.R., Wagstaff L.E., McRae T.R.. **The effect of expectation on transcranial direct current stimulation (tDCS) to suppress food craving and eating in individuals with overweight and obesity**. *Appetite* (2019) **136** 1-7. DOI: 10.1016/j.appet.2018.12.044
39. Ray M.K., Sylvester M.D., Osborn L., Helms J., Turan B., Burgess E.E.. **The critical role of cognitive-based trait differences in transcranial direct current stimulation (tDCS) suppression of food craving and eating in frank obesity**. *Appetite* (2017) **116** 568-574. DOI: 10.1016/j.appet.2017.05.046
40. Riddle J., Alexander M.L., Schiller C.E., Rubinow D.R., Frohlich F.. **Reduction in left frontal alpha oscillations by transcranial alternating current stimulation in major depressive disorder is context dependent in a randomized clinical trial**. *Biological Psychiatry* (2022) **7** 302-311. DOI: 10.1016/j.bpsc.2021.07.001
41. Riddle J., Rubinow D.R., Frohlich F.. **Effect of tACS on prefrontal neural activity is menstrual phase dependent in patients with premenstrual dysphoric disorder**. *Brain Stimulation* (2022) **15** 1088-1090. DOI: 10.1016/j.brs.2022.07.052
42. Sagliano L., D'Olimpio F., Izzo L., Trojano L.. **The effect of bicephalic stimulation of the dorsolateral prefrontal cortex on the attentional bias for threat: A transcranial direct current stimulation study**. *Cognitive, Affective, & Behavioral Neuroscience* (2017) **17** 5. DOI: 10.3758/s13415-017-0532-x
43. Sagliano L., D'Olimpio F., Panico F., Gagliardi S., Trojano L.. **The role of the dorsolateral prefrontal cortex in early threat processing: A TMS study**. *Social Cognitive and Affective Neuroscience* (2016) **11** 12. DOI: 10.1093/scan/nsw105
44. Schutter D.J.L.G., De Weijer A.D., Meuwese J.D.I., Morgan B., Van Honk J.. **Interrelations between motivational stance, cortical excitability, and the frontal electroencephalogram asymmetry of emotion: A transcranial magnetic stimulation study**. *Human Brain Mapping* (2008) **29** 574-580. DOI: 10.1002/hbm.20417
45. Schutter D.J.L.G., Hofman D., Honk J.V.. **Fearful faces selectively increase corticospinal motor tract excitability: A transcranial magnetic stimulation study**. *Psychophysiology* (2008) **45** 3. DOI: 10.1111/j.1469-8986.2007.00635.x
46. Schutter D.J.L.G., Hortensius R.. **Brain oscillations and frequency-dependent modulation of cortical excitability**. *Brain Stimulation* (2011) **4** 97-103. DOI: 10.1016/j.brs.2010.07.002
47. Shacham S.. **A shortened version of the Profile of Mood States**. *Journal of Personality Assessment* (1983) **47** 305-306. DOI: 10.1207/s15327752jpa4703_14
48. Siebner H.R., Funke K., Aberra A.S., Antal A., Bestmann S., Chen R.. **Transcranial magnetic stimulation of the brain: What is stimulated? – A consensus and critical position paper**. *Clinical Neurophysiology* (2022). DOI: 10.1016/j.clinph.2022.04.022
49. Silvanto J., Pascual-Leone A.. **State-dependency of transcranial magnetic stimulation**. *Brain Topography* (2008) **21** 1-10. DOI: 10.1007/s10548-008-0067-0
50. Simis M., Imamura M., de Melo P.S., Marduy A., Pacheco-Barrios K., Teixeira P.E.P.. **Increased motor cortex inhibition as a marker of compensation to chronic pain in knee osteoarthritis**. *Scientific Reports* (2021) **11** 1. DOI: 10.1038/s41598-021-03281-0
51. Spielberger C.D., Sydeman S.J., Owen A.E., Marsh B.J.. *of psychological testing for treatment planning and outcomes assessment* (1999) 993-1021
52. Vanderhasselt M.-.A., Baeken C., Hendricks M., De Raedt R.. **The effects of high frequency rTMS on negative attentional bias are influenced by baseline state anxiety**. *Neuropsychologia* (2011) **49** 7. DOI: 10.1016/j.neuropsychologia.2011.03.006
53. Wang Y., Guo X., Wang M., Kan Y., Zhang H., Zhao H.. **Transcranial direct current stimulation of bilateral dorsolateral prefrontal cortex eliminates creativity impairment induced by acute stress**. *International Journal of Psychophysiology* (2022) **171** 1-11. DOI: 10.1016/j.ijpsycho.2021.11.001
54. Wassermann E.M., Greenberg B.D., Nguyen M.B., Murphy D.L.. **Motor cortex excitability correlates with an anxiety-related personality trait**. *Biological Psychiatry* (2001) **50** 377-382. DOI: 10.1016/S0006-3223(01)01210-0
55. Wendt K., Denison T., Foster G., Krinke L., Thomson A., Wilson S.. **Physiologically informed neuromodulation**. *Journal of the Neurological Sciences* (2022) **434**. DOI: 10.1016/j.jns.2021.120121
|
---
title: 'Global and regional prevalence of multimorbidity in the adult population in
community settings: a systematic review and meta-analysis'
authors:
- Saifur Rahman Chowdhury
- Dipak Chandra Das
- Tachlima Chowdhury Sunna
- Joseph Beyene
- Ahmed Hossain
journal: eClinicalMedicine
year: 2023
pmcid: PMC9971315
doi: 10.1016/j.eclinm.2023.101860
license: CC BY 4.0
---
# Global and regional prevalence of multimorbidity in the adult population in community settings: a systematic review and meta-analysis
## Body
Research in contextEvidence before this studyWe searched PubMed, ScienceDirect, and Google Scholar for peer-reviewed papers and research reports on the prevalence of multimorbidity, using the search words 'prevalence' and 'multimorbidity' and similar terms published between January 1, 2000 and December 31, 2021. One meta-analysis combined 68 studies from 1992 to 2017 and showed that the global pooled prevalence of multimorbidity in community settings was $33.1\%$. In 2021, another meta-study focused on articles that investigated people in community settings from Latin America and the Caribbean. Added value of this studyThis research used studies until 2021 to analyze multimorbidity prevalence in community settings worldwide. South America has the highest prevalence of multimorbidity when comparing prevalence estimates across geographic regions. The prevalence difference was obtained across age groups, gender, country and income level, and study periods. For the first time in a subgroup study, we stratified the number of conditions to estimate the prevalence of multimorbidity. Studies that included mental health in the definition of multimorbidity resulted in a high pooled prevalence. Our research also uses statistical techniques to estimate the pooled prevalence of multimorbidity in adults while capturing heterogeneity in the estimates. This study summarizes the available evidence and encourages policymakers to use more standardized methods to reduce the burden of multimorbidity, which is a critical step toward meeting the sustainable development goal (SDG) goal of reducing premature mortality from non-communicable diseases by one-third through prevention and treatment by 2030.Implications of all the available evidenceOur findings show that the landscape of multimorbidity prevalence has increased in the last two decades though it has remained relatively unchanged since 2010, implying a slow reduction in the burden of multimorbidity. About half of the South American adult population had multimorbidity, and thus these countries should take it as a priority agenda to develop more sustainable and integrated models of care. Research like this is crucial as the world tries to balance lowering the expense of multimorbidity on society and improving healthcare outcomes.
## Summary
### Background
Knowing the prevalence of multimorbidity among adults across continents is a crucial piece of information for achieving Sustainable Development Goal 3.4, which calls for reducing premature death due to non-communicable diseases. A high prevalence of multimorbidity indicates high mortality and increased healthcare utilization. We aimed to understand the prevalence of multimorbidity across WHO geographic regions among adults.
### Methods
We performed a systematic review and meta-analysis of surveys designed to estimate the prevalence of multimorbidity among adults in community settings. We searched PubMed, ScienceDirect, Embase and Google Scholar databases for studies published between January 1, 2000, and December 31, 2021. The random-effects model estimated the pooled proportion of multimorbidity in adults. Heterogeneity was quantified using I2 statistics. We performed subgroup analyses and sensitivity analyses based on continents, age, gender, multimorbidity definition, study periods and sample size. The study protocol was registered with PROSPERO (CRD42020150945).
### Findings
We analyzed data from 126 peer-reviewed studies that included nearly 15.4 million people ($32.1\%$ were male) with a weighted mean age of 56.94 years (standard deviation of 10.84 years) from 54 countries around the world. The overall global prevalence of multimorbidity was $37.2\%$ ($95\%$ CI = 34.9–$39.4\%$). South America ($45.7\%$, $95\%$ CI = 39.0–52.5) had the highest prevalence of multimorbidity, followed by North America ($43.1\%$, $95\%$ CI = 32.3–$53.8\%$), Europe ($39.2\%$, $95\%$ CI = 33.2–$45.2\%$), and Asia ($35\%$, $95\%$ CI = 31.4–$38.5\%$). The subgroup study highlights that multimorbidity is more prevalent in females ($39.4\%$, $95\%$ CI = 36.4–$42.4\%$) than males ($32.8\%$, $95\%$ CI = 30.0–$35.6\%$). More than half of the adult population worldwide above 60 years of age had multimorbid conditions ($51.0\%$, $95\%$ CI = 44.1–$58.0\%$). Multimorbidity has become increasingly prevalent in the last two decades, while the prevalence appears to have stayed stable in the recent decade among adults globally.
### Interpretation
The multimorbidity patterns by geographic regions, time, age, and gender suggest noticeable demographic and regional differences in the burden of multimorbidity. According to insights about prevalence among adults, priority is required for effective and integrative interventions for older adults from South America, Europe, and North America. A high prevalence of multimorbidity among adults from South America suggests immediate interventions are needed to reduce the burden of morbidity. Furthermore, the high prevalence trend in the last two decades indicates that the global burden of multimorbidity continues at the same pace. The low prevalence in Africa suggests that there may be many undiagnosed chronic illness patients in Africa.
### Funding
None.
## Evidence before this study
We searched PubMed, ScienceDirect, and Google Scholar for peer-reviewed papers and research reports on the prevalence of multimorbidity, using the search words 'prevalence' and 'multimorbidity' and similar terms published between January 1, 2000 and December 31, 2021. One meta-analysis combined 68 studies from 1992 to 2017 and showed that the global pooled prevalence of multimorbidity in community settings was $33.1\%$. In 2021, another meta-study focused on articles that investigated people in community settings from Latin America and the Caribbean.
## Added value of this study
This research used studies until 2021 to analyze multimorbidity prevalence in community settings worldwide. South America has the highest prevalence of multimorbidity when comparing prevalence estimates across geographic regions. The prevalence difference was obtained across age groups, gender, country and income level, and study periods. For the first time in a subgroup study, we stratified the number of conditions to estimate the prevalence of multimorbidity. Studies that included mental health in the definition of multimorbidity resulted in a high pooled prevalence. Our research also uses statistical techniques to estimate the pooled prevalence of multimorbidity in adults while capturing heterogeneity in the estimates. This study summarizes the available evidence and encourages policymakers to use more standardized methods to reduce the burden of multimorbidity, which is a critical step toward meeting the sustainable development goal (SDG) goal of reducing premature mortality from non-communicable diseases by one-third through prevention and treatment by 2030.
## Implications of all the available evidence
Our findings show that the landscape of multimorbidity prevalence has increased in the last two decades though it has remained relatively unchanged since 2010, implying a slow reduction in the burden of multimorbidity. About half of the South American adult population had multimorbidity, and thus these countries should take it as a priority agenda to develop more sustainable and integrated models of care. Research like this is crucial as the world tries to balance lowering the expense of multimorbidity on society and improving healthcare outcomes.
## Introduction
Multimorbidity has emerged as a significant public health issue in the world. It is typically defined as the presence of two or more chronic conditions at the same time in one individual.1 Multimorbidity has increased in various population groups due to population aging, lifestyle changes, improved socioeconomic conditions, and improved diagnostic capabilities by health services.2, 3, 4 Due to a lack of data from low-income countries and the use of different definitions of multimorbidity, a recent systematic review highlighted the need to estimate the prevalence of multimorbidity and patterns of multimorbidity.5 The high prevalence of multimorbidity has several negative consequences, including a high mortality rate, increased healthcare utilization, and increased healthcare expenses, influencing overall functioning and quality of life.6, 7, 8, 9, 10 According to a recent review and meta-analysis, those with at least two morbidities have a 1.73 times higher risk of death than people without multimorbidity.8 Moreover, healthcare demands and costs of multimorbidity continue to rise as populations age.11 Although few systematic reviews and meta-analyses on multimorbidity in community settings have been published in recent years, these included fewer studies or are restricted to a specific geographic region.12, 13, 14, 15 According to a systematic review and meta-analysis of studies with data collected between 1992 and 2017, the global pooled prevalence of multimorbidity in community settings was $33.1\%$ ($95\%$ confidence interval: 30.0–$36.3\%$).12 This prior study, however, did not look at how multimorbidity patterns changed over time or gave insight into multimorbidity definitions based on the number of conditions.
In recent years, many studies have been conducted to identify the clinical patterns of chronic conditions.14,16, 17, 18, 19 Two systematic reviews on multimorbidity identified depression, hypertension, and diabetes as the most prevalent co-occurring chronic diseases.5,20 Another study of multimorbidity identified cardiovascular and metabolic diseases as the most common diseases, followed by mental health disorders and musculoskeletal conditions.21 *In a* multi-national cross-sectional study of non-institutionalized adults aged 50 and over in Finland, Poland, Spain, China, Ghana, India, Mexico, Russia, and South Africa, hypertension, cataract, and arthritis were the most prevalent comorbid conditions.22 A study conducted in Germany among health-insured individuals aged 65 and older identified three broad multimorbidity patterns–cardiovascular/metabolic disorders, anxiety/depression disorders, and pain/neuropsychiatric disorders.23 It indicates that mental health disorders were prevalent in the studies, so we examined the prevalence of multimorbidity with and without mental health disorders.
These findings provide an explanation for the clinical patterns as well as the burden of multimorbidity that was observed among the studied people. An accurate and up-to-date prevalence estimation is critical to assess the impact of multimorbidity on public health and project effective and integrative interventions to reduce premature death due to multimorbidity. It is challenging to conduct a meta-analysis to estimate a global prevalence as the different studies used a different number of diseases and disease combinations. There is no gold standard for quantifying multimorbidity; definitions of multimorbidity and statistical approaches for evaluating prevalence differ greatly.24, 25, 26, 27, 28 But the trade-off of generating pooled estimate of multimorbidity exceed the drawbacks of the variability in the data. However, the prevalence of multimorbidity was not thoroughly assessed based on geographic regions, country's economic level, age, study periods, and the number of diseases considered for defining multimorbidity.
Given the growing concern about the rising burden of chronic diseases, understanding the prevalence of multimorbidity in the adult population is critical for developing preventive strategies. As a result, we conducted a systematic review and meta-analysis to examine the global and regional prevalence of multimorbidity and changes in multimorbidity prevalence over time among the adult population in community settings.
## Search strategy
We searched PubMed, Google Scholar, Embase and ScienceDirect online databases to select peer-reviewed papers for our systematic review and meta-analysis. We screened observational studies (cross-sectional and baseline in a cohort) to determine the global prevalence of multimorbidity in the adult population in community settings. Our search included articles published in any language between January 2000 and December 2021, which would help minimize data heterogeneity and provide a more precise estimate of global multimorbidity prevalence. The screening was conducted primarily in English, but we also utilized the Google translation tool for article selection. A description of search terms is given in Appendix A. The search results were compiled using Mendeley citation management software. In addition to the database search, we explored references of selected studies and previously published systematic reviews on similar topics to incorporate all potential pertinent articles to construct our summary estimates. The Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) checklist was followed in this study.29 The protocol was registered in the PROSPERO database (CRD42020150945).
## Selection criteria
Our systematic review included studies that [1] defined multimorbidity as having more than one underlying chronic conditions; [2] documented multimorbidity as the outcome of interest; [3] provided the number of participants in the study, with at least 200; [4] defined multimorbidity in the article, with at least five chronic conditions; [5] were observational studies, either cross-sectional or cohort, including adults 18 years and above; [6] published in years 2000–2021; and [7] were conducted in a community setting. Furthermore, only the recent study was considered if more than one study studied the same population. Only prevalence at baseline was included when the design was a cohort. Studies were excluded if they [1] focused only on comorbidity, [2] defined multimorbidity as more than two diseases [3] studied only inpatients or outpatients in hospital and primary care settings, [4] studied institutional population, i.e., people in nursing home, old home etc., [ 5] included acute conditions in the list of conditions, [6] used less than 5 conditions to define multimorbidity, or [7] were qualitative, interventional studies, opinion articles, conference presentations, books, letters, editorials, reviews, dissertations/theses, or abstracts.
## Data extraction and quality assessment
Using Covidence, two independent reviewers (S.R.C. and D.C.D.) screened the articles. The reviewers examined successively the titles, abstracts, and full texts of all possibly relevant articles identified by our searches. The differences in article selection and data extraction were handled by consensus and, if necessary, discussion with another reviewer (A.H.). Two independent reviewers (S.R.C. and T.C.S.) created a data-extraction form to establish the type of information to be extracted. The reviewers (S.R.C. and T.C.S.) recorded pertinent data on the name of the first author, study settings (e.g., country, year of publication, study period (start-end year), region), and study conduct (e.g., study design, population age and male percentage, number of study participants, data sources, method of ascertainment of morbidity, and minimum number of conditions included in multimorbidity), prevalence of multimorbidity, and number of participants with multimorbidity from the published article only. We further stratified the articles based on the country's income level (World Bank classification by income, GNI per capita).30 Moreover, the study participants were cross tabulated by age group and gender, and multimorbidity was documented whenever possible. If the prevalence of multimorbidity was not directly given, it was manually computed from the data supplied in the articles. In studies providing longitudinal prevalence estimates over a period, we utilized baseline prevalence. After settling any differences, the two reviewers (S.R.C. and T.C.S.) independently extracted the data, discussed the inputs, and revised the extracted data. Unresolved issues were resolved by involving a third reviewer (J.B.).
The Newcastle-Ottawa Scale (NOS), the tool for assessing the quality of non-randomized research, was used to determine the risk of bias for individual studies.31 The eight items of NOS are categorized into three domains of potential bias, namely “selection (representativeness of the sample, sample size, non-respondents, ascertainment of the exposure),” “comparability (the subjects in different outcome groups are comparable, based on the study design or analysis; and confounding factors are controlled),” and “outcome (assessment of the outcome and statistical test)”.31, 32, 33 A few points on the NOS were modified to be relevant to our research question (Supplementary File 1). The articles' methodological stringency, lucidity, and clarity are reflected in the subjective scores. However, we did not eliminate any articles based on their quality scoring. A study can be given one star for each item within the selection and outcome categories. For comparability, a maximum of two stars can be awarded. Thus, a cross-sectional study can be awarded a maximum of 10 stars (10 points), and a cohort study can be awarded a maximum of 9 stars (9 points). Overall, the studies were categorized as “low risk of bias (8–10 stars)”, “moderate risk of bias (6–7 stars)”, and “high risk of bias (0–5 stars)”. Two independent reviewers (S.R.C. and D.C.D.) assessed the quality of the included studies, and the discrepancies were resolved with discussion with the third reviewer (A.H.). The PRISMA statement consists of a 27-item checklist given in Supplementary File 2.
## Statistical analysis
The statistical analysis was performed using meta and metafor packages in the R statistical software (version 4.1.1). Multimorbidity prevalence was estimated as the ratio of the number of people with multimorbidity (numerator) and sample size (denominator). The numerator was derived from the percentage of people with multimorbidity when the numerator was not available. We obtained the pooled prevalence (with $95\%$ CIs) of multimorbidity among the overall population from all studies and subgroups. The pooled prevalence was estimated using a random-effects model that allows the actual effect size to vary from study to study. The calculated proportion from each study and the combined effect estimate with $95\%$ CI were represented graphically using forest plots. We assessed potential publication bias by visually observing the symmetry of funnel plots and using Egger's test. The I2 statistic was used to quantify heterogeneity across the selected studies. The I2 statistic indicates the proportion of overall variation across studies due to heterogeneity rather than chance. Subgroup analysis was carried out to determine the pooled prevalence for each group and look for potential explanations for the heterogeneity. Geographical region (Africa, Asia, Europe, North America, Oceania, and South America); WB/WHO income region (High, Upper-middle, Low- and Lower-middle); Study design (Cross-sectional, Cohort); Multimorbidity (5–9 conditions, 10–19 conditions, ≥20 conditions); Mental health included in the multimorbidity definition (Yes or No); Age groups of study participants (≥30 years, ≥40 years, ≥50 years, ≥60 years) and Gender (male and female) were considered for sub-group analysis. We conducted a trend analysis to see the global multimorbidity prevalence over time (2000–2021). We also conducted sensitivity analyses to assess the findings' robustness in consideration of sample size, multimorbidity prevalence, multimorbidity definitions based on the number of conditions studied, and NOS overall quality of the studies. Two-sided $P \leq .05$ was considered statistically significant.
## Role of the funding source
There was no funding available for this study. All of the study's data was accessible to all of the authors, and the corresponding author had responsibility for publication.
## Identification and selection of studies
A flowchart of the literature search to select the relevant articles is summarized in the PRISMA format and is presented in Fig. 1. The initial search retrieved 8003 studies from the three pre-specified databases. After excluding the duplicates, the titles and abstracts were screened for a further selection of probable articles. Subsequently, the investigators selected 376 articles based on eligibility criteria for full-text review. By manual searching through the included papers’ reference lists and reference lists of previous systematic reviews on similar topics, 12 studies were considered for scrutiny, resulting in the total number of potential articles being 388. After excluding 262 studies in full-text review, finally, 126 studies with a total of 15,400,421 (approximately 15.4 million) people were included in the systematic review and meta-analysis. Sample sizes in the studies range from 264 to 3,759,836.3,27,34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155Fig. 1PRISMA flow diagram for study selection.
## Characteristics of the studies
Table 1 shows the characteristics of the included studies. The 126 population-based studies were conducted across 54 countries. Six of the 126 research included were carried out in multiple countries. The majority of the studies ($$n = 47$$) were conducted in Asia, followed by Europe ($$n = 27$$), South America ($$n = 19$$), Africa ($$n = 10$$), North America ($$n = 14$$), Oceania ($$n = 6$$), and various continents ($$n = 3$$). Between 2000 and 2021, 53 studies were carried out in high-income countries (HICs), 48 in upper middle-income countries (UMICs), and 24 in low- and lower-middle-income countries (Low- and LMICs). Most of the studies (121 studies) were cross-sectional in design, and the remaining five had a cohort design, from which we used data from the baseline assessment. When defining multimorbidity, 37 studies looked at 5–9 diseases, 64 studies at 10–19 diseases, and 24 studies at more than 20 diseases. Table 1Characteristics of the included studies in the meta-analysis (according to the order of year).Author [Ref]CountryWB income countryYear of publicationStudy periodStudy designSource of dataAscertainment of morbiditiesaSample sizeAge, yMean/median age, yGender (male %)Number of conditions includedPrevalence, %Dhungana et al. ,34NepalLow- or LMIC20212016–2018Cross-sectionalNCD (non-communicable diseases) survey 2018 in NepalObjective8931≥2046.742.2714.0Zhang et al. ,35ChinaUMIC20212017Cross-sectionalBeijing Longitudinal Study of Aging (BLSA)Self-reported1837≥60NA44.31253.2Keetile et al. ,36BotswanaUMIC20202016Cross-sectionalSurvey on Chronic Non-Communicable Diseases in Botswana (NCDs survey)Self-reported1178≥15NA30.9105.4Zou et al. ,37ChinaUMIC20202004–2008Cross-sectionalA baseline dataset from China Kadoorie Biobank (CKB) study, a Chinese population-based cohort studySelf-reported and Objective512,88830–79NA41.01615.9Ma et al.38ChinaUMIC20202015–2106Cross-sectionalChina Health and Retirement Longitudinal Study (CHARLS)Self-reported19,656≥4560.248.31454.3Kim et al. ,39KoreaHIC20202016Cross-sectionalKorea National Health and Nutrition Examination Survey (KNHANES)Self-reported68,590≥19NANA3923.7Kshatri et al.40IndiaLow- or LMIC20202019–2020Cross-sectionalA cross-sectional studySelf-reported72560–10670.252.11848.8Kyprianidou et al.41CyprusHIC20202018–2019Cross-sectionalA cross-sectional studySelf-reported1140≥184043.74728.6de Melo et al.42BrazilUMIC20202013–2014Cross-sectionalNational Health Survey databaseSelf-reported11,697≥6070.140.11353.1Zhang et al.43USAHIC20202012–2017Cross-sectionalNational Health Interview Survey (2012–2017) of Asian Indians, Chinese, and NHWs (non-Hispanic whites)Self-reported132,666≥18NA48.51038.2Li et al.44ChinaUMIC20192017Cross-sectionalA community-based cross-sectional health interview and examination surveySelf-reported and Objective4833≥60NA45.5516.1Aminisani et al.45IranUMIC20202017–2018Cross-sectionalProspective Epidemiological Research Studies in Iran (PERSIAN)Self-reported1493≥5061.6383636.6Craig et al.46JamaicaLow- or LMIC20202007–2008Cross-sectionalJamaica Health and Lifestyle Survey $\frac{2007}{2008}$ (JHLS-II)Self-reported255115–74NANA1124.1Vargese et al.47IndiaLow- or LMIC20202017Cross-sectionalA register based cross sectional studySelf-reported525≥1847.446.91216.2Lee et al.48KoreaHIC20202014Cross-sectional2014 Korean Health Panel SurveySelf-reported11,232≥1857.549.6≥2034.8Zhao et al.49ChinaUMIC20202011–2015Cross-sectionalChina Health and Retirement Longitudinal Study (CHARLS) for 2011, 2013, and 2015Self-reported11,817≥5062 (median)48.81161.9Wister et al.50CanadaHIC20202010Cross-sectionalCanadian Longitudinal Study on Aging (CLSA) datasetSelf-reported15,71145–8562492764Yao et al.51ChinaUMIC20192011–2015Cross-sectionalChina Health and Retirement Longitudinal Study (CHARLS)Self-reported19,841≥50NA48.61442.4Zhang et al.52ChinaUMIC20192015Cross-sectionalChina Health and Retirement Longitudinal Survey (CHARLS) 2015Self-reported11,707≥6070.548.71443.6Laires et al.53PortugalHIC20192014Cross-sectionalFifth Portuguese National Health Interview Survey, conducted in 2014Self-reported15,19625–79NA441543.9Ba et al.54VietnamLow- or LMIC20192018Cross-sectionalA cross-sectional studySelf-reported1680≥153850.1916.4Khan et al.55BangladeshLow- or LMIC20192015–2016Cross-sectionalA large-scale cross-sectional studySelf-reported12,338≥3558.548.668.4Singh et al.56South AsiaLow- or LMIC20182010–2011Cross-sectionalCardiometabolic Risk Reduction in South Asia Surveillance StudySelf-reported and Objective16,287≥204147.359.4Lai et al.57Hong KongHIC20192008Cross-sectionalThe Thematic Household Survey (THS) on health-related topicsSelf-reported17,396≥35NA48.5148.8Bao et al.58ChinaUMIC2019NACross-sectionalCross-sectional community health surveySelf-reported18,137≥4561.447.61920.8Hu et al.59TaiwanHIC20192003–2013Cross-sectionalThe National Health Insurance Research DatabaseSelf-reported1,429,527≥20NANA2030.4Park et al.60KoreaHIC20192013–2015Cross-sectionalSixth Korean National Health and Nutrition Examination Survey (KNHANES) conducted in 2013–2015Self-reported8370≥5062.546.31039Hernandez et al.61IrelandHIC2019NACross-sectionalIrish population studySelf-reported6101≥50NA46.33173.3Frolich et al. ,62DenmarkHIC20192012Cross-sectionalDanish national administrative and health registriesObjective1,397,173≥16NA48.41621.6Chang et al. ,63South AfricaUMIC20192014–2015Cross-sectionalPopulation-based survey conducted in The Health and Ageing in Africa: a longitudinal study of an INDEPTH Community in South Africa (HAALSI) ProgrammeSelf-reported and Objective3889≥4061.745.21069.4Nguyen et al. ,64EnglandHIC20192004–2005Cross-sectionalEnglish Longitudinal Study of Aging (ELSA) wave 2Self-reported9171≥5066.444.52680.8dos Santos Costa et al. ,65BrazilUMIC20182014Cross-sectionalCross-sectional population-based studySelf-reported1451≥60NA372992.8Cheung et al. ,66Hong KongHIC20182016–2017Cross-sectionalBaseline well-being assessment of the Jockey Club Community eHealth Care ProjectSelf-reported2618≥60NA47.5741.8Zemedikun et al. ,67UKHIC20182006–2010Cross-sectionalUK Bio-bank, a major collaborative research projectSelf-reported and Objective502,64340–695845.63619El Lawindi et al. ,68EgyptLow- or LMIC20182016–2017Cross-sectionalA community-based cross-sectional studySelf-reported2317≥1836.254.91619.6Stanley et al. ,70New ZealandHIC20182014Cross-sectionalNational-level routine health data on hospital discharges and pharmaceutical dispensingObjective3,489,747≥18NA48.23027.9Araujo et al. ,71BrazilUMIC20182015Cross-sectionalCross-sectional population-based studySelf-reported4001≥18NA47.21229Jankovic et al. ,72SerbiaUMIC20182013Cross-sectional2013 National Health Survey (NHS 2013) of the Serbian populationSelf-reported13,765≥2051.8461330.2Chen et al. ,73ChinaUMIC20182011–2012Cross-sectionalChina Health and Retirement Longitudinal Study 2011Self-reported3737≥45NA51.91645.5Nunes et al. ,74BrazilUMIC20182015–2016Cross-sectionalThe Brazilian Longitudinal Study of Aging (ELSI-Brazil)Self-reported9412≥5062.9461967.8Mondor et al. ,75CanadaHIC20182005–2012Cross-sectionalThe Canadian Community Health Survey (CCHS) (2005–$\frac{2011}{12}$)Objective27,195≥18NA48.61733.5Mounce et al. ,76EnglandHIC20182002–2003CohortThe English Longitudinal Study of Aging (ELSA) cohortSelf-reported4564≥50NA43.71534Ge et al. ,77SingaporeHIC20182015–2016Cross-sectionalPopulation Health Index (PHI) surveyObjective1940≥2151.443.91735Camargo-Casas et al. ,78ColombiaUMIC20182012Cross-sectionalSalud, Bienestery, Envejecimiento Bogota (SABE-B), (Health, Well-being and Ageing Study)Self-reported2000≥6071.136.61240.4Amaral et al. ,79BrazilUMIC20182010Cross-sectionalA project entitled “Conditions of health, quality of life and depression in elderly persons assisted under the Family Health Strategy in Senador Guiomard, Acre”Self-reported26460–102NA391466.3Puth et al. ,80GermanyHIC20172012–2013Cross-sectionalNational telephone health interview survey “German Health Update” (GEDA2012)Self-reported19,294≥18NA48.31739.6Waterhouse et al. ,81South AfricaUMIC20172007–2008Cross-sectionalWave 1 (2007–08) of the South African Study on Global Ageing and Adult HealthSelf-reported and Objective3055≥50NA39.6812.9Alimohammadian et al. ,69IranUMIC20172004–2008Cross-sectionalGolestan cohort dataSelf-reported49,94640–75NA42.4819.4Wang et al. ,82AustraliaHIC20172007Cross-sectional2007 National Survey of Mental Health and Wellbeing (NSMHWB)Self-reported882016–854449.7828.8Kunna et al. ,83ChinaUMIC20172008–2010Cross-sectionalWorld Health Organization Study on Global AGEing and Adult Health (SAGE) Wave 1 (2007–2010)Self-reported and Objective11,814≥50NA46.4829.7Lujic et al. ,84AustraliaHIC20172005–2009CohortThe 45 and Up Study, The PBS (Pharmaceutical Benefits Scheme) database, The NSW (New South Wales) Admitted Patient Data Collection (APDC)Self-reported90,352≥4570.244.3837.4Nunes et al. ,85BrazilUMIC20172013Cross-sectionalPopulation-based data from the Brazilian National Health SurveySelf-reported60,202≥1843.744.92222.2Mini et al. ,86IndiaLow- or LMIC20172011Cross-sectionalUnited Nations Population Fund (UNFPA) in the year 2011 on ‘Building Knowledge Base on Population Ageing in India’Self-reported9852≥6068471230.7Larsen et al. ,87DenmarkHIC20172013Cross-sectionalDanish national health survey conducted in 2013Self-reported162,283≥1647.8491537Gu et al. ,88ChinaUMIC20172013Cross-sectionalA cross-sectional studySelf-reported2452≥6069.251.51349.4Dhalwani et al. ,89EnglandHIC20172008–2013CohortThe English Longitudinal Study of Ageing (ELSA) 4, 5, 6Self-reported5476≥5061 (median)471821.1Nunes et al. ,90BrazilUMIC20162012Cross-sectionalA population-based cross-sectional studySelf-reported2927≥2045.741.11129.1Picco et al. ,91SingaporeHIC20162012–2013Cross-sectionalThe Well-being of the Singapore Elderly (WiSE) studySelf-reported2565≥60NA43.51051.5Palladino et al. ,9216 countriesHIC20162011–2012Cross-sectionalSurvey of Health, Ageing and Retirement in Europe (SHARE) in 2011–12Self-reported56,427≥506644.11337.3Cossec et al. ,93FranceHIC20162012Cross-sectionalHealth, Health Care and Insurance Survey from 2012 (Enquête Santé et Protection Sociale) called ESPSSelf-reported423656–10569.643714.9Vadrevu et al. ,104IndiaLow- or LMIC20162009Cross-sectionalA cross-sectional surveySelf-reported815≥4054.951.3644.1Marengoni et al. ,95SwedenHIC20162001–2004Cross-sectionalSwedish National study on Aging and Care in Kungsholmen (SNAC-K)Objective3155≥6074.435.7≥552.4Jovic et al. ,96SerbiaUMIC20162013Cross-sectional2013 National Health Survey (NHS 2013) of the Serbian populationSelf-reported13,103≥2049.448.11226.9Su et al. ,97ChinaUMIC20162013Cross-sectionalA large-scale survey initiated by Shanghai Health and Family Planning CommissionSelf-reported2058≥80NA42.11049.2Ramond-Roquin et al. ,98CanadaHIC20162010Cross-sectionalThe Program of Research on the Evolution of a Cohort Investigating Health System Effects (PRECISE)Self-reported171025–7551.340.52163.8Lenzi et al. ,99ItalyHIC20162012Cross-sectionalThe hospital discharge record (HDR) database, the mental health information system, residential mental healthcare discharge records, the outpatient pharmaceutical database, the regional mortality register databaseObjective3,759,836≥18NA482615.3Dung et al. ,100VietnamLow- or LMIC20162011Cross-sectionalVietnam Ageing Survey (VNAS)Self-reported2789≥6071.939.71243.9Valadares et al. ,101BrazilUMIC20162012–2013Cross-sectionalCross-sectional population-based studySelf-reported74945–6052.501153Pache et al. ,102SwitzerlandHIC20152003–2006Cross-sectionalPopulation-based studyObjective371435–7549.6472756.3Afshar et al. ,10328 countriesNA20152003Cross-sectionalWorld Health Survey [2003]Self-reported125,404≥18NA48.567.8Roberts et al. ,104CanadaHIC20152011–2012Cross-sectionalCanadian Community Health Survey $\frac{2011}{12}$Self-reported105,406≥20NA44.1912.9Arokiasamy et al. ,1056 CountriesLow- or LMIC20152007–2010Cross-sectionalWorld Health Organization Study on Global AGEing and Adult Health (SAGE) Wave 1 (2007–2010)Self-reported and Objective42,236≥18NA50.7821.9Ha et al. ,106VietnamLow- or LMIC20152010Cross-sectionalPopulation-based studyObjective2400≥6072.634.8639.2Wang et al. ,107ChinaUMIC20152012Cross-sectionalJilin Provincial Chronic Disease SurveySelf-reported21,43518–79NANA1824.7Wang et al. ,108ChinaUMIC20152010–2011Cross-sectionalConfucius Hometown Aging Project in Shandong, China (June 2010–July 2011)Self-reported and Objective1480≥6068.540.61690.5Nunes et al. ,109BrazilUMIC20152008Cross-sectionalA population-based cross-sectional studySelf-reported1593≥60NA37.21781.3Chung et al. ,110Hong KongHIC20152011–2012Cross-sectionalThematic Household Survey (THS) conducted by the Census and Statistics Department (C&SD) of the Hong Kong SAR GovernmentSelf-reported25,780≥15NA47.84612.5Hussain et al. ,3IndonesiaUMIC20152007–2008Cross-sectionalFourth wave of Indonesian Family Life Survey (IFLS-4)Self-reported and Objective9438≥40NA48.41135.7Ruel et al. ,111AustraliaHIC20142000–2002Case-sectionalNorth West Adelaide longitudinal Health Study (NWAHS)Self-reported and Objective1854≥185048832Mahwati et al. ,112IndonesiaUMIC20142007–2008Cross-sectionalThe fourth survey of the Indonesian Family Life Survey (IFLS) which held in 2007Self-reported2960≥60NA46915.8Islam et al. ,27AustraliaHIC20142009Cross-sectionalA cross-sectional surveySelf-reported4574≥5069.3NA1152Banjare et al. ,113IndiaLow- or LMIC20142011–2012Cross-sectionalA cross-sectional surveySelf-reported310≥60NA49.42156.8Hien et al. ,114Burkina FasoLow- or LMIC20142012Cross-sectionalCross-sectional study among community-dwelling elderlyObjective389≥606955.31565Orueta et al. ,115SpainHIC20132007–2011Cross-sectionalPrimary care electronic medical records, hospital admissions, and outpatient care databasesObjective452,698≥65NA42.54761.1Aguiar et al. ,116BrazilUMIC20132011Cross-sectionalA cross-sectional, population-based studySelf-reported622≥5064.101258.2Alaba et al.117South AfricaUMIC20132008Cross-sectionalSouth African National Income Dynamics Survey (SA-NIDS) of 2008Self-reported11,638≥18403964Wu et al. ,118ChinaUMIC20132010Cross-sectionalSAGE-China Wave 1Self-reported and Objective13,157≥5062.648.1818.9Phaswana-Mafuya et al. ,119South AfricaUMIC20132008Cross-sectionalNational population-based cross-sectional surveySelf-reported3638≥50NA42.5822.5Jerliu et al. ,120KosovoUMIC20132011Cross-sectionalA nationwide cross-sectional studySelf-reported1890≥6573.450.2645.2Kiliari et al. ,121CyprusHIC20132008Cross-sectionalA nationally based surveySelf-reported46518–885343.22728.5Fuchs et al. ,122GermanyHIC20122008–2009Cross-sectionalTelephone health interview surveys in representative samples of the German adult population (German Health Update, GEDA)Self-reported21,26218–10048.848.52240.1MacHado et al. ,123BrazilUMIC20122005Cross-sectionalA secondary analysis of a cross-sectional population-based studySelf-reported37740–65NA0539.3Kirchberger et al. ,124GermanyHIC20122008–2009Cross-sectionalThe population-based KORA-Age projectSelf-reported406765–9473.448.81358.6Agborsangaya et al. ,125CanadaHIC20122010Cross-sectionalHealth Quality Council of Alberta (HQCA) 2010 Patient Experience SurveySelf-reported5010≥1846.747.71619Tucker-Seeley et al. ,126USAHIC20112004Cross-sectionalThe Health and Retirement Study (HRS)Self-reported7305≥506546.4635.4Khanam et al. ,127BangladeshLow- or LMIC20112004Cross-sectionalA descriptive cross-sectional studyObjective45260–9269.545.1953.8Taylor et al. ,128AustraliaHIC20102004–2006Cross-sectionalNorth West Adelaide Health Study (NWAHS Stage 2)Self-reported and Objective3206≥20NANA717.1Loza et al. ,129SpainHIC20091999–2000Cross-sectionalA national health surveySelf-reported and Objective2192≥20NA46.3929.7Minh et al. ,1305 countriesLow- or LMIC20082005Cross-sectional2005 cross-site study of 8 sites in 5 Asian countriesSelf-reported18,49425–64NA5077.2Camargo-Casas,78ColumbiaUMIC20182012Cross-sectionalNASelf-reported2000≥6071.136.6NA40.4Wilk et al.131CanadaHIC20212015–2018Cross-sectionalCanadian Community Health Survey (CCHS), 2015–2018Self-reported100,803≥2047.948.958.1Tomita et al.132TanzaniaLow- or LMIC20212017–2018Cross-sectionalThe Dar es Salaam Health and Demographic Surveillance System (HDSS)Self-reported2299≥4053.032.4824.8Smith et al.133IrelandHIC20212009–2013Cross-sectionalIrish Longitudinal Study on Ageing (TILDA) SurveySelf-reported5946≥5062.751.71450.3Delpino et al. ,134BrazilUMIC20212019Cross-sectionalThe Brazilian National Health Survey 2019Self-reported65,80318–59NA47.81422.3Marthias et al. ,135IndonesiaUMIC20212014Cross-sectionalThe Indonesian Family Life Survey 2014 (Wave – 5)Self-reported and Objective3678≥5065 (median)46.11022.0Zhang et al.136ChinaUMIC20212019Cross-sectionalA cross-sectional studySelf-reported and Objective3250≥60NA46.62630.3Lin et al. ,137TaiwanHIC20212017–2019Cross-sectionalA community-based surveySelf-reported373965–8572.942.8727.8Nicholson et al. ,138CanadaHIC20212015Cross-sectionalThe Canadian Longitudinal Study on Aging (CLSA)Self-reported11,16165–85NA47.51575.3Bezerra et al. ,13917 countriesHIC20212015Cross-sectionalSurvey of Health, Aging and Retirement in Europe (SHARE) 2015 (Wave – 6)Self-reported63,844≥50NA44.31333.6Koyanagi et al. ,14048 countriesLow- or LMIC20212002–2004Cross-sectionalThe World Health Survey 2002–2004Self-reported224,842≥1838.349.3103.8Shi et al. ,141BrazilUMIC20211998–2013Cross-sectionalThe National Sample Household and Brazilian National Health SurveySelf-reported795,271≥18NA47.2918.3Wang et al. ,142ChinaUMIC20212018Cross-sectionalA cross-sectional surveySelf-reported1871≥6083.639.03374.3He et al. ,143ChinaUMIC20212014–2019CohortAnnual health examination data set in the Xinzheng electronic health ManagementSelf-reported and Objective50,100≥6569.2 (median)46.1731.4Ballesteros et al. ,144ColombiaUMIC20212015Cross-sectionalColombian population-based survey Health, Wellbeing and Aging (Salud, Bienestar y Envejecimiento—SABE)Self-reported17,571≥6069.244.31062.3Mohamed et al. ,145KenyaLMIC20212003–2015Cross-sectionalNairobi Urban Health & Demographic Surveillance System (NUHDSS)Self-reported and Objective2003≥4048.846.01628.7Kanungo et al. ,146IndiaLow- or LMIC20212017–2019Cross-sectionalLongitudinal Ageing Study in India (LASI), Wave-1Self-reported59,76445–11660.245.91250.4Oh et al. ,147USAHIC20202001–2003Cross-sectionalThe National Survey of American LifeSelf-reported5191≥1842.263.12254.1King et al. ,148USAHIC20192013–2014Cross-sectionalThe National Health and Nutrition Examination Survey (NHANES)Self-reported and Objective5541≥20NA48.21159.6Bowling et al.149USAHIC20192011–2016Cross-sectionalThe National Health and Nutrition Examination Survey (NHANES), 2011–2016Self-reported and Objective4217≥5056.748.71272.4Keats et al.150CanadaHIC20172009–2015CohortAtlantic Partnership for Tomorrow's Health (PATH) studySelf-reported18,709≥35NA30.01838.2Quinaz Romana et al.151PortugalHIC20192013–2016Cross-sectionalThe National Health Examination Survey (INSEF)Objective4911≥25NA47.52038.3de Souza et al.152BrazilUMIC20192001–2002CohortA longitudinal study of municipal technical and administrative employees in Rio de JaneiroSelf-reported and Objective733≥2441.633.81545.6Costa et al.153BrazilUMIC20202013–2014Cross-sectionalBrazilian National SurveySelf-reported and Objective23,329≥2037.947.21410.9Keomma et al.154BrazilUMIC20202015Cross-sectionalThe ISA-Capital health surveySelf-reported and Objective1019≥6067.740.31040Jürisson et al.155EstoniaHIC20212015–2017Cross-sectionalEstonian Health Insurance FundObjective909,477≥2553.445.95539.8aAscertainment of morbidities- Objective: medical records/clinical examinations.
## Global and regional prevalence of multimorbidity
The prevalence of multimorbidity among the adult population ranged from $4.0\%$ to $92.8\%$ in the studies. Prevalence estimates along with confidence intervals for multimorbidity are shown in Fig. 2 by using a forest plot. The random-effects overall pooled estimated (126 studies) prevalence of multimorbidity was $37.2\%$ ($95\%$ CI = $34.9\%$–$39.4\%$, I2 = $99.7\%$). The pooled proportion of multimorbidity was the highest in South America with $45.7\%$ ($95\%$ CI = $39.0\%$–$52.5\%$, I2 = $99.0\%$). On the other hand, the pooled prevalence of multimorbidity was the lowest in Africa with $28.2\%$ ($95\%$ CI = $15.6\%$–$40.8\%$, I2 = $99.0\%$). However, studies from Asia, Europe, North America, and Oceania were calculated to have the pooled prevalence of multimorbidity $35\%$ ($95\%$ CI = $31.4\%$–$38.5\%$, I2 = $99.3\%$), $39.2\%$ ($95\%$ CI = $33.2\%$–$45.2\%$), $43.1\%$ ($95\%$ CI = $32.3\%$–$53.8\%$), and $32.5\%$ ($95\%$ CI = $26.8\%$–$38.2\%$, I2 = $98.9\%$), respectively. Fig. 2Forest Plot of the Overall Prevalence of multimorbidity in community settings.
## Subgroup analysis
The subgroup analysis of the prevalence of multimorbidity by continents, study design, number of diseases included in multimorbidity, age, and gender is shown in Table 2. The forest plots are given in the Supplementary File 3. Of note, 85 studies reported the prevalence of multimorbidity in males and females. According to the table, the pooled prevalence of multimorbidity was higher among female participants ($39.4\%$, $95\%$ CI = 36.4–$42.4\%$, I2 = $99.6\%$) than male participants ($32.8\%$, $95\%$ CI = 30.0–$35.6\%$, I2 = $99.6\%$). The Fig. 3 shows the gender segregation of pooled prevalence of multimorbidity by geographic regions. Female participants from South America (prevalence $50.1\%$ and $95\%$ CI = 39.7–$60.4\%$) appeared to have the most multimorbid conditions in the world. Multimorbid illnesses were notably more prevalent in European and North American women than in male participants. Table 2Summary results of subgroup analysis. SubgroupNo of studiesWeighted *Mean agea* (SE)Pooled prevalence of Multimorbidity$95\%$ CII2 (%)WHO geographic RegionAfrica1049.71 (10.9)0.2820.156–0.40899.9Asia4757.76 (11.6)0.3500.314–0.38599.9Europe2758.16 (9.6)0.3920.332–0.45299.6North America1454.61 (6.1)0.4310.323–0.53899.9Oceania658.38 (13.3)0.3250.268–0.38298.3South America1956.38 (13.4)0.4570.390–0.52599.9WB/WHO income regionHigh5356.61 (9.7)0.3860.353–0.41999.9Upper-middle4860.43 (12.5)0.3870.355–0.41999.9Low and Low-middle2453.19 (11.93)0.3210.243–0.4099.8Study designCross-sectional12156.46 (11.06)0.3740.351–0.39699.3Cohort562.7 (6.71)0.3240.279–0.36996.7Number of conditions included for defining multimorbidity5–9 conditions3757.54 (12.64)0.2500.223–0.27897.910–19 conditions6460.15 (9.96)0.4130.376–0.45099.9≥20 conditions2453.44 (8.47)0.4570.393–0.50099.9GenderFemale85–0.3940.364–0.42499.9Male85–0.3280.300–0.35699.2Mental health included in Multimorbidity definitionYes9157.62 (11.02)0.3840.359–0.41099.3Nob2861.12 (11.56)0.3320.271–0.39298.9Age of the study participants≥30 years7665.2 (6.26)0.4440.393–0.49499.9≥40 years7165.86 (5.69)0.4570.402–0.51299.9≥50 years5867.42 (4.63)0.4720.420–0.52599.9≥60 years3370.91 (2.01)0.5100.441–0.58098.3Overall12656.95 (10.85)0.3730.349–0.39499.0aThe weighted mean age and standard error (SE) were calculated based on the available study sample size and the study participant's mean/median age.bBecause the disease list was not mentioned in a few of the articles, we assumed these articles may not contain mental health. Fig. 3Regional differences of pooled prevalence of multimorbidity by gender.
Based on the continents of the studies, the estimated pooled prevalence of multimorbidity was found $38.6\%$ ($95\%$ CI = $35.3\%$–$41.9\%$, I2 = $99.2\%$) in high-income countries, $38.7\%$ ($95\%$ CI = 35.5–$41.9\%$, I2 = $99.2\%$) in upper middle-income countries (UMICs), and $32.1\%$ ($95\%$ CI = 24.3–$40.0\%$, I2 = $99.5\%$) in Low- and LMICs. In the case of the number of diseases included in the multimorbidity, the prevalence was found $44.7\%$ ($95\%$ CI = $39.5\%$–$50.0\%$, I2 = $99.3\%$) among the studies that considered ≥20 diseases. The prevalence of multimorbidity was $25.0\%$ ($95\%$ CI = 22.3–$27.8\%$, I2 = $99.0\%$) for studies with 5–9 diseases, and $41.3\%$ ($95\%$ CI = $37.6\%$–$45.0\%$, I2 = $99.0\%$) for studies with 10–19 diseases. When mental health is included in the multimorbidity definition, the prevalence ($38.4\%$, $95\%$ CI = 35.9–$41.0\%$, I2 = $99.0\%$) was higher than without inclusion of mental health ($33.2\%$, $95\%$ CI = 27.1–$39.2\%$, I2 = $99.1\%$).
Among the different age groups of the study participants, the highest prevalence was found in the studies that included the respondents more than 60 years with $51.0\%$ ($95\%$ CI = $44.1\%$–$58.0\%$). The pooled prevalence was $44.4\%$ ($95\%$ CI = $39.3\%$–$49.4\%$, I2 = $99.1\%$) among the participants with 30 years and above. When the study participants were ≥40 years and ≥50 years, the pooled proportion of multimorbidity was $45.7\%$ ($95\%$ CI = $40.2\%$–$51.2\%$, I2 = $99.0\%$) and $47.2\%$ ($95\%$ CI = $42.0\%$–$52.5\%$, I2 = $99.1\%$), respectively.
There was a difference in the prevalence of multimorbidity by study design among the studies. The pooled prevalence of multimorbidity was $37.4\%$ ($95\%$ CI = $35.1\%$–$39.6\%$, I2 = $99.0\%$) for cross-sectional studies, and $32.4\%$ ($95\%$ CI = $27.9\%$–$36.9\%$, I2 = $96.7\%$) for cohort studies.
## Trends of global multimorbidity prevalence over time
The global prevalence of multimorbidity by 5-year interval is displayed in Fig. 4, considering studies that contains 10 or more diseases. The five-year span was categorized based on the year in which investigations were done. If a study was completed between 2013 and 2016, we assumed it was conducted between 2011 and 2015 because the majority of years fell within the interval. The study was removed from the analysis if it did not belong to any of the groups. We excluded papers that reported a multimorbidity prevalence of less than $10\%$ or greater than $80\%$ in order to minimize variability in trend analysis. The prevalence of multimorbidity has been on the rise globally since 2000, but it has remained rather stable since 2011. The trend analysis with the studies that considered ten or more illnesses in multimorbidity classifications, showed that the global prevalence of multimorbidity remained high, exceeding $40\%$.Fig. 4Pooled prevalence of multimorbidity by year.
## Sensitivity analysis for global prevalence
We conducted sensitivity analyses including studies with more than 1000 participants, removing studies from Africa, and removing studies that showed prevalence of less than $20\%$ and more than $80\%$. The reasons for removing studies with less than 1000 participants are to increase estimate reliability and precision of the estimate with the studies with a larger sample size. Furthermore, we excluded papers with extreme prevalence estimates of less than $20\%$ and more than $80\%$ because these values could lead to heterogeneity in predicting worldwide prevalence. Forest plots are reported in Supplementary File 4. When considering studies of more than 1000 participants, the global prevalence among participants tends to be $36.1\%$ ($95\%$ CI = 33.7–$38.4\%$, I2 = $98.8\%$), which is in line with the findings of the meta-analysis with 126 studies. After excluding African studies, the prevalence was $37.9\%$ ($95\%$ CI = $35.4\%$–$40.2\%$), which is comparable to the meta-analysis with 126 studies. We also found the global prevalence was higher than the overall pooled prevalence after removing studies with extreme prevalence. The results showed the prevalence $42.3\%$ ($95\%$ CI = 39.8–$44.7\%$, I2 = $98.8\%$) after excluding studies with extreme prevalence. The findings excluding studies with extreme prevalence are, therefore, higher than the meta-analysis of 126 studies. With high-quality papers (minimal bias according to NOS), we found the prevalence to be $36.6\%$ ($95\%$ CI = 33.6–$39.5\%$, I2 = 99.8), which imply a similar result that we analyzed in the meta-analysis of 126 studies. Moreover, the studies using self-reported multimorbid data indicate a prevalence of $38.3\%$ ($95\%$ CI = 35.1–$41.5\%$), but the studies with data from medical records indicate a prevalence of $34.3\%$ [$95\%$ CI = 30.3–$38.2\%$].
## Publication bias
The Egger test found that there was no statistically significant publication bias ($P \leq .05$) among the 83 population-based studies evaluating the relationship between gender and multimorbidity status. However, the Egger test revealed a statistically significant publication bias among the 126 population-based studies for proportion (Supplementary File 5). We also have applied trim-and-fill method to adjust for this publication bias in the analysis. We see that the procedure identified and trimmed 42 added studies. The overall effect estimated by the trim-and-fill is $26.71\%$ ($95\%$ CI = 0.2350–0.2799). Our initial estimate with 126 studies was $37.1\%$, which is substantially larger than the bias-corrected effect. If we assume that publication bias affected our findings, the trim-and-fill method allows us to hypothesize that our initial results were overstated because of publication bias, and the global estimate when controlling for selective publication might be $26.71\%$. Moreover, considering the odds ratio in a funnel plot we found a high existence of publication bias in our study. Consequently, publication bias may be a cause of heterogeneity in investigating overall proportion.
## Discussion
This study analyzed data from 126 studies that involved nearly 15.4 million people from 54 countries, providing an up-to-date global multimorbidity prevalence of $37.2\%$ ($95\%$ CI = 34.9–$39.4\%$). A previous meta-analysis with studies until 2017 found that $33.1\%$ had multimorbidity in the adult population aged 18 and older living in the community.12 In comparison to that meta-analysis including studies in community settings, we found a higher prevalence of multimorbidity. Another meta-analysis that included studies from both community and healthcare settings estimated the overall prevalence of multimorbidity was $42.4\%$ ($95\%$ CI = 38.9–$46.0\%$) among adults.156 The inclusion of studies from primary care and health care settings in the meta-analysis resulted in a higher pooled prevalence than ours.
The sub-group analysis by region showed significant differences in the pooled prevalence of multimorbidity. Our analysis showed that the prevalence of multimorbidity was highest in South America. The result is consistent with a meta-analysis that found that the pooled proportion of multimorbidity in Latin America and the Caribbean was as high as $43\%$ ($95\%$ CI: 35–$51\%$).157 Africa had the lowest prevalence of multimorbidity, according to our analysis. The result could be attributable to the low age group of participants in the African studies compared to other geographic regions. The lowest rate of multimorbidity in Africa should be interpreted with caution because it raises the possibility that many people living with chronic illnesses in Africa are going undiagnosed.
In subgroup analysis, the prevalence of multimorbidity was lower in Low- and LMICs than in UMICs and HICs. The prevalence of multimorbidity was highest in UMICs. This difference is consistent with another study's findings, where a meta-analysis in community settings found that the pooled multimorbidity prevalence was higher in HICs than LMICs.12,156 The majority of the survey included in the meta-analysis were from HICs and UMICs, with a few studies conducted in Low-income countries. It may reflect the differences in diagnostic and data management systems among HICs, UMICs, and Low- and LMICs. According to a study, the disparity in prevalence estimates between HICs and LMICs could be due to the fewer publications on multimorbidity prevalence in LMICs because of limited understanding and importance of multimorbidity in LMICs compared to HICs.158 People in low-income countries may be less likely to seek treatment for diseases than those in high-income countries. Therefore, the prevalence in low-income countries may be underestimated if diseases are defined using medical records.
The pooled prevalence of multimorbidity was higher for the cross-sectional study design than for the cohort study type in this meta-analysis. This disparity in multimorbidity prevalence could be due to study designs with varying levels of methodological differences, such as various study populations, sampling procedures, sample coverage, sample sizes, data collection, and so on. Besides, we considered the baseline sample for a cohort study design that might contribute to the lower prevalence.
For included studies, the more the number of diseases evaluated for multimorbidity, the higher the prevalence. When examining 20 or more conditions for multimorbidity, the prevalence was $44.7\%$, but it was lowered to $41.3\%$ for 10–19 diseases and $25.0\%$ for 5–9 diseases to define multimorbidity. According to a study, the different combinations of illnesses may cause the prevalence of multimorbidity to differ significantly.156,159 A range of different combinations of multimorbidity definitions has been proposed in the literature, ranging from a list of 16 chronic diseases to 291 diseases.156,158, 159, 160, 161 Furthermore, the pooled estimate of multimorbidity prevalence with the studies those included mental health in the definition of multimorbidity was greater. Previous studies identified a correlation between multimorbidity and mental health.20,162,163 Our findings, the higher prevalence of multimorbidity with the studies that included mental health, reveal consistency with the findings of previous research.
Our study showed that prevalence estimates varied substantially according to age and gender. Our research showed that females had a higher pooled prevalence of multimorbidity than males. It indicates an association between gender and multimorbidity (evidence of which was provided in multiple studies).69,162,163 According to our findings, multimorbidity increases with age. While the prevalence estimates varied between and within age groups, our meta-analysis indicated that a large proportion of individuals over 60 had multimorbidity. It is well established that the prevalence of multimorbidity increases in very old persons.164, 165 The calculation of the global prevalence of multimorbidity based on the study's publication interval of 5-year is one of the most important findings of our research. According to our findings, the prevalence of multimorbidity has changed considerably over the previous two decades but has remained relatively consistent since 2011. This suggests a gradual decline in the global burden of multimorbidity. The plateau observed in multimorbidity prevalence since 2011 may be attributable to a handful of the 19 studies that showed low prevalence in 2016–2021. Therefore, this conclusion should be studied further. Over the years, the global prevalence of multimorbidity among adults has exceeded 40 percent, indicating a high burden of multimorbidity exists over years.
One of the study's strengths was its strong study selection and screening protocols. Because of our rigorous search approach and inclusion criteria, we were able to conduct the largest systematic review of multimorbidity prevalence in community settings to date. The majority of the papers included in the review were of high quality. The comprehensive subgroup analyses demonstrate that our findings are applicable to a wide range of contexts. One important finding of our study is the estimation of the global prevalence of multimorbidity by year of publication. This review did, however, have several limitations. To report multimorbidity prevalence, the majority of the studies in our sample used self-reported data. As a result, such research findings were prone to response bias. High heterogeneity between studies in our meta-analysis implies that the prevalence of multimorbidity varies between studies. To overcome this constraint, we used a random-effects model and performed subgroup analyses. Furthermore, considerable heterogeneity may indicate that the prevalence of multimorbidity varies significantly by geographical region, country income classification, gender, age group, number of diseases considered for multimorbidity, or study methodology.
The high prevalence of multimorbidity highlights the need for healthcare reforms and improvements in several continents. Policymakers should commit to increasing multimorbidity awareness, particularly in relation to mental health management, supporting innovation, maximizing the use of existing resources, and coordinating the efforts of multiple countries to reduce the burden and fatal effects of multimorbidity. More than half of the global adult population over the age of 60 has multimorbid illnesses, and female adults are more prone to develop multimorbidity than male adults. Therefore, management should incorporate these findings into healthcare policies, and countries, particularly in South America, should aim to increase their preventative efforts and build more integrated care models to reduce the burden.
## Contributors
A.H., S.R.C., D.C.D., and T.C.S. contributed to the study concept, literature search, and design. A.H., S.R.C., D.C.D., T.C.S. and J.B. contributed to the data acquisition. A.H. and S.R.C. accessed the data and contributed to the data analysis. A.H., S.R.C., and J.B. contributed to the data interpretation. A.H., S.R.C. and D.C.D. drafted the manuscript. All authors contributed to the critical revision of the manuscript.
## Data sharing statement
Because this meta-analysis was based on data extracted from previously published research, most of the data and study materials are available in the public domain. For further discussions, we invite interested parties to contact the corresponding author.
## Declaration of interests
All other authors declare no competing interests.
## Appendix
.Appendix ASearch strategy. A. PubMed #1(“Prevalence” OR “Surveillance” OR “Surveys” OR “Epidemiology”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)3734 #2(“Risk factors” OR “Determinants” OR “Predictors”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)1500 #3(“Aging” OR “Gender”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)1708Searching date starting from $\frac{2000}{01}$/01 to $\frac{2021}{12}$/31All the entries were under ‘All fields’ categoriesB. Google Scholar #1(“Prevalence” OR “Surveillance” OR “Surveys” OR “Epidemiology”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)18,913 #2(“Risk factors” OR “Determinants” OR “Predictors”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)16,500 #3(“Aging” OR “Gender”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)17,103Searching date starting from $\frac{2000}{01}$/01 to $\frac{2021}{12}$/31C. ScienceDirect #1(“Prevalence” OR “Surveys” OR “Epidemiology”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)4104 #2(“Risk factors” OR “Determinants” OR “Predictors”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)4133 #3(“Aging” OR “Gender”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)4391Searching date starting from $\frac{2000}{01}$/01 to $\frac{2021}{12}$/31D. Embase #1(“Prevalence” OR “Surveillance” OR “Surveys” OR “Epidemiology”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)8713 #2(“Risk factors” OR “Determinants” OR “Predictors”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)3616 #3(“Aging” OR “Gender”) AND (“Multimorbidity” OR “Multi-morbidity” OR “Multimorbidities” OR “Multi-morbidities” OR “Multi morbidity” OR “Multi morbidities”)7138Searching date starting from $\frac{2000}{01}$/01 to $\frac{2021}{12}$/31
## Supplementary data
Supplementary File 1S1. Study quality assessment details for cohort, and cross-sectional studies by New-Castle Ottawa Scale. Supplementary File 2S2. Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) checklist. Supplementary File 3Forest plots of subgroup analysis. Supplementary File 4Sensitivity analysis results. Supplementary File 5Funnel plot and publication bias results.
## References
1. Moffat K., Mercer S.W.. **Challenges of managing people with multimorbidity in today's healthcare systems**. *BMC Fam Pract* (2015) **16** 129. DOI: 10.1186/s12875-015-0344-4
2. **Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980-2015: a systematic analysis for the Global Burden of Disease Study 2015**. *Lancet* (2016) **388** 1459-1544. DOI: 10.1016/S0140-6736(16)31012-1
3. Hussain M.A., Huxley R.R., Al Mamun A.. **Multimorbidity prevalence and pattern in Indonesian adults: an exploratory study using national survey data**. *BMJ Open* (2015) **5**. DOI: 10.1136/bmjopen-2015-009810
4. Ryan A., Wallace E., O'Hara P., Smith S.M.. **Multimorbidity and functional decline in community-dwelling adults: a systematic review**. *Health Qual Life Outcomes* (2015) **13** 168. DOI: 10.1186/s12955-015-0355-9
5. Zhang X., Padhi A., Wei T.. **Community prevalence and dyad disease pattern of multimorbidity in China and India: a systematic review**. *BMJ Glob Health* (2022 Sep) **7**. DOI: 10.1136/bmjgh-2022-008880
6. McPhail S.M.. **Multimorbidity in chronic disease: impact on health care resources and costs**. *Risk Manag Healthc Policy* (2016) **9** 143-156. DOI: 10.2147/RMHP.S97248
7. Makovski T.T., Schmitz S., Zeegers M.P., Stranges S., van den Akker M.. **Multimorbidity and quality of life: systematic literature review and meta-analysis**. *Ageing Res Rev* (2019) **53** 100903. DOI: 10.1016/j.arr.2019.04.005
8. Nunes B.P., Flores T.R., Mielke G.I., Thumé E., Facchini L.A.. **Multimorbidity and mortality in older adults: a systematic review and meta-analysis**. *Arch Gerontol Geriatr* (2016) **67** 130-138. DOI: 10.1016/j.archger.2016.07.008
9. Soley-Bori M., Ashworth M., Bisquera A.. **Impact of multimorbidity on healthcare costs and utilisation: a systematic review of the UK literature**. *Br J Gen Pract* (2021) **71** e39-e46. DOI: 10.3399/bjgp20x713897
10. **WHO MiNDbank - multiple chronic conditions: a strategic framework - optimum health and quality of life for individuals with multiple chronic conditions**
11. Barnett K., Mercer S.W., Norbury M., Watt G., Wyke S., Guthrie B.. **Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study**. *Lancet* (2012) **380** 37-43. DOI: 10.1016/S0140-6736(12)60240-2
12. Nguyen H., Manolova G., Daskalopoulou C., Vitoratou S., Prince M., Prina A.M.. **Prevalence of multimorbidity in community settings: a systematic review and meta-analysis of observational studies**. *J Comorb* (2019) **9**. DOI: 10.1177/2235042x19870934
13. Fortin M., Stewart M., Poitras M.E., Almirall J., Maddocks H.. **A systematic review of prevalence studies on multimorbidity: toward a more uniform methodology**. *Ann Fam Med* (2012) **10** 142-151. DOI: 10.1370/afm.1337
14. Violan C., Foguet-Boreu Q., Flores-Mateo G.. **Prevalence, determinants and patterns of multimorbidity in primary care: a systematic review of observational studies**. *PLoS One* (2014) **9**. DOI: 10.1371/journal.pone.0102149
15. Pati S., Swain S., Hussain M.A.. **Prevalence and outcomes of multimorbidity in South Asia: a systematic review**. *BMJ Open* (2015) **5**. DOI: 10.1136/bmjopen-2014-007235
16. Gruneir A., Bronskill S.E., Maxwell C.J.. **The association between multimorbidity and hospitalization is modified by individual demographics and physician continuity of care: a retrospective cohort study**. *BMC Health Serv Res* (2016) **16** 154. DOI: 10.1186/s12913-016-1415-5
17. Schäfer I., von Leitner E.C., Schön G.. **Multimorbidity patterns in the elderly: a new approach of disease clustering identifies complex interrelations between chronic conditions**. *PLoS One* (2010) **5**. DOI: 10.1371/journal.pone.0015941
18. Pati S., Swain S., Knottnerus J.A., Metsemakers J.F.M., Van Den Akker M.. **Magnitude and determinants of multimorbidity and health care utilization among patients attending public versus private primary care: a cross-sectional study from Odisha, India**. *Int J Equity Health* (2020) **19** 57. DOI: 10.1186/s12939-020-01170-y
19. Sinnige J., Braspenning J., Schellevis F., Stirbu-Wagner I., Westert G., Korevaar J.. **The prevalence of disease clusters in older adults with multiple chronic diseases--a systematic literature review**. *PLoS One* (2013) **8**. DOI: 10.1371/journal.pone.0079641
20. Prados-Torres A., Calderón-Larrañaga A., Hancco-Saavedra J., Poblador-Plou B., van den Akker M.. **Multimorbidity patterns: a systematic review**. *J Clin Epidemiol* (2014) **67** 254-266. DOI: 10.1016/j.jclinepi.2013.09.021
21. Garin N., Koyanagi A., Chatterji S.. **Global multimorbidity patterns: a cross-sectional, population-based, multi-country study**. *J Gerontol A Biol Sci Med Sci* (2016) **71** 205-214. DOI: 10.1093/gerona/glv128
22. van Oostrom S.H., Picavet H.S., de Bruin S.R.. **Multimorbidity of chronic diseases and health care utilization in general practice**. *BMC Fam Pract* (2014) **15** 61. DOI: 10.1186/1471-2296-15-61
23. Wallace E., Salisbury C., Guthrie B., Lewis C., Fahey T., Smith S.M.. **Managing patients with multimorbidity in primary care**. *BMJ* (2015) **350** h176. DOI: 10.1136/BMJ.H176
24. Laux G., Kuehlein T., Rosemann T., Szecsenyi J.. **Co- and multimorbidity patterns in primary care based on episodes of care: results from the German CONTENT project**. *BMC Health Serv Res* (2008) **8** 14. DOI: 10.1186/1472-6963-8-14
25. Larsen F.B., Pedersen M.H., Friis K., Glümer C., Lasgaard M.. **A latent class analysis of multimorbidity and the relationship to socio-demographic factors and health-related quality of life. A national population-based study of 162,283 Danish adults**. *PLoS One* (2017) **12**. DOI: 10.1371/JOURNAL.PONE.0169426
26. Diederichs C., Berger K., Bartels D.B.. **The measurement of multiple chronic diseases--a systematic review on existing multimorbidity indices**. *J Gerontol A Biol Sci Med Sci* (2011) **66** 301-311. DOI: 10.1093/gerona/glq208
27. Islam M.M., Valderas J.M., Yen L., Dawda P., Jowsey T., McRae I.S.. **Multimorbidity and comorbidity of chronic diseases among the senior Australians: prevalence and patterns**. *PLoS One* (2014) **9**. DOI: 10.1371/journal.pone.0083783
28. Ofori-Asenso R., Chin K.L., Curtis A.J., Zomer E., Zoungas S., Liew D.. **Recent patterns of multimorbidity among older adults in high-income countries**. *Popul Health Manag* (2019) **22** 127-137. DOI: 10.1089/pop.2018.0069
29. Moher D., Liberati A., Tetzlaff J., Altman D.G.. **Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement**. *PLoS Med* (2009) **6**. DOI: 10.1371/journal.pmed.1000097
30. **World Bank Country and Lending Groups – World Bank data help desk**
31. Wells G., Shea B., O'Connell D.. **The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses**. (2013)
32. Lo C.K., Mertz D., Loeb M.. **Newcastle-Ottawa Scale: comparing reviewers' to authors' assessments**. *BMC Med Res Methodol* (2014) **14** 45. DOI: 10.1186/1471-2288-14-45
33. Hossain A., Nasrullah S.M., Tasnim Z., Hasan M.K., Hasan M.M.. **Seroprevalence of SARS-CoV-2 IgG antibodies among health care workers prior to vaccine administration in Europe, the USA and East Asia: a systematic review and meta-analysis**. *eClinicalMedicine* (2021) **33** 100770. DOI: 10.1016/j.eclinm.2021.100770
34. Dhungana R.R., Karki K.B., Bista B., Pandey A.R., Dhimal M., Maskey M.K.. **Prevalence, pattern and determinants of chronic disease multimorbidity in Nepal: secondary analysis of a national survey**. *BMJ Open* (2021) **11** 1-7. DOI: 10.1136/bmjopen-2020-047665
35. Zhang L., Sun F., Li Y., Tang Z., Ma L.. **Multimorbidity in community-dwelling older adults in Beijing: prevalence and trends, 2004–2017**. *J Nutr Health Aging* (2021) **25** 116-119. DOI: 10.1007/s12603-020-1467-4
36. Keetile M., Navaneetham K., Letamo G.. **Prevalence and correlates of multimorbidity among adults in Botswana: a cross-sectional study**. *PLoS One* (2020) **15**. DOI: 10.1371/journal.pone.0239334
37. Zou S., Wang Z., Bhura M., Zhang G., Tang K.. **Prevalence and associated socioeconomic factors of multimorbidity in 10 regions of China: an analysis of 0.5 million adults**. *J Public Health (Oxf)* (2022) **44** 36-50. DOI: 10.1093/pubmed/fdaa204
38. Ma X., He Y., Xu J.. **Urban-rural disparity in prevalence of multimorbidity in China: a cross-sectional nationally representative study**. *BMJ Open* (2020) **10** e038404. DOI: 10.1136/bmjopen-2020-038404
39. Kim J., Keshavjee S., Atun R.. **Trends, patterns and health consequences of multimorbidity among South Korea adults: analysis of nationally representative survey data 2007-2016**. *J Glob Health* (2020) **10** 20426. DOI: 10.7189/JOGH.10.020426
40. Kshatri J.S., Palo S.K., Bhoi T., Barik S.R., Pati S.. **Prevalence and patterns of multimorbidity among rural elderly: findings of the AHSETS study**. *Front Public Health* (2020) **8** 582663. DOI: 10.3389/fpubh.2020.582663
41. Kyprianidou M., Panagiotakos D., Faka A., Kambanaros M., Makris K.C., Christophi C.A.. **Prevalence of multimorbidity in the Cypriot population; a cross-sectional study (2018-2019)**. *PLoS One* (2020) **15** e0239835. DOI: 10.1371/journal.pone.0239835
42. Melo L.A., Lima K.C.. **Prevalence and factors associated with multimorbidities in Brazilian older adults**. *Cien Saude Colet* (2020) **25** 3869-3877. DOI: 10.1590/1413-812320202510.34492018
43. Zhang Y., Misra R., Sambamoorthi U.. **Prevalence of multimorbidity among Asian Indian, Chinese, and non-Hispanic White adults in the United States**. *Int J Environ Res Public Health* (2020) **17** 3336. DOI: 10.3390/ijerph17093336
44. Li X., Cai L., Cui W.L.. **Association of socioeconomic and lifestyle factors with chronic non-communicable diseases and multimorbidity among the elderly in rural southwest China**. *J Public Health (Oxf)* (2020) **42** 239-246. DOI: 10.1093/pubmed/fdz020
45. Aminisani N., Rastgou L., Shamshirgaran S.M., Sarbakhsh P., Ghaderi S., Hyde M.. **Predictors of multimorbidity among the Kurdish population living in the Northwest of Iran**. *BMC Public Health* (2020) **20** 1094. DOI: 10.1186/s12889-020-09214-2
46. Craig L.S., Hotchkis D.R., Theal K.P., Cunningham-Myrie C., Hernande J.H., Gustat J.. **Prevalence and patterns of multimorbidity in the Jamaican population: a comparative analysis of latent variable models**. *PLoS One* (2020) **15** e0236034. DOI: 10.1371/journal.pone.0236034
47. Vargese S.S., Mathew E., Johny V., Kurian N., Gayathri A.V., Raju A.S.. **Prevalence and pattern of multimorbidity among adults in a primary care rural setting**. *Clin Epidemiol Glob Health* (2020) **8** 482-485. DOI: 10.1016/j.cegh.2019.10.014
48. Lee Y., Kim H., Jeong H., Noh Y.. **Patterns of multimorbidity in adults: an association rules analysis using the Korea health panel**. *Int J Environ Res Public Health* (2020) **17** 2618. DOI: 10.3390/ijerph17082618
49. Zhao Y., Atun R., Oldenburg B.. **Physical multimorbidity, health service use, and catastrophic health expenditure by socioeconomic groups in China: an analysis of population-based panel data**. *Lancet Glob Health* (2020) **8** e840-e849. DOI: 10.1016/S2214-109X(20)30127-3
50. Wister A., Rosenkrantz L., Shashank A., Walker B.B., Schuurman N.. **Multimorbidity and socioeconomic deprivation among older adults: a cross-sectional analysis in five Canadian cities using the CLSA**. *J Aging Environ* (2020) **34** 435-454. DOI: 10.1080/26892618.2020.1734138
51. Yao S.S., Cao G.Y., Han L.. **Prevalence and patterns of multimorbidity in a nationally representative sample of older Chinese: results from the China health and retirement longitudinal study**. *J Gerontol A Biol Sci Med Sci* (2020) **75** 1974-1980. DOI: 10.1093/gerona/glz185
52. Zhang R., Lu Y., Shi L., Zhang S., Chang F.. **Prevalence and patterns of multimorbidity among the elderly in China: a cross-sectional study using national survey data**. *BMJ Open* (2019) **9** 1-9. DOI: 10.1136/bmjopen-2018-024268
53. Laires P.A., Perelman J.. **The current and projected burden of multimorbidity: a cross-sectional study in a Southern Europe population**. *Eur J Ageing* (2019) **16** 181-192. DOI: 10.1007/s10433-018-0485-0
54. Ba N.V., Minh H.V., Quang L.B.. **Prevalence and correlates of multimorbidity among adults in border areas of the Central Highland Region of Vietnam, 2017**. *J Comorb* (2019) **9**. DOI: 10.1177/2235042X19853382
55. Khan N., Rahman M., Mitra D., Afsana K.. **Prevalence of multimorbidity among Bangladeshi adult population: a nationwide cross-sectional study**. *BMJ Open* (2019) **9**. DOI: 10.1136/bmjopen-2019-030886
56. Singh K., Patel S.A., Biswas S.. **Multimorbidity in South Asian adults: prevalence, risk factors and mortality**. *J Public Health* (2019) **41** 80-89. DOI: 10.1093/pubmed/fdy017
57. Lai F.T.T., Guthrie B., Wong S.Y.S.. **Sex-specific intergenerational trends in morbidity burden and multimorbidity status in Hong Kong community: an age-period-cohort analysis of repeated population surveys**. *BMJ Open* (2019) **9**. DOI: 10.1136/bmjopen-2018-023927
58. Bao X.Y., Xie Y.X., Zhang X.X.. **The association between multimorbidity and health-related quality of life: a cross-sectional survey among community middle-aged and elderly residents in southern China**. *Health Qual Life Outcomes* (2019) **17** 107. DOI: 10.1186/s12955-019-1175-0
59. Hu R.H., Hsiao F.Y., Chen L.J., Huang P.T., Hsu W.W.Y.. **Increasing age- and gender-specific burden and complexity of multimorbidity in Taiwan, 2003-2013: a cross-sectional study based on nationwide claims data**. *BMJ Open* (2019) **9** 1-10. DOI: 10.1136/bmjopen-2018-028333
60. Park B., Lee H.A., Park H.. **Use of latent class analysis to identify multimorbidity patterns and associated factors in Korean adults aged 50 years and older**. *PLoS One* (2019) **14**. DOI: 10.1371/journal.pone.0216259
61. Hernández B., Reilly R.B., Kenny R.A.. **Investigation of multimorbidity and prevalent disease combinations in older Irish adults using network analysis and association rules**. *Sci Rep* (2019) **9** 1-12. DOI: 10.1038/s41598-019-51135-7
62. Frølich A., Ghith N., Schiøtz M., Jacobsen R., Stockmarr A.. **Multimorbidity, healthcare utilization and socioeconomic status: a register-based study in Denmark**. *PLoS One* (2019) **14**. DOI: 10.1371/journal.pone.0214183
63. Chang A.Y., Gómez-Olivé F.X., Payne C.. **Chronic multimorbidity among older adults in rural South Africa**. *BMJ Glob Health* (2019) **4**. DOI: 10.1136/bmjgh-2018-001386
64. Nguyen H., Chua K.C., Dregan A.. **Factors associated with multimorbidity patterns in older adults in England: findings from the English Longitudinal Study of Aging (ELSA)**. *J Aging Health* (2020) **32** 1120-1132. DOI: 10.1177/0898264319891026
65. Costa C.D.S., Flores T.R., Wendt A.. **Inequalities in multimorbidity among elderly: a population-based study in a city in Southern Brazil**. *Cad Saúde Pública* (2018) **34**. DOI: 10.1590/0102-311X00040718
66. Cheung J.T.K., Yu R., Wu Z., Wong S.Y.S., Woo J.. **Geriatric syndromes, multimorbidity, and disability overlap and increase healthcare use among older Chinese**. *BMC Geriatr* (2018) **18** 1-8. DOI: 10.1186/s12877-018-0840-1
67. Zemedikun D.T., Gray L.J., Khunti K., Davies M.J., Dhalwani N.N.. **Patterns of multimorbidity in middle-aged and older adults: an analysis of the UK Biobank data**. *Mayo Clin Proc* (2018) **93** 857-866. DOI: 10.1016/j.mayocp.2018.02.012
68. El Lawindi M.I., Salem M.R., Razik M.M.A., Balas E.A.. **Socioeconomic predictors of morbidities in a rural setting: a community based study**. *Internet J Epidemiol* (2018) **15** 1-7. DOI: 10.5580/IJE.53712
69. Alimohammadian M., Majidi A., Yaseri M.. **Multimorbidity as an important issue among women: results of a gender difference investigation in a large population-based cross-sectional study in West Asia**. *BMJ Open* (2017) **7**. DOI: 10.1136/bmjopen-2016-013548
70. Stanley J., Semper K., Millar E., Sarfati D.. **Epidemiology of multimorbidity in New Zealand: a cross-sectional study using national-level hospital and pharmaceutical data**. *BMJ Open* (2018) **8** 1-12. DOI: 10.1136/bmjopen-2018-021689
71. Araujo M.E.A., Silva M.T., Galvao T.F., Nunes B.P., Pereira M.G.. **Prevalence and patterns of multimorbidity in Amazon Region of Brazil and associated determinants: a cross-sectional study**. *BMJ Open* (2018) **8**. DOI: 10.1136/bmjopen-2018-023398
72. Jankovic J., Mirkovic M., Jovic-Vranes A., Santric-Milicevic M., Terzic-Supic Z.. **Association between non-communicable disease multimorbidity and health care utilization in a middle-income country: population-based study**. *Public Health* (2018) **155** 35-42. DOI: 10.1016/j.puhe.2017.11.014
73. Chen H., Cheng M., Zhuang Y., Broad J.B.. **Multimorbidity among middle-aged and older persons in urban China: prevalence, characteristics and health service utilization**. *Geriatr Gerontol Int* (2018) **18** 1447-1452. DOI: 10.1111/ggi.13510
74. Nunes B.P., Batista S.R.R., Andrade F.B., Souza Junior P.R.B., Lima-Costa M.F., Facchini L.A.. **Multimorbidity: the Brazilian Longitudinal Study of Aging (ELSI-Brazil)**. *Rev Saude Publica* (2018) **52Suppl 2** 10s. DOI: 10.11606/S1518-8787.2018052000637
75. Mondor L., Cohen D., Khan A.I., Wodchis W.P.. **Income inequalities in multimorbidity prevalence in Ontario, Canada: a decomposition analysis of linked survey and health administrative data**. *Int J Equity Health* (2018) **17** 90. DOI: 10.1186/s12939-018-0800-6
76. Mounce L.T.A., Campbell J.L., Henley W.E., Tejerina Arreal M.C., Porter I., Valderas J.M.. **Predicting incident multimorbidity**. *Ann Fam Med* (2018) **16** 322-329. DOI: 10.1370/afm.2271
77. Ge L., Yap C.W., Heng B.H.. **Sex differences in associations between multimorbidity and physical function domains among community-dwelling adults in Singapore**. *PLoS One* (2018) **13**. DOI: 10.1371/journal.pone.0197443
78. Camargo-Casas S., Suarez-Monsalve S., Zepeda M.U.P., García-Peña C., Cano-Gutierrez C.A.. **Multimorbidity, depressive symptoms, and self-reported health in older adults: a secondary analysis of the Sabe Bogota Study**. *Rev Invest Clin* (2018) **70** 192-197. DOI: 10.24875/RIC.18002478
79. Amaral T.L.M., Amaral C.D.A., de Lima N.S., Herculano P.V., Do Prado P.R., Monteiro G.T.R.. **Multimorbidity, depression and quality of life among elderly people assisted in the family health strategy in Senador Guiomard, Acre, Brazil**. *Ciên Saúde Colet* (2018) **23** 3077-3084. DOI: 10.1590/1413-81232018239.22532016
80. Puth M.T., Weckbecker K., Schmid M., Münster E.. **Prevalence of multimorbidity in Germany: impact of age and educational level in a cross-sectional study on 19,294 adults**. *BMC Public Health* (2017) **17** 1-7. DOI: 10.1186/s12889-017-4833-3
81. Waterhouse P., Van Der Wielen N., Banda P.C., Channon A.A.. **The impact of multi-morbidity on disability among older adults in South Africa: do hypertension and socio-demographic characteristics matter?**. *Int J Equity Health* (2017) **16** 1-10. DOI: 10.1186/s12939-017-0537-7
82. Wang L., Palmer A.J., Cocker F., Sanderson K.. **Multimorbidity and health-related quality of life (HRQoL) in a nationally representative population sample: implications of count versus cluster method for defining multimorbidity on HRQoL**. *Health Qual Life Outcomes* (2017) **15** 1-12. DOI: 10.1186/s12955-016-0580-x
83. Kunna R., San Sebastian M., Stewart Williams J.. **Measurement and decomposition of socioeconomic inequality in single and multimorbidity in older adults in China and Ghana: results from the WHO study on global AGEing and adult health (SAGE)**. *Int J Equity Health* (2017) **16** 1-17. DOI: 10.1186/s12939-017-0578-y
84. Lujic S., Simpson J.M., Zwar N., Hosseinzadeh H., Jorm L.. **Multimorbidity in Australia: comparing estimates derived using administrative data sources and survey data**. *PLoS One* (2017) **12** e0183817. DOI: 10.1371/journal.pone.0183817
85. Nunes B.P., Chiavegatto Filho A.D.P., Pati S.. **Contextual and individual inequalities of multimorbidity in Brazilian adults: a cross-sectional national-based study**. *BMJ Open* (2017) **7** e015885. DOI: 10.1136/bmjopen-2017-015885
86. Mini G.K., Thankappan K.R.. **Pattern, correlates and implications of non-communicable disease multimorbidity among older adults in selected Indian states: a cross-sectional study**. *BMJ Open* (2017) **7** 1-6. DOI: 10.1136/bmjopen-2016-013529
87. Larsen F.B., Pedersen M.H., Friis K., Gluèmer C., Lasgaard M.. **A latent class analysis of multimorbidity and the relationship to socio-demographic factors and health-related quality of life. A national population-based study of 162,283 Danish adults**. *PLoS One* (2017) **12** e0169426. DOI: 10.1371/journal.pone.0169426
88. Gu J., Chao J., Chen W.. **Multimorbidity in the community-dwelling elderly in urban China**. *Arch Gerontol Geriatr* (2017) **68** 62-67. DOI: 10.1016/j.archger.2016.09.001
89. Dhalwani N.N., Zaccardi F., O'Donovan G.. **Association between lifestyle factors and the incidence of multimorbidity in an older English population**. *J Gerontol A Biol Sci Med Sci* (2017) **72** 528-534. DOI: 10.1093/gerona/glw146
90. Nunes B.P., Camargo-Figuera F.A., Guttier M.. **Multimorbidity in adults from a southern Brazilian city: occurrence and patterns**. *Int J Public Health* (2016) **61** 1013-1020. DOI: 10.1007/s00038-016-0819-7
91. Picco L., Achilla E., Abdin E.. **Economic burden of multimorbidity among older adults: impact on healthcare and societal costs**. *BMC Health Serv Res* (2016) **16** 173. DOI: 10.1186/s12913-016-1421-7
92. Palladino R., Tayu Lee J., Ashworth M., Triassi M., Millett C.. **Associations between multimorbidity, healthcare utilisation and health status: evidence from 16 European countries**. *Age Ageing* (2016) **45** 431-435. DOI: 10.1093/ageing/afw044
93. Le Cossec C., Perrine A.L., Beltzer N., Fuhrman C., Carcaillon-Bentata L.. **Pre-frailty, frailty, and multimorbidity: prevalences and associated characteristics from two French national surveys**. *J Nutr Health Aging* (2016) **20** 860-869. DOI: 10.1007/s12603-016-0802-2
94. Vadrevu L., Kumar V., Kanjilal B.. **Rising challenge of multiple morbidities among the rural poor in India: a case of the Sundarbans in West Bengal**. *Int J Med Sci Public Health* (2016) **5** 343. DOI: 10.5455/ijmsph.2016.25082015129
95. Marengoni A., Angleman S., Meinow B.. **Coexisting chronic conditions in the older population: variation by health indicators**. *Eur J Intern Med* (2016) **31** 29-34. DOI: 10.1016/j.ejim.2016.02.014
96. Jovic D., Vukovic D., Marinkovic J.. **Prevalence and patterns of multi-morbidity in Serbian adults: a cross-sectional study**. *PLoS One* (2016) **11**. DOI: 10.1371/journal.pone.0148646
97. Su P., Ding H., Zhang W.. **The association of multimorbidity and disability in a community-based sample of elderly aged 80 or older in Shanghai, China**. *BMC Geriatr* (2016) **16** 178. DOI: 10.1186/s12877-016-0352-9
98. Ramond-Roquin A., Haggerty J., Lambert M., Almirall J., Fortin M.. **Different multimorbidity measures result in varying estimated levels of physical quality of life in individuals with multimorbidity: a cross-sectional study in the general population**. *Biomed Res Int* (2016) **2016**. DOI: 10.1155/2016/7845438
99. Lenzi J., Avaldi V.M., Rucci P., Pieri G., Fantini M.P.. **Burden of multimorbidity in relation to age, gender and immigrant status: a cross-sectional study based on administrative data**. *BMJ Open* (2016) **6**. DOI: 10.1136/bmjopen-2016-012812
100. Dung L.D., Long G.T.. **Gender differences in prevalence and associated factors of multi-morbidity among older persons in Vietnam**. *Int J Ageing Dev Ctries* (2016) **1** 113-132
101. Valadares A.L., Lui-Filho J.F., Costa-Paiva L., Pinto-Neto A.M.. **Middle-aged female sexual dysfunction and multimorbidity: a population-based study**. *Menopause* (2016) **23** 304-310. DOI: 10.1097/GME.0000000000000533
102. Pache B., Vollenweider P., Waeber G., Marques-Vidal P.. **Prevalence of measured and reported multimorbidity in a representative sample of the Swiss population disease epidemiology - chronic**. *BMC Public Health* (2015) **15** 1-8. DOI: 10.1186/s12889-015-1515-x
103. Afshar S., Roderick P.J., Kowal P., Dimitrov B.D., Hill A.G.. **Multimorbidity and the inequalities of global ageing: a cross-sectional study of 28 countries using the World Health Surveys**. *BMC Public Health* (2015) **15** 1-10. DOI: 10.1186/s12889-015-2008-7
104. Roberts K.C., Rao D.P., Bennett T.L., Loukine L., Jayaraman G.C.. **Prevalence and patterns of chronic disease multimorbidity and associated determinants in Canada**. *Health Promot Chronic Dis Prev Can* (2015) **35** 87-94. DOI: 10.24095/hpcdp.35.6.01
105. Arokiasamy P., Uttamacharya U., Jain K.. **The impact of multimorbidity on adult physical and mental health in low- and middle-income countries: what does the study on global ageing and adult health (SAGE) reveal?**. *BMC Med* (2015) **13** 178. DOI: 10.1186/s12916-015-0402-8
106. Ha N.T., Le N.H., Khanal V., Moorin R.. **Multimorbidity and its social determinants among older people in southern provinces, Vietnam**. *Int J Equity Health* (2015) **14** 50. DOI: 10.1186/s12939-015-0177-8
107. Wang S.B., D'Arcy C., Yu Y.Q.. **Prevalence and patterns of multimorbidity in northeastern China: a cross-sectional study**. *Public Health* (2015) **129** 1539-1546. DOI: 10.1016/j.puhe.2015.06.013
108. Wang R., Yan Z., Liang Y.. **Prevalence and patterns of chronic disease pairs and multimorbidity among older Chinese adults living in a rural area**. *PLoS One* (2015) **10**. DOI: 10.1371/journal.pone.0138521
109. Nunes B.P., Thumé E., Facchini L.A.. **Multimorbidity in older adults: magnitude and challenges for the Brazilian Health System**. *BMC Public Health* (2015) **15** 1172. DOI: 10.1186/s12889-015-2505-8
110. Chung R.Y., Mercer S., Lai F.T.T., Yip B.H.K., Wong M.C.S., Wong S.Y.S.. **Socioeconomic determinants of multimorbidity: a population-based household survey of Hong Kong Chinese**. *PLoS One* (2015) **10** e0140040. DOI: 10.1371/journal.pone.0140040
111. Ruel G., Lévesque J.F., Stocks N.. **Understanding the evolution of multimorbidity: evidences from the North West Adelaide Health Longitudinal Study (NWAHS)**. *PLoS One* (2014) **9**. DOI: 10.1371/journal.pone.0096291
112. Mahwati Y.. **Determinants of multimorbidity among the elderly determinants of multimorbidity among the elderly population in Indonesia**. *Natl Public Health J* (2014) **9** 187-193. DOI: 10.21109/kesmas.v9i2.516
113. Banjare P., Pradhan J.. **Socio-economic inequalities in the prevalence of multi-morbidity among the rural elderly in Bargarh district of Odisha (India)**. *PLoS One* (2014) **9**. DOI: 10.1371/journal.pone.0097832
114. Hien H., Berthé A., Drabo M.K.. **Prevalence and patterns of multimorbidity among the elderly in Burkina Faso: cross-sectional study**. *Trop Med Int Health* (2014) **19** 1328-1333. DOI: 10.1111/tmi.12377
115. Orueta J.F., Nuño-Solinís R., García-Alvarez A., Alonso-Morán E.. **Prevalence of multimorbidity according to the deprivation level among the elderly in the Basque Country**. *BMC Public Health* (2013) **13** 918. DOI: 10.1186/1471-2458-13-918
116. Aguiar L.B., Baccaro L.F., de Souza Santos Machado V., Pinto-Neto A.M., Costa-Paiva L.. **Disability and multimorbidity in women older than 50 years: a population-based household survey**. *Menopause* (2015) **22** 660-666. DOI: 10.1097/GME.0000000000000355
117. Alaba O., Chola L.. **The social determinants of multimorbidity in South Africa**. *Int J Equity Health* (2013) **12** 1. DOI: 10.1186/1475-9276-12-63
118. Wu F., Guo Y., Kowal P.. **Prevalence of major chronic conditions among older Chinese adults: the Study on Global AGEing and adult health (SAGE) wave 1**. *PLoS One* (2013) **8**. DOI: 10.1371/journal.pone.0074176
119. Phaswana-Mafuya N., Peltzer K., Chirinda W.. **Self-reported prevalence of chronic non-communicable diseases and associated factors among older adults in South Africa**. *Glob Health Action* (2013) **6** 20936. DOI: 10.3402/gha.v6i0.20936
120. Jerliu N., Toçi E., Burazeri G., Ramadani N., Brand H.. **Prevalence and socioeconomic correlates of chronic morbidity among elderly people in Kosovo: a population-based survey**. *BMC Geriatr* (2013) **13** 22. DOI: 10.1186/1471-2318-13-22
121. Kiliari N., Theodosopoulou E., Papanastasiou E.. **Multimorbidity and unmet citizens' needs and expectations urge for reforms in the health system of Cyprus: a questionnaire survey**. *JRSM Open* (2014) **5**. DOI: 10.1177/2042533313515860
122. Fuchs J., Busch M., Lange C., Scheidt-Nave C.. **Prevalence and patterns of morbidity among adults in Germany. Results of the German telephone health interview survey German Health Update (GEDA) 2009**. *Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz* (2012) **55** 576-586. DOI: 10.1007/s00103-012-1464-9
123. De Souza Santos MacHado V., Valadares A.L.R., Da Costa-Paiva L.S., Moraes S.S., Pinto-Neto A.M.. **Multimorbidity and associated factors in Brazilian women aged 40 to 65 years: a population-based study**. *Menopause* (2012) **19** 569-575. DOI: 10.1097/gme.0b013e3182455963
124. Kirchberger I., Meisinger C., Heier M.. **Patterns of multimorbidity in the aged population. Results from the KORA-Age study**. *PLoS One* (2012) **7**. DOI: 10.1371/journal.pone.0030556
125. Agborsangaya C.B., Lau D., Lahtinen M., Cooke T., Johnson J.A.. **Multimorbidity prevalence and patterns across socioeconomic determinants: a cross-sectional survey**. *BMC Public Health* (2012) **12** 201. DOI: 10.1186/1471-2458-12-201
126. Tucker-Seeley R.D., Li Y., Sorensen G., Subramanian S.. **Lifecourse socioeconomic circumstances and multimorbidity among older adults**. *BMC Public Health* (2011) **11** 313. DOI: 10.1186/1471-2458-11-313
127. Khanam M.A., Streatfield P.K., Kabir Z.N., Qiu C., Cornelius C., Wahlin Å.. **Prevalence and patterns of multimorbidity among elderly people in rural Bangladesh: a cross-sectional study**. *J Health Popul Nutr* (2011) **29** 406-414. DOI: 10.3329/jhpn.v29i4.8458
128. Taylor A.W., Price K., Gill T.K.. **Multimorbidity: not just an older person's issue**. *BMC Public Health* (2010) **10** 1-10. PMID: 20043862
129. Loza E., Jover J.A., Rodriguez L., Carmona L.. **Multimorbidity: prevalence, effect on quality of life and daily functioning, and variation of this effect when one condition is a rheumatic disease**. *Semin Arthritis Rheum* (2009) **38** 312-319. DOI: 10.1016/j.semarthrit.2008.01.004
130. Van Minh H., Ng N., Juvekar S.. **Self-reported prevalence of chronic diseases and their relation to selected sociodemographic variables: a study in INDEPTH Asian sites, 2005**. *Prev Chronic Dis* (2008) **5** A86. PMID: 18558036
131. Wilk P., Stranges S., Bellocco R.. **Multimorbidity in large Canadian urban centres: a multilevel analysis of pooled 2015-2018 cross-sectional cycles of the Canadian Community Health Survey**. *J Comorb* (2021) **11**. DOI: 10.1177/26335565211058037
132. Tomita A., Leyna G.H., Kim H.Y.. **Patterns of multimorbidity and their association with hospitalisation: a population-based study of older adults in urban Tanzania**. *Age Ageing* (2021) **50** 1349-1360. DOI: 10.1093/ageing/afab046
133. Smith L., Shin J.I., Ghayda R.A.. **Physical multimorbidity and incident urinary incontinence among community-dwelling adults aged ≥50 years: findings from a prospective analysis of the Irish Longitudinal Study on Ageing**. *Age Ageing* (2021) **50** 2038-2046. DOI: 10.1093/ageing/afab151
134. Delpino F.M., Wendt A., Crespo P.A.. **Occurrence and inequalities by education in multimorbidity in Brazilian adults between 2013 and 2019: evidence from the National Health Survey**. *Rev Bras Epidemiol* (2021) **24**. DOI: 10.1590/1980-549720210016.supl.2
135. Marthias T., Anindya K., Ng N.. **Impact of non-communicable disease multimorbidity on health service use, catastrophic health expenditure and productivity loss in Indonesia: a population-based panel data analysis study**. *BMJ Open* (2021) **11**. DOI: 10.1136/bmjopen-2020-041870
136. Zhang C., Xiao S., Shi L.. **Urban-rural differences in patterns and associated factors of multimorbidity among older adults in China: a cross-sectional study based on Apriori algorithm and multinomial logistic regression**. *Front Public Health* (2021) **9** 707062. DOI: 10.3389/fpubh.2021.707062
137. Lin M.H., Chang C.Y., Wu D.M., Lu C.H., Kuo C.C., Chu N.F.. **Relationship of multimorbidity, obesity status, and grip strength among older adults in Taiwan**. *Int J Environ Res Public Health* (2021) **18** 7540. DOI: 10.3390/ijerph18147540
138. Nicholson K., Griffith L.E., Sohel N., Raina P.. **Examining early and late onset of multimorbidity in the Canadian Longitudinal Study on Aging**. *J Am Geriatr Soc* (2021) **69** 1579-1591. DOI: 10.1111/jgs.17096
139. Bezerra de Souza D.L., Oliveras-Fabregas A., Espelt A.. **Multimorbidity and its associated factors among adults aged 50 and over: a cross-sectional study in 17 European countries**. *PLoS One* (2021) **16**. DOI: 10.1371/journal.pone.0246623
140. Koyanagi A., Smith L., Shin J.I.. **Multimorbidity and subjective cognitive complaints: findings from 48 low- and middle-income countries of the World Health Survey 2002-2004**. *J Alzheimers Dis* (2021) **81** 1737-1747. DOI: 10.3233/JAD-201592
141. Shi X., Lima S.M.D.S., Mota C.M.M., Lu Y., Stafford R.S., Pereira C.V.. **Prevalence of multimorbidity of chronic noncommunicable diseases in Brazil: population-based study**. *JMIR Public Health Surveill* (2021) **7**. DOI: 10.2196/29693
142. Wang X., Yao S., Wang M.. **Multimorbidity among two million adults in China**. *Int J Environ Res Public Health* (2020) **17** 3395. DOI: 10.3390/ijerph17103395
143. He K., Zhang W., Hu X.. **Relationship between multimorbidity, disease cluster and all-cause mortality among older adults: a retrospective cohort analysis**. *BMC Public Health* (2021) **21** 1080. DOI: 10.1186/s12889-021-11108-w
144. Ballesteros S.M., Moreno-Montoya J., Grooten W.J.A., Barrera-López P., de la Hoz-Valle J.A.. **Socioeconomic variation of multimorbidity in Colombian older adults**. *Sci Rep* (2021) **11** 22738. DOI: 10.1038/s41598-021-02219-w
145. Mohamed S.F., Haregu T.N., Uthman O.A.. **Multimorbidity from chronic conditions among adults in urban slums: the AWI-Gen Nairobi Site Study findings**. *Glob Heart* (2021) **16** 6. DOI: 10.5334/gh.771
146. Kanungo S., Ghosal S., Kerketta S.. **Association of oral health with multimorbidity among older adults: findings from the longitudinal ageing study in India, wave-1, 2017-2019**. *Int J Environ Res Public Health* (2021) **18** 12853. DOI: 10.3390/ijerph182312853
147. Oh H., Glass J., Narita Z., Koyanagi A., Sinha S., Jacob L.. **Discrimination and multimorbidity among Black Americans: findings from the National Survey of American Life**. *J Racial Ethn Health Disparities* (2021) **8** 210-219. DOI: 10.1007/s40615-020-00773-z
148. King D.E., Xiang J., Pilkerton C.S.. **Multimorbidity trends in United States adults, 1988–2014**. *J Am Board Fam Med* (2018) **31** 503-513. DOI: 10.3122/jabfm.2018.04.180008
149. Bowling C.B., Deng L., Sakhuja S., Morey M.C., Jaeger B.C., Muntner P.. **Prevalence of activity limitations and association with multimorbidity among US adults 50 to 64 years old**. *J Gen Intern Med* (2019) **34** 2390-2396. DOI: 10.1007/s11606-019-05244-8
150. Keats M.R., Cui Y., DeClercq V.. **Multimorbidity in Atlantic Canada and association with low levels of physical activity**. *Prev Med* (2017) **105** 326-331. DOI: 10.1016/j.ypmed.2017.10.013
151. Quinaz Romana G., Kislaya I., Salvador M.R., Cunha Gonçalves S., Nunes B., Dias C.. **Multimorbilidade em Portugal: Dados do Primeiro Inquérito Nacional de Saúde com Exame Físico [Multimorbidity in Portugal: results from the first National Health Examination Survey]**. *Acta Med Port* (2019) **32** 30-37. DOI: 10.20344/amp.11227
152. Souza A.S.S., Faerstein E., Werneck G.L.. **Multimorbidade e uso de serviços de saúde em indivíduos com restrição de atividades habituais: Estudo Pró-Saúde [Multimorbidity and use of health services by individuals with restrictions on habitual activities: the Pró-Saúde Study]**. *Cad Saúde Pública* (2019) **35**. DOI: 10.1590/0102-311X00155118
153. Costa Â.K., Bertoldi A.D., Fontanella A.T.. **Does socioeconomic inequality occur in the multimorbidity among Brazilian adults?**. *Rev Saude Publica* (2020) **54** 138. DOI: 10.11606/s1518-8787.2020054002569
154. Keomma K., Bousquat A., César C.L.G.. **Prevalence of multimorbidity in older adults in São Paulo, Brazil: a study with ISA-Capital**. *Rev Saude Publica* (2022) **56** 69. DOI: 10.11606/s1518-8787.2022056004252
155. Jürisson M., Pisarev H., Uusküla A., Lang K., Oona M., Kalda R.. **Prevalence of chronic conditions and multimorbidity in Estonia: a population-based cross-sectional study**. *BMJ Open* (2021) **11**. DOI: 10.1136/bmjopen-2021-049045
156. Ho I.S., Azcoaga-Lorenzo A., Akbari A.. **Variation in the estimated prevalence of multimorbidity: systematic review and meta-analysis of 193 international studies**. *BMJ Open* (2022) **12**. DOI: 10.1136/bmjopen-2021-057017
157. Huaquía-Díaz A.M., Chalán-Dávila T.S., Carrillo-Larco R.M., Bernabe-Ortiz A.. **Multimorbidity in Latin America and the Caribbean: a systematic review and meta-analysis**. *BMJ Open* (2021) **11**. DOI: 10.1136/BMJOPEN-2021-050409
158. Xu X., Mishra G.D., Jones M.. **Mapping the global research landscape and knowledge gaps on multimorbidity: a bibliometric study**. *J Glob Health* (2017) **7** 010414. DOI: 10.7189/jogh.07.010414
159. Harrison C., Britt H., Miller G., Henderson J.. **Examining different measures of multimorbidity, using a large prospective cross-sectional study in Australian general practice**. *BMJ Open* (2014) **4**. DOI: 10.1136/BMJOPEN-2013-004694
160. Ferrer A., Formiga F., Sanz H., Almeda J., Padrós G.. **Multimorbidity as specific disease combinations, an important predictor factor for mortality in octogenarians: the Octabaix study**. *Clin Interv Aging* (2017) **12** 223-231. DOI: 10.2147/CIA.S123173
161. Quiñones A.R., Markwardt S., Botoseneanu A.. **Multimorbidity combinations and disability in older adults**. *J Gerontol A Biol Sci Med Sci* (2016) **71** 823-830. DOI: 10.1093/gerona/glw035
162. van den Bussche H., Koller D., Kolonko T.. **Which chronic diseases and disease combinations are specific to multimorbidity in the elderly? Results of a claims data based cross-sectional study in Germany**. *BMC Public Health* (2011) **11** 101. DOI: 10.1186/1471-2458-11-101
163. Schulze T., Maercker A., Horn A.B.. **Mental health and multimorbidity: psychosocial adjustment as an important process for quality of life**. *Gerontology* (2014) **60** 249-254. DOI: 10.1159/000358559
164. Marengoni A., Angleman S., Melis R.. **Aging with multimorbidity: a systematic review of the literature**. *Ageing Res Rev* (2011) **10** 430-439. DOI: 10.1016/j.arr.2011.03.003
165. Wang Z., Peng W., Li M.. **Association between multimorbidity patterns and disability among older people covered by long-term care insurance in Shanghai, China**. *BMC Public Health* (2021) **21**. DOI: 10.1186/s12889-021-10463-y
|
---
title: Differences in nutritional risk assessment between NRS2002, RFH-NPT and LDUST
in cirrhotic patients
authors:
- Peiyan Zhang
- Qi Wang
- Mengran Zhu
- Pingping Li
- Yuzhen Wang
journal: Scientific Reports
year: 2023
pmcid: PMC9971362
doi: 10.1038/s41598-023-30031-1
license: CC BY 4.0
---
# Differences in nutritional risk assessment between NRS2002, RFH-NPT and LDUST in cirrhotic patients
## Abstract
Nutritional status is an independent predictor of outcome in cirrhosis patients. Nutritional Risk Screening 2002 (NRS2002), Royal Free Hospital-Nutritional Prioritizing Tool (RFH-NPT), and Liver Disease Undernutrition Screening Tool (LDUST) were employed to detect cirrhosis with malnutrition risk in this work. Meanwhile, their diagnostic performances were compared to find the best screening method. This work aimed to establish the sarcopenia cut-off value of the transversal psoas thickness index (TPTI), and identify the risk factors for malnutrition. Cirrhosis patients who were admitted to Heibei Gerneral hospital from April 2021 to October 2021 and underwent abdominal CT examination were enrolled. 78 patients were assessed by NRS2002, RFH-NPT, and LDUST. The Global Leadership Initiative for Malnutrition (GLIM) criteria were selected as the gold standard for the diagnosis of malnutrition. Meanwhile the cut-off value of sarcopenia was established based on the TPTI of malnourished patients. Logistic regression analysis was adopted to assess the influencing factors of malnutrition risk and malnutrition. The prevalence of malnutrition was $42.31\%$. The prevalence of malnutrition risk was $32.1\%$, $61.5\%$, and $62.8\%$ with NRS2002, RFH-NPT, and LDUST, respectively. NRS2002 presented the best specificity compared with the other methods, while RFH-NPT showed the highest sensitivity. The optimal gender-specific TPTI cut-off value for diagnosing sarcopenia was determined as TPTI < 14.56 mm/m (male) and TPTI < 8.34 mm/m (female). In the multivariate analysis, ascites was associated with malnutrition risk, while sarcopenia showed a significant risk for malnutrition. NRS2002 and RFH-NPT were superior to LDUST at detecting the malnutrition in cirrhosis patients diagnosed according to GLIM criteria. The gender-specific TPTI cut-off value was TPTI < 14.56 mm/m (male) and TPTI < 8.34 mm/m (female). Malnutrition risk should be screened for patients with ascites as soon as possible. In addition, it was important to evaluate malnutrition in sarcopenia patients in time.
## Introduction
Malnutrition is a common complication of cirrhosis. It can prolong the hospital stays, especially the time in intensive care unit (ICU), accelerate the decompensation, and increase mortality1. Early detection and diagnosis are vitally associate to prognosis of cirrhosis patients2.
However, nutrition assessment and nutrition screening are challenges to this group of patients in China. It is caused by multiple factors, including the absence of gold standard for nutrition assessment, unavailable validated screening tools, and few studies research on the difference of screening tools.
The prevalence of malnutrition in cirrhosis ranges from 5 to $99\%$, depending on the severity of liver cirrhosis and the diagnostic criteria3–5. Multiple scientific nutrition societies have established the GLIM criteria as a global diagnostic reference for malnutrition6. As far as we know, the GLIM criteria for cirrhosis have been investigated in some studies.
Various malnutrition risk screening tools have been proposed with different advantages and disadvantages, but their results vary among different studies. NRS20027 is most widely used in the world due to some limitations in the application of cirrhosis because of the absence of ascites parameters. RFH-NPT8 andLDUST9 are established for patients with liver cirrhosis, while their data in China are few.
Cirrhosis is complicated by sarcopenia10. TPTI is an easy, reliable, and repeatable parameter for assessing the sarcopenia since it is computed tomography (CT)-based and just needs to measure the psoas diameter11. However, data on TPTI in cirrhosis are still lacking in China and the cut-off value of sarcopenia is rare.
In this work, the GLIM was undertake as the gold standard to diagnose malnutrition and three malnutrition risk screening tools (NRS2002, RFH-NPT and LDUST) were evaluated. This work aimed to compare the differences in these 3 tools and identify their superiorities in live cirrhosis. Meanwhile, the best cut-off value of sarcopenia was identified by comparing the TPTI in nutrition group and malnutrition group.
## Material and methods
This work included 78 cirrhosis patients who were admitted to Heibei General Hospital from April 2021 to October 2021 and underwent abdominal CT examination. All patients met the diagnostic criteria of the ‘Chinese Guidelines on the Management of Liver Cirrhosis’. Patients with any of below conditions had to be excluded: severe cardiopulmonary disease, cancer, inflammatory bowel disease, and other metabolic disorders. In addition patients over 80 and under 18 years old were excluded.
This work was approved by the ethics committee of the Heibei General Hospital [2021-267] and was performed in accordance with the Declaration of Helsinki. Written informed consents were obtained from all the participants prior to the enrollment.
The data included gender, age, height, body weight, body mass index (BMI), routine blood examination, biochemical parameters and other laboratory tests. The body weight was adjusted according to water content retention. The weight was reduced by $5\%$,$10\%$, and $15\%$ in mild, moderate, and severe ascites, respectively and it was reduced by $5\%$ in patients who had combined peripheral edema. The classical nutritional markers included calf circumference, arm circumference, and arm muscle circumference.
The abdominal CT images of all subjects were obtained within 2 weeks. The axial psoas muscle thickness (APMT) was defined as the largest antero-posterior diameter of the right psoas muscle at the L3 level and recorded in millimeters(mm), and the longest transverse diameter which was perpendicular to APMT was defined as the psoas muscle transversal psoas muscle thickness (TPMT), TPTI = TPMT / height (mm/m)12 (Appendix 1).
The NRS2002, RFH-NPT, and LDUST were used for malnutrition risk screening, and the GLIM criteria were selected for the assessment of malnutrition. Patients who were assessed to be at risk of malnutrition were included in the risk group. According to the GLIM criteria, the 78 patients were divided into a malnutrition group and a nutrition group. All tools were performed by an experienced clinical physician.
The NRS2002 contained nutritional impairment scores, severity of disease scores, and an age adjustment in which over 70 years adds one point13. Nutritional impairment ranged from 1 to 3 according to the weight loss over $5\%$ in 3 to 1 months or the food intake reduced < $50\%$, 50–$75\%$ and > $75\%$ compared with the values in previous week. It can be recorded as 3 points if the BMI < 18.5 kg/m2. The severity of disease was assessed as 1, 2, and 3 points, respectively, in different situations. Cirrhosis was recorded as 1 point. The total score of three parts classified patients into a no risk (< 3 points) and a malnutrition risk (≥ 3 points) group.
RFH-NPT was performed in three steps14. First, the alcoholic hepatitis patients or those who needed tube feeding were assessed as high risk. Those who failed to meet the conditions above needed to assess ascites or edema and its impacts on food intake and weight loss. If ascites or edema was found, 1 point was assigned. The food intake impact on ascites or edema was recorded as 0 to 2 points, which corresponded to no influence, occasional influent, and influent. If the food intake was reduced by half in 5 days, 2 points were given, or otherwise 0 points were assigned. The weight loss in the past 3–6 months ranged from 0 to 2 points, which meant for no, being unable to evaluate, and yes, respectively. Those who did not have fluid overload were estimated using BMI, unplanned weight loss, and dietary intake. BMI was recorded as 0 to 2 points according to > 20 kg/m2, 18.5–20 kg/m2 and < 18.5 kg/m2, respectively. The unplanned weight loss in the past 3–6 months ranged from 0 to 2 points according to < $5\%$, 5–$10\%$, and > $10\%$, respectively. It should assess whether the patient is in acute exacerbation and has or may have no nutritional intake for more than 5 days. It was recorded as 2 points if yes. Patients with a score of 2–7, were at high risk, and they were at low risk with a score of 0 and moderate risk with a score of 1.
The LDUST included six questions. Food intake, weight loss, muscle loss, swelling or fluid, and daily activities were determined as grades A, B and C15. 5 or more points were included in grade A, which was identified as no risk, while 2–5 points in grade B and < 2 points in grade C were identified as malnutrition risk.
There were two components in the GLIM criteria: phenotypic criterion and etiologic criterion6. Phenotypic criteria included weight loss, low BMI, and reduced muscle mass, while etiologic criterion contained reduced food intake or assimilation and inflammation. All the cirrhosis patients were considered to present chronic disease-related inflammation, so they all met the etiologic criterion. Weight loss which went down $5\%$ in 6 months or $10\%$ over 6 months was assessed in this work. Meanwhile, the low BMI (BMI < 18.5 kg/m2 for patients < 70 years old or BMI < 20 kg/m2 for patients > 70 years old) was assessed. If the weight and BMI were in normal ranges, their calf circumference can be assessed to check whether the muscle mass was reduced. According to a Japanese standard, reduced muscle mass was assessed by calf circumference ≤ 30 cm (male) and ≤ 29 cm (female) since there was no sarcopenia standard in China16. The patients who met one phenotypic criterion and one etiologic criterion were diagnosed with malnutrition in the GLIM criteria.
The TPTI of the malnutrition group based on the GLIM criteria was employed to establish the cut-off value of sarcopenia in cirrhosis.
## Statistical analysis
Data were analyzed using SPSS 25 and Graphpad Prism 8.3.0. Data that which conformed to the normal distribution were expressed as the mean ± standard deviation (SD). The comparison between the two groups was performed by using the t test. Non-normally distributed data were represented by Median (P25 ~ P75) and the comparison between two groups was performed using the Mann–Whitney U test. The count data was expressed as percentages (%), and the two groups were compared by the Chi-square test. Correlation and consistency were assessed by spearman correlation coefficient analysis and Kappa consistency test, respectively. Receiver operating characteristic (ROC) curves were adopted to evaluate the ability of the three screening tools to distinguish the malnourished patients. The GLIM criteria were selected as the reference, ROC curves were generated for TPTI, and the best cut-off values to diagnose sarcopenia were determined using the Youden Index. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were estimated to test the quality of screening tools to predict malnutrition. Logistic regression analysis was employed to assess the influencing factors of malnutrition risk, malnutrition, and sarcopenia. P value less than 0.05 meant that the difference was significant.
## Characteristics of patients
The characteristics of the 78 included subjects were detailed in Table 1. The mean age was 54.5 ± 12.54 years old, and 50 ($64.1\%$) of the patients were male. The most frequent etiology of cirrhosis was hepatitis B virus (HBV) ($43.6\%$), followed by alcohol ($24.4\%$). Based on the stage of the disease, $82.1\%$ of patients was decompensated, $44.9\%$ of patients had Child-A, $38.5\%$ had Child-B, and $16.7\%$ had Child-C. According to the GLIM criteria, malnutrition was diagnosed in 33 patients ($42.3\%$). Comparison between the malnutrition group and the nutrition group revealed that there were significant differences in TPTI, arm circumference, NRS2002, RFH-NPT, and LDUST. Malnourished patients were observed with reduced TPTI and arm circumference. Table 1Baseline data of total patients and per nutritional state, stratified by GLIM criteria. CharacteristicsTotal ($$n = 78$$)Malnutrition ($$n = 33$$)No malnutrition ($$n = 45$$)P-valueAge54.50 ± 12.5451.60 ± 12.2756.62 ± 12.430.081Gender, n (%) Male50 ($64.1\%$)24 ($72.7\%$)26 ($57.8\%$)0.174Weight (Kg)67.28 ± 14.0265.47 ± 15.6268.60 ± 12.740.332BMI (Kg/m2)24.64 ± 4.3623.66 ± 4.9825.36 ± 3.740.088Etiology, n (%) HBV34 ($43.6\%$)16 ($48.5\%$)18 ($40.0\%$)0.677 Alcohol19 ($24.4\%$)10 ($30.3\%$)9 ($20.0\%$) HCV4 ($5.1\%$)1 ($3.0\%$)3 ($6.7\%$) Cholestatic7 ($9.0\%$)3 ($9.1\%$)4 ($8.9\%$) MAFLD6 ($7.7\%$)1 ($3.0\%$)5 ($11.1\%$) Drug1 ($1.3\%$)0 ($0\%$)1 ($2.2\%$) Unknown7 ($9.0\%$)2 ($6.1\%$)5 ($11.1\%$)Decompensated, n (%)64 ($82.1\%$)25 ($75.8\%$)39 ($86.7\%$)0.244Ascites, n (%)31 ($39.7\%$)15 ($45.5\%$)16 ($35.6\%$)0.483MELD sore10.08 ± 5.1310.55 ± 5.029.74 ± 5.240.495Child–Pugh, n (%) Child-A/B64 ($82.1\%$)23 ($69.7\%$)41 ($91.11\%$)0.019 Child-C14 ($17.9\%$)10 ($30.3\%$)4 ($8.89\%$)TPTI (mm/m)14.15 ± 4.4112.98 ± 4.1515.00 ± 4.450.045Arm circumference (cm)27.71 ± 4.2426.55 ± 4.3828.56 ± 3.960.037Tricipital skinfold thickness (cm)1.25 (0.85–1.85)1.05 (0.66–1.75)1.45 (0.95–1.95)0.051Arm muscle circumference (cm)23.27 ± 3.5822.73 ± 3.3823.88 (20.94–25.88)0.255Calf circumference (cm)34.63 ± 4.1334.12 ± 4.9035.00 ± 3.470.382Total protein (g/L)65.25 (59.70–70.78)63.86 ± 11.4665.00 ± 9.450.634Albumin (g/L)34.40 (27.15–38.55)32.01 ± 7.1133.96 ± 6.710.220Prealbumin (g/L)9.70 (6.83–13.53)9.00 (6.05–11.90)10.83 ± 4.120.077LDUST, n (%) No risk29 ($37.2\%$)5 ($15.2\%$)24 ($53.3\%$) < 0.001 Risk49 ($62.8\%$)28 ($84.8\%$)21 ($46.7\%$)RFH-NPT, n (%) No risk30 ($38.5\%$)3 ($9.1\%$)27 ($60.0\%$) < 0.001 Risk48 ($61.5\%$)30 ($90.9\%$)18 ($40.0\%$)NRS2002, n (%) No risk53 ($67.9\%$)11 ($33.3\%$)42 ($93.3\%$) < 0.001 Risk25 ($32.1\%$)22 ($66.7\%$)3 ($6.7\%$)BMI body mass index, HBV hepatitis B virus, HCV hepatitis C virus, MAFLD metabolic dysfunction-associated fatty liver disease, MELD the model for end-stage liver disease, TPTI transversal psoas thickness index, LDUST liver disease undernutrition screening tool, RFH-NPT royal free hospital-nutritional prioritizing tool, NRS2002 nutritional risk screening 2002.
## Comparison between screening tools and GLIM criteria
In this work, NRS2002 exhibited the highest correlation with the GLIM criteria based on spearman correlation coefficient (Spearman’s r value of 0.635), followed by RFH-NPT and GLIM criteria (Spearman’s r value of 0.517) (Table 2). Besides, the highest consistency was observed between the NRS2002 and GLIM criteria based on the Kappa consistency test (Kappa = 0.62) (Table 3).Table 2Spearman correlation coefficients of different tools. NRS2002RFH-NPTLDUSTGLIMNRS200210.4860.4150.635RFH-NPT0.48610.3730.517LDUST0.4150.37310.390GLIM0.6350.5170.3901NRS2002 nutritional risk screening 2002, RFH-NPT royal free hospital-nutritional prioritizing tool, LDUST liver disease undernutrition screening tool, GLIM global leadership initiative for malnutrition. Table 3Kappa coefficients of different tools. NRS2002RFH-NPTLDUSTGLIMNRS20021.000.410.340.62RFH-NPT0.411.000.370.48LDUST0.340.371.000.36GLIM0.620.480.361.00NRS2002 nutritional risk screening 2002, RFH-NPT royal free hospital-nutritional prioritizing tool, LDUST liver disease undernutrition screening tool, GLIM global leadership initiative for malnutrition.
## Validation of screening tools
Compared to the GLIM criteria as the benchmark for malnutrition diagnosis, NRS2002 presented the highest accuracy in detecting malnutrition based on the area under the ROC curve (AUC), followed by RFH-NPT (AUC, 0.800 and AUC, 0.755, respectively) (Fig. 1, Table 4). RFH-NPT showed the highest sensitivity ($90.91\%$) and the lowest specificity ($60.00\%$), while NRS2002 presented the highest specificity ($93.33\%$) and the lowest sensitivity ($66.67\%$). LDUST showed a moderate sensitivity and the lowest specificity. It was noticeable that the accuracy of NRS2002 and RFH-NPT was higher than that of LDUST.Figure 1ROC curves of the screening tools for the diagnosis of malnutrition using GLIM criteria as a benchmark. Table 4Measures of diagnostic validity of screening tools. Validity criteriaNRS2002RFH-NPTLDUSTAUC ($95\%$ CI)0.800 (0.692–0.908)0.755 (0.646–0.863)0.691 (0.573–0.809)Accuracy, %82.0573.0866.67Sensitivity, %66.6790.9184.85Specificity, %93.3360.0053.33Youden index0.600.510.38PPV, %88.0062.5057.14NPV, %79.2590.0082.76PLR10.002.271.82NLR0.360.150.28AUC the area under the ROC curves, PPV positive predictive value, NPV negative predictive value, PLR positive likelihood ratio, NLR negative likelihood ratio, NRS2002 nutritional risk screening 2002, RFH-NPT royal free hospital-nutritional prioritizing tool, LDUST liver disease undernutrition screening tool.
## Defining cut-off values for sarcopenia that correlate with malnutrition
Using GLIM criteria as the reference, the cut-off values were calculated and derived from ROC curves and the Youden index (Fig. 2). The TPTI-AUC was 0.713 (0.569–0.858, $P \leq 0.01$) for men, and 0.754 (0.533–0.976, $P \leq 0.05$) for women. Cut-off values for sarcopenia were TPTI < 14.56 mm/m (male) and < 8.34 mm/m (female). With the threshold of sarcopenia, sarcopenia was diagnosed in 31($39.74\%$) cirrhosis patients, of which 22 ($70.97\%$) were malnourished and 9 ($29.03\%$) were not malnourished (X2 = 17.313, $P \leq 0.001$). Regardless of gender, the rate of sarcopenia in the malnutrition group was significantly higher than that in the nutrition group ($P \leq 0.01$). The incidence showed no differences in Child-A, Child-B, and Child-C, and gender. Figure 2Predictive results of TPTI on the diagnosis of malnutrition using GLIM criteria as a benchmark.
## Logistic regression analyses for malnutrition risk and malnutrition
The logistic regression analyses revealed that albumin, prealbumin, lactate dehydrogenase, alkaline phosphatase, ascites, and sarcopenia were influencing factors for malnutrition risk. Among them, only ascites was an independent risk factor (Table 5). TPTI and sarcopenia were the interfering factors to malnutrition, while only sarcopenia was an independent factor, which was revealed by the multivariate analyses (Table 6).Table 5Univariate and multivariate logistic regression analyses for malnutrition risk. ParameterUnivariate logistic regressionMultivariate logistic regressionOR$95\%$ CIP-valueOR$95\%$ CIP-valueAlbumin (g/L)0.8980.821–0.982 < 0.05Prealbumin (g/L)0.8350.729–0.955 < 0.05Lactate dehydrogenase (U/L)1.0111.001–1.021 < 0.05Alkaline phosphatase (U/L)1.0191.003–1.035 < 0.05Ascites4.3751.146–16.697 < 0.055.6651.284–24.994 < 0.05Sarcopenia4.3751.146–16.697 < 0.05Table 6Univariate and multivariate logistic regression analyses for malnutrition. ParameterUnivariate logistic regressionMultivariate logistic regressionOR$95\%$ CIP-valueOR$95\%$ CIP-valueTPTI (mm/m)0.880.778–0.989 < 0.05Sarcopenia8.002.861–22.370 < 0.0018.7392.543–30.038 < 0.001TPTI transversal psoas thickness index.
## Discussion
There are various malnutrition methods to assess cirrhosis, such as SGA and RFH-SGA. These two methods are time-consuming and poorly applied in China. In 2018, the global nutrition societies established the GLIM criteria to form a global consensus on the diagnosis of malnutrition. After that, the use of GLIM criteria was in many crowds, such as cancer or older adults17,18.
At present, few studies have examined on the application of the GLIM criteria for malnutrition in cirrhosis. In addition, some studies have found that compared with SGA and RFH-GA, GLIM can better predict the mortality of patients with chronic liver disease19,20. Meanwhile, there are increasingly more studies selecting the GLIM standard as the gold standard in patients with liver cirrhosis5,21.
Boulhosa RSSB and Diego were the first batch researchers to focus on this field5,22. With the use of GLIM criteria, $42.3\%$ of cirrhosis patients were identified to be complicated with malnutrition in this work, which is similar to data from Boulhosa RSSB and Diego ($57.2\%$ and $38.1\%$, respectively)5,22.
The assessment using the GLIM criteria is performed in two steps. First, whether patients were at risk of malnutrition should be assessed using nutritional screening tools. Second, the GLIM criteria are adopted to assess the malnutrition risk patients with malnutrition. This work aimed to determine which nutritional screening tools could obtain findings consistent findings with the results of the GLIM criteria, so that they could be applied in the first step of Glim in the follow-up clinical work. This work validated three screening tools in predicting malnutrition when diagnosed based on the GLIM criteria.
NRS2002, RFH-NPT, and LDUST were validated and effective methods for nutritional screening of patients with liver cirrhosis. The GLIM group recommend the NRS2002, but this work also found that the RFH-NPT showed the same correlation and consistency as NRS2002.
This work confirmed that NRS2002 and RFH-NPT were the optimal tools applied in patients with the best specificity and sensitivity, respectively. This conduction is in line with previous studies that revealed superiorities of NRS2002 and RFH-NPT in specificity and sensitivity23,24.
Compared with NRS2002, GLIM takes BMI, liver cirrhosis, and food intake as evaluation items. The difference is that the BMI of GLIM is more detailed according to patients’ ages, unlike NRS2002, where is the BMI of all patients is 18.5 kg/m2 as a reference. In terms of food intake, the NRS2002 pays more attention to changes in a short time (within 1 week), while the GLIM expands the range to changes within two weeks. Besides, the NRS2002 assesses the weight loss within 1–3 months, while the GLIM only evaluates weight loss within 6 months of malnutrition grading. There are so many similarities between them that the good consistency between them is easy to understand.
RFH-NPT focuses on fluid retention, and patients are divided into two groups according to different results, and receive different assessment steps for BMI and weight loss are carried out. GLIM is proposed for patients with various diseases. It is not included in the item of fluid retention, but in the assessment of muscle loss. For patients with liver cirrhosis, the assessment of muscle loss is more objective and accurate than fluid retention.
LSUST is a subjective evaluation form for patients. Many Chinese patients will ignore their weight, muscle loss, and diet changes, so the consistency between LSUST and GLIM is not as good as that of the other two methods.
NRS2002 score needs over 3 when it is considered as malnutrition risk, RFH-NPT score reaches 1 point, while LDUST needs two of the six subjective questions to get B or C. NRS2002 is relatively more rigid to confirm the diagnose malnutrition risk. Therefore, patients who were evaluated as having no malnutrition risk by NRS2002 can be more accurately diagnosed as having no malnutrition. RFH-NPT considers patients with ascites or peripheral edema as malnutrition risk, while LDUST needs to merge the fluid overload with another entry. NRS2002 does not involve fluid retention. This is the reason that RFH-NPT shows the highest sensitivity. Among the three methods, NRS2002 and RFH-NPT can be better applied to screen the malnutrition risk in cirrhosis, while LDUST is more subjective and can be used for self-screening management in a short time, so it is applicable to clinical practice.
TPTI is a simple and rapid method of diagnosing sarcopenia. In contrast with the L3 skeletal muscle index which is commonly used abroad25,26, TPTI can be measured by abdominal CT but requires no special measurement software. This work found the cut-off value of sarcopenia as TPTI < 14.56 mm/m (male) and < 8.34 mm/m (female), which was higher than the value mentioned in a foreign study27, which may be related to the distinct way to get thecut off values. Psternosrto27 obtained his threshold by the mortality of patients in a follow-up period, but the malnutrition of patients was adopted in this work. The average TPTI of deaths was lower than that of malnourished patients. The threshold in men is higher than that in women, which is similar all over the world, because of the influence of gender hormones that men have more muscle groups. Even after adjustment for height, the amount of psoas muscle still shows a significant gender difference28.
Because nutritional status is an intervention that can impact prognosis, it is expect to identify the risk factors for different nutritional stages and to take patients with risk factors seriously. The data in this work suggested that the malnutrition risk of patients with ascites was 5.665 times as that of patients without ascites. Patients with ascites are more likely to suffer from anorexia and induced intake due to severe gastrointestinal symptoms and early satiety29,30. In addition, rest energy expenditure increases significantly in these patients31. The nutrition assessment for most ascites patients with increased weight is often ignored by clinicians. According to the literature, Xiaoyu Wang and their colleagues used RFH-NPT to assess 135 cirrhosis patients and found that ascites increased the malnutrition risk, which was similar to the findings of this work32. Malnutrition can accelerate the progression of ascites and significantly increase the probability of refractory ascites33,34. However, ascites was not proven to be a risk factor for malnutrition in this work, which may be because of the small sample size and the use of GLIM criteria as a standard for malnutrition assessment.
The impact factors of sarcopenia have been studied for many years. Studies have shown that Child–Pugh, arm circumference, arm muscle circumference, age, myostatin, albumin, and prealbumin are the impact factors35–37. The research of Hsu CS found that the prevalence of sarcopenia increased in men and patients with ascites, liver failure, and kidney failure38. A similar conclusion was reached by this work, in which prealbumin was an independent factor.
On the other hand, there were some limitations in this work. First, it may be subject to population bias due to the small sample size and single center study may cause population bias. Second, the effects of nutritional support therapy on the prognosis of patients with malnutrition or sarcopenia have not been studied.
## Conclusion
As far as we know, this work was the first study in China to compare the NRS2002, RFH-NPT, and LDUST with GLIM criteria in patients with cirrhosis. In the light of GLIM criteria, the prevalence of malnutrition was $42.3\%$, which confirmed that NRS2002 and RFH-NPT could better screen for malnutrition risk in clinical application in patients with cirrhosis. Moreover, TPTI < 14.56 mm/m (male) and < 8.34 mm/m (female) were proven to be the cut-off values of male and female patients with sarcopenia, respectively. In conclusion, this work verified that clinicians should attach importance to the patients with ascites and sarcopenia. The datasets generated or analyzed in this work were available from the corresponding author on reasonable request.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30031-1.
## References
1. Kuftinec G, Ram Bhamidimarri K, Pearlman M. **Malnutrition in cirrhosis: Frequent but overlooked**. *Liver Transplant.* (2019) **25** 1743-4. DOI: 10.1002/lt.25660
2. Palmer LB, Kuftinec G, Pearlman M. **Nutrition in Cirrhosis**. *Curr. Gastroenterol. Rep.* (2019) **21** 38. DOI: 10.1007/s11894-019-0706-5
3. Juakiem W, Torres DM, Harrison SA. **Nutrition in cirrhosis and chronic liver disease**. *Clin. Liver Dis.* (2014) **18** 179-190. DOI: 10.1016/j.cld.2013.09.004
4. Maharshi S, Sharma BC, Srivastava S. **Malnutrition in cirrhosis increases morbidity and mortality**. *J. Gastroenterol. Hepatol.* (2015) **30** 1507-1513. DOI: 10.1111/jgh.12999
5. Casas Deza D, Betoré Glaria ME, Sanz-parís A. **Mini nutritional assessment—Short form is a useful malnutrition screening tool in patients with liver cirrhosis, using the Global Leadership Initiative for malnutrition criteria as the gold standard**. *Nutr. Clin. Prac.* (2021) **36** 1003-10. DOI: 10.1002/ncp.10640
6. Cederholm T, Jensen GL, Correia M. **GLIM criteria for the diagnosis of malnutrition—A consensus report from the global clinical nutrition community**. *J. Cachexia Sarcopenia Muscle* (2019) **10** 207-217. DOI: 10.1002/jcsm.12383
7. Kondrup J, Rasmussen HH, Hamberg O. **Nutritional risk screening (NRS 2002): A new method based on an analysis of controlled clinical trials**. *Clin. Nutr.* (2003) **22** 321-336. DOI: 10.1016/S0261-5614(02)00214-5
8. Borhofen SM, Gerner C, Lehmann J. **The royal free hospital-nutritional prioritizing tool is an independent predictor of deterioration of liver function and survival in cirrhosis**. *Dig. Dis. Sci.* (2016) **61** 1735-1743. DOI: 10.1007/s10620-015-4015-z
9. Booi AN, Menendez J, Norton HJ. **Validation of a screening tool to identify undernutrition in ambulatory patients with liver cirrhosis**. *Nutr. Clin. Pract.* (2015) **30** 683-689. DOI: 10.1177/0884533615587537
10. Kim G, Kang SH, Kim MY. **Prognostic value of sarcopenia in patients with liver cirrhosis: A systematic review and meta-analysis**. *PLoS One* (2017) **12** e0186990. DOI: 10.1371/journal.pone.0186990
11. Huguet A, Latournerie M, Debry PH. **The psoas muscle transversal diameter predicts mortality in patients with cirrhosis on a waiting list for liver transplantation: A retrospective cohort study**. *Nutrition* (2018) **51–52** 73-9. DOI: 10.1016/j.nut.2018.01.008
12. Paternostro R, Bardach C, Hofer BS. **Prognostic impact of sarcopenia in cirrhotic patients stratified by different severity of portal hypertension**. *Liver Int.* (2021) **41** 799-809. DOI: 10.1111/liv.14758
13. Kondrup J, Allison SP, Elia M. **ESPEN guidelines for nutrition screening 2002**. *Clin. Nutr.* (2003) **22** 415-421. DOI: 10.1016/S0261-5614(03)00098-0
14. Amodio P, Bemeur C, Butterworth R. **The nutritional management of hepatic encephalopathy in patients with cirrhosis: International Society for Hepatic Encephalopathy and Nitrogen Metabolism Consensus**. *Hepatology* (2013) **58** 325-336. DOI: 10.1002/hep.26370
15. White JV, Guenter P, Jensen G. **Consensus statement: Academy of Nutrition and Dietetics and American Society for Parenteral and Enteral Nutrition: Characteristics recommended for the identification and documentation of adult malnutrition (undernutrition)**. *JPEN J. Parenter. Enter. Nutr.* (2012) **36** 275-283. DOI: 10.1177/0148607112440285
16. Maeda K, Ishida Y, Nonogaki T. **Reference body mass index values and the prevalence of malnutrition according to the Global Leadership Initiative on malnutrition criteria**. *Clin. Nutr.* (2020) **39** 180-184. DOI: 10.1016/j.clnu.2019.01.011
17. Verstraeten LMG, van Wijngaarden JP, Pacifico J. **Association between malnutrition and stages of sarcopenia in geriatric rehabilitation inpatients: RESORT**. *Clin. Nutr.* (2021) **40** 4090-4096. DOI: 10.1016/j.clnu.2021.02.007
18. Lengelé L, Bruyère O, Beaudart C. **Malnutrition, assessed by the Global Leadership Initiative on Malnutrition (GLIM) criteria but not by the mini nutritional assessment (MNA), predicts the incidence of sarcopenia over a 5-year in the SarcoPhAge cohort**. *Aging Clin. Exp. Res.* (2021) **33** 1507-17. DOI: 10.1007/s40520-021-01880-5
19. Allard JP, Keller H, Gramlich L. **GLIM criteria has fair sensitivity and specificity for diagnosing malnutrition when using SGA as comparator**. *Clin. Nutr.* (2020) **39** 2771-2777. DOI: 10.1016/j.clnu.2019.12.004
20. Miwa T, Hanai T, Nishimura K. **Usefulness of the Global Leadership Initiative on malnutrition criteria to predict sarcopenia and mortality in patients with chronic liver disease**. *Hepatol. Res.* (2022). DOI: 10.1111/hepr.13816
21. Sánchez-Torralvo FJ, Porras N, Ruiz-García I. **Usefulness of muscle ultrasonography in the nutritional assessment of adult patients with cystic fibrosis**. *Nutrients* (2022) **14** 3377. DOI: 10.3390/nu14163377
22. Boulhosa R, Lourenço RP, Côrtes DM. **Comparison between criteria for diagnosing malnutrition in patients with advanced chronic liver disease: GLIM group proposal versus different nutritional screening tools**. *J. Hum. Nut. Diet.* (2020) **33** 862-8. DOI: 10.1111/jhn.12759
23. Georgiou A, Papatheodoridis GV, Alexopoulou A. **Evaluation of the effectiveness of eight screening tools in detecting risk of malnutrition in cirrhotic patients: The KIRRHOS study**. *Br. J. Nutr.* (2019) **122** 1368-1376. DOI: 10.1017/S0007114519002277
24. Wu Y, Zhu Y, Feng Y. **Royal Free Hospital-Nutritional Prioritizing Tool improves the prediction of malnutrition risk outcomes in liver cirrhosis patients compared with nutritional risk screening 2002**. *Br. J. Nutr.* (2020) **124** 1293-1302. DOI: 10.1017/S0007114520002366
25. Cruz-Jentoft AJ, Baeyens JP, Bauer JM. **Sarcopenia: European consensus on definition and diagnosis: Report of the European Working Group on Sarcopenia in older people**. *Age Ageing* (2010) **39** 412-423. DOI: 10.1093/ageing/afq034
26. Chen LK, Liu LK, Woo J. **Sarcopenia in Asia: Consensus report of the Asian Working Group for Sarcopenia**. *J. Am. Med. Dir. Assoc.* (2014) **15** 95-101. DOI: 10.1016/j.jamda.2013.11.025
27. Paternostro R, Lampichler K, Bardach C. **The value of different CT-based methods for diagnosing low muscle mass and predicting mortality in patients with cirrhosis**. *Liver Int.* (2019) **39** 2374-2385. DOI: 10.1111/liv.14217
28. Baumgartner RN, Waters DL, Gallagher D. **Predictors of skeletal muscle mass in elderly men and women**. *Mech. Ageing Dev.* (1999) **107** 123-136. DOI: 10.1016/S0047-6374(98)00130-4
29. Cheung K, Lee SS, Raman M. **Prevalence and mechanisms of malnutrition in patients with advanced liver disease, and nutrition management strategies**. *Clin. Gastroenterol. Hepatol.* (2012) **10** 117-125. DOI: 10.1016/j.cgh.2011.08.016
30. Aqel BA, Scolapio JS, Dickson RC. **Contribution of ascites to impaired gastric function and nutritional intake in patients with cirrhosis and ascites**. *Clin. Gastroenterol. Hepatol.* (2005) **3** 1095-1100. DOI: 10.1016/S1542-3565(05)00531-8
31. **Clinical practice guidelines on nutrition in chronic liver disease**. *J. Hepatol.* (2019) **70** 172-93. DOI: 10.1016/j.jhep.2018.06.024
32. Wang X, Feng H, Hui Y. **Neutrophil-to-lymphocyte ratio is associated with malnutrition risk estimated by the Royal Free Hospital-Nutritional Prioritizing Tool in hospitalized cirrhosis**. *JPEN J. Parenter. Enter. Nutr.* (2021). DOI: 10.1002/jpen.2097
33. Henkel AS, Buchman AL. **Nutritional support in patients with chronic liver disease**. *Nat. Clin. Pract. Gastroenterol. Hepatol.* (2006) **3** 202-209. DOI: 10.1038/ncpgasthep0443
34. Pérez-reyes E, Rivera-sánchez J, Servín-caamaño A. **Malnutrition is related to a higher frequency of serious complications in patients with cirrhosis**. *Revista Médica del Hospital General de México* (2016) **79** 11-6. DOI: 10.1016/j.hgmx.2015.04.003
35. Alexopoulos T, Vasilieva L, Kontogianni MD. **Myostatin in combination with creatine phosphokinase or albumin may differentiate patients with cirrhosis and sarcopenia**. *Am. J. Physiol. Gastrointest. Liver Physiol.* (2021) **321** G543-G551. DOI: 10.1152/ajpgi.00184.2021
36. Tandon P, Low G, Mourtzakis M. **A model to identify sarcopenia in patients with cirrhosis**. *Clin. Gastroenterol. Hepatol.* (2016) **14** 1473-80.e3. DOI: 10.1016/j.cgh.2016.04.040
37. Xiu S, Sun L, Mu Z. **Low prealbumin levels are associated with sarcopenia in older men with type 2 diabetes mellitus: A cross-sectional study**. *Nutrition* (2021) **91–92** 111415. DOI: 10.1016/j.nut.2021.111415
38. Hsu CS, Kao JH. **Sarcopenia and chronic liver diseases**. *Expert Rev. Gastroenterol. Hepatol.* (2018) **12** 1229-1244. DOI: 10.1080/17474124.2018.1534586
|
---
title: Integrating oculomics with genomics reveals imaging biomarkers for preventive
and personalized prediction of arterial aneurysms
authors:
- Yu Huang
- Cong Li
- Danli Shi
- Huan Wang
- Xianwen Shang
- Wei Wang
- Xueli Zhang
- Xiayin Zhang
- Yijun Hu
- Shulin Tang
- Shunming Liu
- Songyuan Luo
- Ke Zhao
- Ify R. Mordi
- Alex S. F. Doney
- Xiaohong Yang
- Honghua Yu
- Xin Li
- Mingguang He
journal: The EPMA Journal
year: 2023
pmcid: PMC9971392
doi: 10.1007/s13167-023-00315-7
license: CC BY 4.0
---
# Integrating oculomics with genomics reveals imaging biomarkers for preventive and personalized prediction of arterial aneurysms
## Abstract
### Objective
Arterial aneurysms are life-threatening but usually asymptomatic before requiring hospitalization. Oculomics of retinal vascular features (RVFs) extracted from retinal fundus images can reflect systemic vascular properties and therefore were hypothesized to provide valuable information on detecting the risk of aneurysms. By integrating oculomics with genomics, this study aimed to (i) identify predictive RVFs as imaging biomarkers for aneurysms and (ii) evaluate the value of these RVFs in supporting early detection of aneurysms in the context of predictive, preventive and personalized medicine (PPPM).
### Methods
This study involved 51,597 UK Biobank participants who had retinal images available to extract oculomics of RVFs. Phenome-wide association analyses (PheWASs) were conducted to identify RVFs associated with the genetic risks of the main types of aneurysms, including abdominal aortic aneurysm (AAA), thoracic aneurysm (TAA), intracranial aneurysm (ICA) and Marfan syndrome (MFS). An aneurysm-RVF model was then developed to predict future aneurysms. The performance of the model was assessed in both derivation and validation cohorts and was compared with other models employing clinical risk factors. An RVF risk score was derived from our aneurysm-RVF model to identify patients with an increased risk of aneurysms.
### Results
PheWAS identified a total of 32 RVFs that were significantly associated with the genetic risks of aneurysms. Of these, the number of vessels in the optic disc (‘ntreeA’) was associated with both AAA (β = −0.36, $$P \leq 6.75$$e−10) and ICA (β = −0.11, $$P \leq 5.51$$e−06). In addition, the mean angles between each artery branch (‘curveangle_mean_a’) were commonly associated with 4 MFS genes (FBN1: β = −0.10, $$P \leq 1.63$$e−12; COL16A1: β = −0.07, $$P \leq 3.14$$e−09; LOC105373592: β = −0.06, $$P \leq 1.89$$e−05; C8orf81/LOC441376: β = 0.07, $$P \leq 1.02$$e−05). The developed aneurysm-RVF model showed good discrimination ability in predicting the risks of aneurysms. In the derivation cohort, the C-index of the aneurysm-RVF model was 0.809 [$95\%$ CI: 0.780–0.838], which was similar to the clinical risk model (0.806 [0.778–0.834]) but higher than the baseline model (0.739 [0.733–0.746]). Similar performance was observed in the validation cohort, with a C-index of 0.798 (0.727–0.869) for the aneurysm-RVF model, 0.795 (0.718–0.871) for the clinical risk model and 0.719 (0.620–0.816) for the baseline model. An aneurysm risk score was derived from the aneurysm-RVF model for each study participant. The individuals in the upper tertile of the aneurysm risk score had a significantly higher risk of aneurysm compared to those in the lower tertile (hazard ratio = 17.8 [6.5–48.8], $$P \leq 1.02$$e−05).
### Conclusion
We identified a significant association between certain RVFs and the risk of aneurysms and revealed the impressive capability of using RVFs to predict the future risk of aneurysms by a PPPM approach. Our finds have great potential to support not only the predictive diagnosis of aneurysms but also a preventive and more personalized screening plan which may benefit both patients and the healthcare system.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s13167-023-00315-7.
## Biomarkers of arterial aneurysm are needed in the realm of prediction, prevention and personalized medicine (PPPM)
Arterial aneurysm is one of the most common diseases affecting the artery after atherosclerosis and represents a severe condition due to the increased risk of dissection or rupture which has a mortality rate greater than $80\%$ [1, 2]. According to data reported by the ‘2019 Global Burden of Disease’ study, the global number of aneurysm-related deaths is expected to increase by $42\%$ to achieve 244,685 in 2030, compared to 172,426 in 2019 [3]. Irrespective of severity, the onset of aneurysms is usually asymptomatic. Therefore, current medical care strategies for aneurysms need to be improved, especially in terms of predictive diagnosis and targeted screening, such improvement may permit early prevention and personalized management [4].
Effective and safe biomarkers are needed for predicting aneurysms or the development of preventive screening protocols. According to current international guidelines, only a limited number of general biomarkers, such as age and sex, have been applied to facilitate routine aneurysm screening [5, 6]. However, these biomarkers neglect the heterogeneous characteristics of the disease [7–11] and are not precise enough to benefit individuals who do not meet the current screening criteria. A recent study proposed an effective and accurate screening approach for abdominal aortic aneurysm (AAA) based on a broader spectrum of clinical biomarkers [12], thereby providing a foundation for developing and testing new biomarker-based screening approaches for aneurysm. Therefore, the integration of numerous biological data, such as genetics and medical images, and the identification of non-invasive and convenient biomarkers, could facilitate the prediction of aneurysm and provide options for personalized medical prevention strategies, which is an essential step in the context of PPPM [4].
## Genomics is a fundamental consideration for PPPM
Exploring biomarkers associated with aneurysms identified from routinely collected electronic health records might suffer from misclassification bias due to the asymptomatic nature of the disease. Genomics can provide unique data about an individual’s unique risk of disease [13, 14] and genetic risks of aneurysms were stable and precise, which might assist in identifying robust biomarkers [15, 16]. There have been several large-scale genome-wide association studies (GWAS) that report a potential shared genetic aetiology among the main types of aneurysms [17–21], including AAA, thoracic aortic aneurysm (TAA), intracranial aneurysm (ICA) and Marfan syndrome (MFS) in which the main complication is TAA [22, 23]. Therefore, the genetic risk of aneurysm can be a reasonably precise proxy for aneurysms since this risk can (i) better reflect the biological mechanism of aneurysms and (ii) better represent an individual’s risk and reduce misclassification bias due to the delay or lack of aneurysm detection.
## Oculomics features imaging biomarkers of aneurysms and is an emerging tool for PPPM
Identifying medical imaging biomarkers is a key component in PPPM since they are essential to patient-tailored disease prediction or therapy [24]. For aneurysm screening or prediction, the use of image biomarkers derived from ultrasound or computer tomography (CT) is not cost-effective and can expose patients to radioactive agents, thus limiting the broad application of these techniques as routine healthcare processes. Oculomics is an emerging research area that utilizes ocular information derived from non-invasive and easily accessible ocular examinations to gain insights into systemic health. Some researchers have reported the use of oculomics in the context of PPPM [25], especially for the use of oculomics of retinal vascular morphological features (RVFs) as prediction biomarkers for cardiovascular diseases (CVDs) [26–29]. However, only a few studies have investigated their relationship with aneurysms [30, 31]. Since aneurysms mainly demonstrate irreversible changes in the vascular morphology, we hypothesized that the oculomics of RVF may contains more imaging biomarkers for aneurysm than for other CVDs and these RVFs may facilitate the early identification of patients with a high risk of aneurysm from the perspectives of PPPM. This strategy will benefit vulnerable populations, especially those that find it difficult to reach advanced medical resources [32].
## Working hypothesis
Compared to clinical phenotypic data, oculomics derived from retinal fundus image is considered more convenient and efficient, non-invasive, with lower costs (for all resources), and more importantly, reflects systemic vascular properties. Previously, we developed a deep learning algorithm, the Retinal-based Microvascular Health Assessment System (RMHAS) [33] that can generate data on the omics scale from RVFs, thus providing more comprehensive oculomics of RVFs. In this study, personalized genetic risk of aneurysm was integrated with the oculomics by phenome-wide association analysis (PheWAS) to identify biological meaningful aneurysm-RVF biomarkers. Followed by testing their capability in differentiating aneurysms, this study aimed to investigate the value of oculomics in (i) providing imaging biomarkers for the predictive diagnosis of arterial aneurysms and (ii) identifying subjects at high risk and supporting early detection of arterial aneurysms. This could be particularly useful as it might allow us to develop a more refined and targeted screening approach for aneurysm that aligns with the aims of PPPM.
## Study population
The UK *Biobank is* a large-scale and prospective cohort study with over 500,000 participants aged 40–69 years that were recruited between 2006 and 2010. This study collected extensive phenotypic and genotypic data from each participant with their informed consent. Further details of the UK *Biobank data* and the protocols involved have been described elsewhere [34]. In brief, a total of 502,505 individuals in 22 assessment centres across the UK agreed to participate in this study (a response rate of $5.5\%$). During baseline assessment, participants completed comprehensive questionnaires, provided a range of physical measurements and provided biological samples. Detailed health-related events were achieved through linkage to national electronic health record datasets. Ophthalmic assessments, including retinal fundus photography, were introduced to the baseline assessment in 2009 for six assessment centres. Participants without fundus imaging or genetic data were excluded.
This study was approved by the National Information Governance Board for Health and Social Care and the NHS Northwest Multicentre Research Ethics Committee (Reference: 11/NW/0382) and the *Biobank consortium* (Application number: 62489). Since de-identified data in a public dataset was used, the Medical Research Ethics Committee of Guangdong Provincial People’s Hospital waived the requirements to obtain ethical approval.
## Genetic risk of aneurysm
In total, 488,377 participants were genotyped by the UK Biobank Axiom genotyping array. Stringent quality control was performed and genotype imputation was carried out using the Haplotype Reference Consortium (HRC) reference panel; further details relating to the genotype and quality control information were described by Bycroft et al. [ 35].
Genetic information for abdominal, thoracic and intracranial aneurysms, and MFS, were taken from genome-wide association studies (GWAS) for the corresponding traits. For AAA, 12 genetic loci associated with AAA were identified from a meta-analysis of 4972 cases and 99,858 controls [17]; for TAA, we used three single nucleotide polymorphisms (SNPs) that were found to be associated with TAA in a previous GWAS of 1351 affected individuals and 18,295 controls [19]; and for intracranial aneurysms, we used 17 SNPs that were previously identified by a GWAS of 10,754 cases and 306,882 controls [18]. These SNPs were used to generate different weighted genetic risk scores (GRSs) for each participant using the --score function implemented in PLINK 2.0 [36]. The magnitude of the association with different aneurysms (GWAS beta coefficient) was used as the weighting factor for each variant included in the GRS.
In the analysis of MFS, we selected five MFS SNPs that were allocated to different genes [20, 21]. rs10519177 was located in the region of FBN1, a common MFS gene, while the other four SNPs rs2297676 (located within COL16A1), rs1432302 (located within LOC105373592), rs3020167 (located between C8orf81 and LOC441376) and rs2278601 (located within SMAD6) were considered to be associated with the extreme arterial phenotype of MFS such as thoracic aneurysm or dissection. The genotypes of these SNPs were extracted by PLINK 2.0 for each participant using the --extract and --recode functions.
## Oculomics of retinal vascular features
The 45-degree non-mydriatic retinal fundus and optical coherence tomography (OCT) images of the optic disc and macular were captured using a spectral domain OCT (Topcon 3D OCT 1000 Mk2, Topcon Corp, Tokyo, Japan) for each eye. A total of 131,238 fundus images were obtained from 66,500 participants. Only the retinal fundus images from the right eye were used for analysis. A machine learning system, referred to as the Retinal-based Microvascular Health Assessment System (RMHAS), was previously developed and validated to automatically and quickly extract and quantify retinal microvascular features [33]. For each image, pan-retinal vessel geometric parameters, such as calibre, complexity, length, tortuosity and branching angle were quantified. A demonstration of the algorithm is publicly available at: https://www.retinavessel.com/ (Supplementary Method 1).
Compared to previous retinal vessel analysis tools, RMHAS can extract pan-retinal vessel features and generate 91 RVFs for each fundus image. Details of the analysing pipeline and the 91 RVFs are given in Supplementary Figure 1 and Supplementary Table 1.
## Characteristics associated with the risk of aneurysm
According to previous prediction models for aneurysm [12, 37], modifiable clinical risk factors, demographic information and social economic status were considered as characteristics that can be used to predict aneurysm events. Baseline characters included age, sex, systolic (SBP) and diastolic (DBP) blood pressure, blood lipid, glycated haemoglobin, smoking status, body mass index (BMI), baseline cardiovascular disease (defined by International Classification of Diseases [ICD-10] codes I20–I25 and I60–I69 excluding I67.0 and I67.1), baseline diabetes (self-reported type 1 or type 2 diabetes), self-reported hyperlipidaemia and the use of blood pressure- or cholesterol-lowering medications or anti-diabetic medications. Refractive error was measured using a Tomey RC 5000 autorefractor and data from the right eye was used for analysis (Supplementary Method 2 and Supplementary Table 2).
## Identifying aneurysm-RVF associations by PheWAS
PheWAS was performed between RVFs and each aneurysm GRS or SNP. *In* general, each RVF was regressed against a GRS or a SNP while adjusting for age and sex in the main analysis and additionally adjusted for SBP, Townsend deprivation score, smoking and refractive error in the sensitivity analysis. To adjust for multiple testing (we run 91 regressions for each PheWAS), Bonferroni correction was applied thus a P value less than 5.50e−04 ($\frac{0.05}{91}$) was considered statistically significant. The R package ‘PheWAS’ [38] was used to perform statistical analysis. Venn diagrams were used to identify the overall and shared aneurysm-RVFs driven by the effect of aneurysm or arterial dissection genes.
## Assessment of aneurysm risk
The outcome was an 8-year risk of hospital admission due to aneurysm or symptomatic aneurysm defined by the earliest recorded event of fatal or non-fatal aneurysm since recruitment. Subjects were selected by their ICD10 code, OPCS4 (Classification of Interventions and Procedures) code, death record and self-report disease information. From the UK *Biobank data* set, data field ‘41270’, ‘40001’ and ‘40002’ were selected to define participants as ‘aneurysm dissection’, ‘thoracic aneurysm’, ‘abdominal aneurysm’, ‘thoracoabdominal aneurysm’, ‘brain aneurysm’ or ‘other aneurysm’ by ICD-10 code ‘I71-I72’, ‘I671’ or ‘I670’. Data field ‘41272’, ‘41200’ and ‘41210’ were also selected to define participants who had aneurysm surgery by OPCS4 code defined as ‘L18’, ‘L19’, ‘L27’, ‘L28’, ‘L424’, ‘L464’, ‘L254’, ‘L33’, ‘L48’, ‘L49’, ‘L533’, ‘L56’, ‘L57’, ‘L623’, ‘L624’ and ‘L705’. Finally, data fields ‘20002’ and ‘20004’ were selected to define self-reported aneurysm by choosing codes ‘1425’, ‘1492’ and ‘1591’ and ‘1592’ (Supplementary Table 3).
To derive the survival model, the recruitment date served as the beginning of the time at risk for each participant, and the period at risk terminated on the earlier of the first qualifying aneurysm event or the end of follow-up. The retinal images were taken between 2009 and 2010, and the disease records ended in 2018. Hence, follow-up was limited to a maximum of 8 years for each participant for the 8-year aneurysm risk.
## Model derivation and validation
At first, participants with missing clinical risk data or RVFs were excluded; this resulted in 26,964 participants with a complete dataset in the cohort. Then, the complete cohort was divided into a derivation cohort ($70\%$ of the cohort, $$n = 18$$,954) and a validation cohort ($30\%$ of the cohort, $$n = 8010$$). We developed three models, one was the ‘baseline model’, including baseline age and sex as covariates; and the other two models with two distinct covariate sets, one including an additional 20 clinical risk factors, the other with the aneurysm-RVFs identified by previous PheWAS (Supplementary Method 3). We constructed a Cox model for these three models in the derivation cohort, and then run these separately in the validation cohort. The performance of the Cox models, over 8 years of follow-up, was tested by Harrell’s C-index separately in both the derivation and validation cohorts, using 2000 bootstraps (performed by the R package ‘rms’). The $95\%$ CIs were also calculated based on the bootstrapping runs.
## The predictive capability of aneurysm-RVF score
Finally, based on the metric of the aneurysm-RVF model in the derivation cohort, the survival probability was calculated for each participant in the complete cohort and used as an aneurysm-RVF score. This score was categorized into tertiles and the difference in the time-to-aneurysm probability across each tertile of the score was evaluated using Kaplan-Meier curve; log-rank tests were used to calculate P values. For sensitivity analysis, participants who have ever been in hospital due to aneurysm were removed, and the association of the aneurysm-RVF score with the first aneurysm was assessed.
## Statistical analysis
Continuous variables are presented as mean and standard deviation (SD) if approximately symmetrically distributed, and median and interquartile range (IQR) if skewed. Categorical variables are presented as counts and percentages. All analyses were two-sided and a P value of <0.05 was considered statistically significant. All analyses were performed using R 4.0.4 software or Stata14.
## Baseline population
The baseline cohort contained 51,597 participants with retinal vascular measurements and genetic information; these data were analysed by PheWAS. At the time the retinal images were taken, the median age of the participants was 56.0 (IQR: 14) years, $54.9\%$ were female and $86.2\%$ were white. With regard to clinical measurements, the median SBP was 135 (IQR: 24.5) mmHg, and DBP was 81.5 (IQR: 14). Median total cholesterol was 5.65 (IQR: 1.49) mmol/L, while for HDL, LDL and triglycerides, the mean values were 1.43 (IQR: 0.512) mmol/L, 3.50 (IQR: 1.15) mmol/L and 1.41 (IQR: 1.02) mmol/L, respectively. Mean HbA1C was 35.1 (IQR: 5.1) mmol/mol. The population had a relatively elevated BMI (26.6 [IQR: 5.71] kg/m2) and $56.3\%$ had never been a smoker. Regarding medication history, $20.3\%$ had never taken blood pressure-lowering drugs, $3.5\%$ had taken blood glucose-lowering medication and $16.9\%$ had taken lipid-lowering medication. On average, 2244 ($4.3\%$) of the participants had a history of CVD, 548 ($1.1\%$) reported a history of aneurysm, 374,485 ($72.6\%$) had been diagnosed with hypertension, 2505 ($4.8\%$) had diabetes and 23,145 ($44.9\%$) had hyperlipidaemia. The median refractive status was 0.05 (IQR: 2.23) dioptre. Out of the total number of participants, 201 had an aneurysm event after recruitment (Table 1).Table 1Baseline characteristics of the participantsContinuous variableN (all)MedianIQRAge (years)51,5975714BMI (kg/m2)51,36026.65.71Townsend deprivation51,527−1.684.33SBP (mmHg)51,42213524DBP (mmHg)51,42281.514HbA1c (mmol/mol)47,81035.15.1Total cholesterol (mmol/L)48,2195.651.49HDL (mmol/L)45,9801.430.51LDL (mmol/L)48,1383.501.15Triglycerides (mmol/L)48,1731.411.02Refractive error (D)51,0960.052.23Category variableN (all)NPercentageSex51,597 Female28,$30354.90\%$ Male23,$29445.10\%$Ethnicity51,484 Non-white$711813.80\%$ White44,$36686.20\%$Smoker51,294 Never29,$06456.3\%$ Former17,$43433.8\%$ Current$47969.3\%$Blood pressure lowing drugs (yes)51,59710,$46320.3\%$Blood glucose lowing drugs (yes)51,$59718263.5\%$Blood lipid lowing drugs (yes)51,$597874216.9\%$Previous history of CVD (yes)51,$59722444.3\%$Previous history of aneurysm (yes)51,$5975481.1\%$Previous history of hyperlipidaemia (yes)51,59723,$14544.9\%$Previous history of hypertension (yes)51,59737,$48572.6\%$Previous history of diabetes (yes)51,$59725024.8\%$
## PheWAS results relating to the genetic risks of artery aneurysms
First, we performed three PheWASs to identify RVFs that were associated with the GRS of AAA, TAA and ICA, respectively. Nine RVFs that represent the vessel calibre (maximum, mean, stander deviation of central or overall vessel calibre), complexity (number of vascular trees) and tortuosity (vessel curvature) were associated with AAA GRS after Bonferroni correction. Of these, ‘ntreeA’, which refers to the number of artery vessels passing through the optic disc, demonstrated the strongest association with AAA GRS (β = −0.36, $$P \leq 6.75$$e−10) (Fig. 2A). In the sensitivity analysis, after additional adjusting for SBP, smoking, deprivation score and refractive status, four RVFs representing vessel calibres were no longer significant and only five RVFs remained significant (Supplementary Table 4). For TAA and ICA, only the mean artery calibre (‘w_mean_mean_a’) (β = 0.05, $$P \leq 8.37$$e−05) and the number of vascular trees in the optic disc (‘ntreeA’) (β = −0.11, $$P \leq 5.51$$e−06) were associated with the genetic risk of thoracic and intracranial aneurysms, respectively. The ‘ntreeA’ was common for both AAA and ICA (Fig. 2C). The association between ‘w_mean_neam_a’ and TAA GRS disappeared after adjusting for more covariates in the sensitivity analysis (Supplementary Table 4).
In summary, in the main analysis, we found that ten unique RVFs were associated with the genetic risk of the three types of aneurysms (Fig. 1). The majority of the RVFs were associated with AAA GRS, and ‘ntreeA’ was associated with both AAA and ICA.Fig. 1Manhattan plot for the PheWAS of genetic aneurysm risks with RVFs. The x-axis represents the 91 RVFs, y-axis represent the Z-score (Z-score = β/SE) of the PheWAS findings. Different symbols represent different genetic risks of aneurysm, the blue/red colour represents whether the P value is passing the Bonferroni correction. AAA, abdominal aortic aneurysm; TAA, thoracic aortic aneurysm; ICA, intracranial aneurysm
## PheWAS results relating to the genetic risks of Marfan syndrome
We then performed five PheWASs to identify RVFs that were associated with five different MFS genes, respectively. For the most common MFS gene FBN1, 26 RVFs were associated with the mutation of the corresponding SNP, and the number of artery segments (‘nseg_a’, reflecting vessel complexity) showed the strongest association (β = −1.67, $$P \leq 5.07$$e−17). After adjusting for covariates, especially refractive status, 21 RVFs remained significant (Fig. 2B, D, Supplementary Table 5). For the remaining MFS SNPs, three RVFs for rs2297676 (COL16A1), three RVFs for rs1432302 (LOC105373592), three RVFs for rs3020167 (C8orf81/LOC441376) and ten RVFs for rs2278601 (SMAD6) were identified in the main PheWAS analysis (Fig. 1). Sensitivity analysis revealed similar findings (Supplementary Table 5).Fig. 2RVFs associated with aneurysm GRSs, and MFS SNPs identified by PheWAS. A Forest plot demonstrating the significant RVFs that were identified by PheWAS of AAA/TAA/ICA GRSs; B Forest plot demonstrating the most common RVFs that were identified by PheWAS of MFS SNPs; C Venn diagram of the PheWAS results showing the common RVFs associated with AAA/TAA/ICA GRSs; ‘A/B/C’ reflects the corresponding blocks of RVFs shown in A; D Venn diagram of the PheWAS results showing the common RVFs associated with MFS SNPs; ‘A/B’ reflects the corresponding blocks of RVFs shown in B For all significant RVFs identified by different MFS genes, there were some overlaps. For example, ‘curveangle_mean_a’, the mean segment angles between each artery branch at a length of ten pixels, was common for four MFS SNPs (FBN1_rs10519177: β = −0.10, $$P \leq 1.63$$e−12; COL16A1_rs2297676: β = −0.07, $$P \leq 3.14$$e−09; LOC105373592_rs1432302: β = −0.06, $$P \leq 1.89$$e−05; C8orf81/LOC441376_rs3020167: β = 0.07, $$P \leq 1.02$$e−05). And ‘ntreeA’ was shared by four other MFS SNPs (FBN1_rs10519177: β = −0.06, $$P \leq 2.68$$e−07; COL16A1_rs2297676: β = −0.06, $$P \leq 3.14$$e−05; LOC105373592_ rs1432302: β = −0.05, $$P \leq 5.02$$e−05; SMAD6_rs2278601: β = −0.07, $$P \leq 2.54$$e−10) (Fig. 2B and D).
Taking all the PheWAS results together, in the main analysis, 32 unique RVFs were associated with the genetic risks of aneurysms and MFS (we referred to these RVFs as aneurysm-RVF) (Fig. 1). In the sensitivity analysis, 26 aneurysm-RVFs were identified (Supplementary Table 4–5). Although each aneurysm GRS or MFS gene was associated with some specific aneurysm-RVFs, we still observed many aneurysm-RVFs that were commonly shared by different aneurysm risks (Supplementary Figure 2).
## Derivation of an aneurysm-RVF risk model and comparison with a clinical risk model
In the PheWAS stage, 32 RVFs were considered as aneurysm-RVFs. To investigate whether these RVFs were capable of predicting future aneurysm events, an aneurysm-RVF risk model was developed and compared with a clinical risk model and a baseline model. Details of the three models are listed in Supplementary Method 3.
In the derivation cohort, $54.2\%$ were women, the mean age was 55.2 years, and the Townsend deprivation score was −1.71 (IQR: 4.28). The majority of the participants never smoked ($57.1\%$) and had a slightly elevated BMI (27.2 ± 4.68 kg/m2). The summary statistics for clinical risk and the 32 aneurysm-RVFs are shown in Supplementary Table 6 and 7. There were 59 ($0.3\%$) incident cases of aneurysm during the follow-up period. The incidence of aneurysm was 4.06 ($95\%$ CI: 3.09–5.24) per 10,000 person-years, 2.16 ($95\%$ CI: 1.26–3.15) per 1000 person-years in women and 6.31 ($95\%$ CI: 4.55–8.54) in men per 10,000 person-years.
In all three models, older age and male gender were associated with an increased risk of hospital admission due to aneurysm over the follow-up period. In the clinical risk model, both taking blood pressure-lowering medication (HR: 2.78, $95\%$ CI: 1.43–5.39) and a previous history of aneurysm were identified as significant risk factors (HR: 18.75, $95\%$ CI: 9.93–35.41). In the aneurysm-RVF model, after adjustment for age, sex and previous history of aneurysm, both ‘curveangle_mean_v’ (HR: 1.44, $95\%$ CI: 1.08–1.93) and ‘nseg_v’ (HR: 1.73, $95\%$ CI: 1.03–2.90) were associated with an increased risk of aneurysm. Analysis also indicated that ‘curveangle_sd_v’ (HR: 0.68, $95\%$ CI: 0.50–0.93) was associated with a reduction of risk (Supplementary Table 8–10).
After 2000 rounds of bootstrapping within the internal tests, the C-index for the baseline model was 0.739 ($95\%$ CI: 0.739–0.746); for the clinical risk model, the C-index was 0.806 ($95\%$ CI: 0.778–0.834) and for the aneurysm-RVF model, the C-index was 0.809 ($95\%$ CI: 0.780–0.838) (Fig. 3A). *In* general, in the derivation cohort, the aneurysm-RVF model we developed demonstrated a good performance in predicting the future risk of aneurysm compared to the baseline or clinical risk models. Fig. 3The average C-index of the baseline, clinical risks and aneurysm-RVF model from the derivation and validation cohort. A Each coloured violin plot represents the average and distribution of C-index of different models from the derivation cohort, the C-indexes were derived from 2000 bootstrapping: B the average and distribution of C-index of different models from the validation cohort
## Validation of the aneurysm-RVF risk model and comparison with a clinical risk model
The discriminative capability of these models was further examined in a validation cohort. In the validation cohort, there were 28 ($0.3\%$) incident cases; the incidence of aneurysm was 4.55 ($95\%$ CI: 3.03–6.58) per 10,000 person-years. The demographic factors, clinical risk factors and aneurysm-RVFs were similar when compared between the derivation and validation cohort (Supplementary Table 6 and 7).
Over 8 years of follow-up, the baseline model yielded a C-index of 0.719 ($95\%$ CI: 0.620–0.816), the clinical risk model yielded a C-index of 0.795 ($95\%$ CI: 0.718–0.871), and the aneurysm-RVF model yielded a C-index that was similar to the clinical model at 0.798 ($95\%$ CI: 0.727–0.869) (Fig. 3B). Although the overall discriminative capabilities were slightly reduced, the aneurysm-RVF model showed equally good performance when compared to the clinical risk model. These results validated the findings in the derivation cohort.
## Predictive capability of aneurysm-RVF score
Finally, to investigate the capability of aneurysm-RVFs to identify individuals at risk of aneurysm, an aneurysm-RVF score was predicted for each participant in the complete cohort based on metrics from the derivation cohort. The aneurysm-RVF score was divided into tertiles and a Cox proportional hazard analysis was performed. In the main analysis, compared to the lowest tertile, the estimated hazard ratio was 3.00 ($95\%$ CI: 0.97–9.30) for participants in the medium tertile and was 17.80 ($95\%$ CI: 6.50–48.75) for participants in the highest tertile. The log-rank test demonstrated that the overall difference in survival rate was statistically significant among the three tertiles ($P \leq 0.01$); only the difference between the medium and low score groups was boundary significant ($$P \leq 0.05$$) (Fig. 4A). In the sensitivity analysis, where participants with previously reported aneurysm were removed, the aneurysm-RVF score yielded slightly smaller estimates: HR was 3.0 ($95\%$ CI: 0.97–9.30) for participants in the medium tertile and 12.52 ($95\%$ CI: 4.523–34.677) for participants in the highest tertile. The difference was significant among the three tertiles except for comparisons between the medium and low score groups ($$P \leq 0.05$$) (Fig. 4B). These findings demonstrated that the aneurysm-RVF score we developed can precisely discriminate subjects at risk of aneurysm. Fig. 4Kaplan-Meier plots for aneurysm risk according to the aneurysm-RVF score. A Aneurysm-RVF score in the complete cohort; B Aneurysm-RVF score in the participants with first diagnose of the aneurysm
## Summary of the findings
To our knowledge, this is the first study to innovatively integrate the genetic risk of aneurysm and oculomics of pan-retinal vascular geometry to investigate the value of RVFs in predictive diagnosis of systemic aneurysms. Our PheWAS analysis identified 32 RVFs that were associated with the genetic risks of aneurysms. The identified RVFs were considered to share common biological aetiologies with systemic aneurysms and were used to develop an aneurysm-RVF model to predict the future risk of aneurysm. The aneurysm-RVF model exhibited an equally good performance when compared to a clinical risk model (C-index = 0.798 [aneurysm-RVF model] vs. 0.795 [clinical risk model]). The aneurysm risk score derived from our model can successfully stratify subjects at different levels of aneurysm risk (upper vs. lower tertile, HR = 17.80 [6.50–48.75]).
The existing aneurysm screening guidelines only consider a few biomarkers [12, 39] and are generalized for ‘average patients’. In contrast, PPPM aims to improve the outcomes of medical interventions for each individual by developing new medical policies for clear target patients on the basis of population heterogeneity [40]; however, this strategy relies on advances in biomarker discovery [41, 42]. Our study identified new oculomic biomarkers and provided evidence that these biomarkers can be used in the predictive diagnosis of aneurysms. Our strategy can also help to produce a tailored and targeted screening approach to benefit both patients and the healthcare system, thus contributing to the paradigm of PPPM.
## The advantages and rationale of applying an individual aneurysm genetic risk for PPPM
Applying an individual’s genetic information is one of the fundamental steps of the implementation of PPPM although the clinical prospects of this strategy are still being explored. Previously, one of the challenges of applying precision medicine in aneurysm was to translate genomic information to the clinical settings and to develop better diagnostic tools or management setups for aneurysms [43]. The results of our study provided one potential application. There are several advantages of using the genetic risk of aneurysms rather than applying the phenotype itself. Aneurysms are associated with a low prevalence in the general population and are mostly asymptomatic initially. Cases identified from routinely collected phenotypic data would usually underestimate the true prevalence and render the study extremely underpowered. For example, when using phenotypic data, the Atherosclerosis Risk in Communities (ARIC) Study failed to identify any significant association between retinal vasculature and the incidence of AAA [30]. However, in our study, we found that 10 RVFs were associated with the genetic risk of AAA, even after adjusting for additional confounding factors. Of note, the SNPs selected for aneurysms often reveal the biological mechanism of the disease; for example, SNPs selected for AAA encode genes that exert inflammatory and immune function (IL6R and CDKN2BAS1/ANRIL) or participate in lipoprotein metabolism (SORT1 and LDLR). These processes are important for the development of AAA [44]. The use of aneurysm GRSs would better reflect its biological mechanisms [17]; hence, its association with RVFs would be more biological rather than statistical.
We also note that more RVFs were found to be associated with AAA GRS than TAA. This could be due to the better performance of AAA GRS which included 12 SNPs compared to three or four SNPs included in TAA or ICA GRS. In addition, in terms of embryological origin, abdominal and thoracic aneurysms are different; this causes differing pathologies [45]. The Vascular Smooth Muscle Cells (VSMCs) in the ascending aorta and arch vessels are derived from neural crest stem cells and progenitor cells in the second heart field, while those in the descending aorta are derived from the mesodermal somatic precursor’s layer [46, 47]. It is also known that the retinal vasculature is derived from the mesoderm, which is similar to the descending aorta. Hence, the common embryonic origin between AAA and retinal vessels might also lead to the better performance of AAA GRS.
## Oculomics can reflect systemic disorders
As an emerging research area, oculomics has been applied to predict systemic diseases under the framework of PPPM. The ocular system is enriched with connective (e.g. the sclera), neuron (e.g. the retinal neuron layer) and vascular (e.g. the retinal vessel and choroid) tissues; hence, oculomics can reflect systemic disorders from a variety of aspects. One recent study conducted by Evsevieva et al. demonstrated that connective tissue dysfunction can be manifested as disorders in both systemic vascular system as well as ocular system and cause diseases such as myopia and glaucoma [48]. Similarly, MFS is an autosomal dominant genetic disorder of the connective tissue. In a manner that differs from polygenic diseases in which each SNP only exerts a small effect on the outcome [22, 23], the mutation of MFS genes would cause relatively severe dysfunction of the connective tissues. The corresponding clinical manifestations include arterial aneurysms and high myopia: this could be a potential application of using oculomics of high myopia to predict systemic disease [49]. Furthermore, connective tissue disorder can influence retinal vascular geometry [50]: we suspect that this would lead to another potential application of using oculomics of RVFs to predict aneurysms. In a previous case report, an MFS patient with an FBN1 mutation developed an aneurysm and was detected with abnormal retinal vascular morphology in both eyes [51]. Another study investigated the retinal vasculature alterations in genetically confirmed MFS and found that the severity of MFS was significantly correlated with impairments in the retinal vasculature [52]. Consistent with these findings, in our study, we observed that [1] the genes with clear biological effects on connective tissue disorders like MFS were also associated with oculomics of RVFs (Supplementary Figure 2); [2] the identified oculomics of RVFs had a capability in predicting systemic aneurysm, which was more precise than baseline information. Our study provides more evidence to support the application of oculomics in systemic disease interventions under the PPPM framework.
One of the concerns is that MFS patients also suffer from high myopia [49]; this might influence the morphology of the retinal vessels and induce a spurious association between RVF and MFS [33]. Hence, in this study, apart from clinical risk factors, refractive error was also carefully adjusted in the model to reduce potential bias caused by myopic status.
## Biological interpretation of aneurysm RVF
One of our key findings is that ‘curveangle_mean’, a parameter that reflects the tortuosity of the retinal vessels, was commonly identified by different aneurysms and MFSs, thus indicating its importance for aneurysm. Previous studies have described variable associations between retinal vessel tortuosity and cardiovascular diseases, albeit with conflicting evidence on occasion. For example, it was reported that increasing retinal arteriolar tortuosity was associated with an increased risk of stroke in type 2 diabetic patients [53] and might also be associated with hypertension and hyperlipidaemia [54]. Sasongko et al. [ 55] found no association between retinal vessel tortuosity and a range of clinical risk factors, including blood lipids in diabetic patients. However, Cheung and Taarnhøj reported that flatter retinal vessels (smaller in tortuosity) were associated with older age, higher blood pressure as well as higher BMI and these are all risk factors for aneurysm [56, 57]. Similar to Cheung and Taarnhøj’s findings, we found that smaller values of ‘curveangle_mean’ (flatter vessels) were associated with an increased risk of aneurysm. In previous studies, general tortuosity was calculated as the arc-chord ratio of a vessel; this measures a wider range of the vessel arch and is insensitive to the frequency with which a vessel ‘wiggles’ [58]. To overcome this problem, ‘curveangle’ focuses on smaller regions of the vessels by measuring vessel branches sampled at a length of 10 pixels [33]. For simplicity, tortuosity and ‘curveangle’ were standardized to the same direction: the greater the value, the curvier the vessel. We suspect that compared to tortuosity, the ‘curveangle’ might be more sensitive in reflecting the early degeneration of the retinal vessels. However, to what extent this feature can be used for detecting systemic aneurysms needs to be further investigated.
Another interesting finding is that ‘ntreeA’ was negatively associated with arterial aneurysms. This may be relevant to its geometric capability of reducing the blood flow shear stress. Common features of arterial aneurysms include the dysfunction and loss of vascular smooth muscle cells, extracellular matrix degradation and inflammation [59], which can disrupt arterial wall structural integrity, weaken vessels, remodel the arterial wall and subsequent dilate the artery [60]. A high wall shear stress caused by abnormal flow conditions can activate pro-inflammatory signalling in endothelial cells and disrupt the internal elastic lamina and collagen matrix, thereby leading to a focal bulge of the wall and the initiation of arterial aneurysm. It is possible that the increased number of vascular trees may play an important role in diverting and therefore decreasing the shear stress; however, further hemodynamic studies are needed.
## Strength and limitations
The main strength of the present study is that this large and prospective cohort features extensive phenotypic, genotypic and oculomics detail about its participants. In addition, this study used genetic information relating to aneurysms to identify new diagnostic biomarkers that better represented the population; furthermore, these biomarkers were highly robust. We also used a deep learning system to analyse large quantities of retinal images which can automatically measure a wide range of RVF parameters. However, the findings of this study need to be interpreted with caution due to the following of the following limitations. Firstly, participants involved in the UK Biobank study might not be fully representative because extremely poor-health individuals could not participate in this study. Secondly, the gene panel for MFS included a myopia gene; although the refractive error of the eye was adjusted, the pleiotropic effect of the genes can still lead to potential bias. Finally, since patients that were diagnosed with aneurysms in the UK *Biobank data* set were rare, and participants who had retinal images were younger on average, replication in other studies will give better clinical indications of these findings.
## Conclusion and expert recommendations for the framework of PPPM
Our study showed that RVFs, quantified based on retinal images, were associated with the genetic risks of aneurysms, and that the aneurysm-RVF score can efficiently identify patients with risks of aneurysms. Our findings support the further development and application of PPPM in the medical intervention of aneurysms from different perspectives. Firstly, from the aspect of disease prediction, conventional prediction models mainly rely on clinical risk factors while our study innovatively implemented oculomics and identified non-invasive biomarkers for aneurysm prediction, thus enabling the construction of more practical prediction models. Secondly, from the aspect of primary aneurysm prevention, the current screening strategies for AAA are mainly based on age and sex, while for other types of aneurysms, effective screening strategies are still required. The aneurysm-RVF score can effectively stratify patients at risk. Furthermore, these imaging biomarkers are safe and inexpensive, thus helping us to monitor the progression of aneurysms in a long-term and dynamic manner. Thirdly, from the aspect of personalization, as the oculomics of RVF is easily accessible compared to other medical examinations, applying oculomics for the detection of aneurysm would benefit the majority of the population, especially for younger patients or those living in areas with limited medical resources.
Finally, although the aneurysm-RVFs identified from our study demonstrated significant potential as reliable biomarkers, their biological causes remain unclear. Further studies are now warranted to confirm the clinical value of RVFs in the screening and early diagnosis of arterial aneurysms.
## Supplementary information
ESM 1ESM 2
## Data and code availability
Data are available in a public, open access repository (https://www.ukbiobank.ac.uk/). The data and R code used in this current study are available from the corresponding authors on a reasonable request.
## References
1. 1.Erbel R, Aboyans V, Boileau C, Bossone E, Bartolomeo RD, Eggebrecht H, Evangelista A, Falk V, Frank H, Gaemperli O, Grabenwoger M, Haverich A, Iung B, Manolis AJ, Meijboom F, Nienaber CA, Roffi M, Rousseau H, Sechtem U, et al. 2014 ESC Guidelines on the diagnosis and treatment of aortic diseases: document covering acute and chronic aortic diseases of the thoracic and abdominal aorta of the adult The Task Force for the Diagnosis and Treatment of Aortic Diseases of the European Society of Cardiology (ESC). Eur Heart J. 35(41):2873–926.
2. Reimerink JJ, van der Laan MJ, Koelemay MJ, Balm R, Legemate DA. **Systematic review and meta-analysis of population-based mortality from ruptured abdominal aortic aneurysm**. *Br J Surg.* (2013.0) **100** 1405-1413. DOI: 10.1002/bjs.9235
3. Huang X, Wang Z, Shen Z, Lei F, Liu YM, Chen Z, Qin JJ, Liu H, Ji YX, Zhang P, Zhang XJ, Yang J, Cai J, She ZG, Li H. **Projection of global burden and risk factors for aortic aneurysm - timely warning for greater emphasis on managing blood pressure**. *Ann Med.* (2022.0) **54** 553-564. DOI: 10.1080/07853890.2022.2034932
4. Flores M, Glusman G, Brogaard K, Price ND, Hood L. **P4 medicine: how systems medicine will transform the healthcare sector and society**. *Per Med.* (2013.0) **10** 565-576. DOI: 10.2217/pme.13.57
5. Force USPST, Owens DK, Davidson KW, Krist AH, Barry MJ, Cabana M, Caughey AB, Doubeni CA, Epling JW, Kubik M, Landefeld CS, Mangione CM, Pbert L, Silverstein M, Simon MA, Tseng CW, Wong JB. **Screening for abdominal aortic aneurysm: US Preventive Services Task Force recommendation statement**. *JAMA.* (2019.0) **322** 2211-2218. DOI: 10.1001/jama.2019.18928
6. Thompson BG, Brown RD, Amin-Hanjani S, Broderick JP, Cockroft KM, Connolly ES, Duckwiler GR, Harris CC, Howard VJ, Johnston SC, Meyers PM, Molyneux A, Ogilvy CS, Ringer AJ, Torner J. **Guidelines for the management of patients with unruptured intracranial aneurysms: a guideline for healthcare professionals from the American Heart Association/American Stroke Association**. *Stroke.* (2015.0) **46** 2368-2400. DOI: 10.1161/STR.0000000000000070
7. Molacek J, Treska V, Zeithaml J, Hollan I, Topolcan O, Pecen L, Slouka D, Karlikova M, Kucera R. **Blood biomarker panel recommended for personalized prediction, prognosis, and prevention of complications associated with abdominal aortic aneurysm**. *EPMA J.* (2019.0) **10** 125-135. DOI: 10.1007/s13167-019-00173-2
8. Mizrak D, Feng H, Yang B. **Dissecting the heterogeneity of human thoracic aortic aneurysms using single-cell transcriptomics**. *Arterioscler Thromb Vasc Biol.* (2022.0) **42** 919-930. DOI: 10.1161/ATVBAHA.122.317484
9. Nieuwkamp DJ, Algra A, Velthuis BK, Rinkel GJ. **Clinical and radiological heterogeneity in aneurysmal sub-arachnoid haemorrhage according to risk-factor profile**. *Int J Stroke.* (2014.0) **9** 1052-1056. DOI: 10.1111/ijs.12274
10. 10.Lu HS, Sawada H, Wu C. Aortic aneurysm and dissection: heterogeneity and molecular mechanisms. Biomolecules. 2022;12(10):1536.
11. Kent DM, Saver JL, Kasner SE, Nelson J, Carroll JD, Chatellier G, Derumeaux G, Furlan AJ, Herrmann HC, Juni P, Kim JS, Koethe B, Lee PH, Lefebvre B, Mattle HP, Meier B, Reisman M, Smalling RW, Soendergaard L, Song JK, Mas JL, Thaler DE. **Heterogeneity of treatment effects in an analysis of pooled individual patient data from randomized trials of device closure of patent foramen ovale after stroke**. *JAMA.* (2021.0) **326** 2277-2286. DOI: 10.1001/jama.2021.20956
12. Welsh P, Welsh CE, Jhund PS, Woodward M, Brown R, Lewsey J, Celis-Morales CA, Ho FK, MacKay DF, Gill JMR, Gray SR, Katikireddi SV, Pell JP, Forbes J, Sattar N. **Derivation and validation of a 10-year risk score for symptomatic abdominal aortic aneurysm: cohort study of nearly 500 000 individuals**. *Circulation.* (2021.0) **144** 604-614. DOI: 10.1161/CIRCULATIONAHA.120.053022
13. Pena SDJ, Tarazona-Santos E. **Clinical genomics and precision medicine**. *Genet Mol Biol.* (2022.0) **45** e20220150. DOI: 10.1590/1678-4685-gmb-2022-0150
14. Ginsburg GS, Willard HF. **Genomic and personalized medicine: foundations and applications**. *Transl Res.* (2009.0) **154** 277-287. DOI: 10.1016/j.trsl.2009.09.005
15. Isselbacher EM, Lino Cardenas CL, Lindsay ME. **Hereditary influence in thoracic aortic aneurysm and dissection**. *Circulation.* (2016.0) **133** 2516-2528. DOI: 10.1161/CIRCULATIONAHA.116.009762
16. 16.Strianese O, Rizzo F, Ciccarelli M, Galasso G, D'Agostino Y, Salvati A, Del Giudice C, Tesorio P, Rusciano MR. Precision and personalized medicine: how genomic approach improves the management of cardiovascular and neurodegenerative disease. Genes (Basel). 2020;11(7):747.
17. Jones GT, Tromp G, Kuivaniemi H, Gretarsdottir S, Baas AF, Giusti B, Strauss E, Van't Hof FN, Webb TR, Erdman R, Ritchie MD, Elmore JR, Verma A, Pendergrass S, Kullo IJ, Ye Z, Peissig PL, Gottesman O, Verma SS, Malinowski J, Rasmussen-Torvik LJ, Borthwick KM, Smelser DT, Crosslin DR, de Andrade M, Ryer EJ, McCarty CA, Bottinger EP, Pacheco JA, Crawford DC, Carrell DS, Gerhard GS, Franklin DP, Carey DJ, Phillips VL, Williams MJ, Wei W, Blair R, Hill AA, Vasudevan TM, Lewis DR, Thomson IA, Krysa J, Hill GB, Roake J, Merriman TR, Oszkinis G, Galora S, Saracini C, Abbate R, Pulli R, Pratesi C, Saratzis A, Verissimo AR, Bumpstead S, Badger SA, Clough RE, Cockerill G, Hafez H, Scott DJ, Futers TS, Romaine SP, Bridge K, Griffin KJ, Bailey MA, Smith A, Thompson MM, van Bockxmeer FM, Matthiasson SE, Thorleifsson G, Thorsteinsdottir U, Blankensteijn JD, Teijink JA, Wijmenga C, de Graaf J, Kiemeney LA, Lindholt JS, Hughes A, Bradley DT, Stirrups K, Golledge J, Norman PE, Powell JT, Humphries SE, Hamby SE, Goodall AH, Nelson CP, Sakalihasan N, Courtois A, Ferrell RE, Eriksson P, Folkersen L, Franco-Cereceda A, Eicher JD, Johnson AD, Betsholtz C, Ruusalepp A, Franzen O, Schadt EE, Bjorkegren JL, Lipovich L, Drolet AM, Verhoeven EL, Zeebregts CJ, Geelkerken RH, van Sambeek MR, van Sterkenburg SM, de Vries JP, Stefansson K, Thompson JR, de Bakker PI, Deloukas P, Sayers RD, Harrison SC, van Rij AM, Samani NJ, Bown MJ. **Meta-analysis of genome-wide association studies for abdominal aortic aneurysm identifies four new disease-specific risk loci**. *Circ Res* (2017.0) **120** 341-353. DOI: 10.1161/CIRCRESAHA.116.308765
18. Bakker MK, van der Spek RAA, van Rheenen W, Morel S, Bourcier R, Hostettler IC, Alg VS, van Eijk KR, Koido M, Akiyama M, Terao C, Matsuda K, Walters RG, Lin K, Li L, Millwood IY, Chen Z, Rouleau GA, Zhou S, Rannikmae K, Sudlow CLM, Houlden H, van den Berg LH, Dina C, Naggara O, Gentric JC, Shotar E, Eugene F, Desal H, Winsvold BS, Borte S, Johnsen MB, Brumpton BM, Sandvei MS, Willer CJ, Hveem K, Zwart JA, Verschuren WMM, Friedrich CM, Hirsch S, Schilling S, Dauvillier J, Martin O, Stroke HA-I, Jones GT, Bown MJ, Ko NU, Kim H, Coleman JRI, Breen G, Zaroff JG, Klijn CJM, Malik R, Dichgans M, Sargurupremraj M, Tatlisumak T, Amouyel P, Debette S, Rinkel GJE, Worrall BB, Pera J, Slowik A, Gaal-Paavola EI, Niemela M, Jaaskelainen JE, von Und Zu Fraunberg M, Lindgren A, Broderick JP, Werring DJ, Woo D, Redon R, Bijlenga P, Kamatani Y, Veldink JH, Ruigrok YM. **Genome-wide association study of intracranial aneurysms identifies 17 risk loci and genetic overlap with clinical risk factors**. *Nat Genet.* (2020.0) **52** 1303-1313. DOI: 10.1038/s41588-020-00725-7
19. Roychowdhury T, Lu H, Hornsby WE, Crone B, Wang GT, Guo DC, Sendamarai AK, Devineni P, Lin M, Zhou W, Graham SE, Wolford BN, Surakka I, Wang Z, Chang L, Zhang J, Mathis M, Brummett CM, Melendez TL, Shea MJ, Kim KM, Deeb GM, Patel HJ, Eliason J, Eagle KA, Yang B, Ganesh SK, Brumpton B, Asvold BO, Skogholt AH, Hveem K, Program VAMV, Pyarajan S, Klarin D, Tsao PS, Damrauer SM, Leal SM, Milewicz DM, Chen YE, Garcia-Barrio MT, Willer CJ. **Regulatory variants in TCF7L2 are associated with thoracic aortic aneurysm**. *Am J Hum Genet.* (2021.0) **108** 1578-1589. DOI: 10.1016/j.ajhg.2021.06.016
20. LeMaire SA, McDonald ML, Guo DC, Russell L, Miller CC, Johnson RJ, Bekheirnia MR, Franco LM, Nguyen M, Pyeritz RE, Bavaria JE, Devereux R, Maslen C, Holmes KW, Eagle K, Body SC, Seidman C, Seidman JG, Isselbacher EM, Bray M, Coselli JS, Estrera AL, Safi HJ, Belmont JW, Leal SM, Milewicz DM. **Genome-wide association study identifies a susceptibility locus for thoracic aortic aneurysms and aortic dissections spanning FBN1 at 15q21.1**. *Nat Genet.* (2011.0) **43** 996-1000. DOI: 10.1038/ng.934
21. Aubart M, Gazal S, Arnaud P, Benarroch L, Gross MS, Buratti J, Boland A, Meyer V, Zouali H, Hanna N, Milleron O, Stheneur C, Bourgeron T, Desguerre I, Jacob MP, Gouya L, Genin E, Deleuze JF, Jondeau G, Boileau C. **Association of modifiers and other genetic factors explain Marfan syndrome clinical variability**. *Eur J Hum Genet.* (2018.0) **26** 1759-1772. DOI: 10.1038/s41431-018-0164-9
22. Pinard A, Jones GT, Milewicz DM. **Genetics of thoracic and abdominal aortic diseases**. *Circ Res.* (2019.0) **124** 588-606. DOI: 10.1161/CIRCRESAHA.118.312436
23. 23.Hof FN v ‘t, Ruigrok YM, Lee CH, Ripke S, Anderson G, de Andrade M, Baas AF, Blankensteijn JD, Bottinger EP, Bown MJ, Broderick J, Bijlenga P, Carrell DS, Crawford DC, Crosslin DR, Ebeling C, Eriksson JG, Fornage M, Foroud T, et al. Shared genetic risk factors of intracranial, abdominal, and thoracic aneurysms. J Am Heart Assoc. 2016;5(7):e002603.
24. **Medical imaging in personalised medicine: a white paper of the research committee of the European Society of Radiology (ESR)**. *Insights Imaging.* (2015.0) **6** 141-155. DOI: 10.1007/s13244-015-0394-0
25. Kim BR, Yoo TK, Kim HK, Ryu IH, Kim JK, Lee IS, Kim JS, Shin DH, Kim YS, Kim BT. **Oculomics for sarcopenia prediction: a machine learning approach toward predictive, preventive, and personalized medicine**. *EPMA J.* (2022.0) **13** 367-382. DOI: 10.1007/s13167-022-00292-3
26. Flammer J, Konieczka K, Bruno RM, Virdis A, Flammer AJ, Taddei S. **The eye and the heart**. *Eur Heart J.* (2013.0) **34** 1270-1278. DOI: 10.1093/eurheartj/eht023
27. Betzler BK, Sabanayagam C, Tham YC, Cheung CY, Cheng CY, Wong TY, Nusinovici S. **Retinal vascular profile in predicting incident cardiometabolic diseases among individuals with diabetes**. *Microcirculation.* (2022.0) **29** e12772. DOI: 10.1111/micc.12772
28. Cheung CY, Xu D, Cheng CY, Sabanayagam C, Tham YC, Yu M, Rim TH, Chai CY, Gopinath B, Mitchell P, Poulton R, Moffitt TE, Caspi A, Yam JC, Tham CC, Jonas JB, Wang YX, Song SJ, Burrell LM, Farouque O, Li LJ, Tan G, Ting DSW, Hsu W, Lee ML, Wong TY. **A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre**. *Nat Biomed Eng.* (2021.0) **5** 498-508. DOI: 10.1038/s41551-020-00626-4
29. Mordi IR, Trucco E, Syed MG, MacGillivray T, Nar A, Huang Y, George G, Hogg S, Radha V, Prathiba V, Anjana RM, Mohan V, Palmer CNA, Pearson ER, Lang CC, Doney ASF. **Prediction of major adverse cardiovascular events from retinal, clinical, and genomic data in individuals with type 2 diabetes: a population cohort study**. *Diabetes Care.* (2022.0) **45** 710-716. DOI: 10.2337/dc21-1124
30. Folsom AR, Lutsey PL, Klein R, Klein BE, Tang W. **Retinal microvascular signs and incidence of abdominal aortic aneurysm: the Atherosclerosis Risk in Communities Study**. *Ophthalmic Epidemiol.* (2018.0) **25** 246-249. DOI: 10.1080/09286586.2017.1418387
31. Moss HE. **Retinal vascular changes are a marker for cerebral vascular diseases**. *Curr Neurol Neurosci Rep.* (2015.0) **15** 40. DOI: 10.1007/s11910-015-0561-1
32. Bertele N, Karabatsiakis A, Buss C, Talmon A. **How biomarker patterns can be utilized to identify individuals with a high disease burden: a bioinformatics approach towards predictive, preventive, and personalized (3P) medicine**. *EPMA J.* (2021.0) **12** 507-516. DOI: 10.1007/s13167-021-00255-0
33. Shi D, Lin Z, Wang W, Tan Z, Shang X, Zhang X, Meng W, Ge Z, He M. **A deep learning system for fully automated retinal vessel measurement in high throughput image analysis**. *Front Cardiovasc Med.* (2022.0) **9** 823436. DOI: 10.3389/fcvm.2022.823436
34. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, Downey P, Elliott P, Green J, Landray M, Liu B, Matthews P, Ong G, Pell J, Silman A, Young A, Sprosen T, Peakman T, Collins R. **UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age**. *PLoS Med.* (2015.0) **12** e1001779. DOI: 10.1371/journal.pmed.1001779
35. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, Motyer A, Vukcevic D, Delaneau O, O'Connell J, Cortes A, Welsh S, Young A, Effingham M, McVean G, Leslie S, Allen N, Donnelly P, Marchini J. **The UK Biobank resource with deep phenotyping and genomic data**. *Nature.* (2018.0) **562** 203-209. DOI: 10.1038/s41586-018-0579-z
36. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. **Second-generation PLINK: rising to the challenge of larger and richer datasets**. *Gigascience.* (2015.0) **4** 7. DOI: 10.1186/s13742-015-0047-8
37. Backes D, Vergouwen MD, Tiel Groenestege AT, Bor AS, Velthuis BK, Greving JP, Algra A, Wermer MJ, van Walderveen MA, terBrugge KG, Agid R, Rinkel GJ. **PHASES score for prediction of intracranial aneurysm growth**. *Stroke.* (2015.0) **46** 1221-1226. DOI: 10.1161/STROKEAHA.114.008198
38. Carroll RJ, Bastarache L, Denny JC, PheWAS R. **data analysis and plotting tools for phenome-wide association studies in the R environment**. *Bioinformatics.* (2014.0) **30** 2375-2376. DOI: 10.1093/bioinformatics/btu197
39. Hozawa A. **Screening for aortic aneurysm: further evidence is required to clarify the issue**. *J Atheroscler Thromb.* (2021.0) **28** 319. DOI: 10.5551/jat.ED143
40. Wang X. **New strategies of clinical precision medicine**. *Clin Transl Med.* (2022.0) **12** e135. DOI: 10.1002/ctm2.135
41. Ho D, Quake SR, McCabe ERB, Chng WJ, Chow EK, Ding X, Gelb BD, Ginsburg GS, Hassenstab J, Ho CM, Mobley WC, Nolan GP, Rosen ST, Tan P, Yen Y, Zarrinpar A. **Enabling technologies for personalized and precision medicine**. *Trends Biotechnol.* (2020.0) **38** 497-518. DOI: 10.1016/j.tibtech.2019.12.021
42. Vargas AJ, Harris CC. **Biomarker development in the precision medicine era: lung cancer as a case study**. *Nat Rev Cancer.* (2016.0) **16** 525-537. DOI: 10.1038/nrc.2016.56
43. Tromp G, Kuivaniemi H. **Developments in genomics to improve understanding, diagnosis and management of aneurysms and peripheral artery disease**. *Eur J Vasc Endovasc Surg.* (2009.0) **38** 676-682. DOI: 10.1016/j.ejvs.2009.08.010
44. Paige E, Clement M, Lareyre F, Sweeting M, Raffort J, Grenier C, Finigan A, Harrison J, Peters JE, Sun BB, Butterworth AS, Harrison SC, Bown MJ, Lindholt JS, Badger SA, Kullo IJ, Powell J, Norman PE, Scott DJA, Bailey MA, Rose-John S, Danesh J, Freitag DF, Paul DS, Mallat Z. **Interleukin-6 receptor signaling and abdominal aortic aneurysm growth rates**. *Circ Genom Precis Med.* (2019.0) **12** e002413. DOI: 10.1161/CIRCGEN.118.002413
45. Wang G, Jacquet L, Karamariti E, Xu Q. **Origin and differentiation of vascular smooth muscle cells**. *J Physiol.* (2015.0) **593** 3013-3030. DOI: 10.1113/JP270033
46. Petsophonsakul P, Furmanik M, Forsythe R, Dweck M, Schurink GW, Natour E, Reutelingsperger C, Jacobs M, Mees B, Schurgers L. **Role of vascular smooth muscle cell phenotypic switching and calcification in aortic aneurysm formation**. *Arterioscler Thromb Vasc Biol.* (2019.0) **39** 1351-1368. DOI: 10.1161/ATVBAHA.119.312787
47. Pfaltzgraff ER, Bader DM. **Heterogeneity in vascular smooth muscle cell embryonic origin in relation to adult structure, physiology, and disease**. *Dev Dyn.* (2015.0) **244** 410-416. DOI: 10.1002/dvdy.24247
48. Evsevieva M, Sergeeva O, Mazurakova A, Koklesova L, Prokhorenko-Kolomoytseva I, Shchetinin E, Birkenbihl C, Costigliola V, Kubatka P, Golubnitschaja O. **Pre-pregnancy check-up of maternal vascular status and associated phenotype is crucial for the health of mother and offspring**. *EPMA J.* (2022.0) **13** 351-366. DOI: 10.1007/s13167-022-00294-1
49. Ha HI, Seo JB, Lee SH, Kang JW, Goo HW, Lim TH, Shin MJ. **Imaging of Marfan syndrome: multisystemic manifestations**. *Radiographics.* (2007.0) **27** 989-1004. DOI: 10.1148/rg.274065171
50. Konieczka K, Ritch R, Traverso CE, Kim DM, Kook MS, Gallino A, Golubnitschaja O, Erb C, Reitsamer HA, Kida T, Kurysheva N, Yao K. **Flammer syndrome**. *EPMA J.* (2014.0) **5** 11. DOI: 10.1186/1878-5085-5-11
51. Miyata M, Ooto S, Uji A, Tsujikawa A. **Retinal artery tortuosity in Marfan’s syndrome**. *QJM.* (2021.0) **114** 601-602. DOI: 10.1093/qjmed/hcab079
52. Di Marino M, Cesareo M, Aloe G, Nucci C, Giannini C, Martucci A, Aiello F, Pisano C, Ruvolo G, Mancino R. **Retinal and choroidal vasculature in patients with Marfan syndrome**. *Transl Vis Sci Technol.* (2020.0) **9** 5. DOI: 10.1167/tvst.9.9.5
53. Sandoval-Garcia E, McLachlan S, Price AH, MacGillivray TJ, Strachan MWJ, Wilson JF, Price JF. **Retinal arteriolar tortuosity and fractal dimension are associated with long-term cardiovascular outcomes in people with type 2 diabetes**. *Diabetologia.* (2021.0) **64** 2215-2227. DOI: 10.1007/s00125-021-05499-z
54. Owen CG, Rudnicka AR, Nightingale CM, Mullen R, Barman SA, Sattar N, Cook DG, Whincup PH. **Retinal arteriolar tortuosity and cardiovascular risk factors in a multi-ethnic population study of 10-year-old children; the Child Heart and Health Study in England (CHASE)**. *Arterioscler Thromb Vasc Biol.* (2011.0) **31** 1933-8. DOI: 10.1161/ATVBAHA.111.225219
55. Sasongko MB, Wong TY, Nguyen TT, Cheung CY, Shaw JE, Kawasaki R, Lamoureux EL, Wang JJ. **Retinal vessel tortuosity and its relation to traditional and novel vascular risk markers in persons with diabetes**. *Curr Eye Res.* (2016.0) **41** 551-557. PMID: 26086266
56. Cheung CY, Zheng Y, Hsu W, Lee ML, Lau QP, Mitchell P, Wang JJ, Klein R, Wong TY. **Retinal vascular tortuosity, blood pressure, and cardiovascular risk factors**. *Ophthalmology.* (2011.0) **118** 812-818. DOI: 10.1016/j.ophtha.2010.08.045
57. Taarnhoj NC, Munch IC, Sander B, Kessel L, Hougaard JL, Kyvik K, Sorensen TI, Larsen M. **Straight versus tortuous retinal arteries in relation to blood pressure and genetics**. *Br J Ophthalmol.* (2008.0) **92** 1055-1060. DOI: 10.1136/bjo.2007.134593
58. Bullitt E, Gerig G, Pizer SM, Lin W, Aylward SR. **Measuring tortuosity of the intracerebral vasculature from MRA images**. *IEEE Trans Med Imaging.* (2003.0) **22** 1163-1171. DOI: 10.1109/TMI.2003.816964
59. 59.Gurung R, Choong AM, Woo CC, Foo R, Sorokin V. Genetic and epigenetic mechanisms underlying vascular smooth muscle cell phenotypic modulation in abdominal aortic aneurysm. Int J Mol Sci. 2020;21(17):6334.
60. Lu H, Du W, Ren L, Hamblin MH, Becker RC, Chen YE, Fan Y. **Vascular smooth muscle cells in aortic aneurysm: from genetics to mechanisms**. *J Am Heart Assoc.* (2021.0) **10** e023601. DOI: 10.1161/JAHA.121.023601
|
---
title: 'Low Evidence for Tinnitus Risk Factors: A Systematic Review and Meta-analysis'
authors:
- Roshni Biswas
- Eleni Genitsaridi
- Natalia Trpchevska
- Alessandra Lugo
- Winfried Schlee
- Christopher R. Cederroth
- Silvano Gallus
- Deborah A. Hall
journal: 'JARO: Journal of the Association for Research in Otolaryngology'
year: 2022
pmcid: PMC9971395
doi: 10.1007/s10162-022-00874-y
license: CC BY 4.0
---
# Low Evidence for Tinnitus Risk Factors: A Systematic Review and Meta-analysis
## Abstract
### Aims/Hypothesis
Identifying risk factors for tinnitus could facilitate not only the recommendations for prevention measures, but also identifying potential pathways for new interventions. This study reports the first comprehensive systematic review of analytical observational studies able to provide information about causality (i.e., case–control and cohort designs).
### Methods
A literature search of four electronic databases identified epidemiological studies published on tinnitus and different exposures. Independent raters screened all studies, extracted data, and evaluated study quality using the Newcastle–Ottawa Scale. Reported relative risks (RR), hazard ratios (HR), odds ratios (OR), and prevalence ratios (PR) with $95\%$ confidence intervals (CI) were used to compute crude estimates of RR for tinnitus risk factors.
### Results
From 2389 records identified, a total of 374 articles were read as full text (24 reviews, 301 cross-sectional studies, 42 cohort studies, and 7 case–control studies). However, from 49 case–control and cohort studies, only 25 adequately reported risk ratios. Using the findings from these studies, positive causal associations were found for various hearing-related factors (i.e., unspecified hearing loss, sensorineural hearing loss, occupational noise exposure, ototoxic platinum therapy, and otitis media). Evidence was also found for a number of non-otological risk factors including temporo-mandibular joint disorder, depression, chronic obstructive pulmonary disease, and hyperlipidemia. Negative associations indicating preventative effects were found for diabetes and high alcohol consumption. No associations were found for low alcohol consumption, body mass index, head injury, heart failure, hypertension, leisure noise exposure, migraine, rheumatoid arthritis, sex, smoking, stroke, and whiplash. However, with the exception of unspecified hearing loss, these findings resulted from pooling no more than 4 studies, illustrating that the vast majority of the associations still remain inconclusive.
### Conclusions
These systematic review and meta-analysis confirm a number of otological and non-otological risk factors for tinnitus. By highlighting major gaps in knowledge, our synthesis can help provide direction for future research that will shed light on the pathophysiology, improve management strategies, and inform more effective preventions.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s10162-022-00874-y.
## Introduction
Subjective tinnitus is a common condition among adults with a point prevalence of almost $15\%$, as recently measured in Europe [1]. While recent years have seen remarkable progress in understanding tinnitus heterogeneity [2], the risk factors for tinnitus as well as the mechanisms of tinnitus generation and maintenance are not well understood [3]. This gap in knowledge is considered one of the major roadblocks in the pathway to tinnitus cure. Thus, identifying and quantifying the relationship between tinnitus and related exposures (risks) could shed some light on underlying pathophysiology, improve management strategies, and inform preventive interventions [3, 4].
The most widely reported risk factor for tinnitus is hearing loss [3]. Environmental influences that damage the auditory system and lead to hearing loss, such as the exposure to loud noise and ototoxic medications, can also trigger tinnitus [5]. A wide number of non-otologic risk factors have been described, but variability in study methodology and quality makes synthesis of findings challenging, and a systematic review of these factors has managed to provide only a narrative synthesis of study results [6]. Potential non-otologic factors include whiplash and neck trauma, blast-injury and traumatic brain injury, and stress [5, 7]. Neurological conditions like tension-type headaches, physical conditions such as temporomandibular joint disorders, and audiological conditions like hyperacusis have been related with tinnitus [8–10], but the direction of these relationships remains rather uncertain. Lifestyle factors like diet, smoking, alcohol consumption, hypertension, and obesity have also been hypothesized to be related to tinnitus [3, 11]. Studies in twins, adoptees, and familial aggregation have also suggested that tinnitus runs in families due to genetics [12–14], something that has recently been confirmed in genome-wide association and whole genome sequencing studies including replication cohorts [15, 16]. In 2012, patients and professionals prioritized unmet research questions which included uncertainties around the role of dietary factors, electromagnetic energy waves, sex hormones, and allergies in tinnitus [17]. It is unclear whether that call has driven the research agenda in these directions.
Since published evidence on exposures is primarily from cross-sectional studies, it is difficult to conclude if the relationship between the exposure and tinnitus is correlational or causal [6, 18]. Moreover, most exposures have been assessed either in specific population groups such as patients attending hearing or tinnitus clinics or from specific geographical areas or demographic characteristics. For example, the epidemiology of hearing loss study provides valuable information on tinnitus. However, the study population is limited to older adults who are residents of the township of Beaver Dam, Wisconsin, USA [19, 20]. This information lacks generalizability as risk factors do not necessarily affect all population groups in the same way [21, 22]. For conditions in which aging is a known risk factor, risks in older adults are different from the risk in the general population which reflects the effects across the life span. Difficulties also arise when the underlying biological mechanism of interaction between an exposure and outcome is complicated and unclear. For example, studies have found people with normal audiometric threshold having self-reported hearing difficulty and vice versa [23–25]. These complexities preclude finding a straightforward association.
From a methodological point of view, to infer causality, exposure to the variable of interest should ideally occur before the onset of tinnitus symptoms, or before an intervention to alleviate tinnitus symptoms or before a prevention strategy to avert tinnitus symptoms. Unfortunately, most tinnitus-related studies measure the risk factor and tinnitus simultaneously in cross-sectional designs and this provides the lowest level of evidence for causal inference [6, 26]. Instead, analytical observational studies (i.e., case–control and cohort designs) are required for the highest level of evidence for causal inference. Through the identification of cross-sectional, case–control, and cohort studies of tinnitus which investigated any potentially relevant exposures, the present study aims (i) to identify which exposures have been reported by analytical observational studies and hence have a high level of evidence for causal inference and (ii) to determine the strength of evidence for those exposures using meta-analysis.
## Methods
A systematic literature review was conducted to identify all publications providing information on the relationship between various exposures and any tinnitus. The review protocol was pre-registered in PROSPERO [27]. Reporting follows the meta-analysis of observational studies in epidemiology (MOOSE) guidelines (Supplementary Material 1). The systematic review did not require ethics committee or Institutional Review Board approval since human subjects per se were not studied, and the data reviewed are in the public domain.
## Search Strategy
The initial search string was developed for MEDLINE using a combination of Medical Subjects Headings (MeSH) and text words related to tinnitus, related exposures, and prevalence. The common terminologies were identified by reviewing MeSH terms and keywords used in MEDLINE to describe a pre-defined set of key publications. Different versions of these syntax were piloted to design the final search string that successfully recovered all of the pre-defined articles. The final MEDLINE search string was adapted accordingly to fit the other medical/healthcare databases used in our search, namely Embase, Cochrane Database of Systematic Reviews (CDSR), and Cumulated Index to Nursing and Allied Health Literature (CINAHL) (Table 1). Multiple databases were searched in an effort to identify all available studies and all authors contributed to the development of the search syntax, with AL, SG, CC, and DAH having conducted previous systematic reviews. The literature search was conducted to identify all studies published on tinnitus and potential risk factors on $\frac{29}{11}$/2017 and updated on $\frac{06}{11}$/2019. No restriction was applied on the start date. Table 1Search syntax for each databaseName of databaseSearch string usedCDSRtinnitus:ti (among Cochrane Reviews)CINAHL (Cumulated Index to Nursing and Allied Health Literature)MH tinnitus AND “risk factor**” OR epidemiol* OR cohort OR “case control” OR “cross sectional” OR survey* OR longitudinal OR “pooled analysis" OR “meta analysis” OR representative*EmbaseTinnitus:ti,ab,kw AND (prevalence:kw OR incidence:kw OR “risk factors” OR case–control:ti,ab OR cohort,ti:ab,kw OR cross-sectional:ti,ab,kw OR meta-analysis:ti,ab OR pooled-analysis:ti,ab OR survey:ti,ab OR representative:ti,ab OR longitudinal:ti,ab) AND [embase]/lim NOT [MEDLINE]/lim AND (“article”/it OR “article in press”/it OR “review”/it)MEDLINE (Pubmed)(“tinnitus” [MeSH Terms] OR “tinnitus”[All Fields]) AND (“prevalence"[MeSH Terms] OR “incidence” [MeSH Terms] OR prevalence[OT] OR incidence[OT] OR “risk factors” [MeSH Terms] OR “risk factors” [All fields] OR “case–control” [Tiab] OR “cohort” [Tiab] OR “cohort” [OT] OR “cross-sectional” [Tiab] OR “cross-sectional” [OT] OR meta-analysis[tiab] OR pooled-analysis[tiab] OR survey[tiab] OR representative[tiab] OR longitudinal[tiab])CDSR Cochrane Database of Systematic Reviews Studies were included in the present review if they satisfied the eligibility criteria mentioned in the following sections and were published as a full text in English language. The team did not have adequate resources to translate articles that are not published in English.
## Study Design
Eligible study designs included cross-sectional (including population-based surveys on specific sub-populations), case–control or cohort studies, meta-analyses, pooled-analyses, and systematic reviews. Case reports, case series, letters to the editor, book chapters, conference proceedings, dissertations, and these were excluded. No restrictions in settings were applied.
## Study Population
Human subjects providing data on tinnitus (including pulsatile tinnitus) and associated factors and population-based tinnitus surveys were included. Animal studies were excluded. No restrictions in participant age were applied.
## Exposures
All potential exposures related to tinnitus were considered, such as socio-economic characteristics (e.g., level of education), lifestyle habits (including tobacco smoking, alcohol drinking, dietary patterns, obesity, physical activity, mobile use), comorbidities (including depression, anxiety, vertigo, selected cardiovascular diseases, history of trauma), any medications (e.g., aspirin, ototoxic drugs), and otologic conditions. Studies assessing treatment efficacy for tinnitus therapies and with concomitant medical conditions like Meniere’s Disease were excluded.
## Comparators/Controls
Subjects not exposed to tinnitus-associated factors were treated as comparators. Relative risks (RR), hazard ratios (HR), odds ratios (OR), and prevalence ratios (PR) were used as the measures of effect to ascertain the excess risk of tinnitus incidence or prevalence for subjects exposed to specific risk factors as compared to those who were not exposed. Studies without a clear comparator group were excluded.
## Outcomes
All studies reported information on either tinnitus incidence or prevalence, or some measure of effect for tinnitus-related exposures; namely RR, HR, OR, or PR with $95\%$ confidence intervals (CI).
## Article Screening
An EndNote library was populated with all articles retrieved from the electronic database search and this library was updated to include the additional articles retrieved from the updated search. Duplicate records were deleted leaving a total of 2389 publications. No further searches (e.g., hand searches) were performed. A stepwise screening method was performed. Any discrepancies were resolved by discussion, with an arbitration provided by a third member of the team.
Titles and abstract screening was conducted independently by two of the authors (RB, EG, NT, and AL). The publications were scored 1–5 where 1, publication not pertinent or of limited interest for our review; 2, publication probably not pertinent or of limited interest; 3, not possible to evaluate on the basis of title/abstract/keywords, only; 4, publication probably pertinent or of interest; and 5, publication pertinent or of clear interest for our review. The two reviewer scores were summed to give total scores ranging between 2 and 10. Only those publications with a combined score of 5 or more were passed to the next screening step.
Full-text screening of potentially eligible studies was again conducted independently by two reviewers (RB, EG, NT). Of the 847 articles eligible for full-text screening, 15 were excluded because the full text could not be obtained and 392 were excluded because they did not fit the eligibility criteria or did not report relevant information. A further 81 articles reported data only on the prevalence or incidence of tinnitus without considering risk factors. This left a total of 374 articles for data extraction, but only 49 of those were analytical observational studies. Some of these referred to the same study, published in multiple articles.
## Data Extraction
A data extraction form in MS Excel was used to extract the following information: first author, journal, year of publication, country, type of study, name of source database (if any), tinnitus definition, adjustments in the statistical modeling, reporting of prevalence and incidence, followed by the type of exposure assessed and their measure of association (RR, HR, OR, PR, or raw data) (Supplementary Material 2). Data extraction was conducted independently by two of the authors (RB, EG, and NT) and information consolidated through discussion. Authors of included articles were not contacted for further information at this stage.
The included studies used three operational definitions of tinnitus; self-report of “any tinnitus” (AT), self-report of tinnitus that is moderately severe or causes problems getting to sleep (SignificantT), and professionally diagnosed tinnitus (MedT). Given the paucity of analytical observational studies, this variability in tinnitus definition was not taken into account when pooling data for meta-analysis, but the pooled estimates are also reported separately for greater transparency. Pooling was deemed acceptable since we do not interpret specific causal inference from the meta-analysis results. They are primarily to summarize the exposure data.
For summarizing the data, there is currently no recommended classification system for tinnitus-related exposures. Baguley et al. [ 3] suggested a taxonomy for putative tinnitus risk factors that primarily focused on ototoxic medications and other comorbid conditions related to tinnitus. We expanded these categories to accommodate other tinnitus-related exposures identified through the literature search. This resulted in a taxonomy comprising six domains each with categories and sub-categories. The six domains were (i) socio-demographic, (ii) hearing related, (iii) otological, (iv) potentially ototoxic medications, (v) lifestyle, (vi) comorbidities, and (v) other (Table 2). The data extraction form was expanded iteratively to include the additional exposure categories as they emerged during the data extraction process. Table 2Taxonomy of tinnitus-related exposures in six main broad domains (bold)CategorySub-categoryIllustrative examplesHearing relatedTypes of measurementClinically measuredHearing loss (unspecified), sensorineural hearing lossSelf-reportedHearing difficulty, hyperacusisNoise exposureNoiseOccupational and leisure noiseOtologicInfectiousOtitis externa, otitis mediaLabyrinthineCochlear implants, vestibular dysfunctionNeoplasticAcoustic neuroma, SchwannomaOtherCerumen, otosclerosis, presbycusisOtotoxic medicationsAntibioticsErythromycin, MacrolidesAnti-neoplasticPlatinum-based therapyOther drugsNonsteroidal anti-inflammatory drugsLifestyleLifestyle exposuresChemicalChemical solventsPhysical/otherComputer use, electromagnetic fieldNutritionAnthropometricBody mass index, waist to hip ratioDiet and nutritionDietary patterns, deficiencyPhysical activityHours of exerciseSubstance useAlcoholDrinking status, consumptionCoffeeDrinking statusDrug addictionsDrug abuseSmokingCurrent, exposure to second-hand smokeSocio-demographicAgeFamily historyMarital statusRace/ethnicityRegionGeographic region, urbanization levelSocioeconomic statusEducationSchool years, education levelEmployment statusIncomeYearly income, income levelSexSocial environmentFamily situation, working conditionsOther demographicsBenefitsDisability benefits, health insuranceDeprivation scoreDeprivation indexComorbiditiesCardiovascularHeart failure, hypertension, strokeEndocrine and metabolicDiabetes, hyperlipidemia, thyroid disordersENT, otherNasal disordersRhinitis, sinusitisHepatologicalTransplant, cirrhosis, hepatitis BImmune-mediatedSystemic lupus erythematosusRheumatoid arthritisInfections (systemic)Human immunodeficiency virus, Human T-cell lymphotropic virus type 1MusculoskeletalOsteoporosisNeurologicalHeadache, meningitis, migraineNeoplasticCancer survivors, childhood cancerOrofacialDental conditionsTeeth grinding, clenchingMusculoskeletalTemporomandibular joint disorderMentalAnxiety, dementia, depression, stressRenalChronic kidney diseaseRespiratoryChronic obstructive pulmonary diseaseSleep disorderHypersomnia, insomnia, sleep apnea,TraumaticEar injuryTympanic membrane perforationHead and neck injuryTraumatic brain injury, WhiplashOther bodily injuryBodily injury scoreOther systemic conditionsAllergy, chronic illness, disability, menstruation, pregnancyTreatmentsRadiotherapyRadiation (gamma)OtherGeneticsGenotype, specific allelesENT ear, nose and throat; COPD chronic obstructive pulmonary disease
## Quality Assessment
All included articles were evaluated for study quality. Our first-level thresholds for quality required the study to have a case–control or cohort study design, as well as to report the measures of effect as risk ratios (e.g., RR, HR, OR, or PR). Analytical observational study designs (i.e., case–control and cohort studies) passed the quality threshold since their findings provide information on likely causal inference as well as the degree of association. Report of risk ratios was also a requirement since these estimates can then be pooled in a data synthesis. Although crude estimates of risk ratios can be calculated from studies providing raw data (numbers and proportions) for exposures and tinnitus, this does not give an unbiased picture as there is risk of confounding and so where raw data was reported, it was not analyzed further.
Those studies passing the first-level quality benchmark were then subjected to a further quality appraisal using the Newcastle Ottawa Scale (NOS). This risk of bias assessment tool for observational studies is recommended by the Cochrane Collaboration [28]. The NOS assigns up to a maximum of nine points for the least risk of bias in three domains: [1] selection of study groups (four points), [2] comparability of groups (two points), and [3] ascertainment of exposure and outcomes (three points) for case–control and cohort studies, respectively. Scores in domain [3] were calculated separately for each exposure, depending on how that exposure had been measured. A total score of zero indicates very low quality and nine very high quality, but there is no explicit cutoff to distinguish high from low quality.
## Data Synthesis
For most exposures, very few studies met the first-level quality thresholds. Exposures with evidence from two or more studies were identified. For each such tinnitus-associated factor, meta-analyses were conducted to quantify the difference in risk in individuals exposed to the factor when compared to the unexposed comparator. Heterogeneity between studies was assessed using the Q and I2 tests. The pooled RRs from random-effects models using the DerSimonian and Laird moment estimator of the between-study variance component was evaluated. All statistical analyses were conducted with R (version 3.5.2) software.
In some cases, the published reports were derived from the same study data during the same period and so this paragraph describes how these duplications were handled in the meta-analyses. Two out of the six articles on unspecified hearing loss came from the Epidemiology of Hearing Loss Study; one 5-year follow-up [20] and one 10-year follow-up [29]. Since the individuals followed up for 5 years were included in the 10-year analyses, the most recently reported data was used for the meta-analysis. For the remaining four articles on hearing loss, two came from the Blue Mountains Hearing Study and two from the Taiwan National Health Insurance Resource Database, again from the same study period. Meta-analysis therefore used data reported by Gopinath et al. [ 30] and Lee et al. [ 31], respectively. Two out of the three articles on sensorineural hearing loss came from the Taiwan National Health Insurance Resource Database from the same study period. Therefore, the one with the complete information was used [32]. Risk factor information for hypertension, diabetes, and stroke was available from Taiwan National Health Insurance Resource Database for two articles reporting the same study period [33, 34]. Since Chen et al. [ 33] analyzed information for women and Shih et al. [ 34] for both sexes, the latter was entered into the meta-analyses.
Data from a number of other studies did not contribute to the meta-analyses because the reference group was not appropriate. The prospective cohort study by Aarhus et al. [ 35] compared adults with a history of childhood hearing loss to a reference group of adults with normal childhood hearing, and so risk ratios for sensorineural hearing loss and otitis media were not considered. Additional otitis media risk ratios were also omitted as they came from adults with chronic suppurative otitis media in childhood compared to a reference group of adults without a history of otitis media [36]. The study by Dougherty et al. [ 37] compared military personnel wearing hearing protection at the time of injury compared with a reference group without hearing protection, and so their risk ratios on occupational noise exposure were omitted. Full details from these studies are reported in the Supplementary Material 1.
## Results
Of the 374 included studies, 49 were analytical observational studies (7 case–control and 42 cohorts), 301 were cross-sectional studies, and there were 24 reviews (Fig. 1). From those analytical observational studies, only 25 met the first-level quality threshold (i.e., reported risk ratios) and are reported further. Three of the included analytical observational studies evaluated data from case–control studies, and 22 analyzed data from cohort studies. Among the cohort studies, seven were prospective designs, while the rest were from retrospective. All case–control studies were retrospective in design (Supplementary material 1). Table 3 confirms that the average NOS quality assessment scores were at least 6 out of 9 for the exposures analyzed in the meta-analyses. We consider this to indicate high quality. Fig. 1Flow diagramTable 3Exposures analyzed using meta-analytic approach along with number of high-quality studies and quality assessment scoresRisk factorAnalytical study designTotalNOS score mean (SD)Case–controlCohortHearing relatedHearing loss (unspecified)066a8.3 (0.8)Hearing loss (sensorineural)213a7.7 (1.0)Leisure noise exposure1127.9 (0.3)Occupational noise exposure123c7.1 (0.4)Otitis media224b7.3 (0.7)Platinum (ototoxic)0446.0 (1.4)Lifestyle and socio-demographicAlcohol consumption1127.8 (1.0)Body mass index1128.0 (0.0)Sex0448.5 (0.5)Smoking (current)1127.5 (1.0)ComorbiditiesDepression1128.0 (1.4)Diabetesb213a8.0 (1.0)Heart failure1128.0 (1.0)Hyperlipidaemia1127.7 (1.2)Hypertension134a7.5 (1.3)Migraine1127.3 (0.5)Stroke123a8.0 (1.0)Temporo-mandibular joint disorder1128.3 (1.0)Chronic obstructive pulmonary disease1127.5 (0.7)Rheumatoid arthritis1127.5 (0.7)Head injury1348.2 (0.8)Whiplash1127.0 (0.0)All risk factors in bold indicate a statistically significant association with tinnitusCOPD, chronic obstructive pulmonary diseaseaNot all data from included articles were included in the meta-analyses. See text for detailsbNegative association with tinnitus
## Description of Included Studies
The 25 analytical observational studies were conducted in Western and Northern Europe, North America (USA), Asia (Taiwan), and Australia. A short description of the 10 prospective cohort studies is given as follows. European studies were from the UK, Germany, and Scandinavia. One UK study evaluated hearing-related information from participants of the Oxford-Family Planning Association contraceptive study — prospective cohort of 17,000 women recruited at 17 clinics in England and Scotland between 1968 and 1974 [38]. One German study was derived from the Study of Health in Pomerania, a population-based prospective cohort study [39]. In Nord-Trøndelag Norway, participants were assessed at 7, 10, or 13 years old during a school hearing investigation and also screened at 20 and 56 years old during Nord-Trøndelag Health Study, specifically for hearing loss from 1996 to 1998 [35, 36]. Another study in Sweden and Finland analyzed the amount of mobile phone use at baseline with self-reported weekly tinnitus [40]. In the USA, one study considered tinnitus data from the Nurses’ Health Study II, a prospective cohort study on 25 to 42-year-old nurses from 14 US states to study conditions and risk factors related to women’s health [41]. Two articles reported follow-up information from the Epidemiology of Hearing Loss Study, a study on adults aged 48–92 years, residing in Beaver Dam, Wisconsin [20, 29]. Finally, in Australia, two studies reported data from the Blue Mountains Hearing Study, a population-based survey of age-related hearing loss in 55–99-year-olds residing in two postcodes of Sydney, Australia [30, 42].
A short description of the 12 retrospective cohort studies is given as follows. European studies were from Germany and Sweden. One study in Sweden assessed preschool teachers’ risks based on retrospectively reported symptom onset [43]. One study in Germany used medical health insurance data for retrospective analyses [44]. In Asia, six articles from Taiwan reported retrospective analyses of medical health insurance data [31–34, 45, 46]. Taiwan’s National Health Insurance system was established in 1995 and covers almost $100\%$ of the population. The database includes records from primary outpatient departments and inpatient hospital care settings and it is maintained and regulated by the Data Science Centre of the Ministry of Health and Welfare of Taiwan. Finally, four studies were from the USA. Two studies reported retrospective analyses of health records of military personnel from military hospital–based databases, namely the Expeditionary Medical Support System, Defense Medical Surveillance System, the Pharmacy Data Transaction Service, and the Theater Medical Data Store [37, 47]. Civilian data comprised two studies reporting the Childhood Cancer Survivor Study dataset. This is a North American multi-institutional collaborative retrospective study of individuals who survived at least 5 years after diagnosis of cancer during childhood or adolescence, and a cohort of siblings not affected by cancer [48, 49].
The three retrospective case–control studies came from a variety of sources. One study again came from the Taiwan National Health Insurance Resource Database [50]. Here, cases comprised patients with medically diagnosed tinnitus and controls matched for age, sex, and year of index date. Another considered tinnitus cases attending an ENT outpatients clinic in an Austrian hospital with ENT controls matched for age, gender, ethnicity, and 3 week index date [51]. The final study examined records from the UK Clinical Practice Research Datalink, which is an anonymized database created in 1987 with ongoing medical records from over 11 million patients provided by approximately 700 general practices [52].
There was a large heterogeneity in how studies have performed their adjustment for important covariates. Of the statistical modeling used, $92\%$ of studies adjusted the risk ratio estimates for age, and $60\%$ for sex/gender. Some studies corrected for neither [32, 50–52]. Hearing loss was used as a covariate in only $16\%$ of studies, despite tinnitus being a primary outcome in 17 of 25 studies.
## Strength of Evidence for Tinnitus Risk Factors
Cox regression was used as a statistical model in 8 of the included studies. Table 3 reports those exposures reported within more than one included article, for which meta-analysis is possible. In summary, hearing loss [both unspecified (6 studies) and sensorineural (3 studies)], occupational noise exposure (3 studies), otitis media (4 studies), diabetes (3 studies), temporomandibular disorder (2 studies), and ototoxic platinum exposure (4 studies) were the most reliable risk factors for tinnitus.
Supplementary Figs. 1–6 present the forest plots for all hearing-related conditions. Tinnitus was, on average, significantly associated with unspecified hearing loss (RR, 1.94; $95\%$ CI, 1.41–2.67; I2 = $69\%$; $$p \leq 0.02$$; Supplementary Fig. 1), sensorineural hearing loss (RR, 3.68; $95\%$ CI, 2.93–7.04; I2 = $99\%$; $p \leq 0.01$; Supplementary Fig. 2), occupational noise exposure (RR, 1.70; $95\%$ CI, 1.49–1.94; I2 = $48\%$; $$p \leq 0.17$$; Supplementary Fig. 3), otitis media (RR, 1.63; $95\%$ CI, 1.61–1.65; I2 = $0\%$; $$p \leq 0.77$$; Supplementary Fig. 4), and platinum therapy (RR, 2.81; $95\%$ CI, 1.81–4.36; I2 = $26\%$; $$p \leq 0.25$$; Supplementary Fig. 5)]. In contrast, leisure noise exposure was not associated with developing tinnitus (RR, 1.36; $95\%$ CI, 0.70–2.62; I2 = $55\%$; $$p \leq 0.14$$; Supplementary Fig. 6).
Of the lifestyle or socio-demographic factors, only high alcohol consumption was negatively associated with tinnitus (RR, 0.94; $95\%$ CI, 0.91–0.96; I2 = $0\%$; $$p \leq 0.50$$; Supplementary Fig. 7) whereas this was not the case for low alcohol consumption (RR, 1.00; $95\%$ CI, 0.85–1.19; I2 = $42\%$; $$p \leq 0.19$$; Supplementary Fig. 8), smoking (RR, 1.15; $95\%$ CI, 0.81–1.62; I2 = $88\%$; $p \leq 0.01$; Supplementary Fig. 9), or sex (RR, 1.06; $95\%$ CI, 0.92–1.22; I2 = $77\%$; $$p \leq 0.01$$; Supplementary Fig. 10].
Regarding comorbidities, temporo-mandibular joint disorder was associated with tinnitus (RR, 2.06; $95\%$ CI, 1.30–3.27; I2 = $97\%$; $p \leq 0.01$; Supplementary Fig. 11), as well as depression (RR, 1.31; $95\%$ CI, 1.28–1.34; I2 = $0\%$; $$p \leq 0.76$$; Supplementary Fig. 12). In contrast, diabetes (RR, 0.85; $95\%$ CI, 0.82–0.88; I2 = $0\%$; $$p \leq 1$$; Supplementary Fig. 13) was negatively associated with tinnitus.
There was no evidence for an association between tinnitus and the following comorbidities: leisure noise exposure (RR, 1.36, $95\%$ CI, 0.70–2.62; I2 = $55\%$; $$p \leq 0.14$$; Supplementary Fig. 14); body mass index (RR, 0.83; $95\%$ CI, 0.62–1.11; I2 = $78\%$; $$p \leq 0.01$$; Supplementary Fig. 15); heart failure (RR, 0.73; $95\%$ CI, 0.44–1.20; I2 = $90\%$; $p \leq 0.01$; Supplementary Fig. 16), hypertension (RR, 0.98; $95\%$ CI, 0.96–1.00; I2 = $0\%$; $$p \leq 0.39$$; Supplementary Fig. 17), stroke (RR, 0.94; $95\%$ CI, 0.73–1.21; I2 = $0\%$; $$p \leq 0.69$$; Supplementary Fig. 18); hyperlipidemia (RR, 1.18; $95\%$ CI, 1.00–1.40; I2 = $28\%$; $$p \leq 0.24$$; Supplementary Fig. 19); rheumathoid arthritis (RR, 1.19; $95\%$ CI, 0.97–1.47; I2 = $67\%$; $$p \leq 0.08$$; Supplementary Fig. 20); migraine (RR, 2.11; $95\%$ CI, 0.93–4.79; I2 = $93\%$; $p \leq 0.01$; Supplementary Fig. 21); Chronic obstructive pulmonary disease (COPD) (RR, 0.95; $95\%$ CI, 0.90–1.00; I2 = $0\%$; $$p \leq 0.71$$; Supplementary Fig. 22); head injury (RR, 1.21; $95\%$ CI, 0.97–1.52; I2 = $91\%$; $p \leq 0.01$; Supplementary Fig. 23); and whiplash (RR, 1.40; $95\%$ CI, 0.95–2.07; I2 = $60\%$; $$p \leq 0.11$$; Supplementary Fig. 24).
## Discussion
The present study reveals the very limited knowledge on exposures causally related to tinnitus. Using a systematic literature evaluation, 374 articles reported information on exposures related to tinnitus. From this pool, only $13\%$ of articles reported data collected from analytical observational studies (i.e., case–control and cohort studies), of which only half met the quality threshold defined for this review. Thus, from the original set of 374 articles, only $6.7\%$ met our criteria for consideration into the meta-analysis. Our findings confirm the known role of hearing loss in increasing the risk of tinnitus, and this also includes other otological conditions known to affect hearing. Interestingly, our findings also confirm a causal link between temporo-mandibular joint disorder and tinnitus, consistent with previous suggestions [6]. Evidence was also found for a number of other non-otological risk factors including depression, COPD, and hyperlipidemia. Negative associations indicating preventative effects were found for diabetes and high alcohol consumption. These were unexpected findings, but we note that in each case, the pooled estimates come from only two studies and so we have limited confidence that these results are reliable. In the case of diabetes, there is reasonable evidence for a positive association between type 1 and type 2 diabetes and sensorineural hearing loss, and it is suggested that there may be shared risk factors such as glucose processing abnormalities and aging, in addition to some pathologies created by diabetes leading to hearing loss [53]. It is noted that neither of the two included diabetes studies [34, 52] adjusted their models for hearing loss and so this confound needs to be examined in further research. Furthermore, at least one cross-sectional study has suggested an interaction between exposures, such that comorbidity of diabetes and hypertension poses a risk factor for tinnitus [54]. However multi-comorbidities are rarely assessed and are not considered in the current review. This review highlighted little or no evidence for the numerous exposures that have been previously suggested as potential risk factors in previous cross-sectional studies (e.g., migraine, head injury, and whiplash, [6]) or identified as a research priority (e.g., caffeine, [17]).
## Quality and Availability of Evidence
The interpretation of results from systematic reviews depends on the quality of evidence and risk of bias [55]. The degree of reliability of analytical observational studies can be from excellent to very poor depending on selection of study groups, comparability of groups, and ascertainment of exposure and outcomes (three points) for case–control and cohort studies, respectively [28]. Almost half of the cohort studies had to be excluded as they did not report sufficient information, such as risk ratios, and so the quality appraisal was not conducted. However, for those included studies that were evaluated using NOS, the quality assessment score was high. There is an evident need for well-conducted analytical observational studies in the tinnitus field, such that there is enough validity in the results both when studies are considered individually or as a part of a systematic review.
## Gaps in Knowledge
The two most striking gaps highlighted in this review include the lack of analytical observational studies and gaps in knowledge about some known risk factors. It is impossible to elicit the cause-and-effect relationship between potential exposures and tinnitus without analytical observational data. For example, previously Deklerck et al. [ 6] suggested links between various cardiovascular disorders (e.g., dyslipidaemia, peripheral vascular disease and ischaemic heart disease, and stroke) and tinnitus. However, excluding the cross-sectional studies and taking into account only the high quality studies, our review shows that the current evidence is not reliable. In the case of anxiety disorders that have consistently been related to tinnitus [3, 56], and for which treatments reduce the impact of tinnitus [56], only one case–control study reported a positive association [52]. Similarly, knowledge on the association of otosclerosis and tinnitus results from a single case–control study [20] that has not been replicated elsewhere. Finally, factors like diet and physical activity, for which suggestive evidence of a relationship exists from cross-sectional datasets, have never been investigated in analytical observational studies.
## Self-Reported Tinnitus Versus Medically Diagnosed Tinnitus
Another consideration that arose during the present study is that while most population-based studies rely on self-reports of tinnitus, case–control or retrospective cohorts are often based on electronic health records using medically diagnosed tinnitus (e.g., tagged with an ICD code, International Classification of Diseases) as the outcome of interest. Examples include the Taiwan National Health Insurance Resource Database and Clinical Practice Research Datalink. It is indeed possible to conduct hospital-based case–control studies from ENT clinics or population-based case–control studies from existing healthcare or health insurance databases by selecting tinnitus cases and tinnitus-free controls matched for index dates. However, to obtain accurate results, in both instances, reliable coding of tinnitus is essential. One limitation is that ENT doctors may not reliably record tinnitus codes as this is often a secondary symptom associated with a more primary otological disorder. Another limitation concerns the lack of etiologically meaningful tinnitus subtypes. For example, ICD-10 subcodes for tinnitus specify only left ear, right ear, bilateral, or unspecified. While updates to the ICD codes are possible, this would require global consensus by lead experts in order to have these approved by the World Health Organization and further implemented in national medical systems.
Here, nearly half of the analytical observational studies ($\frac{13}{25}$) relied on ICD coding from medical or insurance registry data. The forest plots presented in the Supplementary Figures provide pooled estimates for self-reported AT separate from that for MedT and significantT, in addition to the overall pooled estimate. This provides optimal transparency for the reader. Such healthcare or health insurance databases can also be used to conduct retrospective cohort studies. For example, Martinez et al. [ 57] conducted a retrospective cohort study using UK NHS healthcare records (i.e., Clinical Practice Research Datalink and Hospital Episode Statistics) to determine incidence. Recently, Lugo et al. [ 58] used medical registry data within the Stockholm Public Health Cohort to establish that women with a previous medical evaluation at the specialty care had lower risk for suicidal attempts. These are convenient methods for hypothesis generation and quantify some degrees of association between an exposure and tinnitus. However, tinnitus in a help-seeking individual is qualitatively different from tinnitus in an individual not seeking medical support, and while the former is evaluated by a healthcare professional, the latter is self-reported [59]. In addition, caution should prevail as individuals that may have been assessed by a physician may suffer less than individuals with self-reported severe tinnitus that have not benefited from medical care. As exemplified by the study from Lugo et al. [ 58], the relationship between severe tinnitus and suicide attempts may have been underestimated if solely focusing on ICD-coded data. This may be particularly relevant for any psychological or psychiatric oriented study related to tinnitus. Moreover, individuals that have been labeled an ICD code for tinnitus may not necessarily have a severe tinnitus, as it may have been diagnosed as a secondary symptom accompanying for instance hearing loss, or it may have been an acute tinnitus remitting shortly after the auscultation (occasional tinnitus). Similar to other psychiatric traits that cannot be reliably objectively diagnosed, we recommend that a pattern of help-seeking behavior (such as at least one referral to specialty care for a primary tinnitus complaint) may provide increased reliability over primary care diagnoses when performing retrospective studies.
## Limitations
We acknowledge that an inherent limitation of our study is the paucity of risk ratio data available for meta-analysis. The results presented here were conducted on data reported from an average of two or three sources, thereby limiting the generalizability and the validity of the findings. This precludes the possibility of firm causal inferences on the association between tinnitus and exposures. Moreover, the few high-quality studies available have explored a selected set of risk factors with a clear preference for auditory conditions, resulting in major gaps in knowledge for many relevant associated conditions. The latest update was performed on $\frac{06}{11}$/2019, and it is possible that new tinnitus results may have been reported since.
Another second limitation is the incapacity to draw conclusion on sex-specific mechanisms, which is acknowledged as being under investigated in the tinnitus field [60]. Whereas we have found no association between tinnitus and sex in the present meta-analysis, a number of studies suggest that there could be influences of sex on, for instance, tinnitus severity. Previous work from Schlee et al. [ 61] reported greater stress and anxiety in women with constant tinnitus. These findings are consistent with another report showing that women with severe tinnitus have greater odds of suicide attempts, something that is not found in men [58]. It is possible that such mechanisms are, at least in part, due to genetics since woman with tinnitus have almost ten times the risk of having a sibling with tinnitus [λs = 10.25; $95\%$ CI (7.14–13.61)] whereas for men, this risk is fivefold [λs = 5.03; $95\%$ CI (3.22–7.01)] [12]. A recent cross-sectional study by Basso et al. [ 62] has found that woman with bothersome tinnitus more often report cardiovascular diseases, thyroid diseases, epilepsy, fibromyalgia, and burnout, whereas in men, bothersome tinnitus is related to alcohol consumption, meniere’s disease, anxiety, and panic disorders. The direction of these relationships needs to be assessed in well-designed prospective studies. Some studies report also differences in therapeutic outcome with women being more responsive than men to some treatments as for instance transcranial magnetic stimulation [63], acoustic stimulation [64], or high definition transcranial direct current stimulation [65]. While sex often has an impact on a primary outcome, it does not mean the outcome is sexually dimorphic (i.e., significantly different between the sexes). Future studies will need to address these aspects, as it may also lead to therapeutic benefits tailored to one or the other of the sexes.
A third limitation is the large heterogeneity in the use of adjustment factors, which may also account for the between-study heterogeneity which was high for factors such as unspecified hearing loss, sensorineural hearing loss, smoking, sex, temporomandibular joint disorder, body mass index, heart failure, migraine, and head injury (I2 = 69–$99\%$, $p \leq 0.05$). Quite obviously, meta-analysis would be optimized when pooling studies with compatible statistical measures. Of note, it is interesting to note that there was also a broad heterogeneity in the selection of the models for statistical analyses. Cox regression analyses were used in 8 of 25 studies, three of them with tinnitus being self-reported. Cox regression is traditionally used as a model for when a disease status is achieved, and it also remains (e.g., HIV, mortality, cancer). Whereas in medically assessed tinnitus, there is an increased likelihood of tinnitus being more severe than in self-reported tinnitus; there is a large degree of uncertainty in both cases whether tinnitus was perceived occasionally or constant. Indeed, Edvall et al. [ 66] recently reported using longitudinal data that the more often occasional tinnitus is perceived, the more likely it will become constant and that constant tinnitus increases the odds of tinnitus becoming permanent. In this study, given the highly dynamic transitions of tinnitus states in individuals with occasional tinnitus, a decision was taken not to use a Cox regression model, and instead rely on generalized estimating equation (GEE) to circumvent these issues (Cederroth, personal communication). This study also reveals that individuals with constant tinnitus differ from those with occasional tinnitus and non-tinnitus controls in that they display increased latencies of Wave V of the auditory brainstem response (ABR). The ABR from individuals with occasional tinnitus was indistinguishable from the non-tinnitus controls. These findings overall suggest, as mentioned above in the case of medically assessed tinnitus, that all methods to ensure that tinnitus was at least chronic and constant may help refine the current picture and reinforce the evidence on reliable risk factors for tinnitus. In this case, then Cox regression models may be justified.
## Future Directions
Prospective cohort studies provide the strongest evidence in risk factor analysis and so there is scope for more of these study designs. Some large population-based prospective cohorts have explored the relationship between tinnitus and various exposures. Conducting new, large, and well-designed prospective cohort studies would be ideal — however, we acknowledge this is a major effort that tinnitus alone may not justify. It would also be worthwhile and resource-friendly to explore the existing databases not only to look for associations between tinnitus and unexplored exposures, but also to validate and replicate previous findings. Establishing a resource with a list of databases including tinnitus phenotypes may be extremely useful to the research community, but the benefits to the society may only derive from high-quality studies. In addition, a broad communication endeavor will be required to improve the phenotypic definitions of tinnitus in existing prospective cohorts such that high-quality data may be acquired in the near future.
## Concluding Remarks
Data identification, synthesis, and reporting have emerged as necessary steps in dissemination and translation of raw data into clinical practice and policy decisions. High-quality primary data acquisition is a pre-requisite for high-quality data synthesis. For exposure assessment, this implies conducting well-designed analytical observational studies to provide stronger evidence. Findings from this review show that tinnitus is related to multiple exposures. However, for most of them, there are insufficient data to conclude a causal relationship. The knowledge of risk factors is crucial for understanding tinnitus etiology as it could also provide insights for pathophysiological mechanisms, contribute to improved tinnitus management, and individualize treatments based on the underlying cause, instead of alleviating the tinnitus percept alone.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (PDF 262 KB)Supplementary file1 (PDF 262 KB)Supplementary file2 (XLSX 232 KB)
## References
1. Jarach CM, Lugo A, Scala M, van den Brandt PA, Cederroth CR, Odone A, Garavello W, Schlee W, Langguth B, Gallus S. **Global prevalence and incidence of tinnitus: A systematic review and meta-analysis**. *JAMA Neurol* (2022.0) **79** 888-900. DOI: 10.1001/jamaneurol.2022.2189
2. Cederroth CR, Gallus S, Hall DA, Kleinjung T, Langguth B, Maruotti A, Meyer M, Norena A, Probst T, Pryss RC. **Editorial: towards an Understanding of Tinnitus Heterogeneity**. *Front Aging Neurosci* (2019.0) **11** 53. DOI: 10.3389/fnagi.2019.00053
3. Baguley DM, McFerran DJ, Hall DA. **Tinnitus**. *Lancet* (2013.0) **382** 1600-1607. DOI: 10.1016/s0140-6736(13)60142-7
4. McFerran DJ, Stockdale D, Holme R, Large CH, Baguley DM. **Why is there no cure for tinnitus?**. *Front Neurosci* (2019.0) **13** 802. DOI: 10.3389/fnins.2019.00802
5. Eggermont JJ. **Tinnitus and neural plasticity (Tonndorf lecture at XIth International Tinnitus Seminar, Berlin, 2014)**. *Hear Res* (2015.0) **19** 1-11. DOI: 10.1016/j.heares.2014.10.002
6. 6.Deklerck AN, Debacker JM, Keppler H, Dhooge I (2020) Identifying non-otologic risk factors for tinnitus: a systematic review. Clin Otolaryngol: official J ENT-UK; official J Netherlands Soc Oto-Rhino-Laryngology & Cervico-Facial Surg 45(5):775–787. 10.1111/coa.13592
7. Haider HF, Hoare DJ, Costa R, Potgieter I, Kikidis D, Lapira A, Nikitas C, Caria H, Cunha NT, Paço JC. **Pathophysiology, diagnosis and treatment of somatosensory tinnitus: a scoping review**. *Front Neurosci* (2017.0) **11** 207. DOI: 10.3389/fnins.2017.00207
8. Lugo A, Edvall NK, Lazar A, Mehraei G, Lopez-Escamez JA, Bulla J, Uhlen I, Canlon B, Gallus S, Cederroth CR. **Relationship between headaches and tinnitus in a Swedish study**. *Sci Rep* (2020.0) **10** 8494. DOI: 10.1038/s41598-020-65395-1
9. Cederroth CR, Lugo A, Edvall NK, Lazar A, Lopez-Escamez JA, Bulla J, Uhlen I, Hoare DJ, Baguley DM, Canlon B, Gallus S. **Association between hyperacusis and tinnitus**. *J Clin Med* (2020.0) **9** 2412. DOI: 10.3390/jcm9082412
10. Edvall NK, Gunan E, Genitsaridi E, Lazar A, Mehraei G, Billing M, Tullberg M, Bulla J, Whitton J, Canlon B, Hall DA, Cederroth CR. **Impact of temporo-mandibular joint complaints on tinnitus-related distress**. *Front Neurosci* (2019.0) **13** 879. DOI: 10.3389/fnins.2019.00879
11. Biswas R, Lugo A, Genitsaridi E, Trpchevska N, Akeroyd MA, Cederroth CR, Liu X, Schlee W, Garavello W, Gallus S. **Modifiable lifestyle-related risk factors for tinnitus in the general population: An overview of smoking, alcohol, body mass index and caffeine intake**. *Prog Brain Res* (2021.0) **263** 1-24. DOI: 10.1016/bs.pbr.2021.04.005
12. Trpchevska N, Bulla J, Prada Hellberg M, Edvall NK, Lazar A, Mehraei G, Uhlen I, Schlee W, Canlon B, Gallus S, Lopez-Escamez JA, Cederroth CR. **Sex-dependent aggregation of tinnitus in Swedish families**. *J Clin Med* (2020.0) **9** 3812. DOI: 10.3390/jcm9123812
13. 13.Cederroth CR, PirouziFard M, Trpchevska N, Idrizbegovic E, Canlon B, Sundquist J, Sundquist K, Zöller B (2019) Association of genetic vs environmental factors in Swedish adoptees with clinically significant tinnitus. JAMA Otolaryngol-- Head & Neck Surg 145(3):222–229. 10.1001/jamaoto.2018.3852
14. Maas IL, Brüggemann P, Requena T, Bulla J, Edvall NK, Hjelmborg J, Szczepek AJ, Canlon B, Mazurek B, Lopez-Escamez JA, Cederroth CR. **Genetic susceptibility to bilateral tinnitus in a Swedish twin cohort**. *Genetics in medicine: official journal of the American College of Medical Genetics* (2017.0) **19** 1007-1012. DOI: 10.1038/gim.2017.4
15. Amanat S, Gallego-Martinez A, Sollini J, Perez-Carpena P, Espinosa-Sanchez JM, Aran I, Soto-Varela A, Batuecas-Caletrio A, Canlon B, May P, Cederroth CR, Lopez-Escamez JA. **Burden of rare variants in synaptic genes in patients with severe tinnitus: an exome based extreme phenotype study**. *EBioMedicine* (2021.0) **66** 103309. DOI: 10.1016/j.ebiom.2021.103309
16. 16.Clifford RE, Maihofer AX, Stein MB, Ryan AF, Nievergelt CM (2020) Novel risk loci in tinnitus and causal inference with neuropsychiatric disorders among adults of European ancestry. JAMA Otolaryngol Head Neck Surg 146(11):1015–1025. 10.1001/jamaoto.2020.2920
17. 17.British Tinnitus Association (2012) Tinnitus priority setting partnership. British Tinnitus Association. https://www.jla.nihr.ac.uk/priority-setting-partnerships/tinnitus/downloads/JLA_Tinnitus_Final_PSP_Workshop_Booklet.pdf
18. 18.Biswas R, Hall DA (2020) Prevalence, incidence, and risk factors for tinnitus. In: Current Topics in Behavioral Neurosciences. Springer, Berlin, Heidelberg. 10.1007/7854_2020_154
19. Cruickshanks KJ, Wiley TL, Tweed TS, Klein BE, Klein R, Mares-Perlman JA, Nondahl DM. **Prevalence of hearing loss in older adults in Beaver Dam, Wisconsin. The epidemiology of hearing loss study**. *Am J Epidemiol* (1998.0) **148** 879-886. DOI: 10.1093/oxfordjournals.aje.a009713
20. Nondahl DM, Cruickshanks KJ, Wiley TL, Klein R, Klein BE, Tweed TS. **Prevalence and 5-year incidence of tinnitus among older adults: the epidemiology of hearing loss study**. *J Am Acad Audiol* (2002.0) **13** 323-331. DOI: 10.1055/s-0040-1715975
21. Offord DR, Kraemer HC. **Risk factors and prevention**. *Evid Based Ment Health* (2000.0) **3** 70. DOI: 10.1136/ebmh.3.3.70
22. Westreich D, Edwards JK, Lesko CR, Cole SR, Stuart EA. **Target validity and the hierarchy of study designs**. *Am J Epidemiol* (2019.0) **188** 438-443. DOI: 10.1093/aje/kwy228
23. 23.Curti SA, Taylor EN, Su D, Spankovich C (2019) Prevalence of and characteristics associated with self-reported good hearing in a population with elevated audiometric thresholds. JAMA Otolaryngol-- Head & Neck Surg 145:626–633. 10.1001/jamaoto.2019.1020
24. 24.Ramage-Morin PL, Banks R, Pineault D, Atrach M (2019) Unperceived hearing loss among Canadians aged 40 to 79. Health Reps 30:11–20. 10.25318/82-003-x201900800002-eng
25. Tremblay KL, Backer KC. **Listening and learning: cognitive contributions to the rehabilitation of older adults with and without audiometrically defined hearing loss**. *Ear Hear* (2016.0) **37** 155s-s162. DOI: 10.1097/aud.0000000000000307
26. Veile A, Zimmermann H, Lorenz E, Becher H. **Is smoking a risk factor for tinnitus? A systematic review, meta-analysis and estimation of the population attributable risk in Germany**. *BMJ Open* (2018.0) **8** e016589. DOI: 10.1136/bmjopen-2017-016589
27. 27.Gallus S, Lugo A, Biswas R, Liu X, Cederroth C, Canlon B, Trpchevska N, Hall D, Genitsaridi E, Schlee W, Langguth B (2018) Systematic literature review and meta-analysis on the risk factors for tinnitus. PROSPERO, CRD42018095610 Available from: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42018095610
28. 28.Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M, Tugwell P (2013) The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp
29. Nondahl DM, Cruickshanks KJ, Wiley TL, Klein BE, Klein R, Chappell R, Tweed TS. **The ten-year incidence of tinnitus among older adults**. *Int J Audiol* (2010.0) **49** 580-585. DOI: 10.3109/14992021003753508
30. Gopinath B, McMahon CM, Rochtchina E, Karpa MJ, Mitchell P. **Risk factors and impacts of incident tinnitus in older adults**. *Ann Epidemiol* (2010.0) **20** 129-135. DOI: 10.1016/j.annepidem.2009.09.002
31. 31.Lee CF, Lin MC, Lin HT, Lin CL, Wang TC, Kao CH (2016) Increased risk of tinnitus in patients with temporomandibular disorder: a retrospective population-based cohort study. Eur Archives of oto-rhino-laryngol: official J Eur Federation of Oto-Rhino-Laryngol Soc (EUFOS) : affiliated with the German Soc Oto-Rhino-Laryngology - Head and Neck Surg 273(1):203–208. 10.1007/s00405-015-3491-2
32. Hwang J-H, Tsai S-J, Liu T-C, Chen Y-C, Lai J-T. **Association of tinnitus and other cochlear disorders with a history of migraines**. *JAMA Otolaryngology-Head & Neck Surgery* (2018.0) **144** 712-717. DOI: 10.1001/jamaoto.2018.0939
33. Chen HC, Chung CH, Chen VCF, Wang YC, Chien WC. **Hormone replacement therapy decreases the risk of tinnitus in menopausal women: a nationwide study**. *Oncotarget* (2018.0) **9** 19807-19816. DOI: 10.18632/oncotarget.24452
34. Shih CP, Lin HC, Chung CH, Hsiao PJ, Wang CH, Lee JC, Chien WC. **Increased risk of tinnitus in patients with chronic kidney disease: a nationwide, population-based cohort study**. *PLoS ONE* (2017.0) **12** e0183192. DOI: 10.1371/journal.pone.0183192
35. 35.Aarhus L, Engdahl B, Tambs K, Kvestad E, Hoffman HJ (2015) Association between childhood hearing disorders and tinnitus in adulthood. JAMA Otolaryngol Head & Neck Surg 141(11):983–989. 10.1001/jamaoto.2015.2378
36. Aarhus L, Homoe P, Engdahl B. **Otitis media in childhood and disease in adulthood: a 40-year follow-up study**. *Ear Hear* (2019.0). DOI: 10.1097/aud.0000000000000729
37. 37.Dougherty AL, MacGregor AJ, Han PP, Viirre E, Heltemes KJ, Galarneau MR (2013) Blast-related ear injuries among U.S. military personnel. J Rehab Res Dev 50(6):893–904. 10.1682/JRRD.2012.02.0024
38. Vessey M, Painter R. **Oral contraception and ear disease: findings in a large cohort study**. *Contraception* (2001.0) **63** 61-63. DOI: 10.1016/s0010-7824(01)00176-7
39. Bernhardt O, Mundt T, Welk A, Köppl N, Kocher T, Meyer G, Schwahn C. **Signs and symptoms of temporomandibular disorders and the incidence of tinnitus**. *J Oral Rehabil* (2011.0) **38** 891-901. DOI: 10.1111/j.1365-2842.2011.02224.x
40. Auvinen A, Feychting M, Ahlbom A, Hillert L, Elliott P, Schuz J, Tettamanti G. **Headache, tinnitus and hearing loss in the international Cohort Study of Mobile Phone Use and Health (COSMOS) in Sweden and Finland**. *Int J Epidemiol* (2019.0). DOI: 10.1093/ije/dyz127
41. Glicksman JT, Curhan SG, Curhan GC. **A prospective study of caffeine intake and risk of incident tinnitus**. *Am J Med* (2014.0) **127** 739-743. DOI: 10.1016/j.amjmed.2014.02.033
42. Gopinath B, McMahon CM, Rochtchina E, Karpa MJ, Mitchell P. **Incidence, persistence, and progression of tinnitus symptoms in older adults: the Blue Mountains Hearing Study**. *Ear Hear* (2010.0) **31** 407-412. DOI: 10.1097/AUD.0b013e3181cdb2a2
43. Fredriksson S, Kim JL, Toren K, Magnusson L, Kahari K, Soderberg M, Persson Waye K. **Working in preschool increases the risk of hearing-related symptoms: a cohort study among Swedish women**. *Int Arch Occup Environ Health* (2019.0) **92** 1179-1190. DOI: 10.1007/s00420-019-01453-0
44. Schink T, Kreutz G, Busch V, Pigeot I, Ahrens W. **Incidence and relative risk of hearing disorders in professional musicians**. *Occup Environ Med* (2014.0) **71** 472-476. DOI: 10.1136/oemed-2014-102172
45. Chen YC, Tsai SJ, Chen JC, Hwang JH. **Risks of tinnitus, sensorineural hearing impairment, and sudden deafness in patients with non-migraine headache**. *PLoS ONE* (2019.0) **14** e0222041. DOI: 10.1371/journal.pone.0222041
46. Yang CC, Tai CJ, Chien SH, Lin CL, Chang SN, Sung FC, Chung CJ, Kao CH. **Risk of ear-associated diseases after zygomaticomaxillary complex fracture**. *J Maxillofac Oral Surg* (2015.0) **14** 646-652. DOI: 10.1007/s12663-015-0744-y
47. 47.Eick-Cost AA, Hu Z, Rohrbeck P, Clark LL (2017) Neuropsychiatric outcomes after mefloquine exposure among U.S. military service members. The Am J Trop Med Hygiene 96(1):159–166. 10.4269/ajtmh.16-0390.
48. King AA, Seidel K, Di C, Leisenring WM, Perkins SM, Krull KR, Sklar CA, Green DM, Armstrong GT, Zeltzer LK, Wells E, Stovall M, Ullrich NJ, Oeffinger KC, Robison LL, Packer RJ. **Long-term neurologic health and psychosocial function of adult survivors of childhood medulloblastoma/PNET: a report from the Childhood Cancer Survivor Study**. *Neuro Oncol* (2017.0) **19** 689-698. DOI: 10.1093/neuonc/now242
49. Whelan K, Stratton K, Kawashima T, Leisenring W, Hayashi S, Waterbor J, Blatt J, Sklar CA, Packer R, Mitby P, Robison LL, Mertens AC. **Auditory complications in childhood cancer survivors: a report from the childhood cancer survivor study**. *Pediatr Blood Cancer* (2011.0) **57** 126-134. DOI: 10.1002/pbc.23025
50. Koo M, Hwang JH. **Risk of tinnitus in patients with sleep apnea: A nationwide, population-based, case-control study**. *Laryngoscope* (2017.0) **127** 2171-2175. DOI: 10.1002/lary.26323
51. Hutter HP, Moshammer H, Wallner P, Cartellieri M, Denk-Linnert DM, Katzinger M, Ehrenberger K, Kundi M. **Tinnitus and mobile phone use**. *Occup Environ Med* (2010.0) **67** 804-808. DOI: 10.1136/oem.2009.048116
52. Stohler NA, Reinau D, Jick SS, Bodmer D, Meier CR. **A study on the epidemiology of tinnitus in the United Kingdom**. *Clin Epidemiol* (2019.0) **11** 855-871. DOI: 10.2147/clep.s213136
53. Samocha-Bonet D, Wu B, Ryugo DK. **Diabetes mellitus and hearing loss: a review**. *Ageing Res Rev* (2021.0) **71** 101423. DOI: 10.1016/j.arr.2021.101423
54. Gibrin PC, Melo JJ, Marchiori LL. **Prevalence of tinnitus complaints and probable association with hearing loss, diabetes mellitus and hypertension in elderly**. *CoDAS* (2013.0) **25** 176-180. DOI: 10.1590/s2317-17822013000200014
55. van Dalen EC, Kremer LCM, Moyer VA. **Tips and tricks for understanding and using SR results—no. 8: quality of studies included in a systematic review and associated risk of bias- ‘garbage in, garbage out’**. *Evidence-Based Child Health: A Cochrane Review Journal* (2007.0) **2** 1321-1324. DOI: 10.1002/ebch.17310.1002/ebch.173</p>
56. Langguth B, Kreuzer PM, Kleinjung T, De Ridder D. **Tinnitus: causes and clinical management**. *Lancet Neurology, The* (2013.0) **12** 920-930. DOI: 10.1016/S1474-4422(13)70160-1
57. Martinez C, Wallenhorst C, McFerran D, Hall DA. **Incidence rates of clinically significant tinnitus: 10-year trend from a cohort study in England**. *Ear Hear* (2015.0) **36** e69-e75. DOI: 10.1097/AUD.0000000000000121
58. Lugo A, Trpchevska N, Liu X, Biswas R, Magnusson C, Gallus S, Cederroth CR. **Sex-specific association of tinnitus with suicide attempts**. *JAMA Otolaryngology-Head & Neck Surgery* (2019.0) **145** 685-687. DOI: 10.1001/jamaoto.2019.0566
59. 59.De Ridder D, Schlee W, Vanneste S, Londero A, Weisz N, Kleinjung T, Shekhawat GS, Elgoyhen AB, Song J-J, Andersson G, Adhia D, de Azevedo AA, Baguley DM, Biesinger E, Binetti AC, Del Bo L, Cederroth CR, Cima R, Eggermont JJ, Figueiredo R, Fuller TE, Gallus S, Gilles A, Hall DA, Van de Heyning P, Hoare DJ, Khedr EM, Kikidis D, Kleinstaeuber M, Kreuzer PM, Lai J-T, Lainez JM, Landgrebe M, Li LP-H, Lim HH, Liu T-C, Lopez-Escamez JA, Mazurek B, Moller AR, Neff P, Pantev C, Park SN, Piccirillo JF, Poeppl TB, Rauschecker JP, Salvi R, Sanchez TG, Schecklmann M, Schiller A, Searchfield GD, Tyler R, Vielsmeier V, Vlaeyen JWS, Zhang J, Zheng Y, de Nora M, Langguth B (2021) Tinnitus and tinnitus disorder: theoretical and operational definitions (an international multidisciplinary proposal). In: Progress in Brain Research. ‘Tinnitus - An Interdisciplinary Approach Towards Individualized Treatment: From Heterogeneity to Personalized Medicine vol. 260. Langguth B, de Ridder D, Kleinjung T, Vanneste S, Schlee W (Eds). Elsevier. pp 1–25. 10.1016/bs.pbr.2020.12.002
60. Cederroth CR, Schlee W. **Editorial: sex and gender differences in tinnitus**. *Front Neurosci* (2022.0) **16** 844267. DOI: 10.3389/fnins.2022.844267
61. Schlee W, Hall DA, Edvall NK, Langguth B, Canlon B, Cederroth CR. **Visualization of global disease burden for the optimization of patient management and treatment**. *Front Med* (2017.0) **4** 86. DOI: 10.3389/fmed.2017.00086
62. Basso L, Boecking B, Brueggemann P, Pedersen NL, Canlon B, Cederroth CR, Mazurek B. **Gender-specific risk factors and comorbidities of bothersome tinnitus**. *Front Neurosci* (2020.0) **14** 706. DOI: 10.3389/fnins.2020.00706
63. De Ridder D, van der Loo E, Van der Kelen K, Menovsky T, van de Heyning P, Moller A. **Theta, alpha and beta burst transcranial magnetic stimulation: brain modulation in tinnitus**. *Int J Med Sci* (2007.0) **4** 237-241. DOI: 10.7150/ijms.4.237
64. Partyka M, Neff P, Bacri T, Michels J, Weisz N, Schlee W. **Gender differentiates effects of acoustic stimulation in patients with tinnitus**. *Prog Brain Res* (2021.0) **263** 25-57. DOI: 10.1016/bs.pbr.2021.04.010
65. Van der Wal A, Luyten T, Cardon E, Jacquemin L, Vanderveken OM, Topsakal V, Van de Heyning P, De Hertogh W, Van Looveren N, Van Rompaey V, Michiels S, Gilles A. **Sex differences in the response to different tinnitus treatment**. *Front Neurosci* (2020.0) **14** 422. DOI: 10.3389/fnins.2020.00422
66. Edvall NK, Mehraei G, Claeson M, Lazar A, Bulla J, Leineweber C, Uhlén I, Canlon B, Cederroth CR. **Alterations in auditory brain stem response distinguish occasional and constant tinnitus**. *J Clin Investig* (2022.0) **132** e155094. DOI: 10.1172/JCI155094
|
---
title: Network pharmacology and experimental validation to elucidate the pharmacological
mechanisms of Bushen Huashi decoction against kidney stones
authors:
- Haizhao Liu
- Min Cao
- Yutong Jin
- Beitian Jia
- Liming Wang
- Mengxue Dong
- Lu Han
- Joseph Abankwah
- Jianwei Liu
- Tao Zhou
- Baogui Chen
- Yiyang Wang
- Yuhong Bian
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9971497
doi: 10.3389/fendo.2023.1031895
license: CC BY 4.0
---
# Network pharmacology and experimental validation to elucidate the pharmacological mechanisms of Bushen Huashi decoction against kidney stones
## Abstract
### Introduction
Kidney stone disease (KS) is a complicated disease with an increasing global incidence. It was shown that Bushen Huashi decoction (BSHS) is a classic Chinese medicine formula that has therapeutic benefits for patients with KS. However, its pharmacological profile and mechanism of action are yet to be elucidated.
### Methods
The present study used a network pharmacology approach to characterize the mechanism by which BSHS affects KS. Compounds were retrieved from corresponding databases, and active compounds were selected based on their oral bioavailability (≥30) and drug-likeness index (≥0.18). BSHS potential proteins were obtained from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) database, whereas KS potential genes were obtained from GeneCards and OMIM, TTD, and DisGeNET. Gene ontology and pathway enrichment analysis were used to determine potential pathways associated with genes. The ingredients of BSHS extract were identified by the ultra‐high‐performance liquid chromatography coupled with quadrupole orbitrap mass spectrometry (UHPLC-Q/Orbitrap MS). The network pharmacology analyses predicted the potential underlying action mechanisms of BSHS on KS, which were further validated experimentally in the rat model of calcium oxalate kidney stones.
### Results
Our study found that BSHS reduced renal crystal deposition and improved renal function in ethylene glycol(EG)+ammonium chloride(AC)-induced rats, and also reversed oxidative stress levels and inhibited renal tubular epithelial cell apoptosis in rats. BSHS upregulated protein and mRNA expression of E2, ESR1, ESR2, BCL2, NRF2, and HO-1 in EG+AC-induced rat kidney while downregulating BAX protein and mRNA expression, consistent with the network pharmacology results.
### Discussion
This study provides evidence that BSHS plays a critical role in anti-KS via regulation of E2/ESR$\frac{1}{2}$, NRF2/HO-1, and BCL2/BAX signaling pathways, indicating that BSHS is a candidate herbal drug for further investigation in treating KS.
## Introduction
Kidney stone disease is a urinary system disease caused by the kidney’s abnormal accumulation of crystalline material such as calcium, oxalic acid, uric acid, and cystine [1]. With a prevalence of $7\%$ to $13\%$ in North America, 5-$9\%$ in Europe, and 1-$5\%$ in Asia, it is one of the most common diseases affecting populations all over the world [1]. Recent data from the United States showed that the prevalence of stones in the United States was $8.8\%$, with $10.6\%$ and $7.1\%$ reported for their prevalence in men and women, respectively [2]. Primary hyperparathyroidism [3], obesity [4], diabetes [5, 6], and hypertension [7, 8] are some of the risk factors for kidney stone formation. Patients with kidney stones are also at high risk of hypertension [7], chronic kidney disease (CKD) [9, 10], and end-stage renal disease (ESRD) [10, 11]. Kidney stones always lead to several complications, such as urinary tract obstruction, hydronephrosis, infection, local damage to the kidney, and renal dysfunction [12]. The formation of kidney stones is a complex process involving urinary supersaturation, nucleation, growth, aggregation, and retention of urinary stone components within the renal tubular cells [13]. Surgical treatment for the removal of kidney and ureteral stones has already achieved mature development over the past several decades [14]. Currently, the most common kidney stone treatments include shock wave lithotripsy, ureteroscopic fragmentation and removal, and percutaneous nephrolithotomy [14]. Although surgical therapies have greatly resolved patients’ pain, postoperative adverse effects and a high recurrence rate of stones are vexing. Therefore, scientists are focusing on exploring new targets and new drugs, which is crucial to reduce the incidence and recurrence rate of kidney stones.
Traditional Chinese medicine (TCM) has been effectively used in treating diverse diseases for a long time in China and is also traditionally used to treat kidney stones [15]. Different from the single-target concept of Western medicine, TCM emphasizes the concept of the body as an organic whole. Generally, a TCM prescription is composed of several herbs. Each herb typically has multiple active compounds that simultaneously act on multi-targets. Due to the complexity of the TCM components, conventional pharmacology research methods are complex to fully elucidate the potential molecular mechanism of TCM compounds in disease treatment.
BSHS is an experienced TCM prescription of the National Famous TCM Doctor, Professor Baogui Chen, commonly used to treat kidney stones and has shown remarkable effectiveness in clinical practice. BSHS is mainly composed of eight Chinese herbs, including Lysimachiae Herba, Lygodii Spora, Plantaginis Semen, Clerodendranthus Spicatus, Cibotii Rhizoma, Dipsaci Radix, *Malva verticillata* seed, and Licorice. It remains relatively unclear, however, what the bioactive components of THCQ are as well as their pharmacological mechanisms. During the past few years, system biology, polypharmacology, and system biology-based network pharmacology have boomed due to the increase in biomedical data. With network pharmacology, target molecules, biological functions, and bioactive compounds can be combined to form complex interaction networks, which are precisely in line with the natural characteristics of TCM and can improve our understanding of TCM’s mechanisms of action [16]. A network pharmacology approach can contribute to our understanding of the multicomponent, multitarget, and multi-pathway nature of TCM. This study aimed to decipher the mechanism of action of BSHS in suppressing kidney stones by integrating network pharmacology and pharmacological evaluation.
## Reagents and materials
All medicinal plants were provided by Wuqing Hospital of Traditional Chinese Medicine affiliated to Tianjin University of Traditional Chinese Medicine (Tianjin, China). The Paishi granule (PSG) was purchased from Nanjing Tongrentang Pharmaceutical Co., Ltd. (Nanjing, China). The Urea (BUN) Assay Kit, Creatinine (Cr) Assay kit, Calcium (Ca) Assay Kit, Alkaline phosphatase (ALP) assay kit, Magnesium (Mg) Assay Kit, Superoxide Dismutase (SOD) assay kit, and Malondialdehyde (MDA) assay kit were purchased from Nanjing Jiancheng Bioengineering Research Institute (Nanjing, China). The oxalate content detection kit and BCA Protein Assay Kit were purchased from Beijing Suolaibao Technology Co., Ltd. (Beijing, China), whiles Von Kossa dye and HE dye were purchased from Wuhan Servicebio Technology Co., Ltd. (Wuhan, China). The ELISA assay kits for estradiol (E2) were purchased from cloud-clone Corp. Wuhan (Wuhan, China). The anti-estrogen receptor alpha (ERα) antibody, anti-NRF2 antibody, anti-BCL-2 antibody, and anti-Bax antibody were purchased from Abcam (USA). The Caspase-3 Antibody and Cleaved Caspase-3 Antibody were purchased from CST (USA). The Estrogen receptor beta (ERβ) Rabbit pAb and HO-1 Rabbit pAb were purchased from Abclonal (Wuhan, China). The HPLC grade acetonitrile and formic acid was purchased from Fisher Chemicals (Fisher Scientific, Waltham, MA, USA). The purified water was purchased from Guangzhou Watsons Food and Beverage Co., Ltd (Guangzhou, China).
## Screening bioactive components and action targets of BSHS
We used the TCMSP database (http://www.tcmspw.com/tcmsp.php) to search for BSHS constituent medicines (Lysimachiae Herba, Lygodii Spora, Plantaginis Semen, Dipsaci Radix, and Licorice) and their chemical and pharmacological data. The remaining herbs, such as Clerodendranthus spicatus, Cibotii Rhizoma, and *Malva verticillata* seed were not retrieved from the database; however, their active compounds were retrieved by reviewing the literature. As parameters for screening the compounds collected, oral bioavailability (OB) and drug-like quality (DL) were selected. The OB represents the percentage of unchanged drug that reaches the systemic circulation after oral administration. DL indexes can be used to optimize pharmacokinetic and pharmaceutical properties, such as solubility and chemical stability [17]. Here, we set OB ≥ 30 and DL ≥ 0.18 as criteria to screen for biologically active compounds. TCMSP was further used to screen the targets of the active ingredients of BSHS, and Uniprot was used to correct and deduplicate the drug targets. ( https://www.uniprot.org/).
## The construction of the drug-active ingredient-target interaction network
The obtained active ingredients and cross-targets were sorted using Microsoft Excel worksheet. After the data were imported into Cytoscape 3.7.2 software, a “drug-active ingredient-target” network model was constructed, in which the nodes represent herbs, ingredients, and targets, while the edges represent the relationship role among the three nodes. We calculated the ‘degree’ value according to the number of associations between each node.
## Screening of potential targets for KS
We used terms such as kidney calculus, kidney stones, Nephrolithiasis, Renal calculus, and Renal stones as keywords related to screened potential targets from Genecards (https://www.genecards.org), Therapeutic Target Database (TTD, http://bidd.nus.edu.sg/group/ttd/ttd.asp), Online Mendelian Inheritance in Man (OMIM, https://www.genecards.org), and DisGeNET (https://www.disgenet.org/home/) databases. After eliminating repetitive targets, the potential targets that correlated with KS were obtained. In order to determine the intersection between BSHS and KS targets, we drew a Venn diagram.
## Construction of the protein-protein interaction network and screening of key targets
In order to clarify the functional interactions between the screened potential proteins, we constructed a protein-protein interaction (PPI) network using the STRING database (https://string-db.org/). The PPI network was inputted into Cytoscape 3.7.2 software using the CytoNCA software to analyze the topology of the intersection network. Further, we took the nodes with the ‘degree’ value greater than twice the median as the basis for screening the key targets and finally got the critical targets of BSHS for treating kidney stone disease.
## KS-related target gene ontology and KEGG pathway enrichment analysis for BSHS
The analysis of GO enrichment and KEGG pathways was conducted by DAVID Bioinformatics Resources 6.8 (http://david.ncifcrf.gov). For functional annotation clustering, terms with thresholds of Count ≥ 2 and Expression Analysis Systematic Explorer (EASE) scores ≤ 0.05 were selected.
## Preparation of BSHS
Eight raw herbs of BSHS were provided from the pharmacy department of the Wuqing Hospital of Traditional Chinese Medicine affiliated with Tianjin University of Traditional Chinese Medicine(Tianjin, China) (Supplementary Figure 1). Lysimachiae Herba, Lygodii Spora, Plantaginis Semen, Clerodendranthus Spicatus, Cibotii Rhizoma, Dipsaci Radix, *Malva verticillata* seed, and Licorice were mixed in proportions of 3:1.5:1.5:3:1.5:1.5:1.5:1(w/w) respectively and then soaked in 12 times the volume of distilled water (v/m) for 1 h, decocted twice, at 1.5 h per decoction. After concentrating the decoction to 1 g/mL, it was stored at -20°C until used.
## Preparation of test solution
Weigh 20 mg of BSHS extract in a 1.5 mL centrifuge tube, add 1 mL of pure water, vortex for 2 min, sonicate for 10 min, dilute 10 times with pure water, centrifuge at 1,4000 rpm for 20 min to extract the supernatant and leave for measurement.
## Chromatographic conditions
The chromatographic column was a Waters ACQUITY UPLC® BEH C18 column (1.8 μm, 2.1 × 100 mm); the mobile phase was $0.1\%$ formic acid in water (A) -acetonitrile (B). Gradient elution (0-2 min, $3\%$ B; 2-6 min, 3-$23\%$ B; 6-10 min, 23-$23.5\%$ B; 10-10.5 min, 23.5-$35\%$ B; 10.5-11 min, 35-$40\%$ B; 11-15 min, 40-$45\%$ B; 15-16 min, 45-$100\%$ B; 16-17 min, $100\%$ B; 17.01 min, $3\%$ B; 18 min; $3\%$ B); Flow rate: 0.4 mL/min. Column temperature: 40°C; The injection volume was 5.0 μL.
## Mass spectrometry conditions
The HESI (heated electrospray ionization probe) parameters were as follows: spray voltage, -3.0 kV/+3.5 kV; sheath gas, (N2) 35 L/h; auxiliary gas (N2), 10 L/h; purge gas (N2), 0 L/h; capillary temperature, 350 °C; auxiliary gas heater temperature, 350 °C. The Full MS/dd-MS2 scan method is used, with simultaneous detection in both positive and negative ion modes. The MS1 full Scan Range was m/z 100-1500 with a resolution of 70000; MS2 mass spectrometry scan was dynamic mass range with a resolution of 17500; automatic gain control (AGC) for MS1 and MS2 were set to 3 × 106 and 1 × 105 respectively; maximum injection time was defined as 100 ms and 50 ms for MS1 and MS2, respectively; collision energy (HCD) was performed at a normalised collision energy (NCE) of $\frac{10}{30}$/50 V; isolation width was set to 4.0 m/z; dynamic exclusion time was 10 s.
## Data processing
The raw mass spectrometry data files of BSHS extracts were imported into Compound Discover software for automatic identification according to the Natural Product database, followed by processing of the analytical results from Compound Discover. Xcalibur software was used to identify and characterize the chemical components of BSHS.
## Establishment and grouping of a rat model of calcium oxalate kidney stones
Male SD rats ($$n = 40$$) weighing 180-220 g were purchased from Beijing Viton Lever Laboratory Animal Technology Ltd. (Beijing, China). All animals were housed under standard laboratory conditions with free access to water and food. After a 7-day environmental adaptation period, the rats were randomly divided into four groups, i.e., the normal control group, model group, PSG group, and BSHS group, with 10 rats in each group. The rats in the normal control group were given free water and were intragastrically injected with saline (2 ml/d) after two weeks. The rats in the model group were given $0.75\%$ EG (v/v) + $0.75\%$ AC (w/v) free water for two weeks and were intragastrically injected with saline (2 ml/d) after two weeks. The rats in the PSG group were given $0.75\%$ EG (v/v) + $0.75\%$ AC (w/v) free water for two weeks and were intragastric injected with PSG (6.17 g/kg/d) after two weeks. The rats in the BSHS group were given $0.75\%$ EG (v/v) + $0.75\%$ AC (w/v) free water for two weeks and were intragastric injected with BSHS herbal concentrate (11.6 g/kg/d) after two weeks. A day before the execution, 24 h urine samples were collected from all the rats and stored at -80°C. On day 28, the animals were treated with anesthesia, and blood samples were taken from the abdominal aorta and centrifuged at 3500 rpm for 15 min at 4°C. The serum supernatant was aspirated and stored at -80°C. The left kidney was fixed in $4\%$ paraformaldehyde and embedded in paraffin for hematoxylin and eosin (H&E) staining, Von Kossa staining, and TUNEL staining. After snap-freezing in liquid nitrogen, the right kidney was stored at -80°C. All animal experiments were performed according to the requirements of the Experimental Animal Ethics Committee of Tianjin University of Traditional Chinese Medicine.
## Renal pathological examination and crystal deposition assay
The left kidney was fixed in $4\%$ paraformaldehyde, dehydrated in gradient alcohol and embedded in paraffin as described by Qian et al [18]. Longitudinal 4-µm paraffin sections were prepared for the H & E and Von Kossa staining. These sections were observed under the fluorescence microscope (Olympus IX2-UCB, Japan) to confirm the presence of crystals in the stained materials. The formed crystals were evaluated using professional image analysis software (ImageJ, U.S.A). Each section was photographed with 20 randomly selected fields of view under a 200× microscope. The calculations of the stone area were determined for each section using Image J software to obtain the sum of stone area under 20 fields of view and the percentage of stone area, i.e., percentage of stone area=stone arealongitudinal section area of kidney∗100.
## Urine volume and renal/body weight index in rats
On day 28, we fed rats in metabolic cages and collected 24 h of urine using $0.02\%$ sodium azide to prevent bacterial growth and recording of 24 h urine volume in rats. The kidneys were removed bilaterally, stripped of peritoneum and fat, and then weighed. Renal/body weight index=renal weightbody weight∗100.
## Renal biochemical examination
The right kidneys were placed in ice-cold phosphate-buffered saline, pH 7.4, and homogenized using a tissue homogenizer. Commercial kits were used strictly according to the manufacturer’s instructions to determine the levels of malondialdehyde (MDA), superoxide dismutase (SOD), oxalate, and calcium (Ca) in the tissue.
## Urine biochemical examination
On day 28, we fed rats in metabolic cages and collected 24 h of urine using $0.02\%$ sodium azide to prevent bacterial growth. The levels of urinary oxalate, Ca, phosphorus (P), and magnesium (Mg) were examined by utilizing the commercial kits on the microplate reader (Varioskan Flash, Thermo Scientific) following the instructions of the manufacturers.
## Serum biological parameters analysis
After 2 hours at room temperature, blood samples were centrifuged for 15 minutes at 3500 rpm (4°C). The supernatants were collected and stored at −80°C until use. The levels of serum Ca, P, Mg, creatinine (Cr), and urea nitrogen (BUN) were examined by utilizing the commercial kits in the microplate reader following the instructions of the manufacturers.
## Terminal deoxynucleotidyl transferase dUTP nick-end labeling staining
We used a TUNEL assay kit according to the instructions of the manufacturer to assay renal apoptosis in left kidney tissues embedded in paraffin and cut into 4mm thick sections. Cells positive for TUNEL were counted in 5 randomly selected fields (400x magnification) under a fluorescence microscope (Olympus IX2-UCB, Japan). The rate of apoptotic cells was analyzed using Image J (USA).
## ELISA assay
After 2 hours at room temperature, blood samples were centrifuged for 15 minutes at 3500 rpm (4°C). The supernatants were collected and stored at -80°C until used. The levels of E2 were measured with commercial ELISA kits following the protocols of the manufacturer.
## Quantitative real time polymerase chain reaction
Total RNA from frozen right renal tissue was isolated using RNA simple Total RNA Kit (Tiangen, Beijing, China) and then reverse-transcribed to cDNA with a Reverse Transcription Kit (Tiangen, Beijing, China). The qPCR was performed using Bio-Rad IQ5 (Bio-Rad, USA) and according to the manufacturer’s protocols for the setup procedure. The housekeeping gene GAPDH was used for normalization. The fold changes were calculated using the method of 2-ΔΔCt. All primer sequences used in this study have been shown in Table 1.
**Table 1**
| Genes | Primer sequence(5’-3’) |
| --- | --- |
| β-actin | Forward : CCTCTATGCCAACACAGTGC |
| β-actin | Reverse : CCTGCTTGCTGATCCACATC |
| Esr1 | Forward : GCACCATCGATAAGAACCGG |
| Esr1 | Reverse : TTCGGCCTTCCAAGTCATCT |
| Esr2 | Forward : AGGATGTACCACCGAATGCCAAGT |
| Esr2 | Reverse : TCCAAGTGGGCAAGGAGACAGAAA |
| Nrf2 | Forward : GCCTTCCTCTGCTGCCATTAGTC |
| Nrf2 | Reverse : TGCCTTCAGTGTGCTTCTGGTTG |
| Ho-1 | Forward : TATCGTGCTCGCATGAACACTCTG |
| Ho-1 | Reverse : GTTGAGCAGGAAGGCGGTCTTAG |
| Bcl2 | Forward : CTTCAGGGATGGGGTGAACT |
| Bcl2 | Reverse : ATCAAACAGAGGTCGCATGC |
| Bax | Forward : GACGCATCCACCAAGAAGCTGAG |
| Bax | Reverse : GCTGCCACACGGAAGAAGACC |
## Western blot analysis
RIPA buffer (Solarbio Co., Ltd. Beijing, China) was used to extract proteins from kidney tissues. Protein needs to be denatured by boiling it in a metal bath at 100°C for 10 minutes. The total protein concentration was determined using the BCA protein assay kit (Solarbio, Beijing, China). An equal amount of protein (20 µg) was separated using $10\%$ sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and then electrophoretically transferred onto PVDF (Millipore. Billerica, MA, USA). The membranes were blotted with $5\%$ fat-free milk in tris buffer saline with tween 20 (TBST) buffer for 2 h at room temperature and then incubated at 4°C overnight with primary antibodies: anti- GAPDH (1:1,0000), anti-ERα (1:1,000), anti- ERβ (1:1,000), anti- NRF2 (1:1,000), anti-HO-1 (1:1,000), anti-BCL2 (1:1,000), anti-BAX (1:1,0000), anti-Caspase-3 (1:1,000) and anti-Cleaved Caspase-3 (1:1,000). Afterward, the membranes were incubated with HRP-conjugated anti-rabbit/mouse IgG. The blots were imaged under an enhanced chemiluminescence (ECL) system. The target band molecular weights and net optical density were analyzed using the multifunctional imager (Jena UVP Chem studio, Germany).
## Statistical analysis
All data are presented as mean values ± SD, and graphs were created and analyzed using Prism Software (GraphPad Prism 7). The one-way analysis of variance (ANOVA) was used to evaluate the differences among the groups. It was deemed statistically significant when the $p \leq 0.05.$
## Results
In this study, we identified BSHS-related active compounds, critical therapeutic targets, and the molecular mechanism of action of BSHS in kidney stone disease treatment by network pharmacology, functional gene pathway analysis, network analysis, and other comprehensive methods. Finally, we predicted the potential molecular mechanisms of BSHS and validated in vivo experiments. A flowchart of this research is shown in Figure 1.
**Figure 1:** *The flowchart illustrates the mechanism of BSHS in KS, from target identification, network construction, and enrichment analysis to experimental validation.*
## Screening of BSHS bioactive ingredients and therapeutic targets for KS
We obtained 10 active ingredients of Lysimachiae Herba, 10 active compounds of Lygodii Spora, 9 active ingredients of Plantaginis Semen, 8 active ingredients of Dipsaci Radix, 92 active ingredients of Licorice, 6 active ingredients of Clerodendranthus spicatus, 8 active ingredients of Cibotii Rhizoma, and 1 active ingredient of *Malva verticillata* seed. After removing duplicate entries, a total of 126 active ingredients and 244 ingredient action targets were obtained. A total of 1970 targets related to KS treatment were obtained from four databases: Genecards, OMIM, DisGeNET, and TTD. Using the online Venn diagram editing website (http://jvenn.toulouse.inra.fr/app/example.html), 140 potential target genes were identified for KS treatment by BSHS by importing the potential targets for KS and the targets for BSHS active ingredients (Figure 2A).
**Figure 2:** *Network pharmacology analysis. (A) Target intersections between BSHS and KS. (B) The network of drug-compound-target included 8 kinds of herbs, 126 active components, and 244 target genes. Purple circle: drug, orange and green hexagon: active ingredients of BSHS; blue quadrilateral: targets. (C) A PPI network of predicted BSHS targets against KS. (D) A list of significant proteins from the PPI network was derived from (C). (E) A list of 23 key proteins of BSHS in KS treatment was derived from (D). (F) Based on the GO enrichment analysis, these are the top 10 indicators of BP, CC, and MF. (G) The top 20 signaling pathways were identified according to KEGG.*
## Construction of drug-compound-target networks
We used Cytoscape 3.7.2 software to construct the drug-compound-target network diagram. The purple circle nodes represented the 8 traditional Chinese medicines of BSHS, the hexagons nodes represented the compounds, the A1 and A2 nodes represented the common compounds of Lysimachiae Herba and Lygodii Spora, the A3 node represented the joint compound of Lysimachiae Herba and Licorice, the A4 node represented the joint compound of Lysimachiae Herba and Plantaginis Semen. The A5 and A6 nodes represented the common compounds of Licorice and Cibotii Rhizoma, the B1 node represented the joint compound of Lysimachiae Herba, Plantaginis Semen, and Licorice. The C1 node represented the joint compound of Lysimachiae Herba, Plantaginis Semen, Dipsaci Radix, and Licorice. The C2 node represented the joint compound of *Malva verticillata* seed, Dipsaci Radix, Lygodii Spora, and Cibotii Rhizoma. The D1 node represented the joint compound of Lysimachiae Herba, Lygodii Spora, Licorice, Clerodendranthus spicatus, and Cibotii Rhizoma, while the blue quadrilateral nodes represented the targets. The drug-compound-target network diagram included 365 nodes and 2713 edges (Figure 2B). The top five active compounds in BSHS, ranked according to the degree value, were quercetin, kaempferol, naringenin, β-sitosterol, and baicalein, which may play an essential role in treating kidney stones.
## PPI network analysis and screening of key targets
In order to obtain the key proteins of BSHS in the treatment of KS, we constructed a PPI network with 239 nodes and 4557 edges based on a string database (Figure 2C). Based on the ‘degree’ value of topological parameters calculated by CytoNCA, 23 pivotal proteins were filtered out, including AKT1, IL6, MAPK3, TP53, VEGFA, CASP3, JUN, TNF, PTGS2, EGF, MAPK8, EGFR, STAT3, MYC, MMP9, MAPK1, ESR1, CXCL8, IL1β, CCND1, CAT, FOS, CCL2, which were strongly linked to KS (Figures 2D, E).
## GO enrichment analysis and KEGG pathway analysis of key targets
We used the DAVID database to perform GO enrichment analysis of the 23 key targets for the identification of the relevant biological functions of BSHS against KS. The analysis uncovered 251 biological pathways, 22 cell localizations, and 35 molecular functions. As shown in Figure 2F, the top 10 terms in the biological process (BP), cellular component (CC), and molecular function (MF) categories that are significantly enriched are demonstrated. BP was found to be primarily associated with the negative regulation of apoptotic process, positive regulation of protein phosphorylation, lipopolysaccharide-mediated signaling pathway, and response to estradiol. The CC mainly included the extracellular space, protein complex, cytosol, nucleoplasm, etc. The MF mainly included enzyme binding, transcription factor binding, identical protein binding, etc.
In order to determine the potential pathway for BSHS in the treatment of KS, we performed a KEGG pathway enrichment analysis and found 102 signal pathways related to BSHS. A total of 20 pathways related to KS were screened, mainly including Tumour Necrosis Factor (TNF), estrogen, Mitogen-Activated Protein Kinase (MAPK), and Toll-like receptor signaling pathways. The enrichment pathway was visualized according to the size of the p-value (Figure 2G). In the pathways with the highest enrichment levels, estrogen and apoptosis signaling pathways were most closely related to KS. In addition, CAT is a key target of BSHS anti-kidney stones according to 2.4.3, which protects cells from oxidative stress by scavenging hydrogen peroxide produced by cellular metabolism [19]. Multiple studies have also shown the harmful effects of oxidative stress on kidney stones [20, 21]. Therefore, we predict that oxidative stress is also a key signaling pathway in the treatment of kidney stones by BSHS.
## Identification and characterization of BSHS components
Total ion flow maps for the BSHS positive and negative ion scan modes were obtained from the data acquisition (Supplementary Figures 2A, B). The acquired data was processed with Xcailbur software. Matches were made in the HMDB and PubChem databases according to retention times, the mass information of the quasi-molecular and fragment ions, while keeping the quasi-molecular ions within ±5 ppm. A total of 52 compounds were eventually identified (Supplementary Table 1).
## BSHS inhibits the formation of calcium oxalate crystals
We investigated the impact of BSHS on renal injury and calcium oxalate crystal deposition after EG+AC induction in vivo. H&E and Von Kossa staining of kidney sections revealed renal tubule severe dilation, tubule destruction, and epithelial cell desquamation induced by EG+AC. A large amount of calcium oxalate crystal deposition was noticed after EG+AC induction, whereas coadministration of BSHS could protect the EG+AC-injured kidney tissue from inflammatory damage and calcium oxalate crystal deposition (Figures 3A, B).
**Figure 3:** *The therapeutic effect of BSHS on KS. (A) The H&E and Von Kossa staining of kidney sections from each group revealed tissue injury and CaOx crystal deposition. (B) The Image J (USA) software calculated the percentage of area positively stained for crystal deposition in each kidney section based on 20 random views at 200× magnification. (C) Oxalate content in the rat urine. (D) Ca content in the rat urine. (E) Oxalate content in the rat kidneys. (F) Ca content in the rat kidneys. (G) Cr content in the rat serum. (H) Blood urea nitrogen (BUN) content in the rat serum. (I) Ca content in the rat serum. (J) P content in the rat serum. (K) Mg content in the rat serum. All values were expressed as mean ± SD.
*p<0.05 vs. the normal control group,
**p < 0.01 vs. the normal control group,
***p < 0.001 vs. the normal control group,
#p < 0.05 vs. the model group,
##p < 0.01 vs. the model group,
$$p < 0.01 vs. the PSG group, ns for p>0.05 vs. the model group.*
## BSHS increases 24 h urine volume and improves renal/body weight index in rats
On day 28, we recorded 24h urine volume and weighed the body weight and renal weight of rats. The results showed that there were significantly decreased 24 h urine volume and increased renal/body weight index in the model group rats compared to the normal control group. Simultaneous treatment with BSHS resulted in a remarkable decrease in renal/body weight index, as well as an increase in 24 h urine volume in the rat (Supplementary Figures 3A, B).
## BSHS regulated urine and renal biological parameters in rats
In this study, we investigated whether BSHS inhibited the formation of oxalate, Ca, and P in an animal model, and increased the level of Mg. As expected, there was a significant increase ($p \leq 0.05$) of oxalate and calcium contents in both the kidneys and urine of rats in the model group compared to those in the normal control group. Simultaneous treatment with BSHS resulted in a significant decrease ($p \leq 0.05$) in oxalate and calcium levels in the kidney of rats (Figures 3C–F). BSHS also decreased the level of P and increased the level of Mg in the urine of rats compared to the model group (Supplementary Figures 3C, D).
## BSHS regulated serum biological parameters in rats
Meanwhile, to examine the effect of BSHS on the renal function and serum biological parameters of rats, we further examined the levels of Cr, BUN, Ca, P, and Mg in the serum of rats. The results showed that there were significantly decreased Mg content and increased Ca, P, Cr, and BUN contents in the serum of model group rats compared to the normal control group. Simultaneous treatment with BSHS resulted in a remarkable decrease in Ca, P, Mg, Cr, and BUN levels, as well as an increase in Mg levels in the rat serum. Worthy of note, there was no significant trend of lowering Cr and BUN in the PSG group compared to the model group, suggesting a better effect of BSHS in improving renal function (Figures 3G–K).
## BSHS inhibited apoptosis induced by EG+AC in rat
Network pharmacological analysis indicated that apoptosis might be involved in BSHS treatment of kidney stones. Several studies have also demonstrated that apoptosis is crucial in kidney stone formation (22–24). Therefore, our study examined the apoptotic effects of BSHS on KS in animal models. In the model group, TUNEL-positive cells were significantly higher than in the normal control group, according to the TUNEL staining results. Contrary to what was observed in the model group, BSHS groups showed fewer apoptotic cells (Figures 4A, B). The qRT-PCR results indicated that Bax levels were significantly increased and Bcl2 was decreased in the model group, while treatment with BSHS reversed the increase of Bax levels and restored Bcl2 expression (Figure 4C). The Western bolting results indicated that BAX and Cleaved Caspase-3/Caspase-3 levels were significantly increased and BCL2 was decreased in the model group, while treatment with BSHS reversed the increase of BAX and Cleaved Caspase-3/Caspase-3 levels and restored BCL2 expression (Figures 4D, E) (Supplementary Figures 4A, B).
**Figure 4:** *Effect of BSHS on EG+AC-induced apoptosis in rat kidney tissue. (A) TUNEL staining was used to assess renal apoptosis. (B) Images J was used to count the percentages of TUNEL-positive cells (green) to total cells (blue). (C) The expression of apoptosis-related genes was evaluated by qRT-PCR. (D) The expression of apoptosis-related proteins was evaluated by Western bolting. (E) A graph showing the semi-quantitative analysis of BAX and BCL2. Data are presented as the mean ± SD and density normalized to GAPDH. *
p<0.05 vs. the normal control group, **
p < 0.01 vs. the normal control group, ***
p < 0.001 vs. the normal control group, #
p < 0.05 vs. the model group, ##
p < 0.01 vs. the model group, ###
p < 0.001 vs. the model group.*
## BSHS improves the imbalance of estrogen levels induced by EG+AC in rat
Network pharmacological analysis suggested estrogen signaling pathways may be involved in BSHS treating kidney stones. Interestingly, lower estrogen levels have also been shown to be strongly associated with the formation of kidney stones (25–27). As part of our study, we examined the effect of BSHS on estrogen and estrogen receptors in this animal model. E2 serum levels were found to be lower in the model group compared with the control group in the ELISA experiment. In contrast, it more dramatically increased in the BSHS groups than in the model group (Figure 5A). The qRT-PCR results illustrated that Esr1 and Esr2 mRNA levels were remarkably decreased in the model group, while treatment with BSHS reversed the decrease in Esr1 and Esr2 levels (Figure 5B). The Western blotting results showed the same trend (Figures 5C, D).
**Figure 5:** *The effect of BSHS on the imbalance of estrogen levels induced by EG+AC in the rat. (A) ELISA evaluated the expression of estrogen levels. (B) The expression of estrogen receptor-related genes was evaluated by qRT-PCR. (C) The expression of estrogen receptor-related proteins was examined by Western blotting. (D) A semi-quantitative analysis of ESR1 and ESR2 is shown. Data are presented as the mean ± SD and density normalized to GAPDH. *
p<0.05 vs. the normal control group, **
p < 0.01 vs. the normal control group, ***
p < 0.001 vs. the normal control group, #
p < 0.05 vs. the model group, ##
p < 0.01 vs. the model group, ###
p< 0.001 vs. the model group, ns for p > 0.05 vs. the model group.*
## BSHS alleviated EG+AC-induced oxidative stress in rats
According to growing evidence, oxidative stress may play an essential role in hyperoxaluria-induced kidney injury, resulting in renal CaOx crystallization (28–30). Herein, we examined the potential antioxidative properties of BSHS in this animal model. As expected, compared to the normal control group, rats in the model group significantly increased MDA content and decreased SOD activity in the kidneys. The simultaneous treatment of rats with BSHS reduced MDA levels in the kidneys and increased SOD activity (Figures 6A, B). The qRT-PCR results illustrated that Nrf2 and Ho-1 mRNA levels were remarkably decreased in the model group, while treatment with BSHS reversed the decrease in Nrf2 and Ho-1 levels (Figure 6C). The Western blotting results showed a similar trend (Figures 6D, E). Interestingly, the PSG group did not show significant antioxidant effects, while the BSHS group had great antioxidant capacity.
**Figure 6:** *A comparison of oxidative stress levels in kidneys among different groups. (A) SOD activity in the kidneys of rats. (B) MDA content in the kidneys of rats. (C) qRT-PCR was used to evaluate the expression of genes related to oxidative stress. (D) The expression of oxidative stress-related proteins was evaluated by Western blotting. (E) A semi-quantitative analysis of NRF2 and HO-1 is shown. Data are presented as the mean ± SD and density normalized to GAPDH. **
p < 0.01 vs. the normal control group, ***
p < 0.001 vs. the normal control group, #
p < 0.05 vs. the model group, ##
p < 0.01 vs. the model group, ns for p > 0.05 vs. the model group.*
## Discussion
Kidney stones are a common and frequently-occurring disease of the urinary system, and their incidence increases annually [31]. According to the most recent epidemiological study conducted in China, kidney stones are prevalent in approximately $5.8\%$ of the population [32]. It is estimated that $12\%$ of men and $6\%$ of women in the world population will have kidney stones at least once in their lifetime, with recurrence rates of 70–$80\%$ for men and 47–$60\%$ for women [33]. Among them, calcium oxalate stones are the most common kidney stones [34, 35], accounting for over $80\%$ of them [36]. Although people have an in-depth understanding of crystallization and stone formation, there is currently a lack of effective treatment methods and drugs due to the slow progress in determining the pathophysiology of stone formation. Therefore, kidney stone disease must be given sufficient attention. The expansion of the treatment model of kidney stone disease based on TCM can provide a reliable solution for the pathogenesis of kidney stone disease that is difficult to cure and easy to relapse. Although under the guidance of the holistic view of TCM, Chinese herbal compound has an excellent therapeutic effect on diseases. Due to their complex components, multi-target, and multi-channel treatment characteristics, it is not easily accessible for an in-depth study of its internal mechanism. In recent years, network pharmacology has become a popular technique for analyzing the mechanism of action of complex TCM prescriptions [37]. The combination of network pharmacology and experimental verification was used in this study in order to clarify the pharmacological mechanism of BSHS against kidney stones.
It is well known that the formation of kidney stones is a complex process involving urinary supersaturation, nucleation, growth, aggregation, and retention of urinary stone components within the renal tubular cells [38]. Multiple studies have shown that kidney stone formation could be attributed to higher supersaturation of urine because of low urine volume and increased secretion of calcium, phosphates, oxalates, uric acid, and cysteine in urine (39–41). The elevated urinary excretion of calcium (hypercalciuria) and oxalate (hyperoxaluria) are the most common risk factor for CaOx kidney stones [42, 43]. However, some scholars supported that oxalate in the urine combines with free calcium to form insoluble CaOx, which induces kidney stones and can lead to a decrease in urinary calcium [44, 45]. Interestingly, we discovered that BSHS can decrease urine oxalates and calcium excretion which may be related to the fact that BSHS increase urinary magnesium levels. Studies have shown that magnesium can compete with calcium to bind oxalate and form insoluble solutes that are excreted in the urine [46]. Low urinary oxalate concentrations lead to a reduction in urinary calcium levels, therefore BSHS can treat kidney stones by reducing urinary oxalate, urinary calcium and increasing urinary magnesium levels.
TCM compounds that lacked proper pharmacokinetic properties would not reach their target organs to exert their biological effects [47]. It has been demonstrated that compounds with OB ≥$30\%$ and DL index ≥0.18 may be absorbed and distributed in the human body and are thus considered pharmacokinetically active [48, 49]. Compounds with high-degree may explain the significant therapeutic effects of BSHS on kidney stones in the compound-key targets network. According to this study, quercetin was the most significant compound, followed by kaempferol, naringenin, β-sitosterol, and baicalein. It is reported that quercetin, a natural flavonoid, has efficient antioxidant properties and can be used to inhibit oxidative damage in renal tubular cells and tissues [50]. In addition, quercetin can inhibit the formation of urinary tract stones induced by oxalate [51]. Kaempferol is one of the most common glycoside forms of aglycon flavonoids, which can increase the level of coenzyme Q in kidney cells to play an antioxidant role [52]. As a naturally occurring flavanone, naringenin inhibits oxidative stress in the kidneys and improves kidney function [53]. β-*Sitosterol is* a phytosterol reported in ancient medicinal history for treating nephritis and prostatitis [54]. β-sitosterol has been reported to inhibit nephrotoxicity and anti-kidney oxidation properties [55, 56]. Baicalein is a member of the flavonoid family, and modern pharmacology proves that baicalein can inhibit inflammation by activating the Nrf2 signaling pathway, thereby alleviating lupus nephritis [57]. Oxidative stress-induced apoptosis of renal tubular epithelial cells is a risk factor for stone formation [20]. All these works demonstrate that BSHS has excellent anti-kidney oxidation and renal protection.
It has been revealed that BSHS acts on multiple targets using multiple signaling pathways when we integrate the topological network parameters with all the network analyses. We finally identified estrogen, apoptosis, and oxidative stress as crucial mechanisms for BSHS treatment of kidney stones based on network pharmacology analysis. Multiple clinical studies have suggested that estrogen has a protective effect during the formation of kidney stones [58, 59]. However, the Women’s Health Initiative Study and the Nurses’ Health Study found no positive correlation between hormone replacement therapy and the prevention of kidney stones [59, 60]. These results have caused scholars to question the relationship between estrogen and kidney stones. For this result, some researchers suggested that the long-term estrogen decline caused by menopause may aggravate the deterioration of normal physiological estrogen receptor function in the kidney [60]. Therefore, the poor effect of hormone replacement therapy on renal calculi may be due to the reduced protein expression of estrogen receptors or its cofactors in these women [60]. We validated the effects of BSHS on E2 and estrogen receptors in vivo. The results showed that BSHS could not only increase the level of E2 but also increase the levels of ESR1 and ESR2. There are growing numbers of studies demonstrating that the adhesion or endocytosis of renal tubular epithelial cells to crystals plays an essential role in forming stones [61, 62]. Moreover, crystal adhesion can be enhanced by injured renal tubular epithelial cells, which can promote kidney stones [34]. Interestingly, the damage to renal tubular epithelial cells is closely related to oxidative stress [63]. As an essential antioxidant pathway for endogenous anti-oxidative stress in cells [64, 65], the NRF2/HO-1 signaling pathway is vital in improving oxidative stress in kidney diseases (66–68). Many studies have confirmed that the inhibitory effect of the estrogen signaling pathway on oxidative stress is closely related to the activation of the NRF2/HO-1 antioxidant pathway [69, 70]. Interestingly, our in vivo studies showed that BSHS could not only increase the expression of NRF2 and HO-1 proteins and genes but also increase SOD activity and decrease MDA levels in the rat kidney. It is suggested that BSHS has a good anti-oxidative stress effect on the kidney. It has been reported that renal tubular epithelial cell apoptosis is an essential factor that causes crystals to adhere to renal tubular epithelial cells [22]. Some scholars suggested that oxidative stress is a risk factor for apoptosis [71]. BCL2/BAX signaling pathway is a pivotal way to regulate cell apoptosis. Our studies demonstrated that BSHS increases the expression of BCL2 and reduces the expression of BAX, thereby reducing the level of apoptosis of renal tubular epithelial cells. Therefore, the therapeutic effect of BSHS on the calcium oxalate stone model in rats may be related to the increase of estrogen receptor levels and the inhibition of apoptosis.
Our findings suggest that BSHS may inhibit kidney stone formation mainly by regulating estrogen and estrogen receptor levels, inhibiting oxidative stress processes, reversing apoptosis, and decreasing CaOx crystals deposition through E2/ESR1/ESR2, NRF2/HO-1, and BCL2/BAX signaling pathways. Overall, BSHS ameliorated KS progression through a multi-ingredient, multi-target, and multi-pathway mode, which is different from chemical drugs that work on a distinct and single target. The understanding of complex interactions between disease and chemical ingredients in TCM could be well accomplished by identifying network targets and signaling pathways. It is important to note, however, that this study has some limitations. First, the results may have been slightly skewed since we only validated part of the core pathways and targets of BSHS. Therefore, further validation of other relevant targets and signaling pathways predicted by network pharmacology would be required in future experiments. Secondly, our study did not demonstrate an association between estrogen, oxidative stress, and apoptotic signaling pathways. In a follow-up experiment, we will examine their connection through in-vitro experiments.
## Conclusions
In summary, network pharmacology analysis coupled with experimental validation was performed to decipher the molecular mechanisms of BSHS in the treatment of KS. The network pharmacology analysis revealed that BSHS exerted anti-KS effects via multi-ingredients, multi-targets, and multi-pathways. The experimental results verified that BSHS improved CaOx crystal deposition in KS by modulating the E2/ESR1/ESR2, NRF2/HO-1, and BCL2/BAX signaling pathways. This study could provide an optimized method to elucidate the pharmacological mechanisms of BSHS and supply a novel candidate for treating KS.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The animal study was reviewed and approved by the experimental animal ethics committee of the Tianjin University of Traditional Chinese Medicine.
## Author contributions
YB and YW conceived this project. HL and MC designed the study, wrote the manuscript, and performed the experiments. YJ performed the network pharmacology and data analysis. BJ, MD, and LH edited the manuscript. LW performed the UHPLC-Q/Orbitrap MS experiments. JA, JL, and TZ revision of the manuscript. BC provides drug prescriptions. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1031895/full#supplementary-material
## References
1. Sorokin I, Mamoulakis C, Miyazawa K, Rodgers A, Talati J, Lotan Y. **Epidemiology of stone disease across the world**. *World J Urol* (2017) **35**. DOI: 10.1007/s00345-017-2008-6
2. Scales CD, Smith AC, Hanley JM, Saigal CS. **Prevalence of kidney stones in the united states**. *Eur Urol* (2012) **62**. DOI: 10.1016/j.eururo.2012.03.052
3. Mollerup CL, Vestergaard P, Frøkjaer VG, Mosekilde L, Christiansen P, Blichert-Toft M. **Risk of renal stone events in primary hyperparathyroidism before and after parathyroid surgery: Controlled retrospective follow up study**. *Br Med J* (2002) **325** 807. DOI: 10.1136/bmj.325.7368.807
4. Taylor EN, Stampfer MJ, Curhan GC. **Obesity, weight gain, and the risk of kidney stones**. *JAMA* (2005) **293**. DOI: 10.1001/jama.293.4.455
5. Aune D, Mahamat-Saleh Y, Norat T, Riboli E. **Body fatness, diabetes, physical activity and risk of kidney stones: a systematic review and meta-analysis of cohort studies**. *Eur J Epidemiol* (2018) **33**. DOI: 10.1007/s10654-018-0426-4
6. Taylor EN, Stampfer MJ, Curhan GC. **Diabetes mellitus and the risk of nephrolithiasis**. *Kidney Int* (2005) **68**. DOI: 10.1111/j.1523-1755.2005.00516.x
7. Strazzullo P, Barba G, Vuotto P, Farinaro E, Siani A, Nunziata V. **Past history of nephrolithiasis and incidence of hypertension in men: a reappraisal based on the results of the Olivetti prospective heart study**. *Nephrol dialysis Transplant* (2001) **16**. DOI: 10.1093/ndt/16.11.2232
8. Obligado SH, Goldfarb DS. **The association of nephrolithiasis with hypertension and obesity: A review**. *Am J Hypertens* (2008) **21**. DOI: 10.1038/ajh.2007.62
9. Rule AD, Krambeck AE and L. **JC: Chronic kidney disease in kidney stone formers**. *Clin J Am Soc Nephrol* (2011) **6**. DOI: 10.2215/CJN.10651110
10. Shoag J, Halpern J, Goldfarb DS, Eisner BH. **Risk of chronic and end stage kidney disease in patients with nephrolithiasis**. *J Urol* (2014) **192**. DOI: 10.1016/j.juro.2014.05.117
11. El-Zoghby ZM, Lieske JC, Foley RN, Bergstralh EJ, Li X, Rd MLJ. **AD: Urolithiasis and the risk of ESRD**. *Clin J Am Soc Nephro* (2012) **77**. DOI: 10.2215/CJN.03210312
12. Singh P, Harris PC, Sas DJ, Lieske JC. **The genetics of kidney stone disease and nephrocalcinosis**. *Nat Rev Nephrol* (2022) **18**. DOI: 10.1038/s41581-021-00513-4
13. Aggarwal KP, Narula S, Kakkar M, Tandon C. **Nephrolithiasis: Molecular mechanism of renal stone formation and the critical role played by modulators**. *BioMed Res Int* (2013) **2013** 1-21. DOI: 10.1155/2013/292953
14. Khan SR, Pearle MS, Robertson WG, Gambaro G, Canales BK, Doizi S. **Kidney stones**. *Nat Rev Dis Primers* (2016) **2** 16008. DOI: 10.1038/nrdp.2016.8
15. Du H, Lei Y, Gao AN, Sun XM, Song R, Yu XD. **Traditional Chinese medicine combined with conventional therapy for female kidney stone: A protocol for systematic review**. *Med (Baltimore)* (2020) **99**. DOI: 10.1097/MD.0000000000019611
16. Luo T, Lu Y, Yan S, Xiao X, Rong X, Guo J. **Network pharmacology in research of Chinese medicine formula: Methodology, application and prospective**. *Chin J Integr Med* (2020) **26** 72-80. DOI: 10.1007/s11655-019-3064-0
17. Zhou S, Ai Z, Li W, You P, Wu C, Li L. **Deciphering the pharmacological mechanisms of taohe-chengqi decoction extract against renal fibrosis through integrating network pharmacology and experimental validation**. *Front Pharmacol* (2020) **11**. DOI: 10.3389/fphar.2020.00425
18. Liu Q, Cai BY, Zhu LX, Xin X, Wang X, An ZM. **Liraglutide modulates gut microbiome and attenuates nonalcoholic fatty liver in db/db mice**. *Life Sci* (2020) **15** 118457. DOI: 10.1016/j.lfs.2020.118457
19. Yang F, Pei R, Zhang Z, Liao J, Yu W, Qiao N. **Copper induces oxidative stress and apoptosis through mitochondria-mediated pathway in chicken hepatocytes**. *Toxicol Vitro* (2019) **54**. DOI: 10.1016/j.tiv.2018.10.017
20. Xi J, Jing J, Zhang Y, Liang C, Hao Z, Zhang L. **SIRT3 inhibited the formation of calcium oxalate-induced kidney stones through regulating NRF2/HO-1 signaling pathway**. *J Cell Biochem* (2018) **120**. DOI: 10.1002/jcb.28109
21. Qin B, Wang Q, Lu Y, Li C, Hu H, Zhang J. **Losartan ameliorates calcium oxalate-induced elevation of stone-related proteins in renal tubular cells by inhibiting NADPH oxidase and oxidative stress**. *Oxid Med Cell Longev* (2018) **2018** 1271864. DOI: 10.1155/2018/1271864
22. Ye QL, Wang DM, Wang X, Zhang ZQ, Tian QX, Feng SY. **Sirt1 inhibits kidney stones formation by attenuating calcium oxalate-induced cell injury**. *Chem Biol Interact* (2021) **347** 109605. DOI: 10.1016/j.cbi.2021.109605
23. Sun Y, Kang J, Tao Z, Wang X, Liu Q, Li D. **Effect of endoplasmic reticulum stress-mediated excessive autophagy on apoptosis and formation of kidney stones**. *Life Sci* (2020) **244** 117232. DOI: 10.1016/j.lfs.2019.117232
24. Yu L, Gan X, Bai Y, An R. **CREB1 protects against the renal injury in a rat model of kidney stone disease and calcium oxalate monohydrate crystals-induced injury in NRK-52E cells**. *Toxicol Appl Pharmacol* (2021) **413** 115394. DOI: 10.1016/j.taap.2021.115394
25. Peerapen P, Thongboonkerd V. **Protective cellular mechanism of estrogen against kidney stone formation: A proteomics approach and functional validation**. *Proteomics* (2019) **19**. DOI: 10.1002/pmic.201900095
26. Peerapen P, Boonmark W, Thongboonkerd V. **Trigonelline prevents kidney stone formation processes by inhibiting calcium oxalate crystallization, growth and crystal-cell adhesion, and downregulating crystal receptors**. *Biomed Pharmacother* (2022) **149** 112876. DOI: 10.1016/j.biopha.2022.112876
27. Heller HJ, Sakhaee K, Moe OW, Pak CY. **Etiological role of estrogen status in renal stone formation**. *J Urol* (2002) **168**. DOI: 10.1097/01.ju.0000033907.21910.be
28. Li Z, Chang L, Ren X, Hu Y, Chen Z. **Modulation of rat kidney stone crystallization and the relative oxidative stress pathway by green tea polyphenol**. *ACS Omega* (2021) **6**. DOI: 10.1021/acsomega.0c05903
29. Park JH, Jo YI, Lee JH. **Renal effects of uric acid: hyperuricemia and hypouricemia**. *Korean J Intern Med* (2020) **35**. DOI: 10.3904/kjim.2020.410
30. Khan SR, Canales BK, Dominguez-Gutierrez PR. **Randall's plaque and calcium oxalate stone formation: role for immunity and inflammation**. *Nat Rev Nephrol* (2021) **17**. DOI: 10.1038/s41581-020-00392-1
31. Li DF, Gao YL, Liu HC, Huang XC, Zhu RF, Zhu CT. **Use of thiazide diuretics for the prevention of recurrent kidney calculi: a systematic review and meta-analysis**. *J Transl Med* (2020) **18** 106. DOI: 10.1186/s12967-020-02270-7
32. Zhang D, Li S, Zhang Z, Li N, Yuan X, Jia Z. **Urinary stone composition analysis and clinical characterization of 1520 patients in central China**. *Sci Rep* (2021) **11** 6467. DOI: 10.1038/s41598-021-85723-3
33. Queiroz MF, Melo K, Sabry DA, Sassaki GL, Rocha H, Costa LS. **Gallic Acid-chitosan conjugate inhibits the formation of calcium oxalate crystals**. *Molecules* (2019) **24** 2074. DOI: 10.3390/molecules24112074
34. Tsujihata M. **Mechanism of calcium oxalate renal stone formation and renal tubular cell injury**. *Int J Urol* (2008) **15**. DOI: 10.1111/j.1442-2042.2007.01953.x
35. Ivanovski O, Drüeke TB. **A new era in the treatment of calcium oxalate stones**. *Kidney Int* (2013) **83** 998-1000. DOI: 10.1038/ki.2013.41
36. Huang Y, Zhang YH, Chi ZP, Huang R, Huang H, Liu G. **The handling of oxalate in the body and the origin of oxalate in calcium oxalate stones**. *Urol. Int* (2020) **104**. DOI: 10.1159/000504417
37. Wang Y, Shi Y, Zou J, Zhang X, Liang Y, Tai J. **Network pharmacology exploration reveals a common mechanism in the treatment of cardio-cerebrovascular disease with salvia miltiorrhiza burge. and carthamus tinctorius l**. *BMC Complement Med Ther* (2020) **20**. DOI: 10.1186/s12906-020-03026-y
38. Chen JY, Sun XY, Ouyang JM. **Modulation of calcium oxalate crystal growth and protection from oxidatively damaged renal epithelial cells of corn silk polysaccharides with different molecular weights**. *Oxid Med Cell Longev* (2020) **2020**. DOI: 10.1155/2020/6982948
39. Flannigan R, Choy WH, Chew B, Lange D. **Renal struvite stones–pathogenesis, microbiology, and management strategies**. *Nat Rev Urol* (2014) **11**. DOI: 10.1038/nrurol.2014.99
40. Mehde AA, Mehdi WA, Yusof F, Raus RA, Abidin Z, Ghazali H. **Determination of alpha-2-MRAP gene polymorphisms in nephrolithiasis patients**. *Int J Biol Macromol* (2017) **105**. DOI: 10.1016/j.ijbiomac.2017.07.167
41. Heneghan JF, Alper SL. **This, too, shall pass–like a kidney stone: a possible path to prophylaxis of nephrolithiasis? focus on "Cholinergic signaling inhibits oxalate transport by human intestinal T84 cells"**. *Am J Physiol Cell Physiol* (2012) **302**. DOI: 10.1152/ajpcell.00389.2011
42. Perry GM, Nehrke KW, Bushinsky DA, Reid R, Lewandowski KL, Hueber P. **Sex modifies genetic effects on residual variance in urinary calcium excretion in rat (Rattus norvegicus)**. *Genetics* (2012) **191**. DOI: 10.1534/genetics.112.138909
43. Zhu W, Liu Y, Lan Y, Li X, Luo L, Duan X. **Dietary vinegar prevents kidney stone recurrence**. *EBioMedicine* (2019) **45**. DOI: 10.1016/j.ebiom.2019.06.004
44. Li X, Wang W, Su Y, Yue Z, Bao J. **Inhibitory effect of an aqueous extract of radix paeoniae alba on calcium oxalate nephrolithiasis in a rat model**. *Ren Fail* (2017) **39**. DOI: 10.1080/0886022X.2016.1254658
45. Karabulut D, Kaymak E, Yalçin B, Ulger H, Keti DB. **A different perspective on the filtration barrier after kidney stone formation: An immunohistochemical and biochemical study**. *Urolithiasis* (2021) **49**. DOI: 10.1007/s00240-020-01227-1
46. Wang Z, Zhang Y, Zhang J, Deng Q, Liang H. **Recent advances on the mechanisms of kidney stone formation (Review)**. *Int J Mol Med* (2021) **48**. DOI: 10.3892/ijmm.2021.4982
47. Mao Y, Peng L, Kang A, Xie T, Xu J, Shen C. **Influence of jiegeng on pharmacokinetic properties of flavonoids and saponins in gancao**. *Molecules* (2017) **22** 1587. DOI: 10.3390/molecules22101587
48. Xu X, Zhang W, Huang C, Li Y, Yu H, Wang Y. **A novel chemometric method for the prediction of human oral bioavailability**. *Int J Mol Sci* (2012) **13**. DOI: 10.3390/ijms13066964
49. Tao W, Xu X, Wang X, Li B, Wang Y, Li Y. **Network pharmacology-based prediction of the active ingredients and potential targets of Chinese herbal radix curcumae formula for application to cardiovascular disease**. *J Ethnopharmacol* (2013) **145** 1-10. DOI: 10.1016/j.jep.2012.09.051
50. Al-Rasheed NM, Faddah LM, Mohamed AM, Abdel BN, Al-Rasheed NM, Mohammad RA. **Potential impact of quercetin and idebenone against immuno- inflammatory and oxidative renal damage induced in rats by titanium dioxide nanoparticles toxicity**. *J Oleo Sci* (2013) **62**. DOI: 10.5650/jos.62.961
51. Park HK, Jeong BC, Sung M, Park M, Choi EY, Kim BS. **Reduction of oxidative stress in cultured renal tubular cells and preventive effects on renal stone formation by the bioflavonoid quercetin**. *J Urol* (2008) **179**. DOI: 10.1016/j.juro.2007.11.039
52. Fernández-Del-Río L, Nag A, Gutiérrez CE, Ariza J, Awad AM, Joseph AI. **Kaempferol increases levels of coenzyme q in kidney cells and serves as a biosynthetic ring precursor**. *Free Radic Biol Med* (2017) **110**. DOI: 10.1016/j.freeradbiomed.2017.06.006
53. Khan TH, Ganaie MA, Alharthy KM, Madkhali H, Jan BL, Sheikh IA. **Naringenin prevents doxorubicin-induced toxicity in kidney tissues by regulating the oxidative and inflammatory insult in wistar rats**. *Arch Physiol Biochem* (2020) **126**. DOI: 10.1080/13813455.2018.1529799
54. Choi S, Kim KW, Choi JS, Han ST, Park YI, Lee SK. **Angiogenic activity of beta-sitosterol in the ischaemia/reperfusion-damaged brain of Mongolian gerbil**. *Planta Med* (2002) **68**. DOI: 10.1055/s-2002-26750
55. Sharmila R, Sindhu G, Arockianathan PM. **Nephroprotective effect of beta-sitosterol on n-diethylnitrosamine initiated and ferric nitrilotriacetate promoted acute nephrotoxicity in wistar rats**. *J Basic Clin Physiol Pharmacol* (2016) **27**. DOI: 10.1515/jbcpp-2015-0085
56. Koc K, Geyikoglu F, Cakmak O, Koca A, Kutlu Z, Aysin F. **The targets of β-sitosterol as a novel therapeutic against cardio-renal complications in acute renal ischemia/reperfusion damage**. *Naunyn Schmiedebergs Arch Pharmacol* (2021) **394**. DOI: 10.1007/s00210-020-01984-1
57. Li D, Shi G, Wang J, Zhang D, Pan Y, Dou H. **Baicalein ameliorates pristane-induced lupus nephritis**. *Arthritis Res Ther* (2019) **21** 105. DOI: 10.1186/s13075-019-1876-0
58. Mattix KH, Grodstein F, Stampfer MJ, Curhan GC. **Menopause and postmenopausal hormone use and risk of incident kidney stones**. *J Am Soc Nephrol* (2003) **14**. DOI: 10.1097/01.asn.0000060682.25472.c3
59. Maalouf NM, Sato AH, Welch BJ, Howard BV, Cochrane BB, Sakhaee K. **Postmenopausal hormone use and the risk of nephrolithiasis: results from the women's health initiative hormone therapy trials**. *Arch Intern Med* (2010) **170**. DOI: 10.1001/archinternmed.2010.342
60. Turner RJ, Kerber IJ. **Renal stones, timing hypothesis, and eu-estrogenemia**. *Menopause* (2012) **19**. DOI: 10.1097/gme.0b013e318221be9b
61. Gill WB, Jones KW, Ruggiero KJ. **Protective effects of heparin and other sulfated glycosaminoglycans on crystal adhesion to injured urothelium**. *J Urol* (1982) **127**. DOI: 10.1016/s0022-5347(17)53650-2
62. Lieske JC, Leonard R, Toback FG. **Adhesion of calcium oxalate monohydrate crystals to renal epithelial cells is inhibited by specific anions**. *Am J Physiol* (1995) **268**. DOI: 10.1152/ajprenal.1995.268.4.F604
63. Khan SR. **Reactive oxygen species as the molecular modulators of calcium oxalate kidney stone formation: evidence from clinical and experimental investigations**. *J Urol* (2013) **189**. DOI: 10.1016/j.juro.2012.05.078
64. Liang F, Fang Y, Cao W, Zhang Z, Pan S, Xu X. **Attenuation of tert-butyl hydroperoxide ( t-BHP)-Induced oxidative damage in HepG2 cells by tangeretin: Relevance of the Nrf2-ARE and MAPK signaling pathways**. *J Agric Food Chem* (2018) **66**. DOI: 10.1021/acs.jafc.8b01875
65. Ali T, Kim T, Rehman SU, Khan MS, Amin FU, Khan M. **Natural dietary supplementation of anthocyanins**. *Mol Neurobiol* (2018) **55**. DOI: 10.1007/s12035-017-0798-6
66. Liu M, Grigoryev DN, Crow MT, Haas M, Yamamoto M, Reddy SP. **Transcription factor Nrf2 is protective during ischemic and nephrotoxic acute kidney injury in mice**. *Kidney Int* (2009) **76**. DOI: 10.1038/ki.2009.157
67. Ma JQ, Ding J, Xiao ZH, Liu CM. **Puerarin ameliorates carbon tetrachloride-induced oxidative DNA damage and inflammation in mouse kidney through ERK/Nrf2/ARE pathway**. *Food Chem Toxicol* (2014) **71**. DOI: 10.1016/j.fct.2014.06.017
68. Ruiz S, Pergola PE, Zager RA, Vaziri ND. **Targeting the transcription factor Nrf2 to ameliorate oxidative stress and inflammation in chronic kidney disease**. *Kidney Int* (2013) **83**. DOI: 10.1038/ki.2012.439
69. Torihata Y, Asanuma K, Iijima K, Mikami T, Hamada S, Asano N. **Estrogen-dependent Nrf2 expression protects against reflux-induced esophagitis**. *Dig Dis Sci* (2018) **63**. DOI: 10.1007/s10620-017-4885-3
70. Chu C, Gao X, Li X, Zhang X, Ma R, Jia Y. **Involvement of estrogen receptor-α in the activation of Nrf2-antioxidative signaling pathways by silibinin in pancreatic β-cells**. *Biomol Ther (Seoul)* (2020) **28**. DOI: 10.4062/biomolther.2019.071
71. An R, Wang X, Yang L, Zhang J, Wang N, Xu F. **Polystyrene microplastics cause granulosa cells apoptosis and fibrosis in ovary through oxidative stress in rats**. *Toxicology* (2021) **449** 152665. DOI: 10.1016/j.tox.2020.152665
|
---
title: Is early or late biological maturation trigger obesity? A machine learning
modeling research in Turkey boys and girls
authors:
- Mehmet Gülü
- Fatma Hilal Yagin
- Hakan Yapici
- Khadijeh Irandoust
- Ali Ahmet Dogan
- Morteza Taheri
- Ewa Szura
- Magdalena Barasinska
- Tomasz Gabrys
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9971504
doi: 10.3389/fnut.2023.1139179
license: CC BY 4.0
---
# Is early or late biological maturation trigger obesity? A machine learning modeling research in Turkey boys and girls
## Abstract
Biological maturation status can affect individual differences, sex, height, body fat, and body weight in adolescents and thus may be associated with obesity. The primary aim of this study was to examine the relationship between biological maturation and obesity. Overall, 1,328 adolescents (792 boys and 536 girls) aged 12.00 ± 0.94–12.21 ± 0.99 years, respectively (measured for body mass, body stature, sitting stature). Body weights were deter-mined with Tanita body analysis system and adolescent obesity status was calculated according to the WHO classification. Biological maturation was determined according to the somatic maturation method. Our results showed that boys mature 3.077-fold later than girls. Obesity was an increasing effect on early maturation. It was determined that being obese, overweight and healthy-weight increased the risk of early maturation 9.80, 6.99 and 1.81-fold, respectively. The equation of the model predicting maturation is: Logit (P) = 1/(1 + exp. ( − (−31.386 + sex-boy * (1.124) + [chronological age = 10] * (−7.031) + [chronological age = 11] * (−4.338) + [chronological age = 12] * (−1.677) + age * (−2.075) + weight * 0.093 + height * (−0.141) + obesity * (−2.282) + overweight * (−1.944) + healthy weight * (−0.592)))). Logistic regression model predicted maturity with $80.7\%$ [$95\%$ CI: 77.2–$84.1\%$] accuracy. In addition, the model had a high sensitivity value ($81.7\%$ [76.2–$86.6\%$]), which indicates that the model can successfully distinguish adolescents with early maturation. In conclusion, sex and obesity are independent predictors of maturity, and the risk of early maturation is increased, especially in the case of obesity and in girls.
## Introduction
Childhood obesity is one of the most serious public health challenges of the 21st century [1]. Overweight and obesity are defined as “abnormal or excessive fat accumulation that presents a risk to health” [1]. Overweight and obesity are associated with metabolic diseases that increase the risk of noncommunicable diseases such as cardiovascular disease and diabetes [2, 3]. Childhood or adolescence obesity is associated with higher risk of weight-related morbidity and premature death in adulthood [4, 5]. Overweight and obesity cause at least 2.6 million deaths each year [1]. The World Obesity Federation reports that there has been a dramatic increase in childhood overweight and obesity over the past 30 years [6]. The overall prevalence rates of overweight/obesity and obesity in Azores adolescents were 31 and $27\%$, respectively [7]. In a study conducted with adolescents in Turkey, the prevalence rate of overweight was found to be $18.2\%$ [8].
Hormonal changes during puberty affect weight gain and body weight. Therefore, puberty leads to changes in body weight. These changes include changes in the amount and distribution of adipose tissue, lean body weight, and bone structure. In this period, besides rapid height increase in both sexes, weight gain also occurs [9]. Childhood overweight and obesity can lead to significant health problems in adulthood; these diseases are diabetes; musculoskeletal disorders, especially osteoarthritis; cardiovascular diseases (mainly heart disease and stroke); and some types of cancer (endometrial, breast, and colon) [1]. Excess adipose tissue causes oxidative stress, inflammation, apoptosis and mitochondrial dysfunctions [10, 11]. That’s why, obesity may cause to the onset of type-2-diabetes, liver steatosis, neurodegenerative and cardiovascular diseases that may thrive early in lifespan (12–17).
Biological maturation is a lifespan natural process which promotes morphophysiological changes in human [18]. The onset of puberty in girls is relationship with an improve in the quantity of fat mass, as a consequence of improved blood concentration of estradiol [19]. According to a study, the relative age effect was found to primarily affect the U13 and U15 categories in body composition [20]. In a study, a relationship was found between biological maturation and muscle strength [21]. Early sexual maturation is relationship with excessiveness body weight in girls and more stature for age in both sexes [22]. There are clearly dissimilarity between boys and girls in fat mass and distribution, particularly in adolescence period [23]. In girls, there is proof which early sexual maturation is relationship with a more prevalence of overweight and obesity [24, 25]. The number of studies in boys is quite scarce, and the evidences are mixed [25, 26].
The prevalence of obesity in children and adolescents in the United *States is* ~$17\%$, posing a risk to health status and life expectancy in adulthood [27, 28]. Obese children and adolescents are 5 times more likely to become obese in adulthood than non-obese children and adolescents [29]. About $55\%$ of obese children become obese in adolescence, about $80\%$ of obese adolescents will be obese in adulthood, and about $70\%$ will be obese over the age of 30 [29]. Overweight and obesity and related diseases may be largely prevented [1]. Therefore, research should focus on reducing and preventing obesity in children and adolescents. Prevention of childhood and adolescent obesity therefore needs high priority. Unlike adults, children do not have the opportunity to choose the environment they live in, the foods they consume, and the choices they make. For this reason, the growth and development of children can be affected by many factors. These factors can also cause the child or adolescent to become obese. Biological maturation status may effect individual differences, gender, height, body fat and body weight [7]. Early sexual maturation is relationship with overweight (girls only) and stature in children aged 8–14 years [22]. The primary aim of this study was to examine the relationship between biological maturation and obesity.
## Participants
The design of this study was cross-sectional. The research was conducted in Kirikkale province of Turkey. There are 8,758 students between the ages of 10–13 in Kirikkale province. G*power software was used to determine the sample size of the study [30]. As a result of the power analysis (alpha value = 0.05 and 1-beta value = 0.80, ηp2 = 0.25), it was found that at least 179 should be included in the study. In this study, 1,328 participants (792 boys; age = 12.00 ± 0.94 years and 536 girls; age = 12.21 ± 0.99 years) were randomly selected. The inclusion criteria of the participants were to attend physical education classes regularly for 2 h 1 day a week. Regarding dietary attitudes, the researchers did not gather any data. Each teenager and their parents received information about the study’s methods and potential hazards before becoming participants. They were informed about the study and given the option to voluntarily select whether to take part.
## Procedures
Permissions were obtained from public and private institutions for the study. Children, their parents and physical education teachers were informed about the measurement protocols and the purpose of the study. The information form about the research was read and signed by the parents. A form file was created for each of the participants who wanted to take part in the research voluntarily. The study was ap-proved by Kirikkale University Non-Interventional Research Ethics Committee (date: 2022-06-08, no: $\frac{2022}{10}$) and was conducted according to the principles stated in the Declaration of Helsinki. Anthropometric measurements were taken by experts in the field. None of the children participating in the measurements were excluded from the study. Within the scope of the study, each child’s age, gender, height, body weight, leg length and sitting height measurements were taken. It was determined based on the references in the WHO child and adolescent weight classification table to determine information on obesity status.
## Anthropometric measurements
Standardized procedures were applied for the anthropometric measurements of the participants [31]. Height and sitting heights were measured with a 0.1 cm long portable stadiometer (Seca 213, Hamburg, Germany). Tanita Body Composition Analyzer device (Tanita, BC-418, Japan) was used for body weight assessment. BMI value calculation was obtained by dividing body weight (kg) by the square of body height (m). To determine the maturity status of the participants, the percentage of estimated adult height (%PAS) included at the time of observation was estimated using [32]. In determining the maturity level of each participant, classification was made according to the %PAS z-score. Subsequently, the maturation status of the participants was classified as early (z-score > 0.5), timely (z-score ± 0.5), and late (z-score < 0.5).
## Sitting height measurement
The sitting heights of the participants were measured with a Holtain brand (stadiometer with 0.1 mm precision) measuring device. After the chair was adjusted according to the height of the participant, they were asked to take a deep breath and sit upright without moving. The resulting value was recorded in centimeters.
## Somatic maturation
Estimated adult body height (PAS) was used as an indicator of maturity (Khamis and Roche, 1994). Height was determined and treated as the estimated percentage of adult height (%PAS). PAS protocol calculation was calculated by age (decimal), body mass, height and mean parental height. The heights of the parents were collected with an informed consent form. The PAS variable was expressed as the percentile of estimated adult height (APAS) [33]. Among children of the same chrono-logical age, individuals with a higher estimated adult height were considered to be in a more advanced state of physical maturation compared with shorter individuals (Khamis and Roche, 1994). The Khamis-Roche method has been applied in many studies in order to predict the biological maturity status [34, 35]. In this study, a grouping was made among children. Using the sample median z-score of the obtained %PAS value, the latest maturation status ($p \leq 50$%) and the earliest maturation status ($p \leq 50$%) are given.
## Obesity classification
The prevalence of overweight and obesity in adolescents is defined according to the WHO growth reference for school-aged children and adolescents (overweight = 1 standard deviation body mass index for age and sex, and obese = 2 standard deviations body mass index for age and sex) [1]. In this direction, body mass index (BMI) values were calculated by measuring the height and weight of the individuals; The 85th and 95th percentiles were considered overweight, and those above the 95th percentile were considered obese [36] determining the BMI values and obesity status of the participants, the child body mass index calculation application on the website of Centers for Disease Control and Prevention (CDC) was used [37].
## Data analysis
The suitability of the quantitative variables to the univariate normal distribution was examined by visual (histogram and probability graphs) and analytical (Shapiro–Wilk test) methods. The Henze-Zirkler test was used to examine the multivariate normal distribution. The assumption of homogeneity of variances was examined with the Levene test. Since quantitative variables were not normally distributed, they were ex-pressed as median, and interquartile range (IQR). The two-way PERMANOVA (Permutational Analysis of Variance) test, with the Euclidean distance as the similarity matrix as the first factor maturity groups and the second factor obesity groups, was used to examine the difference and interaction effect between the groups (Permutation $$n = 9$$,999). In multivariate analysis, possible risk factors were examined by binary logistic regression using independent predictors in early and late maturity. Hosmer-Lemeshow and Omnibus tests were used to evaluate the logistic regression model and its coefficients. Classification performance measures were calculated using the confusion matrix regarding the prediction performance of the logistic regression model. In evaluating the performance of the model, accuracy, F1-score, sensitivity, specificity, positive predictive value, negative predictive value criteria were used. A $p \leq 0.05$ was considered statistically significant in all results. American Psychological Association (APA) 6.0 style was used to report statistical differences [38]. Statistical analyzes were performed using Python 3.9 software and SPSS 28.0 (IBM Corp., Armonk, NY, United States) package program. GraphPad 9.4.1 program was used for graphics.
## Results
The results of PERMANOVA showed that besides the interaction effect of maturity × obesity (F = −81.539; $p \leq 0.001$) at the age of the participants, obesity was the main effect ($F = 8.089$; p2 < 0.001). Interaction results showed that those with early maturation and healthyweight were significantly older. However, the main effect of maturity was not significant ($F = 0.841$; p1 = 0.240) (Table 1; Figure 1).
In addition to the main effect of maturity ($F = 3.682$; p1 = 0.002) and obesity ($F = 529.580$; p2 < 0.001) groups for weight, the interaction effect of maturity x obesity (F = −84.307; $$p \leq 0.02$$) was significant (Table 2; Figure 2).
The effect of maturity x obesity (F = −84.934; $$p \leq 0.043$$) was significant for the height of the participants, and although obesity was the main effect ($F = 9.341$; p2 < 0.001), there was no main effect for maturity ($F = 0.380$; p1 = 0.521), and our interaction results showed that those with healthyweight and early maturity had a higher height (Table 3; Figure 3).
There was no interaction effect of maturity × obesity (F = −86.300; $$p \leq 0.054$$) on sitting height of the participants, also maturity ($F = 0.981$; p1 = 0.195) and obesity ($F = 0.232$; p2 = 0.763) had no main effects (Table 4; Figure 4).
The main effect of maturity ($F = 0.352$; p1 = 0.543) and the interaction effect of maturity x obesity (F = −87.313; $$p \leq 0.165$$) were not detected in the leg length results, but the main effect of obesity was significant for leg length ($F = 6.825$; p2 < 0.001). The obesity, overweight and healthy weight group had a significantly higher leg length compared to the underweight group ($p \leq 0.05$) (Table 5; Figure 5).
Age, weight, height, sitting height, leg length, chronological age, sex and obesity were included in the model as predictive variables in binary logistic regression analysis. Age, weight, height, chronological age, sex, and obesity had significant OR p values for maturity. It was determined that one unit increase in age increased early maturation 7.94 times. Chronological age was an enhancing predictor for early maturation, and the risk of premature maturation was 1,000, 76.92, 5.35 fold higher in the 10, 11, and 12 chronological age groups, respectively, compared to the 13 year age group. Furthermore, boys matured 3.077 fold later than girls. Therefore, being a girl was an enhancing predictor for early maturation. Weight was an important determinant for maturity groups, and a one-unit increase in weight increased late maturity 1.098 fold. Height showed a enhancing effect on early maturation, and it was determined that an increase in height by one unit increased early maturation 1.15 fold. Obesity had an increasing effect on early maturation. It was determined that being obese, overweight and healthyweight increased the risk of early maturation 9.80, 6.99 and 1.81 fold, respectively. As a result, The equation of the model predicting maturation is: Logit (P) = 1/(1+ exp. ( − (−31.386 + boy * (1.124) + [chronological age = 10] * (−7.031) + [chronological age = 11] * (−4.338) + [chronological age = 12] * (−1.677) + age * (−2.075) + weight * 0.093 + height * (−0.141) + obesity * (−2.282) + overweight * (−1.944) + healthy weight * (−0.592)))).
With the developed logistic regression-based equation, it can be quickly determined whether a person matures early or late (Table 6).
**Table 6**
| Maturity | B | SE | Wald | Value of p | OR | 95% CI for OR | 95% CI for OR.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Maturity | B | SE | Wald | Value of p | OR | Lower bound | Upper bound |
| [Sex = boy] | 1.124 | 0.279 | 16.215 | <0.001 | 3.077 | 1.780 | 5.318 |
| [Chronological age = 10] | −7.031 | 0.877 | 64.227 | <0.001 | 0.001 | 0.001 | 0.005 |
| [Chronological age = 11] | −4.338 | 0.604 | 51.627 | <0.001 | 0.013 | 0.004 | 0.043 |
| [Chronological age = 12] | −1.677 | 0.382 | 19.304 | <0.001 | 0.187 | 0.089 | 0.395 |
| [Chronological age = 13] | | | 71.343 | <0.001 | | | |
| Age (years) | −2.075 | 0.256 | 65.645 | <0.001 | 0.126 | 0.076 | 0.207 |
| Weight (kg) | 0.093 | 0.029 | 10.093 | 0.001 | 1.098 | 1.036 | 1.163 |
| Height (cm) | −0.141 | 0.063 | 5.028 | 0.025 | 0.869 | 0.768 | 0.982 |
| Sitting height (cm) | 0.063 | 0.061 | 1.069 | 0.301 | 1.065 | 0.945 | 1.201 |
| Leg length (cm) | 0.063 | 0.059 | 1.162 | 0.281 | 1.065 | 0.950 | 1.195 |
| Obesity | −2.282 | 1.047 | 4.751 | 0.029 | 0.102 | 0.013 | 0.794 |
| Overweight | −1.944 | 0.665 | 8.553 | 0.003 | 0.143 | 0.039 | 0.527 |
| Healthy weight | −0.592 | 0.321 | 3.404 | 0.065 | 0.553 | 0.295 | 1.038 |
| Underweight | | | 9.181 | 0.027 | | | |
| Constant | 31.386 | 4.007 | 61.354 | <0.001 | | | |
Table 7 shows the performance criteria results and confidence intervals regarding the estimation performance of the logistic regression model. Our model predicted maturity with $80.7\%$ [$95\%$ CI: 77.2–$84.1\%$] accuracy. In addition, the model had a high sensitivity value ($81.7\%$ [76.2–$86.6\%$]), which indicates that the model can successfully distinguish adolescents with early maturation.
**Table 7**
| Metric | Value | 95% CI lower limit | 95% CI upper limit |
| --- | --- | --- | --- |
| Accuracy | 0.807 | 0.772 | 0.841 |
| F1-score | 0.795 | 0.76 | 0.83 |
| Sensitivity | 0.817 | 0.762 | 0.864 |
| Specificity | 0.798 | 0.746 | 0.844 |
| Positive predictive value | 0.774 | 0.717 | 0.825 |
| Negative predictive value | 0.837 | 0.787 | 0.88 |
## Discussion
The primary aim of this study was to examine the relationship between biological maturation and obesity. To the best of our knowledge, this is the first study predicting maturity by measuring some obesity parameters based on machine learning approach. Our predictive model showed that obesity increases the risk of early maturation. In addition, early maturation was higher in adolescent girls. Our model predicted maturity with $80.7\%$ [$95\%$ CI: 77.2–$84.1\%$] accuracy. Moreover, the high sensitivity value of the model ($81.7\%$ [76.2–$86.6\%$]) indicates that the model can successfully distinguish early maturing adolescents. Consistent with the current study, it was found in a study that obesity is associated with maturation in both boys and girls (1,525 boys and 1,501 girls aged 8–14), however, the association was differed.
Although a positive correlation was found in girls which is consistent with our study, a negative one was reported in boys unlike the present study [39]. This controversial results would be attributed to the study design, evaluation technique different populations studied and study duration. In justifying the obtained results for boys, it should be noted that there are conflictary results regarding the relationship between obesity and timing of pubertal onset in boys [40, 41] and this can be a reason for low correlation of obesity and maturity in boys rather than girls. In other words, obesity would contribute to early onset of puberty in girls more than boys. One possible reason for this effect refers to higher threshold of BMI for puberty development in boys than girls [41]. The second aim was to examine the effects of anthropometric measurements, gender, and obesity on biological maturation in adolescents. Our results showed that the maturity*obesity interaction effect was significant for age, body weight, and height. Age, weight, height, chronological age, gender, and obesity were important risk factors and predictors for maturity. Interestingly, in a logitudinal study lasting more than 10 years, it was reported that pubertal growth patterns, including earlier puberty onset timing, smaller puberty intensity, and shorter puberty spurt duration, had a positive association with higher obesity risks in late adolescence [42]. Notably, contraversial studies in boys are more observed compared to the girls. For instance, earlier puberty was found in overweight boys compared to normal weight and later puberty in obese compared to overweight (4,131 boys from 2005–2010) [43]. In a study, it was shown that with each unit increase in childhood BMI, the age of peak height velocity (PHV) was earlier by 2 months just in normal weight boys, while the same was not found in overweight ones [6]. In agreement with our study, there are several studies demonstrating the positive correlation between body composition and earlier onset of puberty in obese girls [44, 45]. Based on research evidences, there are some reasons for the early onset of puberty in girls including endocrine-disrupting chemicals [46], psychosocial factors [47], and chronic stress [48]. One possible mechanism involved in earlier puberty would be related to sex hormones changes. For instance, Increased levels of estradiol secretion would cause some changes in body fat distribution in pubescent girls. Accordingly, high estradiol levels are associated with precocious puberty [49]. In reality, the complicated etiology of obesity is influenced by a variety of variables, including biological, behavioral, environmental, physical activity and genetic ones [8, 50]. A wide range of exercise-related health advantages may be promoted by studying the lipidomic profile, which enables complicated biological processes to be adjusted more efficiently in the context of exercise [51]. Physical exercise has numerous repercussions on metabolism and function of different organs and tissues by enhancing whole-body metabolic homeostasis in response to different exercise-related adaptations. In a review, Latino et al. showed as exercise determine significant changes in lipidomic profiles, but they manifested in very different ways depending on the type of tissue examined [52] As healthy lifestyle during pubertal years is highly related to accelerated biological maturation in childhood and adolescence [53, 54], its recommended to consider it in future studies.
There are several limitations in the study. One of the main limitations of this study is that it is a cross-sectional study. The second one refers to the lack of measurement of testicular enlargement as a classic marker of pubertal onset, since it requires invasive palpation. It should be also noted that there are limited reliable self-reported markers for pubertal timing in boys and most data rely solely on visual grading of genital development. Another limitation of this study is the impossibility of laboratory measurements such as sex hormone levels and thelarche evaluation, which is recommended in future studies.
## Conclusion
Today, the use of machine learning technology is one of the advantages that allows the prediction of scientific facts in the best possible way. In conclusion, sex and obesity are independent predictors of maturity, and the risk of early maturation is increased, especially in the case of obesity in girls. This study’s main conclusion was that early biological maturation was strongly associated with obesity, particularly in females. In other words, early biological maturation in girls may result in their going through the menstrual cycle early and, as a result, developing a number of health issues as adults. Preventive programs can aid in their timely entry into the biological maturation process by further examining the fundamental mechanism of obesity. The participation of several public and private partners is necessary to stop the children and adolescent obesity epidemic and reduce the health hazards related to obesity. Governments, international partners, non-governmental organizations, and the corporate sector all have a crucial role to play in encouraging physical activity and better nutrition for children and adolescents.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by The study was approved by Kirikkale University Non-Interventional Research Ethics Committee (date: 2022-06-08, no: $\frac{2022}{10}$) and was conducted according to the principles stated in the Declaration of Helsinki. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
## Author contributions
MG and FY: conceptualization, validation, and supervision. MG, HY, and FY: methodology and investigation. FY: software, formal analysis, resources, and data curation. MG, HY, AD, KI, and MT: writing – original draft preparation. MG, KI, MT, TG, MB, and ES: writing – review and editing. AD: visualization. MG: project administration. MB, ES, and TG: funding acquisition. All authors contributed to the article and approved the submitted version.
## Funding
Published with the financial support of the European Union, as part of the project entitled Development of capacities and environment for boosting the international, intersectoral, and interdisciplinary cooperation At UWB, project reg. no. CZ$\frac{.02.2.69}{0.0}$/$\frac{0.0}{18}$_$\frac{054}{0014627.}$
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. 1.WHO (2020). Noncommunicable Diseases: Childhood Overweight and Obesity. https://www.who.int/news-room/questions-and-answers/item/noncommunicable-diseases-childhood-overweight-and-obesity. (2020)
2. Han TS, Tajar A, Lean MEJ. **Obesity and weight management in the elderly**. *Br Med Bull* (2011) **97** 169-96. DOI: 10.1093/bmb/ldr002
3. Khan MA, Smith JEM. **“Covibesity,” a new pandemic**. *Obesity Med.* (2020) **19** 100282. DOI: 10.1016/j.obmed.2020.100282
4. Park MH, Falconer C, Viner RM, Kinra S. **The impact of childhood obesity on morbidity and mortality in adulthood: a systematic review**. *Obes Rev* (2012) **13** 985-1000. DOI: 10.1111/j.1467-789X.2012.01015.x
5. Reilly JJ, Kelly J. **Long-term impact of overweight and obesity in childhood and adolescence on morbidity and premature mortality in adulthood: systematic review**. *Int J Obes* (2011) **35** 891-8. DOI: 10.1038/ijo.2010.222
6. Lobstein T, Jackson-Leach R, Moodie ML, Hall KD, Gortmaker SL, Swinburn BA. **Child and adolescent obesity: part of a bigger picture**. *Lancet* (2015) **385** 2510-20. DOI: 10.1016/S0140-6736(14)61746-3
7. Coelho ESMJ, Vaz Ronque ER, Cyrino ES, Fernandes RA, Valente-Dos-Santos J, Machado-Rodrigues A. **Nutritional status, biological maturation and cardiorespiratory fitness in Azorean youth aged 11-15 years**. *BMC Public Health* (2013) **13** 495. DOI: 10.1186/1471-2458-13-495
8. Gülü M, Yapici H, Mainer-Pardos E, Alves AR, Nobari H. **Investigation of obesity, eating behaviors and physical activity levels living in rural and urban areas during the covid-19 pandemic era: a study of Turkish adolescent**. *BMC Pediatr* (2022) **22** 405-10. DOI: 10.1186/s12887-022-03473-1
9. Burt Solorzano CM, Mccartney CR. **Obesity and the pubertal transition in girls and boys**. *Reproduction* (2010) **140** 399-410. DOI: 10.1530/REP-10-0119
10. Faienza MF, Francavilla R, Goffredo R, Ventura A, Marzano F, Panzarino G. **Oxidative stress in obesity and metabolic syndrome in children and adolescents**. *Horm Res Paediatr* (2012) **78** 158-64. DOI: 10.1159/000342642
11. Giordano P, Del Vecchio GC, Cecinati V, Delvecchio M, Altomare M, De Palma F. **Metabolic, inflammatory, endothelial and haemostatic markers in a group of Italian obese children and adolescents**. *Eur J Pediatr* (2011) **170** 845-50. DOI: 10.1007/s00431-010-1356-7
12. Ciccone MM, Faienza MF, Altomare M, Nacci C, Montagnani M, Valente F. **Endothelial and metabolic function interactions in overweight/obese children: the role of high-molecular weight Adiponectin**. *J Atheroscler Thromb* (2016) **23** 950-9. DOI: 10.5551/jat.31740
13. Faienza MF, Acquafredda A, Tesse R, Luce V, Ventura A, Maggialetti N. **Risk factors for subclinical atherosclerosis in diabetic and obese children**. *Int J Med Sci* (2013) **10** 338-43. DOI: 10.7150/ijms.5181
14. Faienza MF, Santoro N, Lauciello R, Calabrò R, Giordani L, Di Salvo G. **IGF2 gene variants and risk of hypertension in obese children and adolescents**. *Pediatr Res* (2010) **67** 340-4. DOI: 10.1203/PDR.0b013e3181d22757
15. Faienza MF, Wang DQ, Frühbeck G, Garruti G, Portincasa P. **The dangerous link between childhood and adulthood predictors of obesity and metabolic syndrome**. *Intern Emerg Med* (2016) **11** 175-82. DOI: 10.1007/s11739-015-1382-6
16. Marzano F, Faienza MF, Caratozzolo MF, Brunetti G, Chiara M, Horner DS. **Pilot study on circulating miRNA signature in children with obesity born small for gestational age and appropriate for gestational age**. *Pediatr Obes* (2018) **13** 803-11. DOI: 10.1111/ijpo.12439
17. Nacci C, Leo V, De Benedictis L, Carratù MR, Bartolomeo N, Altomare M. **Elevated Endothelin-1 (ET-1) levels may contribute to Hypoadiponectinemia in childhood obesity**. *J Clin Endocrinol Metabol* (2013) **98** E683-93. DOI: 10.1210/jc.2012-4119
18. Malina RM, Rogol AD, Cumming SP, Coelho e Silva MJ, Figueiredo AJ. **Biological maturation of youth athletes: assessment and implications**. *Br J Sports Med* (2015) **49** 852-9. DOI: 10.1136/bjsports-2015-094623
19. Rogol AD, Roemmich JN, Clark PA. **Growth at puberty**. *J Adolesc Health* (2002) **31** 192-200. DOI: 10.1016/S1054-139X(02)00485-8
20. Toselli S, Mauro M, Grigoletto A, Cataldi S, Benedetti L, Nanni G. **Maturation selection biases and relative age effect in Italian soccer players of different levels**. *Biology (Basel)* (2022) **11** 1559. DOI: 10.3390/biology11111559
21. Yapici H, Gulu M, Yagin FH, Eken O, Gabrys T, Knappova V. **Exploring the relationship between biological maturation level, muscle strength, and muscle power in adolescents**. *Biology* (2022) **11** 1722. DOI: 10.3390/biology11121722
22. Benedet J, Da Silva Lopes A, Adami F, De Fragas Hinnig P, De Vasconcelos Fde A. **Association of sexual maturation with excess body weight and height in children and adolescents**. *BMC Pediatr* (2014) **14** 72. DOI: 10.1186/1471-2431-14-72
23. Sweeting HN. **Gendered dimensions of obesity in childhood and adolescence**. *Nutr J* (2008) **7** 1. DOI: 10.1186/1475-2891-7-1
24. Himes JH, Obarzanek E, Baranowski T, Wilson DM, Rochon J, Mcclanahan BS. **Early sexual maturation, body composition, and obesity in African-American girls**. *Obes Res* (2004) **12** 64s-72s. DOI: 10.1038/oby.2004.270
25. Wang Y. **Is obesity associated with early sexual maturation? A comparison of the association in American boys versus girls**. *Pediatrics* (2002a) **110** 903-10. DOI: 10.1542/peds.110.5.903
26. Ribeiro J, Santos P, Duarte J, Mota J. **Association between overweight and early sexual maturation in Portuguese boys and girls**. *Ann Hum Biol* (2006) **33** 55-63. DOI: 10.1080/00207390500434135
27. Ogden CL, Carroll MD, Lawman HG, Fryar CD, Kruszon-Moran D, Kit BK. **Trends in obesity prevalence among children and adolescents in the United States, 1988-1994 through 2013-2014**. *JAMA* (2016) **315** 2292-9. DOI: 10.1001/jama.2016.6361
28. Styne DM, Arslanian SA, Connor EL, Farooqi IS, Murad MH, Silverstein JH. **Pediatric obesity-assessment, treatment, and prevention: an Endocrine Society clinical practice guideline**. *J Clin Endocrinol Metab* (2017) **102** 709-57. DOI: 10.1210/jc.2016-2573
29. Simmonds M, Llewellyn A, Owen CG, Woolacott N. **Predicting adult obesity from childhood obesity: a systematic review and meta-analysis**. *Obes Rev* (2016) **17** 95-107. DOI: 10.1111/obr.12334
30. Faul F, Erdfelder E, Lang A-G, Buchner A. **G* power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences**. *Behav Res Methods* (2007) **39** 175-91. DOI: 10.3758/BF03193146
31. Ingebrigtsen J, Brochmann M, Castagna C, Bradley PS, Ade J, Krustrup P. **Relationships between field performance tests in high-level soccer players**. *J. Strength Cond. Res* (2014) **28** 942-9. PMID: 23838979
32. Khamis HJ, Roche AF. **Predicting adult stature without using skeletal age: The Khamis-Roche method**. *Pediatrics* (1994) **94** 504-7. PMID: 7936860
33. Malina RM, Bouchards C, Bar-Or O. *Growth, Maturation and Physical Activity* (2004)
34. Malina RM, Cumming SP, Morano PJ, Barron M, Miller SJ. **Maturity status of youth football players: A noninvasive estimate**. *Med. Sci. Sport. Exerc.* (2005) **37** 1044-52
35. Cumming SP, Standage M, Gillison FB, Dompier TP, Malina RM. **Biological maturity status, body size, and exercise behaviour in British youth: A pilot study**. *J. Sport. Sci.* (2009) **27** 677-86
36. Neyzi O, Günöz H, Furman A, Bundak R, Gökçay G, Darendeliler F. **Weight, height, head circunference and body mass index referencesfor Turkish children**. *J. Child Health Dis.* (2008) **51** 1-14
37. 37.Cdc (2021). Centers for Disease Control and Prevention [Online]. Available: https://www.cdc.gov/healthyweight/bmi/calculator.html. *Centers for Disease Control and Prevention [Online]* (2021)
38. Yagin B, Yagin FH, Gozukara H, Colak C. **A web-based software for reporting guidelines of scientific researches**. *J. Cogn. Syst. Res* (2021) **6** 39-43
39. Wang YJP. **Is obesity associated with early sexual maturation?**. *Longit. Stud. Child Health Dev.* (2002b) **110** 903-10. DOI: 10.1542/peds.110.5.903
40. Li W, Liu Q, Deng X, Chen Y, Liu S, Story M. **Association between obesity and puberty timing: a systematic review and meta-analysis**. *Environ Res Public Health* (2017) **14** 1266. DOI: 10.3390/ijerph14101266
41. Liu M, Cao B, Luo Q, Wang Q, Liang X, Wu D. **The critical BMI hypothesis for puberty initiation and the gender prevalence difference: Evidence from an epidemiological survey in Beijing, China**. *Front Endocrinol (Lausanne)* (2022) **13** 1009133. DOI: 10.3389/fendo.2022.1009133
42. Chen L, Su B, Zhang Y, Ma T, Liu J, Yang Z. **Association between height growth patterns in puberty and stature in late adolescence: a longitudinal analysis in Chinese children and adolescents from 2006 to 2016**. *Front Endocrinol (Lausanne)* (2022) **13** 882840. DOI: 10.3389/fendo.2022.882840
43. Lee JM, Wasserman R, Kaciroti N, Gebremariam A, Steffes J, Dowshen S. **Timing of puberty in overweight versus obese boys**. *Pediatrics* (2016) **137** e20150164. DOI: 10.1542/peds.2015-0164
44. Davison KK, Susman EJ, Birch LL. **Percent body fat at age 5 predicts earlier pubertal development among girls at age 9**. *Pediatrics* (2003) **111** 815-21. DOI: 10.1542/peds.111.4.815
45. Lee JM, Appugliese D, Kaciroti N, Corwyn RF, Bradley RH, Lumeng JC. **Weight status in young girls and the onset of puberty**. *Longit. Stud. Child Health Dev.* (2007) **119** e624-30. DOI: 10.1542/peds.2006-2188
46. Uldbjerg C, Koch T, Lim Y-H, Gregersen L, Olesen C, Andersson A-M. **Prenatal and postnatal exposures to endocrine disrupting chemicals and timing of pubertal onset in girls and boys: a systematic review and meta-analysis**. *Hum. Reprod. Update* (2022) **28** 687-716. DOI: 10.1093/humupd/dmac013
47. Glass DJ, Geerkens JT, Martin MA. **Psychosocial and energetic factors on human female pubertal timing: a systematized review**. *Evol Hum Sci* (2022) **4** e28. DOI: 10.1017/ehs.2022.24
48. Ferrari V, Stefanucci S, Ciofi D, Stagi S, Pediatric TM, Gynecology A. **Analysis of the timing of puberty in a recent cohort of Italian girls: evidence for earlier onset compared to previous studies**. *J Pediatr Adolesc Gynecol* (2022) **35** 23-9. DOI: 10.1016/j.jpag.2021.06.007
49. Li D, Zhang B, Cheng J, Chen D, Wu Y, Luo Q. **Obesity‐related genetic polymorphisms are associated with the risk of early puberty in Han Chinese girls**. *Clin Endocr* (2022) **96** 319-27. DOI: 10.1111/cen.14631
50. Stone T., Dipietro L., Stachenfeld N.S.. (2021)
51. Han X. **Lipidomics for studying metabolism**. *Nat Rev Endocrinol* (2016) **12** 668-79. DOI: 10.1038/nrendo.2016.98
52. Latino F, Cataldi S, Carvutto R, De Candia M, D'elia F, Patti A. **The importance of Lipidomic approach for mapping and exploring the molecular networks underlying physical exercise: a systematic review**. *Int J Mol Sci* (2021) **22** 8734. DOI: 10.3390/ijms22168734
53. Beunen G, Malina RM. **“Growth and biologic maturation: relevance to athletic performance”**. *The Young Athlete* (2008) **1** 3-17
54. Cengizel CO, Oz E, Cengizel EJ. **Short-term plyometric and jump rope training effect on body profile and athletic performance in adolescent basketball players**. *Int J Sport Stud Health* (2022) **5** 5. DOI: 10.5812/intjssh-132585
|
---
title: 'SCORE2 cardiovascular risk prediction models in an ethnic and socioeconomic
diverse population in the Netherlands: an external validation study'
authors:
- Janet M. Kist
- Rimke C. Vos
- Albert T.A. Mairuhu
- Jeroen N. Struijs
- Petra G. van Peet
- Hedwig M.M. Vos
- Hendrikus J.A. van Os
- Edith D. Beishuizen
- Yvo W.J. Sijpkens
- Mohammad A. Faiq
- Mattijs E. Numans
- Rolf H.H. Groenwold
journal: eClinicalMedicine
year: 2023
pmcid: PMC9971516
doi: 10.1016/j.eclinm.2023.101862
license: CC BY 4.0
---
# SCORE2 cardiovascular risk prediction models in an ethnic and socioeconomic diverse population in the Netherlands: an external validation study
## Body
Research in contextEvidence before this studyEthnicity and socioeconomic status are associated with higher cardiovascular disease risk. Whether ethnicity and socioeconomic status should explicitly be included in cardiovascular disease risk prediction models on mainland *Europe is* still unclear. In preparation for this manuscript, PubMed was searched from inception until December 31, 2022, for “cardiovascular disease”, “cardiovascular prediction”, “socioeconomic status”, “ethnicity”, “SCORE2”, “social determinants of health” (which yielded 2600 studies).Added value of this studyOn top of traditional risk factors (age, smoking, cholesterol and blood pressure), socioeconomic status and ethnicity further distinguish individuals at higher or lower absolute risk for cardiovascular disease events. Adjusting risk prediction of high-risk subgroups to SCORE2 models intended for use in higher-risk European countries (best fitting model according the observed to expected events per subgroups) would, in our study, have led to an almost twofold increase of individuals eligible for treatment (from $10\%$ to $17\%$, and from $2\%$ to $5\%$ in men and women, respectively).Implications of all the available evidenceAppropriate risk modelling within countries taking ethnicity as well as socioeconomic status into account is necessary for adequate cardiovascular disease risk counselling. The existing information in routine health care data could service the population through research into health disparities to help improve health equity.
## Summary
### Background
Socioeconomic status and ethnicity are not explicitly incorporated as risk factors in the four SCORE2 cardiovascular disease (CVD) risk models developed for country-wide implementation across Europe (low, moderate, high and very-high model). The aim of this study was to evaluate the performance of the four SCORE2 CVD risk prediction models in an ethnic and socioeconomic diverse population in the Netherlands.
### Methods
The SCORE2 CVD risk models were externally validated in socioeconomic and ethnic (by country of origin) subgroups, from a population-based cohort in the Netherlands, with GP, hospital and registry data. In total 155,000 individuals, between 40 and 70 years old in the study period from 2007 to 2020 and without previous CVD or diabetes were included. Variables (age, sex, smoking status, blood pressure, cholesterol) and outcome first CVD event (stroke, myocardial infarction, CVD death) were consistent with SCORE2.
### Findings
6966 CVD events were observed, versus 5495 events predicted by the CVD low-risk model (intended for use in the Netherlands). Relative underprediction was similar in men and women (observed/predicted (OE-ratio), 1.3 and 1.2 in men and women, respectively). Underprediction was larger in low socioeconomic subgroups of the overall study population (OE-ratio 1.5 and 1.6 in men and women, respectively), and comparable in Dutch and the combined “other ethnicities” low socioeconomic subgroups. Underprediction in the Surinamese subgroup was largest (OE-ratio 1.9, in men and women), particularly in the low socioeconomic Surinamese subgroups (OE-ratio 2.5 and 2.1 in men and women). In the subgroups with underprediction in the low-risk model, the intermediate or high-risk SCORE2 models showed improved OE-ratios. Discrimination showed moderate performance in all subgroups and the four SCORE2 models, with C-statistics between 0.65 and 0.72, similar to the SCORE2 model development study.
### Interpretation
The SCORE 2 CVD risk model for low-risk countries (as the Netherlands are) was found to underpredict CVD risk, particularly in low socioeconomic and Surinamese ethnic subgroups. Including socioeconomic status and ethnicity as predictors in CVD risk models and implementing CVD risk adjustment within countries is desirable for adequate CVD risk prediction and counselling.
### Funding
$\frac{10.13039}{501100005039}$Leiden University Medical Centre and $\frac{10.13039}{501100001717}$Leiden University.
## Evidence before this study
Ethnicity and socioeconomic status are associated with higher cardiovascular disease risk. Whether ethnicity and socioeconomic status should explicitly be included in cardiovascular disease risk prediction models on mainland *Europe is* still unclear. In preparation for this manuscript, PubMed was searched from inception until December 31, 2022, for “cardiovascular disease”, “cardiovascular prediction”, “socioeconomic status”, “ethnicity”, “SCORE2”, “social determinants of health” (which yielded 2600 studies).
## Added value of this study
On top of traditional risk factors (age, smoking, cholesterol and blood pressure), socioeconomic status and ethnicity further distinguish individuals at higher or lower absolute risk for cardiovascular disease events. Adjusting risk prediction of high-risk subgroups to SCORE2 models intended for use in higher-risk European countries (best fitting model according the observed to expected events per subgroups) would, in our study, have led to an almost twofold increase of individuals eligible for treatment (from $10\%$ to $17\%$, and from $2\%$ to $5\%$ in men and women, respectively).
## Implications of all the available evidence
Appropriate risk modelling within countries taking ethnicity as well as socioeconomic status into account is necessary for adequate cardiovascular disease risk counselling. The existing information in routine health care data could service the population through research into health disparities to help improve health equity.
## Introduction
Cardiovascular disease (CVD) is the most common cause of death in Europe, causing $45\%$ of all deaths.1,2 After decades of decreasing numbers in Europe, CVD deaths are expected to rise again due to an ageing population and unhealthy lifestyles.2,3 To distinguish seemingly healthy persons in Europe at low and high CVD risk, the SCORE2 was recently developed to predict the 10-year cardiovascular risk of fatal and non-fatal cardiovascular events.4 SCORE2 is based on pooled large prospective cohorts of almost 700,000 patients and externally validated in over one million individuals.4 The SCORE2 risk models are adjusted to background country-level CVD risk, on the basis of standardised CVD mortality rates, to four country-wide CVD risk models (low, moderate, high and very-high).
SCORE2 models are incorporated in the 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice, which provides preventive recommendations for populations (e.g., health policies) and the individual.5 On the individual level, the guideline provides a stepwise approach, with lifestyle and treatment recommendations based on SCORE2 risk prediction, comorbidities, risk modifiers (e.g., family history, ethnicity, stress factors), lifetime benefit and patient preferences.5 The European SCORE2 prediction models are based on traditional risk factors for CVD (age, sex, smoking status, diabetes, blood pressure and cholesterol).4 However, studies showed large differences in CVD in socioeconomic and ethnic subgroups of the population.6, 7, 8, 9, 10, 11, 12, 13, 14 The differences in CVD death found in socioeconomic and ethnic subgroups within European countries, are of similar magnitude compared with the differences in CVD death between Eastern and Western European countries.11,14 These differences might urge targeted absolute risk adjustment for specified subgroups within countries on top of traditional risk factors. The observed differences, however, could still be the result of differences in traditional risk factors (such as diabetes, smoking and hypertension), driving the higher observed risks in these subgroups.5,10,15, 16, 17, 18 In Europe, ASSIGN was the first to incorporate socioeconomic status in their CVD risk model for the Scottish population, and the QRISK CVD risk model has incorporated socioeconomic status as well as ethnicity for use in the UK population.19,20 For the other European countries, it is still unclear whether ethnicity and socioeconomic status need to be incorporated in CVD risk models.5,21 Specifically, the SCORE2 prediction models have not been externally validated in combined ethnic and socioeconomic subgroups yet.5 The primary aim of this study was to evaluate the performance of the SCORE2 prediction models in the population and in different ethnic and socioeconomic subgroups.
## Methods
This external validation study was reported according to the Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis statement (TRIPOD, Appendix A).22
## Design, population and study period
This was a prospective, dynamic, population-based study, from the novel routine health care database of the Extramural LUMC Academic Network (ELAN). For this study data from general practitioners (GPs) and hospitals of the region of The Hague in the Netherlands were linked with medical and registry data from Statistics Netherlands using unique pseudonymised identifiers for households and individuals.23,24 Individuals were included if they were registered between 2007 and 2020 at a GP participating in the ELAN database for at least 6 months. We excluded individuals with a history of diabetes or CVD before cohort entry, defined as non-fatal stroke or myocardial infarction, as well as individuals using statin and diabetes medication before cohort entry (flowchart in Appendix B).
Cohort entry was defined as the date a person was registered with a participating GP, start of 2007 and age between (turning) 40 and 70 years. Follow-up time in person years was calculated from cohort entry until development of an event (CVD death, first CVD event), attaining age 80 (70 plus 10-year follow-up), end of study period (July 2020) or deregistration with a participating GP practice, whichever came first.
## Outcome and competing event
The main outcome of this study was similar to SCORE2, a combined endpoint of CVD death and first CVD event after cohort entry (stroke or myocardial infarction). Competing events were non-CVD deaths. Deaths were derived from the death registry of Statistics Netherlands, coded according to the International Classification of Diseases and Related Health Problems 10th version (ICD-10). ICD-10 death codes have been found to be reliable in $98\%$ of the cases.25 CVD events were extracted from GP files, coded according to the International Classification of Primary Care (ICPC), consistent with SCORE2 ICD-10 codes. The complete list of used CVD codes can be found in Appendix C.
## Predictors of cardiovascular events
For the validation of the SCORE2 prediction model, the variables age, sex, smoking, non-HDL-cholesterol and systolic blood pressure were used. Age and sex were derived at cohort entry. Individual data on smoking, blood pressure and non-HDL cholesterol were derived from medical records (GP when available, otherwise from hospital data, value before or nearest to cohort entry). Medical data from GPs and hospital were linked with individual data from Statistics Netherlands, on deaths, sex, age, ethnicity (country of origin) and socioeconomic status (disposable household income, derived from Dutch tax register). Additional information on smoking status was derived from text mining (string matching) on free text information within the GP data (details in Appendix D). Ethnicity of individuals was based on the country of origin (classification as used by Statistics Netherlands).24 *For this* study, disposable household income which represents “the net amount a household can spend on an annual basis, adjusted for any differences in household size and composition” was used as a proxy for socioeconomic status.24 The information on disposable household level provided by Statistics Netherlands consisted of percentiles of disposable household level as compared to the general population of the Netherlands.24 The quintile cut-points for 2014 were €16,000, €21,100, €26,800 and €34,700.
## Statistical analysis
Analyses were performed in the overall population by sex and stratified by ethnicity and socioeconomic status. For external validation, the original SCORE2 Fine and Gray model formulas, syntax, intercept and regression-coefficients were provided by the SCORE2 research team.4 After 10 years of follow-up, the risk predictions for the low, moderate, high and very high-risk models were calculated. On sample size considerations, external validation was performed on subgroups with at least 200 CVD events (minimal numbers for an appropriate external validation).26 Calibration was assessed with observed/mean expected probability (OE-ratio) at timepoint 10 years and with calibration plots in deciles of the population by predicted risk. Discrimination was assessed using Harrell's C-statistic.
Pre-processing and analyses of the data were performed using R Statistical Computing (version 4.2.1). Data were analysed on the secure data infrastructure of Statistics Netherlands.
On missing data, when information on a binary variable was not reported, for example a CVD diagnosis, it was assumed that particular variable was absent. On variables, percentages of missingness were reported. For disposable household income, blood pressure and cholesterol we assumed these were missing at random and were imputed using multiple imputation (variables in the imputation model: age, blood pressure, cholesterol, fasting glucoses, eGFR, smoking status, disposable household income, ethnicity, follow-up time and CVD events). Five imputed datasets were generated using the MICE algorithm in R, the convergence of imputed values was assessed in 40 imputations, and results were combined with Rubin's rule.27 Several sensitivity analyses were performed. First, to assess differences due to missingness, performance analyses were repeated in part of the population with at least 1 measurement. Second, to assess spatial and socioeconomic trends we analysed part of the population from a smaller, more urban area of The Hague with and without imputation of missing socioeconomic status. Third, performance variation in age groups (40–50 and 50–70 years) was assessed, because SCORE2 treatment thresholds are different under and above 50 years of age (2.5 and $7.5\%$ versus 5 and $10\%$ predicted probability). Fourth, to assess validity of the CVD event outcome, we compared CVD events (ICPC coded) with start of the combination of statins and antithrombotic medication in the same individual.
## Ethics
Routinely collected data were anonymized through a trusted third party and Statistics Netherlands to prevent identification of individuals by researchers. In accordance to Dutch legislation, GPs and hospitals informed individuals about use of their anonymized data for research purposes and individuals could withdraw via an informed opt-out procedure and informed consent from individuals in the study was waived and not obtained. For this waiver, the appropriate approval that the study is not subject to the Medical Examination Act was granted, after evaluation of the research protocol by the authority of the area (Medical Ethical Committee LUMC Leiden, under reference number G18.070).
## Role of the funding source
The ELAN Vascular study is supported by the Department of Public Health & Primary care and the Board of Directors of the Leiden University Medical Centre and by the Leiden University. The funding sources had no role in study design, collection, analysis, interpretation of the data, in the writing of the report, or the decision to submit the paper for publication.
## Results
Data of 74,880 men and 80,133 women from the region of The Hague were included from January 1st, 2007 to July 1st, 2020 (based on the in- and exclusion criteria, from 537,000 individuals in the ELAN Datawarehouse).
In men 4251 CVD events and in women 2715 CVD events occurred. The mean age was 48 years (SD 9). Compared to the income boundaries of the overall Dutch population, our population showed a different distribution, with lower numbers of individuals in the lower disposable household income quintiles. For men, $64\%$ were of Dutch origin, $6\%$ of Surinamese, for women, $62\%$ were Dutch and $7\%$ Surinamese (Table 1). With lower numbers, the other ethnicities were combined as one group ($31\%$ of men and women). Mean systolic blood pressure was 136 mmHg (SD 22, $31\%$ missing), mean total cholesterol was 5.5 mmol/l (SD 1.1, $35\%$ missing), and mean HDL cholesterol 1.4 mmol/L (SD 0.4, $36\%$ missing). The median follow-up time was 9.9 years (interquartile range, IQR 5.7–12.8).Table 1Baseline characteristics routine health care cohort, 2007–2020, 40–70 years of age. MenCVD eventsWomenCVD eventsMissingSCORE2 derivation cohortn (%), mean (SD)nn (%), mean (SD)n%n (%), mean (SD)Total74,88080.133677.684Age at cohort entry (years)48.1 (8.6)48.3 (8.8)57 [9]Systolic blood pressure (mmHg)138 [21]134 [22]31136 [19]Total cholesterol (mmol/l)5.5 (1.1)5.5 (1.1)355.8 (1.1)HDL cholesterol (mmol/l)1.2 (0.3)1.5 (0.4)361.4 (0.4)Smoking24,978 (33.4)25,751 (32.1)a101,211 [15]Follow-up (person-years, median, IQR)9.8 (5.6–12.6)10.1 (5.8–12.8)10.7 (5.0–18.6)Events CVD events4251 (5.7)2715 (3.4)30,121 (4.4) Non-CVD death2100 (2.8)1918 (2.4)33,809 (5.0)Subgroups Socioeconomic status6 1st (lowest)11,121 (15.9)65313,487 (16.8)555 2nd8889 (12.7)55511,110 (13.9)489 3rd12,298 (17.6)70413,857 (17.3)473 4th16,274 (23.3)96816,547 (20.6)544 5th (highest)21,411 (30.6)119620,946 (26.1)555 Ethnicity (by country of origin) Dutch47,901 (64.0)289749,776 (62.1)1893 Surinamese4143 (5.5)3005492 (6.9)237 Other ethnicities: Antilleans1445 (1.9)661675 (2.1)48 Brittons932 (1.2)29802 (1.0)22 Germans1071 (1.4)971216 (1.5)76 Indonesian4122 (5.5)2764466 (5.6)141 Moroccan2256 (3.0)811980 (2.5)22 Turkish2246 (3.0)1181970 (2.5)45 Middle & Eastern Europeans1006 (1.3)331632 (2.0)29 Other countries9758 (13.0)35411,124 (13.9)202SD, standard deviation. CVD events, cardiovascular disease events, combination of CVD death and first CVD event. CVD death, ICD-10 I10-I25, R96, I46, I47-I51, I61-I65 (except I62.0), G45, I67-I69 (except I67.1), I70 tot I72.CVD event, myocardial infarction and stroke (ICPC K75, K90, except K90.1).Socioeconomic status, by disposable household income, in quintiles of the population of the Netherlands.aMissing in $50\%$ of cases, when missing non-smoking was assumed.
## Performance: Total population
At external validation of the SCORE2 low-risk model (intended for use in the Netherlands), we found that for the total population the SCORE2 low-risk model predicted 5495 CVD events in the population included in this study, whereas 6966 CVD events were observed. Calibration showed underprediction both in men and women. Calibration showed an OE-ratio (observed/mean expected ratio) of 1.30 ($95\%$ confidence interval (CI) 1.30–1.30) in men and 1.22 ($95\%$ CI 1.21–1.22) in women (Fig. 1).Fig. 1OE-ratio external validation SCORE2 by ethnicity and socioeconomic status. OE-ratio, observed risk/mean predicted probability (<1 overprediction, >1 underprediction), the OE-ratio is an overall assessment of the correspondence of the predicted probability compared to the actual observed risks. 1st–5th SES, socioeconomic status, by disposable household income, in quintiles of the population of the Netherlands, 1st SES is lowest, 5th SES is highest. Low-, moderate-, high- and very-high-risk model, the four SCORE2 models for different regions of Europe, the low-risk model is the intended model for the Netherlands.
The moderate-risk model showed an OE-ratio of 1.02 ($95\%$ CI 1.02–1.02) in men and women. The high-risk model and very-high-risk model (intended for use in Eastern Europe) showed overprediction in men and women for the total population, with an OE-ratio below 1 (Fig. 1). Discrimination showed moderate performance in all regional SCORE2 risk models, with a C-statistic of 0.70 ($95\%$ CI 0.69–0.71) and 0.72 ($95\%$ CI 0.71–0.73) in men and women, respectively (Table 2).Table 2Discrimination external validation SCORE2 by ethnicity and socioeconomic status. SubgroupsHarrell's C-statistic ($95\%$ CI)MenWomenTotal population 1st SES0.70 (0.68–0.72)0.73 (0.71–0.75) 2nd SES0.70 (0.68–0.72)0.73 (0.69–0.77) 3rd SES0.70 (0.68–0.72)0.72 (0.70–0.74) 4th SES0.70 (0.68–0.71)0.70 (0.68–0.72) 5th SES0.72 (0.70–0.73)0.71 (0.69–0.73) Total0.70 (0.69–0.71)0.72 (0.71–0.73)Dutch 1st SES0.67 (0.64–0.70)0.69 (0.66–0.72) 2nd SES0.70 (0.67–0.72)0.72 (0.68–0.75) 3rd SES0.70 (0.68–0.72)0.72 (0.69–0.74) 4th SES0.69 (0.67–0.71)0.72 (0.69–0.74) 5th SES0.72 (0.71–0.74)0.71 (0.68–0.73) Total0.70 (0.69–0.71)0.72 (0.70–0.73)Surinamese 1st SES0.63 (0.57–0.70)0.71 (0.65–0.77) 2nd SES0.63 (0.54–0.71)0.74 (0.60–0.88) 3rd SES0.64 (0.55–0.73)0.69 (0.62–0.76) 4th SES0.68 (0.61–0.75)0.63 (0.55–0.71) 5th SES0.69 (0.62–0.76)0.71 (0.63–0.80) Total0.65 (0.62–0.69)0.70 (0.67–0.73)Other ethnicities 1st SES0.74 (0.71–0.77)0.76 (0.72–0.80) 2nd SES0.71 (0.66–0.75)0.75 (0.68–0.82) 3rd SES0.73 (0.68–0.77)0.74 (0.69–0.79) 4th SES0.72 (0.68–0.76)0.69 (0.63–0.75) 5th SES0.71 (0.68–0.75)0.72 (0.65–0.78) Total0.72 (0.71–0.74)0.74 (0.72–0.76)Harrell's C-statistic, displays whether the model accurately discriminates individuals with the event from individuals without the event (when 1, discrimination of a model is perfect, when 0.5, the model does not discriminate between individuals with and without the event).Ethnicity, by country of birth.1st–5th SES, socioeconomic status, by disposable household income, in quintiles of the population of the Netherlands, 1st SES is lowest, 5th SES is highest.
## Performance: Socioeconomic subgroups
*In* general, at calibration in socioeconomic subgroups for the total population the SCORE2 models showed more extreme underprediction in subgroups at lower socioeconomic status. The OE-ratios of the low-risk model, ranged in the total population in lowest to highest socioeconomic quintile from 1.54 ($95\%$ CI 1.52–1.57) to 1.17 ($95\%$ CI 1.16–1.18) in men, and from 1.64 ($95\%$ CI 1.63–1.66) to 0.94 ($95\%$ CI 0.92–0.95) in women. Calibration plots by socioeconomic subgroups showed similar results (Appendix E).
Discrimination showed moderate performance in all four risk models and across subgroups. For the low-risk model in socioeconomic subgroups, C-statistics ranged from 0.70 to 0.72, and 0.70 to 0.73 in men and women respectively (Table 2).
## Performance: Ethnic subgroups
Performance was assessed in the Dutch, Surinamese and other ethnicity subgroups. Calibration in the Dutch and other ethnicity subgroups showed underprediction (OE-ratio low-risk model Dutch and 1.21 ($95\%$ CI 1.21–1.22) and 1.15 ($95\%$ CI 1.14–1.16) in men and women, respectively). In “other ethnicities” 1.40 ($95\%$ CI 1.35–1.44) and 1.20 ($95\%$ CI 1.02–1.38) in men and women, respectively (Fig. 1, calibration plots Fig. 2). In Surinamese, calibration showed underprediction to a greater distance (OE-ratio in the low-risk model was 1.92 ($95\%$ CI 1.89–1.95) in men and 1.86 ($95\%$ CI 1.81–1.92) in women (Fig. 1), calibration plots Fig. 2).Fig. 2Calibration plots, SCORE2 low-risk model by ethnicity (LOESS plotted). Calibration plots, asses model fit by graphically comparing predicted probability to the proportion of observed events in decile groups of the population. Loess, locally estimated scatterplot smoothing.
For “other ethnicities” men the high-risk model 1.05 ($95\%$ CI 1.05–1.06) was more nearing one in the OE-ratio, while for Surinamese men and women, the high-risk model still had an OE-ratio of 1.45 ($95\%$ CI 1.42–1.47) and 1.21 ($95\%$ CI 1.20–1.22), respectively (Fig. 1).
Discrimination showed moderate performance in all regional SCORE2 risk models, with a C-statistic in Dutch of 0.70 ($95\%$ CI 0.69–0.71) and 0.72 ($95\%$ CI 0.71–0.73), in “other ethnicities” 0.72 ($95\%$ CI 0.71–0.74) and 0.74 ($95\%$ CI 0.72–0.76) and in Surinamese, 0.65 ($95\%$ CI 0.62–0.69) and 0.70 ($95\%$ CI 0.67–0.73), in men and women respectively (Table 2).
## Performance: Combined socioeconomic and ethnic subgroups
At calibration in the combined socioeconomic and ethnic subgroups, the SCORE2 models showed more extreme underprediction in subgroups at lower socioeconomic status.
In combined subgroups of men, OE-ratios were highest in the Surinamese subgroups, from 2.53 ($95\%$ CI 2.37–2.71) to 1.88 ($95\%$ CI 1.71–2.09) and lowest in the Dutch subgroups, from 1.45 ($95\%$ CI 1.41–1.50) to 1.09 ($95\%$ CI 1.08–1.10). In women, the Surinamese had the highest OE-ratios, from 2.05 ($95\%$ CI 1.96–2.15) to 1.08 ($95\%$ CI 0.86–1.46), and the lowest were the “other ethnicities” subgroup, from 1.35 ($95\%$ CI 1.32–1.38) to 0.88 ($95\%$ CI 0.80–1.35) (Fig. 1).
Discrimination in the combined socioeconomic and ethnic subgroups showed moderate performance in all four risk models (C-statistics ranged from 0.63 to 0.74, and 0.63 to 0.76 in men and women respectively, Table 2).
## Sensitivity analyses
First, analyses of the subjects in whom at least one laboratory measurement was made showed slightly lower or slightly higher observed risks. Second, in the analyses in the smaller more urban area of the city of The Hague, the proportion of individuals in the lowest quintile of household income was higher, and missingness was approximately $5\%$ higher, but calibration and discrimination of the smaller area as well as with and without information on household income, showed similar results in all subgroups. Third, when we stratified on age between forty and fifty and over fifty, calibration showed more underprediction in men 50 and above and in women between 40 and 50 years of age, discrimination was slightly lower in these age subgroups. Fourth, $93\%$ of individuals with a CVD event (first myocardial infarction or stroke) started with the combination of statin and antithrombotic medication.
## Discussion
This study aimed to externally validate the SCORE2 CVD risk prediction models with a focus on its performance in socioeconomic and ethnic subgroups. In the general study population, the performance of the SCORE2 low-risk model (intended for use in the Netherlands) showed underprediction in men and slight underprediction in women. However, in terms of calibration the performance of the SCORE2 low-risk model was particularly poor in low socioeconomic subgroups and in Surinamese subgroups. The moderate or high-risk SCORE2 model appeared to be a much better predictive model for these subgroups.
Our study also showed that information available in contemporary routine health care data combined with national registry databases in the Netherlands can be utilized for research into health disparities, ethnicity and socioeconomic status.
While much evidence exists on the relation between socioeconomic status, ethnicity and the risk for CVD, the SCORE2 CVD risk prediction models do not take these factors into account explicitly.4,5 The observed higher CVD risk in lower socioeconomic subgroups is consistent with literature since the 1970s.6,9 Evidence on the combined effect on absolute risk of socioeconomic status and traditional risk factors was inconclusive for the European main land population.5 In CVD risk models in Scotland and the UK, ASSIGN and QRISK include socioeconomic status since 2007 in their risk models, with socioeconomic area codes (Scottish Index of Multiple Deprivation score SIMD and Townsend score, respectively).19,28 SIMD is based on income, employment, education, health, access to services, crime and housing.19 The Townsend score is based on unemployment, car ownership, home ownership and household overcrowding.29 Individual information on socioeconomic status is however preferable and income is a valid measure for the association between socioeconomic status and CVD.30 One study assessed the performance of SCORE2 in Scotland, and found comparable higher CVD OE-ratios in lowest socioeconomic subgroups.31 Also, in the performed sensitivity analyses in suburban and urban areas, low socioeconomic status subgroups showed a comparable increased absolute risk. With $74\%$ of the population living in a suburban or urban area in the Netherlands, this could indicate a generalizable risk increase for low socioeconomic subgroups in the general population.32 The ESC guideline recommends to use absolute risk multipliers for ethnicities (factor 1.3 for Indian, 1.7 for Pakistan and 0.85 for African Caribbean descendance).5 However, these multipliers were specifically developed for ethnic subgroups in the UK, and might not be applicable for ethnic subgroups in other European regions.5,21 The CVD risk among Surinamese in our The Hague region ($76\%$ of South Asian and $13\%$ of African Caribbean descendance) was 1.9 times higher than the predicted CVD probabilities (men and women), which is comparable with the multiplier found in Pakistan descendants in the UK, but is higher than the UK multiplier for individuals of Indian and African Caribbean descent.32 In the Dutch and the combined “other ethnicities” subgroups we found comparable higher risks for CVD events in lower socioeconomic subgroups. The Surinamese subgroups ($76\%$ South Asian) showed higher observed risks for CVD, with highest OE ratios in lowest socioeconomic Surinamese. It has been hypothesized that higher CVD risk in individuals of low socioeconomic status or of South Asian decent is due to complex intermixing factors, e.g., (epi)genetical, stress and lifestyle differences.15,18,33, 34, 35, 36 These partly known complex intermixing factors probably contribute as to why traditional CVD risk factors combined with socioeconomic status or ethnicity do not account for an appropriate risk prediction.7,35, 36 Given that the observed risk distinctions in the socioeconomic and ethnic subgroups in our study go beyond the risk predicted by traditional risk factors and are as large as those encountered between western and eastern Europe populations, individualized absolute CVD risk adjustment based on both ethnicity and socioeconomic status within countries is warranted.
Overall, the men and women in our population showed a higher observed risk for CVD events compared to expected probability based on the SCORE2 models. Given the individuals in the study population were generally younger compared to the development cohort, this is an unexpected result. This study was performed in a contemporary population. Possibly this unexpected higher CVD risk might be due to the general increase of unhealthier lifestyles in the Netherlands, with for example an increase of obesity from $42\%$ in 2007 to $50\%$ in 2020.37 This study has several limitations. First, the missingness in our cohort in smoking, cholesterol and blood pressure is considerable. Even though the percentage of missingness was similar to similar cohorts (20–$50\%$ missing values versus e.g., 15–$60\%$ missing values across different variables in the QRISK derivation study), but still QRISK yields robust external validation results.21 We performed several sensitivity analyses in parts of the cohort with 20–$50\%$ missingness, which yielded similar results in calibration. We expect that the missingness in this CVD calibration has not affected our performance measures. Second, the study population was not large enough to analyse calibration in more ethnic subgroups. However, we were able to stratify the two largest ethnicity subgroups, and combined the other ethnic subgroups. Third, country of origin was the only available source as a proxy for ethnicity. Information on (self-perceived) ethnicity of the Surinamese subgroup would have been preferable and could have led to more specific results on individuals of South Asian or African Caribbean descendance.
Our study also has strengths. First this study is, to our knowledge, the first to assess the predictive performance of the SCORE2 CVD risk prediction model in a contemporary population stratified by socioeconomic status and ethnicity. The combination of routine health care data from GPs and hospitals, combined with Statistics *Netherlands data* is unique and the only database yet in the Netherlands where data from these 3 sources are combined for research on health inequities/inequalities. Furthermore, traditional cohorts usually have an underrepresentation of individuals with a low socioeconomic background or different ethnicities, whereas routine health care databases are inclusive of all individuals in countries within a universal health care system.
The developed 4 SCORE2 risk prediction models would probably fit the majority of the socioeconomic and ethnic diverse population of Europe. Currently, national CVD mortality data guides the choice for country wide implementation of one of the SCORE2 models. As a short-term future research direction, absolute CVD risk differences in national mortality data in (combined) socioeconomic and ethnic subgroups could be estimated. Based on these CVD risk differences, ESC/SCORE2 could consider to advise implementation of the SCORE2 low, moderate-, high- or very-high-risk models for specific socioeconomic and ethnic subgroups within countries.
In the long-term, the expanding development of population-based research cohorts in Europe, combined with traditional cohorts, could be beneficial for further finetuning risk prediction to existing health disparities in European populations. For population-based CVD prediction research we recommend including information on self-perceived ethnicity, and additional risk elevating predictors, such as body mass index, pregnancy related risk factors and family history of CVD.
Shifting toward applying higher CVD risk models within countries in high-risk subgroups, would in our population have led to an increase of the population meeting treatment thresholds from $10\%$ to $17\%$, and from $2\%$ to $5\%$, in men and women, respectively.
In the Netherlands, as a proxy for low socioeconomic status, eligibility for housing rental benefits or financial health care benefits could be used.
In conclusion, our study showed that the SCORE 2 CVD risk model for low-risk countries (as the Netherlands are) was found to underpredict CVD risk, particularly in low socioeconomic and Surinamese ethnic subgroups. Including socioeconomic status and ethnicity as predictors in CVD risk models and implementing CVD risk adjustment within countries is desirable for adequate CVD risk prediction and counseling.
## Contributors
J.M.K., M.E.N., R.H.H.G., J.N.S., R.C.V., P.G.v. P., A.T.A.M., H.M.M.V., E.D.B., H.J.A.v. O. and Y.W.J.S. contributed to the conception and design of the work. JM.K, M.E.N., J.N.S., A.T.A.M., H.M.M.V., E.D.B. and Y.W.J.S. contributed to the acquisition of the data. The data was analysed, and the underlying data was accessed and verified by J.M.K., M.A.F. and R.H.H.G. J.M.K. drafted the manuscript and figures. All authors contributed to the interpretation of the data analyses, and critically revised the manuscript. All authors read and approved the final version.
## Data sharing statement
The data used for this study is part of a larger study on health equity in the city of The Hague (ELAN Datawarehouse, sub study ELAN Vascular). Data used for this study is prohibited from sharing, although, requests for collaborations on the research project ELAN Vascular, can be addressed to the Health Campus of the Leiden University Medical Centre (https://healthcampusdenhaag.nl/). The research protocol, coding and data dictionary used for this study is available on request via the author ([email protected]).
## Declaration of interests
All authors declare no competing interests.
## Supplementary data
Appendix A–E
## References
1. 1European Cardiovascular Disease StatisticsEuropean Heart Network2017. (2017.0)
2. Salomaa V.. **Worrisome trends in the incidence of coronary artery disease events among young individuals**. *Eur J Prev Cardiol* (2020.0) **27** 1175-1177. PMID: 31941364
3. Roth G.A., Mensah G.A., Johnson C.O.. **Global burden of cardiovascular diseases and risk factors, 1990-2019: update from the GBD 2019 study**. *J Am Coll Cardiol* (2020.0) **76** 2982-3021. PMID: 33309175
4. **SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe**. *Eur Heart J* (2021.0) **42** 2439-2454. PMID: 34120177
5. Visseren F.L.J., Mach F., Smulders Y.M.. **2021 ESC guidelines on cardiovascular disease prevention in clinical practice: developed by the task force for cardiovascular disease prevention in clinical practice with representatives of the European Society of Cardiology and 12 medical societies with the special contribution of the European Association of Preventive Cardiology (EAPC)**. *Eur Heart J* (2021.0) **42** 3227-3337. PMID: 34458905
6. Marmot M.G., Adelstein A.M., Robinson N., Rose G.A.. **Changing social-class distribution of heart disease**. *Br Med J* (1978.0) **2** 1109-1112. PMID: 709255
7. Tunstall-Pedoe H., Woodward M.. **By neglecting deprivation, cardiovascular risk scoring will exacerbate social gradients in disease**. *Heart* (2006.0) **92** 307-310. PMID: 16166099
8. Stringhini S., Carmeli C., Jokela M.. **Socioeconomic status and the 25 × 25 risk factors as determinants of premature mortality: a multicohort study and meta-analysis of 1.7 million men and women**. *Lancet* (2017.0) **389** 1229-1237. PMID: 28159391
9. de Mestral C., Stringhini S.. **Socioeconomic status and cardiovascular disease: an update**. *Curr Cardiol Rep* (2017.0) **19** 115. PMID: 28965316
10. Mullachery P.H., Vela E., Cleries M.. **Inequalities by income in the prevalence of cardiovascular disease and its risk factors in the adult population of Catalonia**. *J Am Heart Assoc* (2022.0) **11**
11. Rafnsson S.B., Bhopal R.S., Agyemang C.. **Sizable variations in circulatory disease mortality by region and country of birth in six European countries**. *Eur J Public Health* (2013.0) **23** 594-605. PMID: 23478207
12. Rabanal K.S., Lindman A.S., Selmer R.M., Aamodt G.. **Ethnic differences in risk factors and total risk of cardiovascular disease based on the Norwegian CONOR study**. *Eur J Prev Cardiol* (2013.0) **20** 1013-1021. PMID: 22642981
13. Razieh C., Zaccardi F., Miksza J.. **Differences in the risk of cardiovascular disease across ethnic groups: UK Biobank observational study**. *Nutr Metab Cardiovasc Dis* (2022.0) **32** 2594-2602. PMID: 36064688
14. Kist J.M., Smit G.W.G., Mairuhu A.T.A.. **Large health disparities in cardiovascular death in men and women, by ethnicity and socioeconomic status in an urban based population cohort**. *eClinicalMedicine* (2021.0) **40** 101120. PMID: 34485880
15. Rabanal K.S., Meyer H.E., Tell G.S.. **Can traditional risk factors explain the higher risk of cardiovascular disease in South Asians compared to Europeans in Norway and New Zealand? Two cohort studies**. *BMJ Open* (2017.0) **7**
16. Perini W., Snijder M.B., Peters R.J.G., Kunst A.E.. **Ethnic disparities in estimated cardiovascular disease risk in Amsterdam, the Netherlands: the HELIUS study**. *Neth Heart J* (2018.0) **26** 252-262. PMID: 29644501
17. Rosenthal T., Touyz R.M., Oparil S.. **Migrating populations and health: risk factors for cardiovascular disease and metabolic syndrome**. *Curr Hypertens Rep* (2022.0) **24** 325-340. PMID: 35704140
18. Devareddy A., Sarraju A., Rodriguez F.. **Health disparities across the continuum of ASCVD risk**. *Curr Cardiol Rep* (2022.0) **24** 1129-1137. PMID: 35788894
19. Woodward M., Brindle P., Tunstall-Pedoe H.. **Adding social deprivation and family history to cardiovascular risk assessment: the ASSIGN score from the Scottish Heart Health Extended Cohort (SHHEC)**. *Heart* (2007.0) **93** 172-176. PMID: 17090561
20. Hippisley-Cox J., Coupland C., Vinogradova Y.. **Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2**. *BMJ* (2008.0) **336** 1475-1482. PMID: 18573856
21. Hippisley-Cox J., Coupland C., Brindle P.. **Development and validation of QRISK3 risk prediction algorithms to estimate future risk of cardiovascular disease: prospective cohort study**. *BMJ* (2017.0) **357** j2099. PMID: 28536104
22. Collins G.S., Reitsma J.B., Altman D.G., Moons K.G.. **Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement**. *BMJ* (2015.0) **350** g7594. PMID: 25569120
23. Bakker B.F.M., van Rooijen J., van Toor L.. **The system of social statistical datasets of Statistics Netherlands: an integral approach to the production of register-based social statistics**. *Stat J IAOS* (2014.0) **30** 411-424
24. 24Statistics Netherlands. Microdata.
25. Harteloh P., de Bruin K., Kardaun J.. **The reliability of cause-of-death coding in The Netherlands**. *Eur J Epidemiol* (2010.0) **25** 531-538. PMID: 20309611
26. Collins G.S., Ogundimu E.O., Altman D.G.. **Sample size considerations for the external validation of a multivariable prognostic model: a resampling study**. *Stat Med* (2016.0) **35** 214-226. PMID: 26553135
27. Buuren van S.. (2018.0)
28. Hippisley-Cox J., Coupland C., Vinogradova Y., Robson J., May M., Brindle P.. **Derivation and validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study**. *BMJ* (2007.0) **335** 136. PMID: 17615182
29. 29UK Office for National Statistics. Townsend score.
30. Schultz W.M., Kelli H.M., Lisko J.C.. **Socioeconomic status and cardiovascular outcomes**. *Circulation* (2018.0) **137** 2166-2178. PMID: 29760227
31. Kimenai D.M., Pirondini L., Gregson J.. **Socioeconomic deprivation: an important, largely unrecognized risk factor in primary prevention of cardiovascular disease**. *Circulation* (2022.0) **146** 240-248. PMID: 35748241
32. **Planbureau van de Leefomgeving. De stad verbeeld**. (2015.0)
33. Oudhof K.. (2008.0)
34. Javed Z., Haisum Maqsood M., Yahya T.. **Race, racism, and cardiovascular health: applying a social determinants of health framework to racial/ethnic disparities in cardiovascular disease**. *Circ Cardiovasc Qual Outcomes* (2022.0) **15**
35. Volgman A.S., Palaniappan L.S., Aggarwal N.T.. **Atherosclerotic cardiovascular disease in South Asians in the United States: epidemiology, risk factors, and treatments: a scientific statement from the American Heart Association**. *Circulation* (2018.0) **138** e1-e34. PMID: 29794080
36. Powell-Wiley T.M., Baumer Y., Baah F.O.. **Social determinants of cardiovascular disease**. *Circ Res* (2022.0) **130** 782-799. PMID: 35239404
37. **Statline**. (2022.0)
|
---
title: Mouse models of surgical and neuropathic pain produce distinct functional alterations
to prodynorphin expressing neurons in the prelimbic cortex
authors:
- Shudi Zhou
- Yuexi Yin
- Patrick L. Sheets
journal: Neurobiology of Pain
year: 2023
pmcid: PMC9971546
doi: 10.1016/j.ynpai.2023.100121
license: CC BY 4.0
---
# Mouse models of surgical and neuropathic pain produce distinct functional alterations to prodynorphin expressing neurons in the prelimbic cortex
## Highlights
•A subset of neurons in the prelimbic cortex (PL) express prodynorphin (Pdyn).•PLPdyn+ neurons consist of excitatory and inhibitory subtypes.•Paw incision induces transient hyperexcitability of excitatory PLPdyn+ neurons.•Nerve injury induces sustained hyperexcitability of excitatory PLPdyn+ neurons.
## Abstract
The medial prefrontal cortex (mPFC) consists of a heterogeneous population of neurons that respond to painful stimuli, and our understanding of how different pain models alter these specific mPFC cell types remains incomplete. A distinct subpopulation of mPFC neurons express prodynorphin (Pdyn+), the endogenous peptide agonist for kappa opioid receptors (KORs). Here, we used whole cell patch clamp for studying excitability changes to Pdyn expressing neurons in the prelimbic region of the mPFC (PLPdyn+ neurons) in mouse models of surgical and neuropathic pain. Our recordings revealed that PLPdyn+ neurons consist of both pyramidal and inhibitory cell types. We find that the plantar incision model (PIM) of surgical pain increases intrinsic excitability only in pyramidal PLPdyn+ neurons one day after incision. Following recovery from incision, excitability of pyramidal PLPdyn+ neurons did not differ between male PIM and sham mice, but was decreased in PIM female mice. Moreover, the excitability of inhibitory PLPdyn+ neurons was increased in male PIM mice, but was with no difference between female sham and PIM mice. In the spared nerve injury model (SNI), pyramidal PLPdyn+ neurons were hyperexcitable at both 3 days and 14 days after SNI. However, inhibitory PLPdyn+ neurons were hypoexcitable at 3 days but hyperexcitable at 14 days after SNI. Our findings suggest different subtypes of PLPdyn+ neurons manifest distinct alterations in the development of different pain modalities and are regulated by surgical pain in a sex-specific manner. Our study provides information on a specific neuronal population that is affected by surgical and neuropathic pain.
## Introduction
Neurons in the medial prefrontal cortex (mPFC) respond to noxious (i.e. painful) stimuli (Condes-Lara et al., 1989) and play an important role in the emotional valence, attentional components and the descending modulation of pain (Porro et al., 2002, Lee et al., 2015, Martinez et al., 2017, Dale et al., 2018, Zhou et al., 2018). The mPFC is comprised of a heterogeneous population of glutamatergic (i.e. excitatory) pyramidal neurons and GABAergic (i.e. inhibitory) neurons that receive, integrate, and relay synaptic information coming from both intracortical (local) and subcortical (long range) origins. Over a decade of work confirms epigenetic, morphological, functional, and circuit changes to mPFC neurons in multiple rodent models of pain (Cao et al., 2009, Metz et al., 2009, Alvarado et al., 2013, Blom et al., 2014, Alvarado et al., 2015, Cordeiro Matos et al., 2015, Zhang et al., 2015, Kelly et al., 2016, Kiritoshi et al., 2016, Cheriyan and Sheets, 2018, Shiers et al., 2018, Huang et al., 2019, Mitric et al., 2019, Cheriyan and Sheets, 2020, Jones and Sheets, 2020). However, these reported changes vary across distinct neuronal subtypes, laminar locations and substructures within the mPFC (Cheriyan and Sheets, 2018, Mitric et al., 2019, Jones and Sheets, 2020, Jefferson et al., 2021). This variability indicates that different forms of pain engage specific neuronal populations across anatomically distinct mPFC circuits.
One method for categorizing cortical neurons is by the expression of distinct neurotransmitters, which can provide insight into their function. The opioid neuropeptide dynorphin (Dyn) is the endogenous ligand for the kappa opioid receptor (KOR) (Chavkin et al., 1982), which is a receptor shown to be involved in mediating the negative affective component of pain. Prodynorphin (Pdyn), the precursor peptide to Dyn, is expressed in the prefrontal cortices of both humans and non-human primates (Khachaturian et al., 1985, Peckys and Hurd, 2001). In mice, a subset of cortical neurons express Pdyn and are mainly located in superficial laminar layers (i.e., laminar 2) including the mPFC (Pina et al., 2020). Increased expression of Pdyn mRNA is detected in the mPFC in both rat and mouse chronic pain models (Candeletti and Ferri, 1995, Palmisano et al., 2018), but is unchanged early (POD 4) after surgical incision (Nwaneshiudu et al., 2019). This indicates that chronic pain may be altering the activity of Pdyn expressing neurons in the mPFC (mPFCPdyn+ neurons). However, functional changes to mPFCPdyn+ neurons in surgical and chronic pain models yet to be identified. In this study, we used whole cell patch clamp electrophysiology in acute brain slices to study how the activity of Pdyn expressing neurons in the prelimbic cortex (PLPdyn+ neurons), a subregion of mPFC, changes in mouse models of both postoperative pain and neuropathic pain.
## Ethical approval and responsible use of animals
The Institutional Animal Use and Care Committee (IACUC) of the Indiana University School of Medicine approved (protocol# 19144) all procedures and experiments presented in this study. Experimenters maximized efforts to reduce animal use and animal suffering. Unfortunately, alternatives to ex-vivo and in-vivo techniques were not feasible for this project.
## Animals
To obtain transgenic mice expressing the red fluorescent protein TdTomato specifically in neurons that share the neuronal lineage marker prodynorphin (Pdyn), female Pdyn-IRES-Cre (B6;129S-Pdyntm1.1(cre)Mjkr/LowIJ; stock #027958; Jackson labs) mice were mated with male Ai14 (B6;129S6-Gt(ROSA)26SORtm14(CAG-tdTomato)Hze/J; stock #007914; Jackson Labs) mice. All experiments were conducted in Pdyn TdTomato offspring of both sexes in accordance with the animal care and use guidelines of Indiana University, the National Institutes of Health, and the Society for Neuroscience. All mice were housed in a temperature controlled (21 ± 2 °C) vivarium with a 12 h/12 h light/dark cycle (lights on at 7 AM). Mice (total $$n = 106$$) of the same sex were housed together before PIM, SNI or sham surgeries with unrestricted access to food and water. Animals were randomly assigned to each group. After surgery, mice were single housed. Experiments were performed when mice reached a minimum age of 8 weeks. Proper pain controls were conducted to minimize animal suffering.
## Intracranial virus injection
For intracranial injection, mice (postnatal day 42–49) were anesthetized with 1.5 % isoflurane in 100 % O2 at a flow rate of 1.0 L/min (SurgiVet Isotech 4, Smith). Artificial tears ointment (Rugby) was applied to the eyes after induction of anesthesia and a feedback controlled heating pad (FHC; Bowdoin, ME) connected with a thermometer was used to maintain body temperature at 37 °C. Mouse head was placed in a stereotaxic apparatus (900 series, Kopf Instruments). The top of the mouse head was shaved and was aseptically prepared by using 3 skin preparation protocols: povidone iodine (7.5 %, Purdue Products LP) and 70 % isopropyl alcohol wipes (Curity). The skull was exposed by incising the scalp using a No.11 blade (Royal-tek), and a Ram Power hand drill (MHC) was used to make a craniotomy for the placement of the injection needle. The coordinates for PL injection were (relative to bregma; in mm): 0.2 lateral, 1.7 rostral and 0.7 deep. Single injections (70 nL/injection; 75nL/min) of AAV1-EF1a-DIO-hChR2(E123A)-EYFP (Addgene, #35507-AAV1) were targeted to right PL using the UltraMicroPump controlling a Gastight 1701 Hamilton syringe paired with a beveled, 2 in., 27 gauge removable needle. After each injection, the Hamilton syringe was left in place for 7 min to prevent backflow before being removed slowly. For local anesthesia, lidocaine hydrochloride jelly USP (2 %, Akorn) was applied to the scalp by sterile cotton swabs before closing the incision with VetBond (3 M, St. Paul, MN). Subcutaneous injection of 5 mg/Kg Meloxicam (0.06 mg/Kg, Norbook) and buprenorphine hydrochloride (0.3 mg/mL 0.06 mg/Kg, PPR, NY) were applied to prevent inflammation and pain. Mice were monitored every day after the injection by a veterinarian from Laboratory Animal Research Center, Indiana University, School of Medicine, and were allowed to recover from the injection for at least 4 days before neuropathic pain model.
## Plantar incision model (PIM) of postoperative pain
Mice (≥8 weeks) were briefly (5–10 s) anesthetized with 1.5 % isoflurane in 100 % O2 at a flow rate of 1.0 L/min (SurgiVet Isotech 4). Anesthetized mice were placed on bite bar in a rebreathing anesthetic circuit with nose cone (Vetamac, Rossville, IN). Artificial tears ointment (Dechra, UK) was applied after induction of anesthesia and a feedback controlled heating pad (FHC) was used to maintain body temperature at 37 °C. The plantar surface of the left hind paw was aseptically prepared by using three skin preparation protocols: povidone iodine (7.5 %, Purdue Products LP) and 70 % isopropyl alcohol wipes (Curity). We implemented the plantar Incision model (PIM) adapted from previously described methods (Pogatzki and Raja, 2003, Cowie and Stucky, 2019). After aseptic preparation, a 5 mm longitudinal Incision was made using a No.11 scalpel blade through the skin and fascia on the plantar surface of the left hind paw, starting 2 mm from the proximal end of the heel and extending toward the toes. The underlying plantaris muscle was elevated using pointed tips tweezers, leaving the muscle intact. A 5 mm longitudinal Incision was made along the center of the exposed plantaris muscle. The skin Incision was closed with two horizontal simple interrupted sutures of 5–0 nylon (McKesson) on the proximal and distal end of the Incision covered with triple antibiotic ointment (Actvis Pharma, Inc). Mice were allowed to recover from anesthesia in their home cages with wet feed and on a heating pad (FHC) before being returned to the vivarium. Sham mice underwent anesthesia, antiseptic preparation and topical triple antibiotic ointment application, but without Incisions on the skin, fascia, or the muscles. Mice were checked every day, and any mouse with evidence of infection or dehiscence, or lose of sutures were excluded from the study. For pain recovery studies, nylon sutures were removed when mice were anesthetized after postoperative day 1 (POD1) behavior measurement.
## Spared nerve injury (SNI) model of neuropathic pain
For each group, the experimenter was blinded to the baseline von Frey results and randomly selected one mouse to be SNI and one to be sham. Mice were weighed and briefly anesthetized in an anesthesia box with 1.5–2.5 % isoflurane in 100 % O2 at a flow rate of 0.8–1.0 L/min. The snout of the mouse was then placed into a flexible nose cone connected to the isoflurane vaporizer allowing for continued anesthesia. Body temperature was maintained at 37 °C using a feedback controlled heating pad. The lateral surface of the left hind leg was shaved and disinfected using betadine and isopropyl alcohol. An approximately 4 mm incision through the skin was made and the underlying muscle layers were separated by blunt dissection using saline moistened sterile wooden dowels. The trifurcation of the left sciatic nerve was visualized. For SNI mice, an approximately 2 mm section of the tibial and common peroneal nerves distal to the trifurcation was removed, leaving the sural nerve intact. For sham mice, the trifurcation was exposed and visualized but not manipulated. The muscle layers of both SNI and sham mice were replaced and the outer skin layers were glued together using Vetbond. All mice recovered in a clean home cage with ad libitum water and wet feed on a heating pad for at least 30 min before being returned to the vivarium. Mice were monitored for 4 days post operation for signs of excessive pain such as reduced eating, drinking, activity, or grooming.
## Assessment of sensory pain behavior
The behavior testing apparatus was an elevated wire mesh platform fabricated by Sheets lab personnel. Mice were acclimated separately inside 6 in. tall, 3 in. diameter Plexiglas tubes for 1 h before baseline testing. On testing day, standard von Frey filaments were used for (Touch Test, VWR) the standard up-down (SUDO) method (Chaplan et al., 1994, Bonin et al., 2014) to assess the mechanical allodynia via 50 % paw withdrawal threshold (PWT). The baseline PWT was acquired the day before the plantar incision. For PIM mice, Von Frey filaments were presented to the area between the two sutures, 1 mm medial to the incision. The filament was applied perpendicular to the plantar aspect of the hind paw until it was slightly bent for approximately 5 s. Filaments were presented at intervals of at least 30 s. Positive responses were considered as sharp withdrawal of the hind paw, flinching, and licking upon the application of the filament. The test was initiated using the 2.00 g filament. If no response was observed, the next stiffer filament was applied until a positive response was evoked; however, if a positive response was observed, a less stiff fiber was applied. The test was terminated four trials after the first positive response per the SUDO method. For the pain recovery studies, von Frey testing was performed at POD1, POD4, and POD7. Starting from POD7, von Frey testing was performed daily until pain behavior for each mouse returned to baseline for two consecutive days. For SNI studies, pain assessments were performed at baseline (before SNI or sham surgery) and POD3 or POD14. Von Frey filaments were presented to the area innervated by the sural nerve of both hind paws and were measured via von Frey filaments to evaluate the development of mechanical allodynia. The experimenter was blinded to the surgical group of the mice throughout the experiment.
## Acute brain slice preparation
After the final behavioral test needed for a particular pain (i.e. PIM or SNI) or control group (i.e. sham), acute brain slices were prepared as described previously. Mice were briefly (∼15 s) anesthetized with 99.9 % isoflurane (Patterson Veterinary) and quickly decapitated. Brains were rapidly dissected and submerged in an ice cold choline solution containing the following (in mM): 110 choline chloride, 25 NaHCO3, 25 d-glucose, 11.6 sodium ascorbate, 7 MgSO4, 3.1 sodium pyruvate, 2.5 KCL, 1.25 NaH2PO4, and 0.5 CaCl2). Coronal brain slices containing mPFC (spine of the blade tilted rostrally 15° off vertical plane, 300 µm thick) were prepared with a vibratome (VT1200S; Leica). Slices were transferred to a 37 °C artificial CSF (ACSF) bath containing the following (in mM): 127 NaCl, 25 NaHCO3, 25 d-glucose, 2.5 KCl, 1MgCl2, 2 CaCl2, and NaH2PO4 for 30 min. Slices were subsequently incubated for at least 45 min in ACSF at room temperature before being transferred to the recording chamber.
## Electrophysiological recordings
Slices were placed in the recording chamber of a SliceScope Pro 6000 (Scientifica) and continuously perfused with the ACSF (30–32 °C) at the rate of ∼ 1 mL per minute. Slices were held in place with a slice anchor (Warner Instruments). Fluorescently labeled PLPdyn+ neurons were identified with a LED illumination system (CoolLED pE-4000) using a 580 nm wavelength LED with a RFP filter (ET FITC/RFP, Olympus). Recording pipets were made from borosilicate capillaries with filaments (G150-F; Warner Instruments) using a horizontal pipet puller (P-97; Sutter Instruments). PLPdyn+ neurons approximately 60 µm or deeper from the surface of the slice were targeted for recording. After establishing a Gigaohm (GΩ) seal between the pipette tip and the cell membrane gentle negative pressure was applied from inside the pipette to open the cell membrane. After the opening of the cell membrane, neurons were allowed to stabilize for 5 min before recording. For intrinsic whole cell recordings, the patch electrode (3–4 MΩ resistance) were filled with (in mM) 128 K-gluconate, 10 HEPES, 1 EGTA, 4 MgCl2, 4 ATP, 0.4 GTP, 10 phosphocreatine, 3 ascorbate plus 3.5 mg/ml biocytin (Sigma-Aldrich). Whole cell patch clamp recordings were performed at 30–32 °C, amplified and filtered at 4 kHz, and digitized at 10 kHz using a Multiclamp 700B amplifier (Molecular Devices). Membrane potential was held at −70 mV in a voltage clamp mode. No synaptic blockers were added during intrinsic electrophysiological recordings. In ex vivo optogenetics, CPP and NBQX (Tocris, Bristol, UK; 5 μM) were added to the ACSF to block glutamatergic (excitatory) inputs; SR 95531 hydrobromide (GABAzine, Tocris, Bristol, UK; 10 μM) was added to the ACSF to block GABAergic (inhibitory) inputs. Pipette capacitance was compensated and the inclusion of data required a series resistance < 35 MΩ. Current clamp recordings were bridge balanced. The experimenter was blinded to the surgical group of the mice throughout the recording.
## Morphological identification of neurons following electrophysiological recordings
After each recording, the patch pipette was slowly withdrawn to allow the cell membrane to reseal. Slices containing successfully resealed neurons were transferred back to the incubating chamber containing ACSF to allow for adequate filling of soma and dendrites with biocytin. Slices were first fixed in 4 % paraformaldehyde/phosphate buffered solution (PFA/PB) overnight, and then transferred to PB. Fixed brain slices were first washed 3 × 10 min in 0.1 % PBST individually before being placed in 1 mL blocking solution containing 0.3 % PBST and 5 % normal goat serum (EMD Millipore Corp, USA) on a rotator for 1 h at room temperature. Then slices were incubated with 1:1000 streptavidin, Alexa Fluor 488 conjugated green fluorescent dye (Invitrogen) for visualizing dendritic morphology for 1 h. Slices were then washed 3 × 10 min in 0.1 % PBST. After rinsing, samples were placed on cover slips and mounted on microscope slides with ProLongTM Gold antifade mountant (Invitrogen, United States). Brain sections were protected from light at every step.
## Tissue preparation and immunohistochemistry
Mice were anesthetized by intraperitoneal administration of a mixture of ketamine (87.5 mg/kg, Henry Schein, Dublin, OH) and xylazine (12.5 mg/kg, Akorn Animal Health, Lake Forest, IL). Then mice were perfused with 1X PBS followed by 4 % PFA/PBS. Brains were removed and post fixed in 4 % PFA/PBS for 24 h before being transferred to 30 % sucrose at 4 °C. Coronal sections (50 µm) were cut in the rostral to caudal direction using a vibratome (VT1200S; Leica). Free floating sections were washed in 0.1 % PBST (3 × 10 min) individually before being blocked with 1 mL block solution on a rotator for 1 h at room temperature. Then brain sections were washed in 0.1 % PBST (3 × 10 min). Slices were first incubated with primary antibody (chicken GAD$\frac{1}{67}$, Synaptic System, Germany, 1:500; rabbit Pdyn polyclonal antibody, ThermoFisher, PA5-22286, 1:500) in block solution overnight at 4 °C, then washed in 0.1 % PBST (3 × 10 min). Brain sections were next incubated with secondary antibody (Alexa FluorTM 488 goat anti-chicken IgG, 1:500; Alexa FluorTM 488 goat anti-rabbit IgG, 1:500, Invitrogen, United States) with block solution for 2 h at room temperature. After rinsing, slices were placed on cover slips and mounted on microscope slides with Fluoromount-GTM, with DAPI (Invitrogen, United States). Brain sections were protected from light at every step. Fluorescent images were captured by using All-in-One Fluorescence Microscope, BZ-X800 (Keyence, Itasca, IL) using 20 X air objective.
## Quantification and statistical analysis
Analysis of recording data was performed offline using Custom MATLAB (The MathWorks, RRID:SCR_001622). All of the statistical details of the experiments can be found in the figure legends. Behavior data for Sham/PIM POD 1 and sham/SNI were analyzed by ordinary-two-way ANOVA. Behavior data for Sham/PIM recovery were analyzed by repeated measurement (RM) two-way ANOVA (GraphPad Prism 8). For the intrinsic data comparison, unpaired Student’s t-test was used for normally distributed data, while Wilcoxon rank-sum test is used for non-parametric data. Differences were considered significant a $p \leq 0.05.$ For excitability comparison, data were analyzed by repeated measurement two-way ANOVA (GraphPad Prism 8). Unless otherwise noted, results are presented as mean ± SD. For data presented as boxplots, the box displays the central 50 % of the data with the central line indicating the median and the lower/upper boundary lines being the 25 %/75 % quantile of the data. The outliers are plotted individually using the '+' marker symbol. Cell counting for IHC was conducted manually.
## Prodynorphin expressing neurons in the prelimbic cortex are heterogeneous
*We* genetically labeled neurons that share the neuronal lineage marker prodynorphin (Pdyn) with the red fluorescent protein tdTomato by crossing Pdyn-Cre female mice with Ai14 floxed tdTomato reporter mice (Fig. 1A; see Methods). Confocal imaging of coronal slices showed that Pdyn+ tdTomato neurons are extensively expressed in laminae $\frac{2}{3}$ (L$\frac{2}{3}$) of the prelimbic (PL) cortex, a subregion with the mPFC, with sparser expression at deeper laminar layers (Fig. 1B). We used confocal microscopy to examine the morphology of L2 PLPdyn+ neurons following whole cell electrophysiological recordings. We found a majority of L2 PLPdyn+ neurons displayed apical dendrites characteristic of pyramidal (glutamatergic) neurons (Fig. 1C), while a minority displayed a smaller soma and lacked a clear apical dendrite indicative of inhibitory (GABAergic) neurons (Fig. 1D). Analysis of recording data showed that L2 PLPdyn+ neurons with a clear apical dendrite displayed both lower firing rates and smaller input resistance compared to those neurons lacking an apical dendrite (Fig. 1E-L, Table 1). L2 PLPdyn+ neurons without a clear apical dendrite exhibit significant fast afterhyperpolarization (Fig. 1E, I, insets). Most notable was that pyramidal L2 PLPdyn+ neurons displayed a significantly larger membrane capacitance (Fig. 1H, L), which correlates with larger size. Neurons with uncertain morphology or poor imaging were excluded from electrophysiological analyses. Combining data from the morphological and electrophysiological analyses for both male and female mice, 186 (55 %) of recorded L2 PLPdyn+ neurons in this study were pyramidal neurons, while 151 (45 %) were inhibitory neurons. We recognize that this recording data is not consistent with our IHC data (24 % PLPdyn+ neurons express GAD1/GAD67). This discrepancy is likely due to recorded neurons that could not be morphologically identified being excluded from analyses. Fig. 1Prodynorphin-expressing neurons in the prelimbic cortex are heterogeneous. A. Schematic showing the generation of Pdyn-cre-tdTomato mice by crossing Pdyn-cre females with Ai14-tdTomato reporter males. B. Representative confocal image (10 X) of a coronal slice showing that Pdyn+ neurons are predominately expressed at L2 of the prelimbic (PL) cortex (bottom left: orientation of the brain slice and location of the PL cortex; D = dorsal; V = ventral). C. Representative confocal image (20 X) of pyramidal PLPdyn+ neurons with typical apical dendrite structure (white arrows). D. Representative confocal image (20 X) of an inhibitory PLPdyn+ neuron (lower) and a pyramidal PLPdyn+ neuron (upper). E-L. Electrophysiological comparison of pyramidal (black) and inhibitory (red) L2 PLPdyn+ neurons recorded from male (E-H) and female (I-L) mice. E. Representative traces showing action potentials of a pyramidal PLPdyn+ neuron (black) and an inhibitory PLPdyn+ neuron (red) from sham POD1 male mice. Inset, afterhyperpolarization comparison between pyramidal and inhibitory PLPdyn+ neuron. ( scale bar, horizontal 100 ms, vertical 100 mV). F. The relationship of AP firing with current step amplitude from pyramidal ($$n = 12$$ from 9 slices in 5 mice) and inhibitory PLPdyn+ neuron ($$n = 8$$ from 4 slices in 3 mice) in sham POD1 male mice (RM two-way ANOVA, Fsurgery [1, 18] = 58.83, *** $P \leq 0.001$, Fcurrent [12,216] = 252.8, $P \leq 0.001$, and Finteraction [12,216] = 29.42, *** $P \leq 0.001$; Bonferroni post hoc, *** $P \leq 0.001$). Error bars = S.E.M.. G. Pyramidal and inhibitory PLPdyn+ neurons are significant different in input resistance in sham POD1 male mice (t[18] = -5.507, *** $P \leq 0.001$; Error bars = S.E.M.; Student’s t-test). H. Pyramidal PLPdyn+ neurons have significant higher membrane capacitance than inhibitory PLPdyn+ neurons in sham POD1 male mice (t[18] = 5.09, *** $P \leq 0.001$; Student’s t-test). I. Representative traces showing action potentials of a pyramidal PLPdyn+ neuron (black) and an inhibitory PLPdyn+ neuron (red) from sham POD1 female mice. Inset, afterhyperpolarization comparison between pyramidal and inhibitory PLPdyn+ neuron. ( scale bar, horizontal 100 ms, vertical 100 mV). J. The relationship of AP firing with current step amplitude from pyramidal ($$n = 15$$ from 10 slices in 6 mice) and inhibitory PLPdyn+ neuron ($$n = 8$$ from 6 slices in 4 mice) in sham POD1 female mice (RM two-way ANOVA, Fsurgery [1, 21] = 9.936, ** $$P \leq 0.005$$, Fcurrent [12,252] = 136.3, *** $P \leq 0.001$, and Finteraction [12,252] = 9.470, *** $P \leq 0.001$; Bonferroni post hoc, *** $P \leq 0.001$). Error bars = S.E.M. K. Pyramidal and inhibitory PLPdyn3+ neurons are significant different in input resistance in sham POD1 female mice (t[21] = -2.2928, * $$P \leq 0.03$$; Error bars = S.E.M.; Student’s t-test). L. Pyramidal PLPdyn+ neurons have significant higher membrane capacitance than inhibitory PLPdyn+ neurons in sham POD1 female mice (t[21] = 3.13, ** $$P \leq 0.005$$; Student’s t-test). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Table 1Comparison of pyramidal and inhibitory PLPdyn+ neurons from male and female sham mice, related to Fig. 1.PLPdyn+ neuronsMaleFemalePYR ($$n = 12$$ neurons; 5 mice)INs ($$n = 9$$ neurons; 4 mice)PYR ($$n = 15$$ neurons; 6 mice)INs ($$n = 8$$ neurons; 4 mice)Subthreshold propertiesResting potential (mV)−86.3 ± 3.81−78.1155 ± 9.4809 #*−87.7 ± 3.93−86.7 ± 2.57Input resistance (mΩ)100 ± 20.42156 ± 32#***89.8 ± 3.1111 ± 11.8#*Firing propertiesThreshold (mV)−39.6 ± 3.19−37.127 ± 4.4839.4 ± 3.51−40.1 ± 4.53Threshold (pA)300 [1 0 0]⊺200 [50]⊺$***350 [50]⊺275 [1 2 5]⊺Frequency/current (Hz/pA)0.1 ± 0.0260.14 ± 0.021$**0.1 ± 0.0160.13 ± 0.021#***APs @400 pA8.5 ± 3.5317.44 ± 3.32#***8.4 ± 3.3614.25 ± 4.74Height (mV)73.5 ± 8.863.73 ± 15.08#*73 ± 9.0374 ± 10.45#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).
It has been shown that Pdyn expression changes during rodent development (Alvarez-Bolado et al., 1990). We therefore stained a subset of brain slices from Pdyn+ tdTomato mice with an anti-Pdyn antibody. We find that approximately 84.3 % of the neurons labeled with tdTomato in the PL cortex are co-labeled with Pdyn antibody (Fig. 2A). Based on this finding, we cannot exclude the possibility that a small percentage of Pdyn TdTomato neurons recorded were not actively expressing Pdyn. Previous studies indicate that Pdyn+ neurons in the cortex are heterogeneous, consisting of both excitatory pyramidal and inhibitory neurons (Sohn et al., 2014, Loh et al., 2017, Smith et al., 2019). Immunohistochemistry showed that 24 % of the PLPdyn+ neurons express glutamate decarboxylase 1 (GAD1/GAD67), which is a molecular marker for GABAergic neurons (Fig. 2B). To test the nature of local synaptic outputs from PLPdyn+ neurons, we injected AAV1-EF1a-DIO-hChR2(E123A)-EYFP into the PL cortex of Pdyn-Cre-tdTomato mice (Fig. 2C), allowing for optogenetic control of PLPdyn+ neurons. Given the difficulty of isolating AAV infection in only L2, these experiments consisted of testing local connections from PLPdyn+ neurons in both L2 and L3 (L$\frac{2}{3}$). After adequate time for hChR2(E123A) expression in L$\frac{2}{3}$ PLPdyn+ neurons (∼14 days), we prepared acute coronal brain slices and recorded from unlabeled L5 neurons in the PL cortex (Fig. 2D, E). We detected both excitatory (inward) and inhibitory (outward) currents following broad field optogenetic activation of hChR2(E123A)-expressing PLPdyn+ neurons (Fig. 2F-H). We found that a small subpopulation of L5 cortical neurons only receive inhibitory inputs from L$\frac{2}{3}$ PLPdyn+ neurons, but not excitatory inputs (Fig. 2G-H). Addition of AMPA receptor antagonist NBQX (5 μM) and NMDA receptor antagonist CPP (5 μM) eliminated evoked inward excitatory postsynaptic currents (Fig. 2G) but not outward inhibitory postsynaptic currents (Fig. 2H), which were could be blocked using the GABA-A receptor antagonist GABAzine (10 μM; Fig. 2H). We did observe a reduction in outward inhibitory postsynaptic currents after treatment with NBQX and CPP (Fig. 2H), which indicates that glutamatergic L$\frac{2}{3}$ PLPdyn+ neurons can drive disynaptic feedforward inhibition of L5 PL neurons. These findings indicate that L$\frac{2}{3}$ PLPdyn+ neurons consist of both glutamatergic and GABAergic phenotypes that have direct local connections onto L5 PL neurons. Fig. 2PLPdyn+ neurons send local excitatory and inhibitory inputs to layer 5 (L5) cortical neurons. A. Confocal images showing Pdyn tdtomato neurons (left), anti-Pdyn (middle), and overlay (right). Quantification from 3 Pdyn tdTomato male mice, 3 slices in totals. B. Confocal images showing Pdyn tdtomato neurons (left), GAD$\frac{1}{67}$ (middle), and overlay (right). Quantification from 3 Pdyn tdTomato male mice, 3 slices in totals. C. Injection of AAV-DIO-ChR2-YFP into the PL of Pdyn-cre-tdTomato mice. D. Schematic showing electrophysiological recording from a L5 unlabeled cortical neuron in PL cortex while photoactivating ChR2 labeled Pdyn+ neuronal terminals. E. Representative coronal brain sections (Left: bright field; Middle: 580 nm; Right: 470 nm). F. Examples of evoked EPSC (black trace; holding at −60 mV) and evoked IPSC (red trace; holding at + 10 mV) recorded during wide-field activation of ChR2 Pdyn+ terminals via blue LED stimulation before (left) and after (right) adding synaptic blockers (CPP and NBQX). Left inset, schematic showing potential sources of local synaptic input to L5 PL unlabeled cortical neurons during photoactivation. Right, examples of evoked EPSC (black trace; holding at −60 mV) and evoked IPSC (red trace; holding at + 10 mV) recorded in the presence of glutamatergic blockers (solid black + red, CPP 5 µM, NBQX 5 µM) and GABAergic blockers (transparent black + red, GABAzine 10 µM). Right inset, schematic showing potential synaptic inputs to L5 PL cortical neurons during photoactivation while blocking excitatory and inhibitory connections. G, H. Excitatory (G) and inhibitory (H) input values from L$\frac{2}{3}$ Pdyn+ neurons to L5 neurons following treatment with glutamatergic blockers (CPP 5 µM, NBQX 5 µM) and GABAergic blockers (GABAzine 10 µM). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
## The intrinsic firing of pyramidal L2 PLPdyn+ neurons increases early after surgical incision
Our next goal was to determine whether surgical incision acutely alters the excitability L2 PLPdyn+ neurons. For this, we applied the plantar incision model (PIM) of postoperative pain following previously established methods (Cowie and Stucky, 2019) (Fig. 3A). We measured mechanical paw withdrawal threshold (PWT) prior to surgery (i.e. baseline) and changes to PWT were measured on postoperative day 1 (POD1). Both male and female mice in the PIM group displayed significant mechanical allodynia on POD1 compared to sham controls (Fig. 3B, C). Following assessment of pain behavior on POD1, we prepared acute brain slices from PIM and sham mice for whole cell electrophysiological recordings of L2 PLPdyn+ neurons. Analyses of recording data from the different neuronal subtypes (i.e. pyramidal and inhibitory) were based on the morphology and capacitance described in Fig. 1. Plantar incision significantly increased the excitability of pyramidal L2 PLPdyn+ neurons in both male and female mice at POD1 (Fig. 3D-O). This increase in excitability manifested as increased action potential (AP) firing in response to depolarizing current steps in both males (Fig. 3D, E) and females (Fig. 3J, K). Further analysis revealed that plantar incision significantly increased input resistance of pyramidal L2 PLPdyn+ neurons from male mice, but not did alter resting membrane potential nor current and voltage threshold for AP firing (Fig. 3F-I). In female mice, plantar incision decreased the current threshold (Fig. 3L, Table 2) and depolarized the resting membrane potential (Fig. 3O, Table 2) for AP firing in pyramidal L2 PLPdyn+ neurons without altering the voltage threshold (Fig. 3M, Table 2) or input resistance (Fig. 3O, Table 2). These data suggest that the mechanisms by which surgical incision increass AP firing in pyramidal L2 PLPdyn+ neurons differ between male and female mice. Interestingly, we found that plantar incision did not alter membrane nor excitable properties of inhibitory L2 PLPdyn+ neurons at POD1 (Fig. 4, Table 3).Fig. 3The intrinsic firing of pyramidal L2 PLPdyn+ neurons is increased early after surgical incision. A. Schematic for plantar incision model of postoperative pain. B. Plantar incision induced mechanical allodynia in male mice (sham = 11; PIM = 10, ordinary-two-way ANOVA, Fsurgery [1,19] = 14.41, P = ** 0.001, Ftime [1, 19] = 77.16, *** $P \leq 0.001$, and Finteraction [1, 19] = 39.15, *** $P \leq 0.001$); Sidak’s post hoc, *** $P \leq 0.001$; Error bars = S.E.M.). C. Plantar incision induced mechanical allodynia in female mice (sham = 9, PIM = 10; ordinary-two-way ANOVA, Fsurgery [1,17] = 26.08, *** $P \leq 0.001$, Ftime [1, 17] = 37.81, *** $P \leq 0.001$, and Finteraction [1, 17] = 25.78, *** $P \leq 0.001$; Sidak’s post hoc, *** $P \leq 0.001$; Error bars = S.E.M.). D. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham POD1 (black) and PIM POD1 (red) male mice. E. The relationship of AP firing with current step amplitude from sham ($$n = 12$$ from 9 slices in 5 mice) and PIM ($$n = 14$$ from 8 slices in 5 mice) POD1 male mice (RM two-way ANOVA, Fsurgery [1, 24] = 5.023, * $$P \leq 0.035$$, Fcurrent [12,288] = 114.6, *** $P \leq 0.001$, and Finteraction [12,288] = 3.866, *** $P \leq 0.001$; Bonferroni post hoc, * $P \leq 0.05$, ** $P \leq 0.01$; Error bars = S.E.M.). F-H. Plantar incision did not alter (F) current threshold (t[24] = 1.4768, $$P \leq 0.15269$$), (G) voltage threshold, (t[24] = -1.466, $$P \leq 0.15564$$) or (H) resting membrane potential, (t[24] = -1.3605, $$P \leq 0.17948$$) of pyramidal L2 PLPdyn+ neurons (Student’s t-test). Error bars = S.E.M.. I. Plantar incision significantly increased input resistance (t[24] = -2.0917, *$$P \leq 0.047228$$; Student’s t-test; Error bars = S.E.M.) of pyramidal L2 PLPdyn+ neurons in male mice. J. representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham POD1 (black) and PIM POD1 (red) female mice. K. The relationship of AP firing with current step amplitude from sham ($$n = 15$$ from 10 slices in 6 mice) and PIM ($$n = 12$$ from 9 slices in 7 mice) POD1 female mice (RM two-way ANOVA, Fsurgery [1,25] = 9.057, ** $$P \leq 0.006$$, Fcurrent [12,300] = 141.7, *** $P \leq 0.001$, and Finteraction [12,300] = 7.217, *** $P \leq 0.001$; Bonferroni post hoc, **$P \leq 0.01$, ***$P \leq 0.001$; Error bars = S.E.M.). L. Plantar incision significantly decreased current threshold (t[25] = 2.5663, *$$P \leq 0.016654$$; Student’s t-test;;Error bars = S.E.M.). M. Plantar incision did not alter voltage threshold for AP firing (t[25] = -1.6148, $$P \leq 0.1190$$; Error bars = S.E.M.). N. Plantar incision significantly increased the resting membrane potential (t[25] = -2.1487, *$$P \leq 0.041538$$; Error bars = S.E.M.). O. Plantar incision did not alter input resistance (t[25] = -1.94, $$P \leq 0.$$ 063734; Error bars = S.E.M.) of pyramidal L2 PLPdyn+ neurons in female mice (Student’s t-test). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Table 2Comparison of pyramidal PLPdyn+ neurons from sham and PIM POD1 male and female mice, related to Fig. 3.PLPdyn+ neuronsMaleFemaleSham ($$n = 12$$ neurons; 5 mice)PIM ($$n = 14$$ neurons; 5 mice)Sham ($$n = 15$$ neurons; 6 mice)PIM ($$n = 12$$ neurons; 7 mice)Subthreshold propertiesResting potential (mV)−86.3 ± 3.81−83.6 ± 5.82−87.7 ± 3.93−84.3 ± 4.15#*Input resistance (mΩ)100 ± 20.42126 ± 37.53#*89.8 ± 3.1109.5 ± 37.14Firing propertiesThreshold (mV)−39.6 ± 3.19−37.6 ± 3.1939.4 ± 3.51−36.2 ± 6.58Threshold (pA)300 [1 0 0]⊺225 [1 5 0]⊺350 [50]⊺250 [1 0 0]⊺#*Frequency/current (Hz/pA)0.1 ± 0.0260.12 ± 0.021$*0.1 ± 0.0160.12 ± 0.026APs @400 pA8.5 ± 3.5312.4 ± 6.168.4 ± 3.3610.6 ± 4.94#*Height (mV)73.5 ± 8.871.7 ± 6.2673 ± 9.0368 ± 8.99#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).Fig. 4Excitability of inhibitory L2 PLPdyn+ neurons remained unchanged one day after plantar incision. A. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and PIM (red) male mice. B. The relationship of AP firing with current step amplitude from sham ($$n = 8$$ from 4 slices in 3 mice) and PIM ($$n = 6$$ from 4 slices in 4 mice) POD1 male mice (RM two-way ANOVA, Fsurgery [1,12] = 0.2866, $$P \leq 0.602$$, Fcurrent [12,144] = 151.7, $P \leq 0.001$, Finteraction [12,144] = 0.3458, $$P \leq 0.979$$; Error bars = S.E.M.). C-F. Plantar incision produced no significant changes to (C) current threshold (Z[11] = 50, $$P \leq 1$$; Wilcoxon rank-sum test), (D) voltage threshold (t[11] = 0.86204, $$P \leq 0.40707$$; Student’s t test), (E) resting membrane potential (t[11] = 0.56753, $$P \leq 0.48175$$; Student’s t-test), or (F) input resistance (t[11] = 0.38175, $$P \leq 0.70991$$; Student’s t-test) of inhibitory L2 PLPdyn+ neurons from male mice. Error bars = S.E.M.. G. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and PIM (red) female mice. H. The relationship of AP firing with current step amplitude from sham ($$n = 8$$ from 6 slices in 4 mice) and PIM ($$n = 8$$ from 5 slices in 4 mice) POD1 (RM two-way ANOVA, Fsurgery [1,14] = 0.4498, $$P \leq 0.513$$; Fcurrent [12,168] = 149.2, $P \leq 0.001$; Finteraction [12,168] = 0.4858, $$P \leq 0.921$$). I-L. Plantar incision produced no significant changes to (I) current threshold (t[14] = 1.0491, $$P \leq 0.3119$$; Students t-test), (J) voltage threshold (t[14] = 0.49749, $$P \leq 0.62656$$; Student’s t-test), (K) resting membrane potential (t[14] = -0.64363, $$P \leq 0.53022$$; Student’s t-test), or (L) input resistance (t[14] = 0.98769, $$P \leq 0.34007$$; Student’s t-test) of inhibitory L2 PLPdyn+ neurons from female mice. Error bars = S.E.M.. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Table 3Comparison of inhibitory PLPdyn+ neurons from sham and PIM POD 1 male and female mice, related to Fig. 4.PLPdyn+ neuronsMaleFemaleSham ($$n = 8$$ neurons; 3 mice)PIM ($$n = 6$$ neurons; 4 mice)Sham ($$n = 8$$ neurons; 4 mice)PIM ($$n = 8$$ neurons; 4 mice)Subthreshold propertiesResting potential (mV)−77.8 ± 10.02 #*−79.3 ± 7.22−86.7 ± 2.57−85.7 ± 3.97Input resistance (mΩ)161 ± 29.19#***155 ± 58.22111 ± 11.8#*120 ± 29.81Firing propertiesThreshold (mV)−36.8 ± 4.65−39 ± 4.58−40.1 ± 4.53−41.3 ± 5.59Threshold (pA)200 [50]⊺$***175 [1 0 0]⊺275 [1 2 5]⊺250 [75]⊺Frequency/current (Hz/pA)0.14 ± 0.022$**0.16 ± 0.06780.13 ± 0.021#***0.13 ± 0.0095APs @400 pA18.1 ± 2.8#***18.8 ± 5.9814.25 ± 4.7412.9 ± 3.27Height (mV)60.6 ± 12.73#*72 ± 11.6174 ± 10.4575.2 ± 16.3#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).
## Sex-specific changes to the excitability of L2 PLPdyn+ neuronal subtypes are detected after the recovery from plantar incision
We next determined whether the initial hyperexcitation of pyramidal L2 PLPdyn+ neurons was either attenuated or maintained following the recovery from plantar incision. For this, we removed sutures from the paw following behavior measurement at POD1 (see Methods). Behavioral testing showed that PWT returned to baseline values approximately 2 weeks after plantar incision in male (Fig. 5A) and female (Fig. 6A) mice, which is consistent with previous studies (Pogatzki and Raja, 2003, Cowie and Stucky, 2019). Following the recovery, data analyses revealed no statistical differences in the excitable properties of pyramidal L2 PLPdyn+ neurons between PIM and sham male mice (Fig. 5B, C; Table 4). Furthermore, no significant difference was detected in the excitability of pyramidal PLPdyn+ neurons from sham POD 1 and recovery mice. However, inhibitory L2 PLPdyn+ neurons from male PIM mice displayed a significantly increased excitability following the recovery (Fig. 5D, E; Table 4). Further analyses revealed that voltage threshold for AP firing in inhibitory L2 PLPdyn+ neurons from male PIM recovery mice was significantly decreased, but no significant changes were detected in the current threshold, resting membrane potential or input resistance (Table 4). Interestingly, pyramidal L2 PLPdyn+ neurons recorded from the female PIM recovery group displayed a significant decrease in the excitability (Fig. 6B, C) with a depolarized RMP and decreased input resistance (Table 5). No statistical differences were detected in the inhibitory L2 PLPdyn+ neurons compared to the female sham recovery group (Fig. 6D, E; Table 5).Fig. 5Excitability of inhibitory L2 PLPdyn+ neurons was increased when postoperative pain behavior subsided in male mice. A. In mice with plantar incision, mean 50 % PTW gradually increased one day after PIM in male. ( PIM = 5, sham = 5; RM two-way ANOVA, Fsurgery [1,8] = 27.70, *** $P \leq 0.001$, Ftime [9, 72] = 3.147, ** $$P \leq 0.003$$, and Finteraction [9, 72] = 3.186, ** $$P \leq 0.03$$); Dunnett post hoc, * $P \leq 0.05$ ** $P \leq 0.01$ *** $P \leq 0.001$; Error bars = S.E.M.). B. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham (black) and PIM (red) recovery male mice. C. The relationship of AP firing with current step amplitude of pyramidal L2 PLPdyn+ neurons from sham ($$n = 15$$ from 7 slices in 5 mice) and PIM ($$n = 12$$ from 10 slices in 5 mice) recovery male mice (RM two-way ANOVA, Fsurgery [1, 25] = 0.4025, $$P \leq 0.532$$, Fcurrent [12,300] = 102.4, *** $P \leq 0.001$, Finteraction [12,300] = 0.5142, $$P \leq 0.905$$); Error bars = S.E.M.). D. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and PIM (red) recovery male mice. E. The relationship of AP firing with current step amplitude of inhibitory L2 PLPdyn+ neurons from sham ($$n = 12$$ from 9 slices in 5 mice) and PIM ($$n = 8$$ from 6 slices in 5 mice) recovery male mice (RM two-way ANOVA, Fsurgery [1, 18] = 4.756, * $$P \leq 0.043$$, Fcurrent [12,216] = 71.79, *** $P \leq 0.001$, Finteraction [12,216] = 3.374, $P \leq 0.001$); Bonferroni post hoc, * $P \leq 0.05$; Error bars = S.E.M.). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 6Excitability of pyramidal PLPdyn+ neurons was decreased when postoperative pain behavior subsided in female mice. A. In mice with plantar incision, mean 50 % PTW gradually increased one day after PIM in female. ( PIM = 5, sham = 5; RM two-way ANOVA, Fsurgery [1,8] = 89.07, *** $P \leq 0.001$, Ftime [12, 96] = 4.037, ** $P \leq 0.01$, and Finteraction [12, 96] = 1.514, $$P \leq 0.132$$; Dunnett post hoc, * $P \leq 0.05$, ** $P \leq 0.01$, *** $P \leq 0.001$; Error bars = S.E.M.). B. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham (black) and PIM (red) recovery female mice. C. The relationship of AP firing with current step amplitude of pyramidal L2 PLPdyn+ neurons recorded from sham ($$n = 10$$ from 7 slices in 5 mice) and PIM ($$n = 13$$ from 8 slices in 5 mice) recovery female mice (RM two-way ANOVA, Fsurgery [1, 21] = 4.333, * $$P \leq 0.05$$, Fcurrent [12,252] = 121.1, *** $P \leq 0.001$, Finteraction [12,252] = 3.67, *** $P \leq 0.001$); Bonferroni post hoc, * $P \leq 0.05$, ** $P \leq 0.01$; Error bars = S.E.M.). D. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and PIM (red) recovery female mice. E. The relationship of AP firing with current step amplitude of inhibitory L2 PLPdyn+ neurons from sham ($$n = 12$$ from 6 slices in 4 mice) and PIM ($$n = 7$$ from 6 slices in 5 mice) recovery female mice (RM two-way ANOVA, Fsurgery [1, 17] = 0.002754, $$P \leq 0.959$$, Fcurrent [12,204] = 102.3, *** $P \leq 0.001$, Finteraction [12,204] = 0.553, $$P \leq 0.877$$); Error bars = S.E.M.). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Table 4Comparison of pyramidal and inhibitory PLPdyn+ neurons from sham and PIM recovery male mice, related to Fig. 5.PLPdyn+ neuronsPYRINsSham Recovery ($$n = 15$$ neurons; 5 mice)PIM Recovery ($$n = 12$$ neurons; 5 mice)Sham Recovery ($$n = 12$$ neurons; 5 mice)PIM Recovery ($$n = 8$$ neurons; 5 mice)Subthreshold propertiesResting potential (mV)−87.4 ± 5.96−85.4 ± 2.97−78.8 ± 5.77−82.6 ± 7.8Input resistance (mΩ)95.8 ± 25.4590.5 ± 19.17$116.5 ± 21.74128.7 ± 52.22Firing propertiesThreshold (mV)−42.6 ± 5.52−40 ± 2.27$−34.5 ± 6.96−45.8 ± 6.93#**Threshold (pA)300 [1 0 0]⊺300 [1 0 0]⊺$250 [75]⊺200 [1 2 5]⊺Frequency/current (Hz/pA)0.11 ± 0.0210.11 ± 0.0210.12 ± 0.010.19 ± 0.1$**APs @400 pA8.7 ± 4.768.4 ± 4.9111.8 ± 3.2219.25 ± 9Height (mV)75.1 ± 6.5872 ± 6.6567.9 ± 11.1781.4 ± 13.78$*#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).Table 5Comparison of pyramidal and inhibitory PLPdyn+ neurons from sham and PIM recovery female mice, related to Fig. 6.PLPdyn+ neuronsPYRINsSham Recovery ($$n = 10$$ neurons; 5 mice)PIM Recovery ($$n = 13$$ neurons; 5 mice)Sham Recovery ($$n = 12$$ neurons; 4 mice)PIM Recovery ($$n = 7$$ neurons; 5 mice)Subthreshold propertiesResting potential (mV)−83.9 ± 2.77−80.9 ± 3.21#*−79.2 ± 3.28−77.1 ± 5.77Input resistance (mΩ)102.26 ± 19.0886 ± 15.95#*143.7 ± 41.36137.3 ± 28.38Firing propertiesThreshold (mV)−35.7 ± 3.26−35.7 ± 3.66−33.8 ± 4.55−35 ± 3.86Threshold (pA)300 [1 5 0]⊺350 [1 2 5]⊺200 [1 5 0]⊺200 [50]⊺Frequency/current (Hz/pA)0.1 ± 0.0170.098 ± 0.0160.13 ± 0.0310.12 ± 0.019APs @400 pA8.3 ± 3.775.1 ± 3.2#*13.5 ± 1.914.3 ± 1.5Height (mV)73.4 ± 9.9276.7 ± 9.3566.6 ± 12.3973.7 ± 12.59#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).
## Nerve injury induces a prolonged hyperexcitation of pyramidal L2 PLPdyn+ neurons
We next applied the spared nerve injury (SNI) model of neuropathic pain, which allowed us to record L2 PLPdyn+ neurons at early (POD3-4) and later (POD14-15) stages of pain development. Male mice displayed significant mechanical allodynia at both POD3 (Fig. 7A) and POD14 (Fig. 7F). At POD3-4, the excitability of pyramidal L2 PLPdyn+ neurons in male SNI was significantly enhanced (Fig. 7B, C) consistent with our findings from POD1 PIM male mice. Further analyses revealed a significant decrease in the current threshold for AP firing, but no difference in voltage threshold for AP firing, RMP, or input resistance (Table 6). Conversely, the excitability of inhibitory L2 PLPdyn+ neurons significantly decreased at the early stages of SNI (Fig. 7D, E), including an increased current threshold for AP firing (Table 6). At POD14-15, the excitability of pyramidal PLPdyn+ neurons remained significantly increased in male SNI mice (Fig. 7G, H, Table 7). However, in contrast to recordings from POD3-4, inhibitory PLPdyn+ neurons recorded from POD14-15 SNI male mice displayed significantly more AP firing in response to depolarizing current steps together with an increased input resistance (Fig. 7I, J, Table 7). Further analyses showed that there was no statistical difference in the excitability of inhibitory PLPdyn+ neurons from the sham group between POD3 and POD14 (POD3, $$n = 17$$, POD14, $$n = 17$$; Two-way ANOVA, Ftime = 0.02851, $$P \leq 0.867$$; Finteraction = 0.2733, $$P \leq 0.993$$).Fig. 7Enhanced excitability of pyramidal L2 PLPdyn+ neurons in the spared nerve injury (SNI) model of neuropathic pain in male mice. A. SNI induced mechanical allodynia in male mice at POD3 (SNI = 5, sham = 5; ordinary-two-way ANOVA, Fsurgery [1,8] = 10.38, * $$P \leq 0.0122$$, Ftime [1, 8] = 33.46, *** $$P \leq 0.0004$$, and Finteraction [1, 8] = 6.538, * $$P \leq 0.0338$$; Sidak’s post hoc, *** $P \leq 0.001$; Error bars = S.E.M.). B. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham (black) and SNI (red) POD3 male mice. C. The relationship of AP firing with current step amplitude of pyramidal L2 PLPdyn+ neurons from sham ($$n = 6$$ from 3 slices in 2 mice) and SNI ($$n = 10$$ from 8 slices in 4 mice) POD3 male mice (RM two-way ANOVA, Fsurgery [1, 14] = 4.795, *$$P \leq 0.46$$, Fcurrent [12,168] = 94.29, *** $P \leq 0.001$, Finteraction [12,168] = 4.472, *** $P \leq 0.001$); Bonferroni post hoc, ** $P \leq 0.01$ *** $P \leq 0.001$; Error bars = S.E.M.). D. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and SNI (red) POD3 male mice. E. The relationship of AP firing with current step amplitude of inhibitory L2 PLPdyn+ neurons from sham ($$n = 15$$ from 7 slices in 4 mice) and SNI ($$n = 11$$ from 7 slices in 5 mice) POD3 male mice (RM two-way ANOVA, Fsurgery [1, 24] = 6.175, * $$P \leq 0.020$$, Fcurrent [12,288] = 263.8, *** $P \leq 0.001$, Finteraction [12,288] = 4.893, *** $P \leq 0.001$; Bonferroni post hoc, * $P \leq 0.05$, ** $P \leq 0.01$; Error bars = S.E.M.). F. SNI induced mechanical allodynia in male mice at POD14 (SNI = 5, sham = 5; ordinary-two-way ANOVA, Fsurgery [1,8] = 57.87, *** $P \leq 0.0001$, Ftime [1, 8] = 370.6, **** $P \leq 0.0001$, and Finteraction [1, 8] = 58.05, **** $P \leq 0.0001$; Sidak’s post hoc, **** $P \leq 0.0001$; Error bars = S.E.M.). G. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham (black) and SNI (red) POD14 male mice. H. The relationship of AP firing with current step amplitude of pyramidal L2 PLPdyn+ neurons from sham ($$n = 13$$ from 9 slices in 5 mice) and SNI ($$n = 13$$ from 8 slices in 5 mice) POD14 male mice (RM two-way ANOVA, Fsurgery [1, 24] = 4.334, * $$P \leq 0.048$$, Fcurrent [12,288] = 4.334, *** $P \leq 0.001$, Finteraction [12,252] = 4.121, *** $P \leq 0.001$; Bonferroni post hoc, * $P \leq 0.05$, ** $P \leq 0.01$; Error bars = S.E.M.). I. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and SNI (red) POD14 male mice. J. The relationship of AP firing with current step amplitude of inhibitory L2 PLPdyn+ neurons from sham ($$n = 13$$ from 7 slices in 5 mice) and SNI ($$n = 8$$ from 6 slices in 4 mice) POD14 male mice (Ordinary-two-way ANOVA, Fsurgery [1, 19] = 4.46, * $$P \leq 0.048$$, Fcurrent [12,228] = 234.8, *** $P \leq 0.001$, Finteraction [12,228] = 3.993, *** $P \leq 0.001$; Bonferroni post hoc, * $P \leq 0.05$; Error bars = S.E.M.). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Table 6Comparison of pyramidal and inhibitory PLPdyn+ neurons from sham and SNI POD3 male mice, related to Fig. 7.PLPdyn+ neuronsPYRINsSham ($$n = 6$$ neurons; 2 mice)SNI ($$n = 10$$ neurons; 4 mice)Sham ($$n = 15$$ neurons; 5 mice)SNI ($$n = 11$$ neurons; 5 mice)Subthreshold propertiesResting potential (mV)−80.7 ± 5.6−79.4 ± 7.1−82 ± 5.07−77.6 ± 6.8Input resistance (mΩ)78.7 ± 9.0688.6 ± 10.65121.7 ± 16.82112 ± 31.81Firing propertiesThreshold (mV)−36.1 ± 5.24−35.7 ± 7.35−37.4 ± 5.25−33.5 ± 7.57Threshold (pA)400 [50]⊺300 [1 0 0]⊺#*200 [50]⊺250 [1 0 0]⊺$*Frequency/current (Hz/pA)0.1 ± 0.0180.1 ± 0.0130.1 ± 0.0160.1 ± 0.0097APs @400 pA3.5 ± 2.747.4 ± 2.95#*13.7 ± 2.7910.1 ± 4.72#*Height (mV)78.1 ± 4.6779.7 ± 9.779.05 ± 8.7771.9 ± 15.65#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).Table 7Comparison of pyramidal and inhibitory PLPdyn+ neurons from sham and SNI POD14 male mice, related to Fig. 7.PLPdyn+ neuronsPYRINsSham ($$n = 13$$ neurons; 5 mice)SNI ($$n = 13$$ neurons; 5 mice)Sham ($$n = 13$$ neurons; 5 mice)SNI ($$n = 8$$ neurons; 4 mice)Subthreshold propertiesResting potential (mV)−79.8 ± 4.25−82 ± 5.16−77.8 ± 4.75−79.4 ± 5.02Input resistance (mΩ)81.6 ± 22.0594.3 ± 27.12121.8 ± 34.75153.2 ± 20.18#*Firing propertiesThreshold (mV)−34.7 ± 4.72−37.4 ± 3.54−36.7 ± 4.95−35.1 ± 2.99Threshold (pA)400 [1 5 0]⊺300 [1 2 5]⊺200 [1 0 0]⊺200 [50]⊺Frequency/current (Hz/pA)0.1 ± 0.0190.1 ± 0.0160.1 ± 0.0180.1 ± 0.029APs @400 pA3.9 ± 47.5 ± 4.27#*13.6 ± 4.5219.4 ± 2.78#Height (mV)80.8 ± 7.9176.6 ± 6.680.2 ± 9.9275 ± 7.45#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).
Female mice also displayed significant mechanical allodynia at both POD3 (Fig. 8A) and POD14 (Fig. 8F). Similar to males, a significant increase and a decrease in the excitability of pyramidal and inhibitory L2 PLPdyn+ neurons were measured in POD3-4 SNI female mice, respectively (Fig. 8B-E). There was no statistical significance in the current or voltage threshold for AP firing, RMP, or input resistance in pyramidal PLPdyn+ neurons (Table 8). However, a significant increase in the current threshold for AP firing and a significant decrease in the input resistance was detected in inhibitory L2 PLPdyn+ neurons 3 days after SNI (Table 8). At POD14, the excitability of pyramidal L2 PLPdyn+ neurons remained significantly increased in female SNI mice (Fig. 8G, H) with a decrease in the voltage threshold for AP firing (Table 9). In contrast to POD3-4, inhibitory L2 PLPdyn+ neurons displayed significant hyperexcitability at POD14 (Fig. 8I, J). Further analyses revealed a decrease in both current and voltage threshold for AP firing and an increase in input resistance recorded from inhibitory L2 PLPdyn+ neurons at POD14 (Table 9).Fig. 8Enhanced excitability of pyramidal L2 PLPdyn+ neurons in the SNI female mice. A. SNI induced mechanical allodynia in female mice at POD3 (SNI = 5, sham = 5; ordinary-two-way ANOVA, Fsurgery [1,8] = 2.248, $$P \leq 0.1722$$, Ftime [1, 8] = 22.09, $$P \leq 0.0015$$, and Finteraction [1, 8] = 12.39, $$P \leq 0.0078$$; Sidak’s post hoc, ** $P \leq 0.01$; Error bars = S.E.M.). B. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham (black) and SNI (red) POD3 female mice. C. The relationship of AP firing with current step amplitude of pyramidal L2 PLPdyn+ neurons from sham ($$n = 7$$ from 5 slices in 4 mice) and SNI ($$n = 9$$ from 5 slices in 4 mice) POD3 female mice (RM two-way ANOVA, Fsurgery [1, 14] = 5.605, * $$P \leq 0.033$$, Fcurrent [12,168] = 55.79, *** $P \leq 0.001$, Finteraction [12,168] = 5.651, *** $P \leq 0.001$; Bonferroni post hoc, ** $P \leq 0.01$ *** $P \leq 0.001$; Error bars = S.E.M.). D. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and SNI (red) POD3 female mice. E. The relationship of AP firing with current step amplitude of inhibitory L2 PLPdyn+ neurons from sham ($$n = 10$$ from 5 slices in 3 mice) and SNI ($$n = 6$$ from 3 slices in 3 mice) POD3 female mice (RM two-way ANOVA, Fsurgery [1, 14] = 13.59, ** $$P \leq 0.002$$, Fcurrent [12,168] = 115.3, *** $P \leq 0.001$, Finteraction [12,168] = 10.72, *** $P \leq 0.001$; Bonferroni post hoc, *** $P \leq 0.001$; Error bars = S.E.M.). F. SNI induced mechanical allodynia in female mice at POD14 (SNI = 6, sham = 6; ordinary-two-way ANOVA, Fsurgery [1,10] = 0.007, $$P \leq 0.94$$, Ftime [1, 10] = 16.77, ** $$P \leq 0.0022$$, and Finteraction [1, 10] = 10.99, ** $$P \leq 0.0078$$; Sidak’s post hoc, *** $P \leq 0.001$; Error bars = S.E.M.). G. Representative traces showing action potentials of pyramidal PLPdyn+ neurons from sham (black) and SNI (red) POD14 female mice. H. The relationship of AP firing with current step amplitude of pyramidal L2 PLPdyn+ neurons from sham ($$n = 8$$ from 6 slices in 5 mice) and SNI ($$n = 14$$ from 9 slices in 6 mice) POD14 female mice (RM two-way ANOVA, Fsurgery [1, 20] = 5.86, * $$P \leq 0.025$$, Fcurrent [12,240] = 214.7, *** $P \leq 0.001$, Finteraction [12,240] = 5.45, *** $P \leq 0.001$; Bonferroni post hoc, ** $P \leq 0.01$, *** $P \leq 0.001$; Error bars = S.E.M.). I. Representative traces showing action potentials of inhibitory PLPdyn+ neurons from sham (black) and SNI (red) POD14 female mice. J. The relationship of AP firing with current step amplitude of inhibitory L2 PLPdyn+ neurons from sham ($$n = 11$$ from 6 slices in 5 mice) and SNI ($$n = 7$$ from 4 slices in 3 mice) POD14 female mice (RM two-way ANOVA, Fsurgery [1, 16] = 6.718, * $$P \leq 0.02$$, Fcurrent [12,192] = 208.1, *** $P \leq 0.001$, Finteraction [12,192] = 4.579, *** $P \leq 0.001$; Bonferroni post hoc, ** $P \leq 0.01$, *** $P \leq 0.001$; Error bars = S.E.M.). ( For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Table 8Comparison of pyramidal and inhibitory PLPdyn+ neurons from sham and SNI POD3 female mice, related to Fig. 8.PLPdyn+ neuronsPYRINsSham ($$n = 7$$ neurons; 4 mice)SNI ($$n = 9$$ neurons; 4 mice)Sham ($$n = 10$$ neurons; 3 mice)SNI ($$n = 6$$ neurons; 3 mice)Subthreshold propertiesResting potential (mV)−81 ± 5.21−78.9 ± 3.96−79.2 ± 4.12−82.8 ± 2.39Input resistance (mΩ)82.1 ± 15.4798.7 ± 23.4131.2 ± 21.299.3 ± 22.7#*Firing propertiesThreshold (mV)−34.5 ± 5.78−35.5 ± 4.69−37 ± 3.95−36.1 ± 3.33Threshold (pA)400 [1 5 0]⊺300 [1 0 0]⊺200 [1 0 0]⊺300 [1 0 0]⊺#**Frequency/current (Hz/pA)0.1 ± 0.00980.12 ± 0.0230.13 ± 0.0210.12 ± 0.021APs @400 pA4.3 ± 3.7310.3 ± 4.87#*16.1 ± 38 ± 6.51Height (mV)79.2 ± 7.4578.4 ± 9.1779.9 ± 10.0664.2 ± 13.44#*#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).Table 9Comparison of pyramidal and inhibitory PLPdyn+ neurons from sham and SNI POD14 female mice, related to Fig. 8.PLPdyn+ neuronsPYRINsSham ($$n = 8$$ neurons; 5 mice)SNI ($$n = 14$$ neurons; 6 mice)Sham ($$n = 11$$ neurons; 5 mice)SNI ($$n = 7$$ neurons; 3 mice)Subthreshold propertiesResting potential (mV)−79.6 ± 5.75−81.7 ± 3.95−78.1 ± 4.33−82.1 ± 6.75Input resistance (mΩ)91 ± 15.5897.7 ± 15.76121.4 ± 21.09157.5 ± 38.95#*Firing propertiesThreshold (mV)−33 ± 5.67−37.7 ± 3.64#*−34.1 ± 4.76−42.1 ± 4.65#**Threshold (pA)300 [1 0 0]⊺300 [50]⊺250 [1 0 0]⊺200 [50]⊺$*Frequency/current (Hz/pA)0.09 ± 0.0130.1 ± 0.0140.13 ± 0.0240.14 ± 0.022APs @400 pA6 ± 2.399.1 ± 2.49#**13.9 ± 4.1617.9 ± 3.44Height (mV)79.2 ± 5.180.5 ± 7.0777.4 ± 12.8476.3 ± 10.29#: Student’s unpaired t-test; Data shown as mean ± standard deviation.$: Mann-Whitney U test; Data shown are median ± standard deviation.*: $p \leq 0.05$; **: $p \leq 0.01$; ***: $p \leq 0.001.$⊺: median (interquartile range).
## Discussion
In this study, we focused on characterizing the effects of surgical and neuropathic pain on the excitability of a subset of laminar layer 2 (L2) neurons in the prelimbic (PL) region of the mPFC that express prodynorphin (PLPdyn+ neurons). Prodynorphin (Pdyn) is the precursor peptide for dynorphin (Dyn), the endogenous opioid ligand for the kappa opioid receptors (KOR), which are receptors involved in mediating sensory (Obara et al., 2003, Xu et al., 2004, Aita et al., 2010) and negative affective components of pain (Cahill et al., 2014, Massaly et al., 2019, Navratilova et al., 2019). We first present data showing that L2 PLPdyn+ neurons are heterogeneous, consisting of both pyramidal (i.e. excitatory) and inhibitory neurons. We also show that L$\frac{2}{3}$ PLPdyn+ neurons send both excitatory and inhibitory inputs to unlabeled L5 neurons in PL. This is consistent with the top-down organization of local circuits found in the mPFC and other cortical regions including motor and somatosensory cortex (Kampa et al., 2006, Weiler et al., 2008, Anderson et al., 2010, Hooks et al., 2011, Ferreira et al., 2015, Cheriyan and Sheets, 2018).
Electrophysiological recordings in acute brain slices revealed that the excitable properties of PLPdyn+ neuronal subtypes are distinctly altered by surgical and nerve injury. One day after plantar incision (PI) of the hind paw, we identified a significant increase in the excitability of pyramidal, but not inhibitory, L2 PLPdyn+ neurons in both male and female mice. Given that we demonstrate PLPdyn+ neurons send projections locally within PL, this result suggest a potential increase in the local release of Dyn in PL soon after surgical injury (Abraham et al., 2021). The relevance of this finding is that KORs are expressed in mPFC (Mansour et al., 1988), and extensive work has shown that activation of KORs contribute to behaviors associated with negative affect such as aversion, fear, stress, and depression (McLaughlin et al., 2006, Knoll et al., 2007, Land et al., 2009, Ebner et al., 2010, Bruchas et al., 2011, Tejeda et al., 2013, Cahill et al., 2014). For example, activation of KORs produces conditioned place aversion (CPA) in rodents and dysphoria in humans (Bals-Kubik et al., 1993, Shippenberg et al., 1993, Knoll and Carlezon, 2010, Chavkin and Koob, 2016). Blocking KOR activity in the mPFC with KOR antagonist, nor-BNI, attenuates CPA induced by systemic administration of KOR agonist U69,593 (Tejeda et al., 2013). Together, this infers that hyperactivity of PLPdyn+ neurons may contribute to an mPFC/KOR mediated aversion associated with acute surgical pain.
When pain behavior subsided approximately 2 weeks after PI, we observed no significant differences in the excitability of pyramidal L2 PLPdyn+ neurons between sham and PIM male mice. This transient hyperexcitability of pyramidal L2 PLPdyn+ neurons from PI supports the notion that the activity of these specific cells contribute to the expression and negative affect of postoperative pain that subsides following recovery from incision. However, we found a significant increase in the excitability of the inhibitory L2 PLPdyn+ neurons in male mice following recovery from PI. One interpretation of these findings is that inhibitory L2 PLPdyn+ neurons serve to dampen hyperexcitability in the PL during recovery from injury. However, the mechanisms that contribute to increased firing of inhibitory L2 PLPdyn+ neurons and the overall effect of this change on PL circuits still need to be resolved. In the female mice recovered from PI, we observed a significant decrease in the excitability of pyramidal PLPdyn+ neurons, but no difference in the inhibitory PLPdyn+ neurons. This implies that the regulation of PLPdyn+ neurons during recovery from surgical injury is different based on sex.
Previous work using a Pdyn knockout mouse indicated that Pdyn contributes to the maintenance but not the initiation of neuropathic pain (Wang et al., 2001). Interestingly, increased Pdyn mRNA levels are detected in the mPFC during the development of chronic pain (Candeletti and Ferri, 1995, Palmisano et al., 2018), but not in acute pain (Nwaneshiudu et al., 2019). We therefore questioned whether a prolonged increase in the excitability of the pyramidal PLPdyn+ population would be detected both 3 days (initiation) and 2 weeks (maintenance) after spared nerve injury (SNI), a robust model of neuropathic pain. We found a significant increase in the excitability of pyramidal L2 PLPdyn+ neurons in both male and female SNI mice both 3 days and 14 days after SNI. Interestingly, there was a dynamic change in the excitability of inhibitory L2 PLPdyn+ neurons. Inhibitory PLPdyn+ neurons were hypoexcitable 3 days after SNI but at 14 days after SNI we observed an increase in excitability of inhibitory PLPdyn+ neurons compared to recordings from sham mice. Previous studies report that PLPdyn+ inhibitory neurons express somatostatin (SOM) (Sohn et al., 2014, Loh et al., 2017, Smith et al., 2019). Our lab has reported that spontaneous excitatory synaptic inputs were reduced to L$\frac{2}{3}$ SOM expressing neurons of the PL in the female mice 7 days after SNI (Jones and Sheets, 2020), which is commonly considered an early time point in chronic neuropathic pain. Together with our current data, this indicates that reduced excitability of L2 Pdyn+/SOM+ GABAergic neurons in PL is a signature of the early stages of neuropathic pain. In addition, these findings suggest that increased Pdyn mRNA levels observed in the mPFC at later time points of neuropathic pain models (Candeletti and Ferri, 1995, Palmisano et al., 2018) result from prolonged hyperexcitability of both pyramidal and inhibitory PLPdyn+ neurons.
A majority of PLPdyn+ neurons identified in this study are located at L$\frac{2}{3}$ of PL, which is a laminar location that integrates a variety of excitatory inputs from regions such as the midline thalamus, contralateral mPFC, basolateral amygdala (BLA) and hippocampus (Hoover and Vertes, 2007, Little and Carter, 2012). Excitatory inputs from the BLA preferentially and monosynaptically target L2 corticoamygdalar (CA) neurons compared to neighboring neurons in the PL (Little and Carter, 2013). In the experiments where we injected AAV1-EF1a-DIO-hChR2(E123A)-EYFP into the PL cortex of Pdyn-Cre-tdTomato mice (Fig. 2C), we found EYFP axonal labeling in the BLA (data not shown) indicating that L2 PLPdyn+ neurons are CA neurons. These findings infer that L2 PLPdyn+ neurons are targeted by the BLA. The relevance of a BLA → L2 PLPdyn+ pathway is that hyperactivity of ascending BLA inputs to the mPFC occurs in arthritic and nerve-injury pain models (Ji et al., 2010, Zhang et al., 2015, Huang et al., 2019). Together, this suggests that the enhanced excitatory input from the BLA is contributing factor to increased excitability of L2 PLPdyn+ neurons observed in the PIM and SNI models.
There are also indications for how pain-induced changes to PLPdyn+ excitability affects local circuit activity within PL. Our lab has shown that L$\frac{2}{3}$ cortical neurons in the mPFC send local descending inputs onto L5 pyramidal neurons that project to the periaqueductal grey (PAG) (Cheriyan and Sheets, 2018). The PAG is a midbrain structure that plays a key role in endogenous analgesia (Reynolds, 1969). Reduced PL output to the PAG is implicated in neuropathic pain (Cheriyan and Sheets, 2018, Huang et al., 2019), and recent work reports that L2 CA activity driven by BLA inputs evokes strong local inhibition of L5 cortico-PAG neurons in the mPFC (Manoocheri and Carter, 2022). Based on these findings, our current working hypothesis to be tested in future studies is that incision or nerve injury activates and/or sensitizes the BLA → L2 PLPdyn+ pathway enhancing local inhibition of L5 cortico-PAG neurons.
## Conclusion
The mPFC play a significant role in regulating the sensory and affective components of pain. There is extensive evidence indicating functional changes within the mPFC in both acute and chronic pain modalities. However, published reports about how neurons with specific biomarker were regulated in both acute and chronic pain were scarce. In our study, we found different subtypes of prodynorphin expressing neurons in the mPFC and different subtypes exhibit distinct alterations in different pain modals and at different time point. The potential dynamic changes of different subtypes of Pdyn+ neurons may play a critical role in pain chronification. Moreover, Pdyn+ neurons produce dynorphin, which activate KORs. Studies have shown that activation of KORs in the mPFC regulates the affective aspect of pain. Thus, Pdyn+ neurons may serve as a potential target preventing pain chronification and attenuating affective aspects of pain.
## Author contributions
P.L.S. and S.Z. designed research and S.Z. performed all experiments. Y.Y. provided assistance with surgical procedures and behavioral analysis. S.Z. and P.L.S. analyzed data and wrote the paper.
## CRediT authorship contribution statement
Shudi Zhou: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Yuexi Yin: Investigation. Patrick L. Sheets: Conceptualization, Methodology, Validation, Formal analysis, Writing – original draft, Writing – review & editing, Visualization, Resources, Supervision, Project administration, Funding acquisition.
## Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
## Data availability
Data will be made available on request.
## References
1. Abraham A.D., Casello S.M., Schattauer S.S., Wong B.A., Mizuno G.O., Mahe K., Tian L., Land B.B., Chavkin C.. **Release of endogenous dynorphin opioids in the prefrontal cortex disrupts cognition**. *Neuropsychopharmacology* (2021.0) **46** 2330-2339. PMID: 34545197
2. Aita M., Byers M.R., Chavkin C., Xu M.. **Trigeminal injury causes kappa opioid-dependent allodynic, glial and immune cell responses in mice**. *Mol. Pain* (2010.0) **6** 8. PMID: 20109235
3. Alvarado S., Tajerian M., Millecamps M., Suderman M., Stone L.S., Szyf M.. **Peripheral nerve injury is accompanied by chronic transcriptome-wide changes in the mouse prefrontal cortex**. *Mol. Pain* (2013.0) **9** 21. PMID: 23597049
4. Alvarado S., Tajerian M., Suderman M., Machnes Z., Pierfelice S., Millecamps M., Stone L.S., Szyf M.. **An epigenetic hypothesis for the genomic memory of pain**. *Front. Cell. Neurosci.* (2015.0) **9** 88. PMID: 25852480
5. Alvarez-Bolado G., Fairén A., Douglass J., Naranjo J.R.. **Expression of the prodynorphin gene in the developing and adult cerebral cortex of the rat: an in situ hybridization study**. *J. Comp. Neurol.* (1990.0) **300** 287-300. PMID: 2266188
6. Anderson C.T., Sheets P.L., Kiritani T., Shepherd G.M.G.. **Sublayer-specific microcircuits of corticospinal and corticostriatal neurons in motor cortex**. *Nat. Neurosci.* (2010.0) **13** 739-744. PMID: 20436481
7. Bals-Kubik R., Ableitner A., Herz A., Shippenberg T.S.. **Neuroanatomical sites mediating the motivational effects of opioids as mapped by the conditioned place preference paradigm in rats**. *J. Pharmacol. Exp. Ther.* (1993.0) **264** 489-495. PMID: 8093731
8. Blom S.M., Pfister J.-P., Santello M., Senn W., Nevian T.. **Nerve injury-induced neuropathic pain causes disinhibition of the anterior cingulate cortex**. *J. Neurosci.* (2014.0) **34** 5754-5764. PMID: 24760836
9. Bonin R.P., Bories C., De Koninck Y.. **A simplified up-down method (SUDO) for measuring mechanical nociception in rodents using von Frey filaments**. *Mol. Pain* (2014.0) **10** 26. PMID: 24739328
10. Bruchas M., Schindler A., Shankar H., Messinger D., Miyatake M., Land B., Lemos J., Hagan C.E., Neumaier J., Quintana A., Palmiter R., Chavkin C.. **Selective p38α MAPK deletion in serotonergic neurons produces stress resilience in models of depression and addiction**. *Neuron* (2011.0) **71** 498-511. PMID: 21835346
11. Cahill C.M., Taylor A.M., Cook C., Ong E., Moron J.A., Evans C.J.. **Does the kappa opioid receptor system contribute to pain aversion?**. *Front. Pharmacol.* (2014.0) **5** 253. PMID: 25452729
12. Candeletti S., Ferri S.. **Cerebrospinal alterations of immunoreactive dynorphin A after unilateral dorsal rhizotomy in the rat**. *Brain Res.* (1995.0) **670** 289-296. PMID: 7743192
13. Cao X.Y., Xu H., Wu L.J., Li X.Y., Chen T., Zhuo M.. **Characterization of intrinsic properties of cingulate pyramidal neurons in adult mice after nerve injury**. *Mol. Pain* (2009.0) **5** 73. PMID: 20015370
14. Chaplan S.R., Bach F.W., Pogrel J.W., Chung J.M., Yaksh T.L.. **Quantitative assessment of tactile allodynia in the rat paw**. *J. Neurosci. Methods* (1994.0) **53** 55-63. PMID: 7990513
15. Chavkin C., James I.F., Goldstein A.. **Dynorphin is a specific endogenous ligand of the kappa opioid receptor**. *Science* (1982.0) **215** 413-415. PMID: 6120570
16. Chavkin C., Koob G.F.. **Dynorphin, Dysphoria, and Dependence: the Stress of Addiction**. *Neuropsychopharmacology* (2016.0) **41** 373-374
17. Cheriyan J., Sheets P.L.. **Altered Excitability and Local Connectivity of mPFC-PAG Neurons in a Mouse Model of Neuropathic Pain**. *J. Neurosci.* (2018.0) **38** 4829-4839. PMID: 29695413
18. Cheriyan J., Sheets P.L.. **Peripheral nerve injury reduces the excitation-inhibition balance of basolateral amygdala inputs to prelimbic pyramidal neurons projecting to the periaqueductal gray**. *Mol. Brain* (2020.0) **13** 100. PMID: 32600466
19. Condés-Lara M., Zapata I.O., León-Olea M., Sánchez-Alvarez M.. **Dorsal raphe and nociceptive stimulations evoke convergent responses on the thalamic centralis lateralis and medial prefrontal cortex neurons**. *Brain Res.* (1989.0) **499** 145-152. PMID: 2804662
20. Cordeiro Matos S., Zhang Z., Seguela P.. **Peripheral Neuropathy Induces HCN Channel Dysfunction in Pyramidal Neurons of the Medial Prefrontal Cortex**. *J. Neurosci.* (2015.0) **35** 13244-13256. PMID: 26400952
21. Cowie A., Stucky C.. **A mouse model of postoperative pain**. *Bio-Protocol* (2019.0) **9**
22. Dale J., Zhou H., Zhang Q., Martinez E., Hu S., Liu K., Urien L., Chen Z., Wang J.. **Scaling up cortical control inhibits pain**. *Cell Rep.* (2018.0) **23** 1301-1313. PMID: 29719246
23. Ebner S.R., Roitman M.F., Potter D.N., Rachlin A.B., Chartoff E.H.. **Depressive-like effects of the kappa opioid receptor agonist salvinorin A are associated with decreased phasic dopamine release in the nucleus accumbens**. *Psychopharmacology (Berl)* (2010.0) **210** 241-252. PMID: 20372879
24. Ferreira A.N., Yousuf H., Dalton S., Sheets P.L.. **Highly differentiated cellular and circuit properties of infralimbic pyramidal neurons projecting to the periaqueductal gray and amygdala**. *Front. Cell. Neurosci.* (2015.0) **9** 161. PMID: 25972785
25. Hooks B.M., Hires S.A., Zhang Y.-X., Huber D., Petreanu L., Svoboda K., Shepherd G.M.G., Petersen C.C.C.H.. **Laminar analysis of excitatory local circuits in vibrissal motor and sensory cortical areas**. *PLoS Biol.* (2011.0) **9** e1000572. PMID: 21245906
26. Hoover W.B., Vertes R.P.. **Anatomical analysis of afferent projections to the medial prefrontal cortex in the rat**. *Brain Struct. Funct.* (2007.0) **212** 149-179. PMID: 17717690
27. Huang J., Gadotti V.M., Chen L., Souza I.A., Huang S., Wang D., Ramakrishnan C., Deisseroth K., Zhang Z., Zamponi G.W.. **A neuronal circuit for activating descending modulation of neuropathic pain**. *Nat. Neurosci.* (2019.0) **22** 1659-1668. PMID: 31501573
28. Jefferson T., Kelly C.J., Martina M.. **Differential Rearrangement of Excitatory Inputs to the Medial Prefrontal Cortex in Chronic Pain Models**. *Front. Neural Circ.* (2021.0) **15**
29. Ji G., Sun H., Fu Y., Li Z., Pais-Vieira M., Galhardo V., Neugebauer V.. **Cognitive impairment in pain through amygdala-driven prefrontal cortical deactivation**. *J. Neurosci.* (2010.0) **30** 5451-5464. PMID: 20392966
30. Jones A.F., Sheets P.L.. **Sex-Specific Disruption of Distinct mPFC Inhibitory Neurons in Spared-Nerve Injury Model of Neuropathic Pain**. *Cell Rep.* (2020.0) **31** 107729. PMID: 32521254
31. Kampa B.M., Letzkus J.J., Stuart G.J.. **Cortical feed-forward networks for binding different streams of sensory information**. *Nat. Neurosci.* (2006.0) **9** 1472-1473. PMID: 17099707
32. Kelly C.J., Huang M., Meltzer H., Martina M.. **Reduced Glutamatergic Currents and Dendritic Branching of Layer 5 Pyramidal Cells Contribute to Medial Prefrontal Cortex Deactivation in a Rat Model of Neuropathic Pain**. *Front. Cell. Neurosci.* (2016.0) **10** 133. PMID: 27252623
33. Khachaturian H., Lewis M.E., Haber S.N., Houghten R.A., Akil H., Watson S.J.. **Prodynorphin peptide immunocytochemistry in rhesus monkey brain**. *Peptides* (1985.0) **6** 155-166
34. Kiritoshi T., Ji G., Neugebauer V.. **Rescue of Impaired mGluR5-Driven Endocannabinoid Signaling Restores Prefrontal Cortical Output to Inhibit Pain in Arthritic Rats**. *J. Neurosci.* (2016.0) **36** 837-850. PMID: 26791214
35. Knoll A.T., Carlezon W.A.. **Dynorphin, stress, and depression**. *Brain Res.* (2010.0) **1314** 56-73. PMID: 19782055
36. Knoll A.T., Meloni E.G., Thomas J.B., Carroll F.I., Carlezon W.A.. **Anxiolytic-like effects of kappa-opioid receptor antagonists in models of unlearned and learned fear in rats**. *J. Pharmacol. Exp. Ther.* (2007.0) **323** 838-845. PMID: 17823306
37. Land B.B., Bruchas M.R., Schattauer S., Giardino W.J., Aita M., Messinger D., Hnasko T.S., Palmiter R.D., Chavkin C.. **Activation of the kappa opioid receptor in the dorsal raphe nucleus mediates the aversive effects of stress and reinstates drug seeking**. *PNAS* (2009.0) **106** 19168-19173. PMID: 19864633
38. Lee M., Manders T.R., Eberle S.E., Su C., D'amour J., Yang R., Lin H.Y., Deisseroth K., Froemke R.C., Wang J.. **Activation of corticostriatal circuitry relieves chronic neuropathic pain**. *J. Neurosci.* (2015.0) **35** 5247-5259. PMID: 25834050
39. Little J.P., Carter A.G.. **Subcellular synaptic connectivity of layer 2 pyramidal neurons in the medial prefrontal cortex**. *J. Neurosci.* (2012.0) **32** 12808-12819. PMID: 22973004
40. Little J.P., Carter A.G.. **Synaptic Mechanisms Underlying Strong Reciprocal Connectivity between the Medial Prefrontal Cortex and Basolateral Amygdala**. *J. Neurosci.* (2013.0) **33** 15333-15342. PMID: 24068800
41. Loh R., Collins S., Galvez R.. **Neocortical prodynorphin expression is transiently increased with learning: Implications for time- and learning-dependent neocortical kappa opioid receptor activation**. *Behav. Brain Res.* (2017.0) **335** 145-150. PMID: 28802836
42. Manoocheri K., Carter A.G.. **Rostral and caudal basolateral amygdala engage distinct circuits in the prelimbic and infralimbic prefrontal cortex**. *Elife* (2022.0) 11
43. Mansour A., Khachaturian H., Lewis M.E., Akil H., Watson S.J.. **Anatomy of CNS opioid receptors**. *Trends Neurosci.* (1988.0) **11** 308-314. PMID: 2465635
44. Martinez E., Lin H.H., Zhou H., Dale J., Liu K., Wang J.. **Corticostriatal Regulation of Acute Pain**. *Front. Cell. Neurosci.* (2017.0) **11** 146. PMID: 28603489
45. Massaly N., Copits B.A., Wilson-Poe A.R., Hipólito L., Markovic T., Yoon H.J., Liu S., Walicki M.C., Bhatti D.L., Sirohi S., Klaas A., Walker B.M., Neve R., Cahill C.M., Shoghi K.I., Gereau R.W., McCall J.G., Al-Hasani R., Bruchas M.R., Morón J.A.. **Pain-Induced Negative Affect Is Mediated via Recruitment of The Nucleus Accumbens Kappa Opioid System**. *Neuron* (2019.0) **102** 564-573.e6. PMID: 30878290
46. McLaughlin J.P., Li S., Valdez J., Chavkin T.A., Chavkin C.. **Social defeat stress-induced behavioral responses are mediated by the endogenous kappa opioid system**. *Neuropsychopharmacology* (2006.0) **31** 1241-1248. PMID: 16123746
47. Metz A.E., Yau H.-J., Centeno M.V., Apkarian A.V., Martina M.. **Morphological and functional reorganization of rat medial prefrontal cortex in neuropathic pain**. *Proc. Natl. Acad. Sci. USA* (2009.0) **106** 2423-2428. PMID: 19171885
48. Mitric M., Seewald A., Moschetti G., Sacerdote P., Ferraguti F., Kummer K.K., Kress M.. **Layer- and subregion-specific electrophysiological and morphological changes of the medial prefrontal cortex in a mouse model of neuropathic pain**. *Sci. Rep.* (2019.0) **9** 9479. PMID: 31263213
49. Navratilova E., Ji G., Phelps C., Qu C., Hein M., Yakhnitsa V., Neugebauer V., Porreca F.. **Kappa opioid signaling in the central nucleus of the amygdala promotes disinhibition and aversiveness of chronic neuropathic pain**. *Pain* (2019.0) **160** 824-832. PMID: 30681985
50. Nwaneshiudu CA, Shi XY & Clark JD. (2019). Incisional Injury Modulates Morphine Reward and Morphine-Primed Reinstatement: A Role of Kappa Opioid Receptor Activation. Anesthesia and analgesia.
51. Obara I., Mika J., Schafer M.K., Przewlocka B.. **Antagonists of the kappa-opioid receptor enhance allodynia in rats and mice after sciatic nerve ligation**. *Br. J. Pharmacol.* (2003.0) **140** 538-546. PMID: 12970097
52. Palmisano M, Caputi FF, Mercatelli D, Romualdi P & Candeletti S. (2018). Dynorphinergic system alterations in the corticostriatal circuitry of neuropathic mice support its role in the negative affective component of pain. Genes Brain Behavior, e12467.
53. Peckys D., Hurd Y.L.. **Prodynorphin and kappa opioid receptor mRNA expression in the cingulate and prefrontal cortices of subjects diagnosed with schizophrenia or affective disorders**. *Brain Res. Bull.* (2001.0) **55** 619-624. PMID: 11576758
54. Pina M.M., Pati D., Hwa L.S., Wu S.Y., Mahoney A.A., Omenyi C.G., Navarro M., Kash T.L.. **The kappa opioid receptor modulates GABA neuron excitability and synaptic transmission in midbrainprojections from the insular cortex**. *Neuropharmacology* (2020.0) **165** 107831. PMID: 31870854
55. Pogatzki E.M., Raja S.N.. **A mouse model of incisional pain**. *Anesthesiology* (2003.0) **99** 1023-1027. PMID: 14508341
56. Porro C.A., Baraldi P., Pagnoni G., Serafini M., Facchin P., Maieron M., Nichelli P.. **Does anticipation of pain affect cortical nociceptive systems?**. *J. Neurosci.* (2002.0) **22** 3206-3214. PMID: 11943821
57. Reynolds D.V.. **Surgery in the rat during electrical analgesia induced by focal brain stimulation**. *Science (New York, N.Y.)* (1969.0) **164** 444-445. PMID: 4887743
58. Shiers S., Pradhan G., Mwirigi J., Mejia G., Ahmad A., Kroener S., Price T.. **Neuropathic Pain Creates an Enduring Prefrontal Cortex Dysfunction Corrected by the Type II Diabetic Drug Metformin But Not by Gabapentin**. *J. Neurosci.* (2018.0) **38** 7337-7350. PMID: 30030404
59. Shippenberg T.S., Bals-Kubik R., Herz A.. **Examination of the neurochemical substrates mediating the motivational effects of opioids: role of the mesolimbic dopamine system and D-1 vs. D-2 dopamine receptors**. *J. Pharmacol. Exp. Ther.* (1993.0) **265** 53-59. PMID: 8386244
60. Smith SJ, Sümbül U, Graybuck LT, Collman F, Seshamani S, Gala R, Gliko O, Elabbady L, Miller JA, Bakken TE, Rossier J, Yao Z, Lein E, Zeng H, Tasic B & Hawrylycz M. (2019). Single-cell transcriptomic evidence for dense intracortical neuropeptide networks. eLife8.
61. Sohn J., Hioki H., Okamoto S., Kaneko T.. **Preprodynorphin-expressing neurons constitute a large subgroup of somatostatin-expressing GABAergic interneurons in the mouse neocortex**. *J. Comp. Neurol.* (2014.0) **522** 1506-1526. PMID: 24122731
62. Tejeda H.A., Counotte D.S., Oh E., Ramamoorthy S., Schultz-Kuszak K.N., Bäckman C.M., Chefer V., O'Donnell P., Shippenberg T.S.. **Prefrontal cortical kappa-opioid receptor modulation of local neurotransmission and conditioned place aversion**. *Neuropsychopharmacology* (2013.0) **38** 1770-1779. PMID: 23542927
63. Wang Z., Gardell L.R., Ossipov M.H., Vanderah T.W., Brennan M.B., Hochgeschwender U., Hruby V.J., Malan T.P., Lai J., Porreca F.. **Pronociceptive actions of dynorphin maintain chronic neuropathic pain**. *J. Neurosci.* (2001.0) **21** 1779-1786. PMID: 11222667
64. Weiler N., Wood L., Yu J., Solla S.A., Shepherd G.M.G.. **Top-down laminar organization of the excitatory network in motor cortex**. *Nat. Neurosci.* (2008.0) **11** 360-366. PMID: 18246064
65. Xu M., Petraschka M., McLaughlin J.P., Westenbroek R.E., Caron M.G., Lefkowitz R.J., Czyzyk T.A., Pintar J.E., Terman G.W., Chavkin C.. **Neuropathic pain activates the endogenous kappa opioid system in mouse spinal cord and induces opioid receptor tolerance**. *J. Neurosci.* (2004.0) **24** 4576-4584. PMID: 15140929
66. Zhang Z., Gadotti V., Chen L., Souza I., Stemkowski P., Zamponi G.. **Role of Prelimbic GABAergic Circuits in Sensory and Emotional Aspects of Neuropathic Pain**. *Cell Rep.* (2015.0) **12** 752-759. PMID: 26212331
67. Zhou H., Martinez E., Lin H.H., Yang R., Dale J.A., Liu K., Huang D., Wang J.. **Inhibition of the Prefrontal Projection to the Nucleus Accumbens Enhances Pain Sensitivity and Affect**. *Front. Cell. Neurosci.* (2018.0) **12** 240. PMID: 30150924
|
---
title: Sex differences in islet stress responses support female β cell resilience
authors:
- George P. Brownrigg
- Yi Han Xia
- Chieh Min Jamie Chu
- Su Wang
- Charlotte Chao
- Jiashuo Aaron Zhang
- Søs Skovsø
- Evgeniy Panzhinskiy
- Xiaoke Hu
- James D. Johnson
- Elizabeth J. Rideout
journal: Molecular Metabolism
year: 2023
pmcid: PMC9971554
doi: 10.1016/j.molmet.2023.101678
license: CC BY 4.0
---
# Sex differences in islet stress responses support female β cell resilience
## Abstract
### Objective
Pancreatic β cells play a key role in maintaining glucose homeostasis; dysfunction of this critical cell type causes type 2 diabetes (T2D). Emerging evidence points to sex differences in β cells, but few studies have examined male-female differences in β cell stress responses and resilience across multiple contexts, including diabetes. Here, we address the need for high-quality information on sex differences in β cell and islet gene expression and function using both human and rodent samples.
### Methods
In humans, we compared β cell gene expression and insulin secretion in donors with T2D to non-diabetic donors in both males and females. In mice, we generated a well-powered islet RNAseq dataset from 20-week-old male and female siblings with similar insulin sensitivity. Our unbiased gene expression analysis pointed to a sex difference in the endoplasmic reticulum (ER) stress response. Based on this analysis, we hypothesized female islets would be more resilient to ER stress than male islets. To test this, we subjected islets isolated from age-matched male and female mice to thapsigargin treatment and monitored protein synthesis, cell death, and β cell insulin production and secretion. Transcriptomic and proteomic analyses were used to characterize sex differences in islet responses to ER stress.
### Results
Our single-cell analysis of human β cells revealed sex-specific changes to gene expression and function in T2D, correlating with more robust insulin secretion in human islets isolated from female donors with T2D compared to male donors with T2D. In mice, RNA sequencing revealed differential enrichment of unfolded protein response pathway-associated genes, where female islets showed higher expression of genes linked with protein synthesis, folding, and processing. This differential expression was physiologically significant, as islets isolated from female mice were more resilient to ER stress induction with thapsigargin. Specifically, female islets showed a greater ability to maintain glucose-stimulated insulin production and secretion during ER stress compared with males.
### Conclusions
Our data demonstrate sex differences in β cell gene expression in both humans and mice, and that female β cells show a greater ability to maintain glucose-stimulated insulin secretion across multiple physiological and pathological contexts.
## Highlights
•Islet β cells from female donors living with T2D have more insulin secretion than males.•Islets from female mice show higher expression of unfolded protein response genes.•Islets from female mice show better recovery of protein synthesis and survival after ER stress.•Islets from female mice maintain better insulin secretion during ER stress.•Islets from female mice islets show distinct transcriptomic and proteomic responses to ER stress.
## Introduction
Pancreatic β cells make and secrete insulin, an essential hormone required to maintain whole-body glucose homeostasis. Emerging evidence from multiple species points to biological sex as an important, but often overlooked, factor that affects β cell biology [[1], [2], [3], [4], [5], [6]]. Large-scale surveys of gene expression in mice and humans show that differences exist between the sexes in the pancreas [[7], [8], [9]], in islets [10], and in β cells specifically [4,11]. Humans also have a sex-specific β cell gene expression response to aging [12], and show male-female differences in pancreatic β cell number [6]. With respect to β cell function, most data from rodent and human studies suggests glucose-stimulated insulin secretion is higher in females than in males [5,10,[13], [14], [15], [16]]. While male-female differences in peripheral insulin sensitivity [15,[17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27]] may contribute to these differences, sex-biased insulin secretion in humans persists in the context of equivalent insulin sensitivity between males and females [5]. Whether sex differences in other aspects of β cell gene expression and function are similarly independent of insulin sensitivity in rodents and humans remains unclear, as insulin sensitivity is not routinely monitored across datasets showing sex differences in β cell biology.
Biological sex also affects the risk of developing T2D. Across many population groups, men are at a higher risk of developing T2D than women [[28], [29], [30], [31]]. Some of this differential risk is explained by lifestyle and cultural factors [[31], [32], [33]]. Biological sex also plays a role; however, as the male-biased risk of developing diabetes-like phenotypes is conserved across multiple animal models [22,[34], [35], [36], [37], [38], [39]]. Despite a dominant role for β cell function in T2D pathogenesis [40,41], T2D- and stress-associated changes to β cell gene expression and function in each sex remain largely unexplored, as most studies on these topics do not include biological sex as a variable in their analysis [[42], [43], [44], [45], [46], [47], [48], [49]]. Collecting detailed knowledge of β cell gene expression and function from each sex under physiological and pathological conditions is therefore a key first step toward understanding whether male-female differences in this important cell type may contribute to T2D risk.
The overall goal of our study was to provide detailed knowledge of β cell gene expression and function in both males and females across multiple contexts to advance our understanding of sex differences in this important cell type. Our data show male-female differences in islet and β cell gene expression and stress responses in both humans and mice. These differences contribute to sex differences in β cell resilience, where we find female β cells show a greater ability to maintain glucose-stimulated insulin secretion in response to stress and T2D in mice and humans, respectively. Given that an insulin tolerance test indicated similar insulin sensitivity between the male and female mice used in our study, our findings suggest biological sex is an important variable to consider in studies on islet and β cell function.
## Animals
Mice were bred in-house or purchased from the Jackson Laboratory. Unless otherwise stated, islets were isolated from C57BL/6J mice aged 20–24 weeks. Animals were housed and studied in the UBC Modified Barrier Facility using protocols approved by the UBC Animal Care Committee and in accordance with international guidelines. Mice were housed on a 12-hour light/dark cycle with food and drinking water ad libitum. Mice were fed a regular chow diet (LabDiet #5053); $24.5\%$ energy from protein, $13.1\%$ energy from fat, and $62.4\%$ energy from carbohydrates.
## Islet isolation, culture, dispersion and treatment
Mouse islet isolations were performed by ductal collagenase injection followed by filtration and hand-picking, using modifications of the protocol described by Salvalaggio [50]. Islets recovered overnight, in islet culture media (RPMI media with 11.1 mM d-glucose supplemented with $10\%$ vol/vol fetal bovine serum (FBS) (Thermo: 12483020) and $1\%$ vol/vol Penicillin-Streptomycin (P/S) (GIBCO: 15140-148)) at 37 °C with $5\%$ CO2. After four washes with Minimal Essential Medium [l-glutamine, calcium and magnesium free] (Corning: 15-015 CV) islets were dispersed with $0.01\%$ trypsin and resuspended in islet culture media. Cell seedings were done as per the experimental procedure (protein synthesis: 20,000 cells per well, live cell imaging: 5,000 cells per well). ER stress was induced by treating islets with the SERCA inhibitor thapsigargin. For assays less than 24 h, we used (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S). For assays greater than 24 h we used (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S, $10\%$ vol/vol FBS).
## Analysis of protein synthesis
Dispersed islets were seeded into an optical 96-well plate (Perkin Elmer) at a density of approximately 20,000 cells per well in islet culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S, $10\%$ vol/vol FBS). 24 h after seeding, treatments were applied in fresh islet culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S). After incubation, fresh culture media was applied (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S), then supplemented with 20 μM OPP (Invitrogen) and respective drug treatments. The assay was performed according to manufacturer's instructions. Cells were imaged at 10× with an ImageXpressMICRO high-content imager and analyzed with MetaXpress (Molecular Devices) to quantify the integrated staining intensity of OPP-Alexa Fluor 594 in cells identified by NuclearMask Blue Stain.
## Live cell imaging
Dispersed islets were seeded into 384-well plates (Perkin Elmer) at a density of approximately 5,000 cells per well. Islet viability was measured with the TC20 Automated Cell Counter (Bio-Rad: 1450102) with Trypan Blue (Bio-Rad: 1450021). Islets were allowed to adhere for 48 h in islet culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S, $10\%$ vol/vol FBS). Cells were stained with Hoechst 33342 (Sigma–Aldrich) (0.05 μg/mL) and propidium iodide (Sigma–Aldrich) (0.5 μg/mL) for 1 h in islet culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S, $10\%$ vol/vol FBS) prior to the addition of treatments and imaging. 384-well plates were placed into the environmentally-controlled (37 °C, $5\%$ CO2) ImageXpressMICRO high content imaging system. To measure cell death, islet cells were imaged every 2 h for 84 h. MetaXpress software was used to quantify cell death, defined as the number of Propidium Iodide-positive/Hoechst 33342-positive cells. To measure *Ins2* gene activity, dispersed islets from Ins2GFP/WT mice aged 21–23 weeks were used [51]. Islet cells were imaged every 30 min for 60 h. MetaXpress analysis software and custom R scripts were used to perform single-cell tracking of Ins2GFP/WT β cells as previously described [51].
## Western blot
After overnight recovery, islets were split equally per mouse into islet culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S, $10\%$ vol/vol FBS) resulting in ∼100–150 islets per condition. Islets were treated for 24 h with DMSO or 1 μM Tg in islet culture media then sonicated in RIPA lysis buffer (150 mM NaCl, $1\%$ Nonidet P-40, $0.5\%$ DOC, $0.1\%$ SDS, 50 mM Tris (pH 7.4), 2 mM EGTA, 2 mM Na3VO4, and 2 mM NaF supplemented with complete mini protease inhibitor cocktail (Roche, Laval, QC)). Equal protein amounts (8–10 μg of protein) in equal volumes were loaded for each experiment. Protein lysates were incubated in Laemmli loading buffer (Thermo, J61337AC) at 95°C for 5 min and resolved by SDS-PAGE. Proteins were then transferred to PVDF membranes (BioRad, CA) and probed with antibodies against HSPA5 (1:1000, Cat. # 3183, Cell Signalling), eIF2α (1:1000, Cat. # 2103, Cell Signalling), phospho-eIF2α (1:1000, Cat. # 3398, Cell Signalling), IRE1α (1:1000, Cat. # 3294, Cell Signalling), phospho-IRE1α (1:1000, Cat. # PA1-16927, Thermo Fisher Scientific), CHOP (1:1000, #ab11419, Abcam), β-actin (1:1000, NB600-501, Novus Biologicals). The signals were detected by secondary HRP-conjugated antibodies (Anti-mouse, Cat. # 7076; Anti-rabbit, Cat. # 7074; CST) and either Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) or Forte (Immobilon). Protein band intensities were quantified using Image Studio (LI-COR).
## Islet secretion and content
Glucose-stimulated insulin/proinsulin production and secretion were assessed using size-matched islets (five islets per well, in triplicate) seeded into 96-well V-bottom Tissue Culture Treated Microplates (Corning: #CLS3894). Islets were allowed to adhere for 48 h in culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S, $10\%$ vol/vol FBS), based on a published method [52], as this protocol permits analysis of large sample numbers and treatments, and minimizes islet loss. Adherent islets were washed with Krebs–Ringer Buffer (KRB; 129 mM NaCl, 4.8 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 2.5 mM CaCl2, 5 mM NaHCO3, 10 mM HEPES, $0.5\%$ bovine serum albumin) containing 3 mM glucose then pre-incubated for 4 h in 3 mM glucose KRB. 1 μM Tg was added to the 3 mM low glucose pre-incubation buffer 4 h prior, 2 h prior, or at the start of the low glucose incubation period. Islets were incubated in KRB with 3 mM glucose then 20 mM glucose for 45 min each. Supernatant was collected after each stimulation. Islet insulin and proinsulin content was extracted by freeze-thawing in 100 μL of acid ethanol, then the plates were shaken at 1200 rpm for 10 min at 4°C to lyse the islets. Insulin was measured by Rodent Insulin Chemiluminescent ELISA (ALPCO: 80-INSMR) and proinsulin by Rat/Mouse Proinsulin ELISA (Mercodia: 10-1232-01). Measurements were performed on a Spark plate reader (TECAN).
## Blood collection and in vivo analysis of glucose homeostasis and insulin secretion
Mice were fasted for 6 h prior to glucose and insulin tolerance tests. During glucose and insulin tolerance tests, tail blood was collected for blood glucose measurements using a glucometer (One Touch Ultra 2 Glucometer, Lifescan, Canada). For intraperitoneal (i.p.) glucose tolerance tests, the glucose dose was 2 g glucose/kg of body mass. For insulin tolerance tests, the insulin dose was 0.75U insulin/kg body mass. For measurements of in vivo glucose-stimulated insulin secretion, femoral blood was collected after i.p. injection of 2 g glucose/kg body mass. Blood samples were kept on ice during collection, centrifuged at 2000 rpm for 10 min at 4°C and stored as plasma at −20 °C. Plasma samples were analysed for insulin using Rodent Insulin Chemiluminescent ELISA (ALPCO: 80-INSMR). Glucose-stimulated insulin secretion from an independent large cohort of Insrf/f:Ins1Cre−/−:nTnG+/− mice reared in our facility was replotted from a published Johnson lab study [53]. Blood glucose was monitored using control and insulin-reduced mice. Control mice were Ins1−/−:Ins2f/f:mTmG + tamoxifen and Ins1−/−:Ins2f/f:Pdx1CreERT:mTmG + corn oil [54]. Because we confirmed there were no significant differences in blood glucose between control genotypes, the data were combined in our analysis. Insulin-reduced mice were generated by injecting Ins1−/−:Ins2f/f:Pdx1CreERT:mTmG mice with tamoxifen (3 mg/40 g body weight, dissolved in corn oil, for 4 consecutive days) at 6–8 weeks [54].
## RNA sequencing
To assess basal transcriptional differences, islets from male and female mice ($$n = 9$$M, 8F) were snap-frozen and stored at −80°C until RNA extraction. To assess Tg-induced transcriptional changes, islets from each mouse were treated with DMSO or 1 μM Tg for 6- or 12-hours in culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S). Four groups per sex (eight groups total) were analyzed: 6 h DMSO, 6 h Tg, 12 h DMSO and 12 h Tg, $$n = 3$$–4 per group, each n represents pooled islet RNA from two mice. Islets were frozen at −80°C in 100 μL of RLT buffer (Qiagen) with beta mercaptoethanol ($1\%$). RNA was isolated using RNeasy Mini Kit (Qiagen #74106) according to manufacturer's instructions. RNA from 43 to 62 islets was pooled from two mice and 19–150 ng of RNA was sequenced per pooled sample. RNA sequencing was performed at the UBC Biomedical Research Centre Sequencing Core. Sample quality control was performed using the Agilent 2100 Bioanalyzer System (RNA Pico LabChip Kit). Qualifying samples were prepped following the standard protocol for the NEBNext Ultra II Stranded mRNA (New England Biolabs). Sequencing was performed on the Illumina NextSeq 500 with Paired End 42bp × 42bp reads. Demultiplexed read sequences were then aligned to the reference sequence (UCSC mm10) using STAR aligner (v 2.5.0b) [55]. Gene differential expression was analyzed using DESeq2 R package [56]. Pathway enrichment analysis were performed using Reactome [57]. Over-representation analysis was performed using NetworkAnalyst3.0 (www.networkanalyst.ca) [58].
## Proteomics
Islets were treated with DMSO or 1 μM Tg for 6 h in islet culture media (11.1 mM d-glucose RPMI, $1\%$ vol/vol P/S). Two groups per sex (four groups total): 6 h DMSO and 6 h Tg, $$n = 5$$–7 per group, each n represents 200–240 islets pooled from two mice. Islet pellets were frozen at −80°C in 100 μL of SDS lysis buffer ($4\%$ SDS, 100 mM Tris, pH 8) and the proteins in each sample were precipitated using acetone. The University of Victoria proteomics service performed non-targeted quantitative proteomic analysis using data-independent acquisition (DIA) with LC-MS/MS on an Orbitrap mass spectrometer using 1 μg of protein. A mouse FASTA database was downloaded from Uniprot (http://uniprot.org). This file was used with the 6 gas phase fraction files from the analysis of the chromatogram library sample to create a mouse islet-specific chromatogram library using the EncyclopeDIA (v 1.2.2) software package (Searle et al., 2018). This chromatogram library file was then used to perform identification and quantitation of the proteins in the samples again using EncyclopeDIA with Overlapping DIA as the acquisition type, trypsin used as the enzyme, CID/HCD as the fragmentation, 10 ppm mass tolerances for the precursor, fragment, and library mass tolerances. The Percolator version used was 3.10. The precursor FDR rate was set to $1\%$. Protein abundances were log2 transformed, imputation was performed for missing values, then proteins were normalized to median sample intensities. Differential expression was analyzed using limma in Perseus [59].
## Analysis of the transcriptome and partial proteome
Tg-induced changes to gene expression and protein levels were compared 6 h post-treatment. Log2-transformed fold change values were used to assess the congruence between our proteomics data and RNAseq data. Genes and proteins that were concordantly altered by Tg treatment at both the mRNA and protein level were searched in PubMed for relevant literature on their role in β cells. The search term used was ((“beta cell”) OR (islet) OR (“β cell”)) AND (Gene_Name). Additional annotations for all mouse proteins were downloaded from Uniprot [Dec 2022].
## Data from HPAP
To compare sex differences in dynamic insulin secretion, data acquired was from the Human Pancreas Analysis Program (HPAP-RRID:SCR_016202) Database (https://hpap.pmacs.upenn.edu), a Human Islet Research Network (RRID:SCR_014393) consortium (UC4-DK-112217, U01-DK-123594, UC4-DK-112232, and U01-DK-123716).
## Statistical analysis
Statistical analyses and data presentation were carried out using GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA) or R (v 4.1.1). Correlation plots were generated using the corrplot R package (v 0.92) with default settings [60]. All R codes are published on github (https://github.com/johnsonlabubc/ER-Stress-in-Mouse-Beta-Cells-Data-Analysis). Student's t-tests or two-way ANOVAs were used for parametric data. A Mann–Whitney test was used for non-parametric data. Statistical tests are indicated in the figure legends. For all statistical analyses, differences were considered significant if the p-value was less than 0.05. ∗: $p \leq 0.05$; ∗∗$p \leq 0.01$; ∗∗∗$p \leq 0.001.$ Data were presented as means ± SEM with individual data points from biological replicates.
## Sex differences in β cell transcriptional and functional responses in ND and T2D human islets
Gene expression studies on human pancreas and islets identify significant sex differences in gene expression [7,10]. Indeed, >1500 genes expressed in the pancreas show sex-biased expression [7]. While our dataset contained fewer individuals, we also note sex differences in mRNA levels of many genes in β cells (Supplementary file 1). Given this differential gene expression, we wanted to define β cell-specific gene expression changes in T2D in each sex. We therefore used a recently-compiled meta-analysis of publicly available scRNAseq datasets from male and female human islets [61]. Our goal was to use sex-based analysis to determine whether β cell gene expression changes in T2D are shared between the sexes. In line with prior reports [12], β cells from non-diabetic (ND) and T2D donors showed significant transcriptional differences. In β cells isolated from female T2D donors, mRNA levels of 127 genes were significantly different from ND female donors (77 downregulated, 50 upregulated in T2D) (Figure 1A–C). In β cells isolated from male T2D donors, 462 genes were differentially expressed compared with male ND donors (138 downregulated, 324 upregulated in T2D) (Figure 1A–C). Of the 660 genes that were differentially regulated in T2D, 71 were differentially regulated in both males and females (15 downregulated, 56 upregulated in T2D) (Figs. S1A–C); however, the fold change for these 71 shared genes was different between males and females (Fig. S1A; Supplementary file 2). This suggests that for shared genes, the magnitude of gene expression changes in T2D was not the same between the sexes. Beyond shared genes, we observed that the majority of differentially expressed genes in T2D ($\frac{589}{660}$) were unique to either males or females (Figs. S1B,C; Supplementary file 2). Indeed, the most prominent gene expression changes in T2D were found in genes that were unique to one sex (Figs. S2A,B; Supplementary file 2). While these data do not address the reasons for the sex-biased risk of T2D, and could reflect differences in medication [[62], [63], [64], [65]], age [12,16,[66], [67], [68]], and body mass index [15,68], our data suggest biological sex influences β cell gene expression in T2D.Figure 1Sex differences in human islet transcriptomic and functional responses in type 2 diabetes. scRNAseq data from male and female human β cells. For donor metadata see Supplementary file 8. ( A–C) Venn diagrams compare the number of significantly differentially expressed genes between ND and T2D donors (p-adj<0.05). All differentially expressed genes (A), downregulated genes (B), upregulated genes (C) in T2D human β cells. For complete gene lists see Supplementary file 1 and 2. ( D–F) Top 10 significantly enriched Reactome pathways (ND vs T2D) from non-sex-specific (D), female (E), or male (F) significantly differentially expressed genes (p-adj< 0.05). Gene ratio is calculated as k/n, where k is the number of genes identified in each Reactome pathway, and n is the number of genes from the submitted gene list participating in any Reactome pathway. For complete Reactome pathway lists see Supplementary file 2. ( G–K) Human islet perifusion data from the Human Pancreas Analysis Program in ND and T2D donor islets in females (F, I) and males (G, H). 3 mM glucose (3 mM G); 16.7 mM glucose (16.7 mM G); 0.1 mM isobutylmethylxanthine (0.1 mM IBMX); 30 mM potassium chloride (30 mM KCl); 4 mM amino acid mixture (4 mM AAM; mM: 0.44 alanine, 0.19 arginine, 0.038 aspartate, 0.094 citrulline, 0.12 glutamate, 0.30 glycine, 0.077 histidine, 0.094 isoleucine, 0.16 leucine, 0.37 lysine, 0.05 methionine, 0.70 ornithine, 0.08 phenylalanine, 0.35 proline, 0.57 serine, 0.27 threonine, 0.073 tryptophan, and 0.20 valine, 2 mM glutamine). ( I–K) Quantification of area under the curve (AUC) is shown for the various stimulatory media in females (I), males (J) and donors with T2D (K). ( I) In females, insulin secretion from ND islets was not significantly higher than T2D islets under any culture condition ($$p \leq 0.4806$$ [AAM + LG], $$p \leq 0.2270$$ [AAM + HG], $$p \leq 0.1384$$ [AAM + HG + IBMX], and $$p \leq 0.1465$$ [KCl]; unpaired Student's t-test). ( J) In males, insulin secretion from ND islets was significantly higher than T2D islets under 4 mM AAM +16.7 mM glucose (HG) + 0.1 mM IBMX stimulation ($$p \leq 0.0442$$ [AAM + HG + IBMX]; unpaired Student's t-test), but not in other conditions ($$p \leq 0.5315$$ [AAM + LG], $$p \leq 0.0818$$ [AAM + HG], and $$p \leq 0.2259$$ [KCl]; unpaired Student's t-test). ( K) Total insulin secretion showed a trend toward lower secretion in T2D male islets than ND male islets ($$p \leq 0.1514$$ and $$p \leq 0.0503$$ for females and males, respectively; unpaired Student's t-test). ∗ indicates $p \leq 0.05$; ns indicates not significant; error bars indicate SEM.Figure 1 To determine which biological pathways were altered in β cells of T2D donors from each sex, we performed pathway enrichment analysis. Genes that were upregulated in β cells isolated from T2D donors included genes involved in Golgi-ER transport and the unfolded protein response (UPR) pathways (Figure 1D–F; Supplementary file 2). While these biological pathways were significantly upregulated in T2D in both males and females, ∼$75\%$ of the differentially regulated genes in these categories were unique to each sex (Table 1). Genes that were downregulated in β cells from T2D donors revealed further differences between the sexes: biological pathways downregulated in β cells from female T2D donors included cellular responses to stress and to stimuli (Figure 1E; Supplementary file 2), whereas β cells from male T2D donors showed downregulation of pathways associated with respiratory electron transport and translation initiation (Figure 1F; Supplementary file 2). Our analysis therefore suggests that sex-biased β cell gene expression responses to T2D may influence different cellular processes in males and females. Table 1Human β cell pathway gene numbers. The number of genes corresponding to each T2D upregulated pathway in males, females or both sexes. Table 1Pathway NameNumber of Pathway GenesUnique MaleCommonUnique FemaleAsparagine N-linked glycosylation1694Cellular responses to stimuli4996Cellular responses to stress4796COPI-dependent Golgi-to-ER retrograde traffic762COPI-mediated anterograde transport752ER to Golgi Anterograde Transport1052Golgi-to-ER retrograde transport762Hedgehog ligand biogenesis1241Hh mutants abrogate ligand secretion1231Hh mutants are degraded by ERAD1231IRE1alpha activates chaperones931Metabolism of proteins692013Signaling by Hedgehog1662The role of GTSE1 in G2/M progression after G2 checkpoint1441Unfolded Protein Response (UPR)1331XBP1(S) activates chaperone genes831 The sex-biased β cell transcriptional response in T2D prompted us to compare glucose-stimulated insulin secretion in each sex from ND and T2D human islets using data from the Human Pancreas Analysis Program database [69]. In ND donors, islets from males and females showed similar patterns of insulin secretion in response to various stimulatory media (Figure 1G,H). In donors with T2D, we found that insulin secretion was impaired to a greater degree in islets from males than in females (Figure 1G–K). This difference cannot be fully attributed to a sex difference in disease severity, as our analysis of donor characteristics revealed no significant correlation between sex and HbA1c (Fig. S3A). Indeed, in male but not female islets, insulin secretion was lower in donors with T2D following stimulation with both high glucose and IBMX (Figure 1I,J), which potentiates insulin secretion by increasing cAMP levels to a similar degree as the incretins [70]. Human islets from female donors with T2D therefore show a greater ability to maintain glucose-stimulated insulin secretion than islets from males with T2D (Figure 1K). Indeed, while diabetes status was the main donor characteristic that correlated with changes in insulin secretion (Fig. S3B), we noted that in T2D sex and age were two donor characteristics showing trends toward an effect on insulin secretion (Fig. S3A). Combined with our β cell gene expression data, these findings suggest that β cell transcriptional and functional responses in T2D are not shared between the sexes.
## Sex differences in UPR-associated gene expression in mouse islets
Our unbiased analysis of human β cell gene expression and function in T2D revealed differences between male and female donors with T2D. Because human β cell gene expression and function can be affected by factors such as peripheral insulin sensitivity, disease processes, and medication [31,33], we investigated sex differences in β cell gene expression and function in another context. *We* generated a well-powered islet RNAseq dataset from 20-week-old male and female C57BL/6J mice. We used an insulin tolerance test (ITT) to show that insulin sensitivity was similar between the sexes at this age (Fig. S4A); however, we acknowledge that the ITT may not be as sensitive as a hyperinsulinemic-euglycemic clamp in detecting modest sex differences in insulin sensitivity.
Principal component analysis and unsupervised clustering clearly separated male and female islets on the basis of gene expression (Figure 2A; Fig. S5A). We found that $17.7\%$ ($\frac{3268}{18938}$) of genes were differentially expressed between the sexes (1648 upregulated in females, 1620 upregulated in males), in line with estimates of sex-biased gene expression in other tissues [71,72]. Overrepresentation and pathway enrichment analysis both identified UPR-associated pathways as a biological process that differed significantly between the sexes, where the majority of genes in this category were enriched in female islets (Figure 2B,C; Supplementary file 3). *Additional* genes that were enriched in female islets were those associated with the gene ontology term “Cellular response to endoplasmic reticulum stress” (GO:0034976), which included many genes involved in regulating protein synthesis (Figure 2D). For example, females showed significantly higher levels of most ribosomal protein genes (Figure 2E). *Further* genes enriched in females included those associated with protein folding, protein processing, and quality control (Figure 2D). Given that protein synthesis, processing, and folding capacity are intrinsically important for multiple islet cell types [[73], [74], [75], [76]], including β cells [77,78], this suggests female islets may have a larger protein production and folding capacity than male islets. Figure 2Sex-biased gene expression in mouse islet bulk RNAseq. ( A) Principal component analysis (PCA) of RNAseq data from male and female mouse islets. ( B) Over-representation analysis (ORA) of all significantly differentially expressed genes (p-adj <0.01) from male and female mouse islets. Top 30 enriched KEGG pathways (large nodes; size = proportional to connections, darker red color = greater significance) and associated genes (small nodes; green = male enriched, yellow = female enriched). ( C) Top significantly enriched Reactome pathways from the top 1000 significantly differentially expressed genes. ( p-adj <0.01) for males and females. Gene ratio is calculated as k/n, where k is the number of genes identified in each Reactome pathway, and n is the number of genes from the submitted gene list participating in any Reactome pathway. For complete Reactome pathway lists see Supplementary file 3. ( D) All transcripts of differentially expressed genes under the gene ontology term “Cellular response to ER stress” (GO:0034976) and genes labeled by their role in transcription, translation, protein processing, protein folding, secretion and protein quality control. ( E) All transcripts of differentially expressed ribosomal genes. Figure 2
## Female islets are more resilient to endoplasmic reticulum stress in mice
The burden of insulin production causes endoplasmic reticulum (ER) stress in β cells [[79], [80], [81]]. ER stress is associated with an attenuation of mRNA translation [82], and, if ER stress is prolonged, can lead to cell death [[83], [84], [85]]. Given that female islets exhibited higher expression of genes associated with protein synthesis, processing, and folding than males, and higher expression of genes associated with the UPR, which is activated in response to ER stress [86], we examined global protein synthesis rates in male and female islets under basal conditions and during ER stress. We incubated islets with O-propargyl-puromycin (OPP), which is incorporated into newly-translated proteins and can be ligated to a fluorophore. Using this technique, we monitored the accumulation of newly-synthesized islet proteins with single-cell resolution (Fig. S6A). In basal culture conditions, male islet cells had significantly greater protein synthesis rates compared with female islet cells (Fig. S6B). To investigate islet protein synthesis under ER stress in each sex, we treated islets with thapsigargin (Tg), a specific inhibitor of the sarcoplasmic/endoplasmic reticulum Ca2+-ATPase (SERCA) that induces ER stress and the UPR by lowering ER calcium levels [83,87]. At 2 h post-Tg treatment, protein synthesis was repressed as expected in both male and female islet cells (Figure 3A,B; Fig. S6C). At 24 h post-Tg treatment, we found that protein synthesis was restored to higher-than basal levels in female islet cells, but not in male islet cells (Figure 3A,B; Fig. S6C). Importantly, a two-way ANOVA showed that recovery from protein synthesis repression was significantly different between males and females (sex:treatment interaction $p \leq 0.0001$). This suggests that while protein synthesis repression associated with ER stress was transient in female islets, this phenotype persisted for longer in male islets. Because insulin biosynthesis accounts for approximately half the total protein production in β cells [88], one potential explanation for the sex-specific recovery from protein synthesis repression is a sex difference in transcriptional changes to insulin. To test this, we quantified GFP levels in β cells isolated from male and female mice with GFP knocked into the endogenous mouse Ins2 locus (Ins2GFP/WT) [51,89]. While ER stress induced a significant reduction in *Ins2* gene activity, this response was equivalent between the sexes. This suggests Ins2 transcriptional changes cannot fully explain the sex difference in recovery from protein synthesis repression during ER stress (Figs. S7A–C).Figure 3Sex differences in mouse islet ER stress-associated phenotypes. ( A, B) *Protein synthesis* was quantified in dispersed islet cells from 20-week-old male and female B6 mice after treatment with 1 μM Tg for 2- or 24-hours. ( A) In female islet cells, protein synthesis was significantly lower after a 2-hour Tg treatment compared to control ($p \leq 0.0001$; one-way ANOVA followed by Tukey HSD test), significantly higher after a 24-hour Tg treatment compared to a 2-hour Tg treatment ($p \leq 0.0001$; one-way ANOVA followed by Tukey HSD test) and recovered to a significantly higher level than control levels $p \leq 0.0001$; one-way ANOVA followed by Tukey HSD test). ( B) In male islet cells, protein synthesis was significantly lower after a 2- and 24-hour Tg treatment compared to control ($p \leq 0.0001$ for both treatments; one-way ANOVA followed by Tukey HSD test) and was not significantly different after a 24-hour treatment compared to a 2-hour Tg treatment $$p \leq 0.3022$$; one-way ANOVA followed by Tukey HSD test). The magnitude of protein synthesis repression and recovery was significantly different in all sex:treatment interactions ($$p \leq 0.0015$$ [DMSO-2hr], $p \leq 0.0001$ [DMSO-24hr], $p \leq 0.0001$ [2hr–24hr]; two-way ANOVA followed by Tukey HSD test). ( C–H) Quantification of propidium iodide (PI) cell death assay of dispersed islets from 20-week-old male and female B6 mice treated with thapsigargin (0.1 μM, 1 μM or 10 μM Tg) or DMSO for 84 h $$n = 4$$–6 mice, >1000 cells per group. Percentage (%) of PI positive cells was quantified as the number of PI-positive/Hoechst 33342-positive cells in female (C) and male (D) islet cells. Relative cell death at 84 h in Tg treatments compared with DMSO treatment in females (E, G) and males (F, H). The control for both 0.1 and 1.0 μM Tg treatments is $0.1\%$ DMSO (E, F). The control for 10 μM Tg treatment is $0.2\%$ DMSO (G, H). In female islet cells, cell death was significantly higher in 10 μM Tg compared to control ($p \leq 0.0001$; unpaired Student's t-test). In male islet cells, cell death was significantly higher in 0.1, 1.0 and 10 μM Tg compared to control ($$p \leq 0.0230$$ [0.1 μM], $p \leq 0.0001$ [1 μM] and $p \leq 0.0001$ [10 μM]; unpaired Student's t-test) (D). For E-H, at 84 h the % of PI positive cells for each treatment was normalized to the DMSO control avg for each sex. ∗ indicates $p \leq 0.05$, ∗∗ indicates $p \leq 0.01$, ∗∗∗∗ indicates $p \leq 0.0001$; ns indicates not significant; error bars indicate SEM.Figure 3 Given the prolonged protein synthesis repression in males following ER stress, we next quantified cell death, another ER stress-associated phenotype [86], in male and female islets. Using a kinetic cell death assay, we observed clear sex differences in Tg-induced cell death at 0.1 μM and 1.0 μM Tg doses throughout the time course of the experiment (Figure 3C,D). Notably, viability prior to the assay was not different between males and females (Fig. S8). After 84 h of Tg treatment, no significant increase in female islet cell apoptosis was observed with either 0.1 μM or 1.0 μM Tg treatment compared with controls (Figure 3E). In contrast, cell death was significantly higher at both the 0.1 μM and the 1.0 μM doses of Tg in male islet cells compared with vehicle-only controls (Figure 3F). Our analysis shows the magnitude of Tg-induced cell death was larger in male islet cells compared with female islet cells (sex:treatment interaction $$p \leq 0.0399$$ [0.1 μM], $$p \leq 0.0007$$ [1.0 μM]). While one possible explanation for these data is that female islets are resistant to Tg-induced cell death, we found a significant increase in apoptosis in both female and male islet cells treated with 10 μM Tg (Figure 3G,H, sex:treatment interaction $$p \leq 0.0996$$ [0.1 μM]; data graphed separately due to different DMSO control). This suggests female islets were more resilient to mild ER stress caused by low-dose Tg than male islets.
To determine whether this increased ER stress resilience was caused by differential UPR signaling, we monitored levels of several protein markers of UPR activation including binding immunoglobulin protein (BiP), phosphorylated inositol-requiring enzyme 1 (pIRE1), phosphorylated eukaryotic initiation factor alpha (peIF2α), and C/EBP homologous protein (CHOP) [90,91] after treating male and female islets with 1 μM Tg for 24 h. We found no sex difference in UPR protein markers between male and female islets without Tg treatment (Figs. S9A–D) and observed a significant increase in levels of pIRE1α and CHOP in islets from both sexes and BiP in female islets after a 24-hour Tg treatment (Figs. S9A–D). Lack of a sex difference in protein markers suggests UPR activation by Tg treatment was similar between male and female islets at 20 weeks of age. This finding differs from the male-biased UPR activation reported in the KINGS mouse model of endogenous ER stress [37]. While one potential explanation for this discrepancy is that Tg treatment induces acute ER stress in contrast to the chronic ER stress in KINGS mice, further experiments will be needed to confirm this possibility. Of note, we reproduced the male-biased induction of BiP in islets isolated from 60-week-old male and female mice (Figs. S9E–G), suggesting that age contributes to the sex difference in UPR activation. Together, our data indicate that despite equivalent UPR activation in male and female islets treated with Tg, significant sex differences exist in ER stress-associated protein synthesis repression and cell death.
## Female islets retain greater β cell function during ER stress in mice
We next examined glucose-stimulated insulin secretion in islets cultured under basal conditions and during Tg treatment (Figure 4A). In all conditions tested, high glucose significantly stimulated insulin secretion in both sexes (Fig. S10A); however, we identified sex differences in how well islets sustained glucose-stimulated insulin secretion during longer Tg treatments (Figure 4B,C, Fig. S10A,B). Female islets, in both low and high glucose, maintained robust insulin secretion during Tg treatment (Figure 4B). Specifically, we observed a significant increase in insulin secretion after short Tg treatment (0 and 2 h post-Tg), with a return to basal secretion levels 4 h post-Tg (Figure 4B). Because *Tg is* a drug that depletes ER calcium stores, it may induce an acute rise in cytosolic calcium that could explain this acute increase in high glucose-stimulated insulin secretion in Tg-treated samples compared with vehicle [83]. In contrast, male islets showed no significant increase in insulin secretion after short Tg treatment, and there was a significant drop in insulin secretion at 4 h post-Tg treatment (Figure 4C). This suggests female islets sustained insulin secretion for a longer period than male islets during ER stress. Given that insulin content measurements showed insulin content significantly increased during the 4-hour Tg treatment in female islets, but not male islets (Figure 4D, Fig. S10C), our data suggest one reason female islets maintain insulin secretion during ER stress is by augmenting islet insulin content. Proinsulin secretion followed similar trends to those we observed with insulin secretion (Figs. S10D,E), but Tg treatment reduced proinsulin content to a greater degree in male islets (Figure 4E). There was no sex difference in the ratio of proinsulin:insulin content at any timepoint (Fig. S10G). This suggests that in addition to a greater ability to maintain glucose-stimulated insulin secretion during ER stress, female islets also show a larger increase in insulin content and a smaller decrease in proinsulin content in this context. Figure 4Sex differences in ex vivo and in vivo insulin secretion. ( A) Experimental workflow of static glucose-stimulated insulin secretion. ( B, C) Relative high glucose (20 mM; high glucose, HG) in treatments compared with DMSO in female (B) and male (C) islets. Female islet HG secretion was significantly higher compared with control after 0- and 2-hour Tg pre-treatments ($$p \leq 0.0083$$ [0-hour] and $$p \leq 0.0371$$ [2-hour]; Mann Whitney test). Male islet HG secretion was significantly lower compared with control after a 4-hour Tg pre-treatment ($$p \leq 0.0013$$; Mann Whitney test). ( D) Insulin content. Female islet insulin content was significantly higher compared with control after a 4-hour Tg pre-treatment ($$p \leq 0.0269$$; Mann Whitney test). ( E) Proinsulin content. Female islet proinsulin content was significantly lower compared with control after a 2-hour Tg pre-treatment ($$p \leq 0.0437$$; Mann Whitney test). Male islet proinsulin content was significantly lower compared with control after 2- and 4-hour Tg pre-treatments ($$p \leq 0.0014$$ [2-hour] and $$p \leq 0.0005$$ [4-hour]; Mann Whitney test). ( F–H) Physiology measurements after a 6-hour fast in 20-week-old male and female B6 mice. ( F, G) Insulin levels from glucose-stimulated insulin secretion tests (F: nM, G: % basal insulin) following a single glucose injection (2 g glucose/kg body weight, i.p). Area under the curve (AUC) calculations ($$n = 13$$ females, $$n = 18$$ males). ( F) Insulin levels were significantly higher in male mice at 0 min and 30 min post injection ($$p \leq 0.0063$$ [0 min] and $$p \leq 0.0009$$ [30 min]; Student's t-test). AUC was significantly higher in males ($$p \leq 0.0159$$; Student's t-test). ( G) Insulin levels (% baseline). Glucose-stimulated insulin secretion was significantly higher in female mice 15 min post injection ($$p \leq 0.0279$$; Student's t-test). ( H) Glucose levels from glucose tolerance tests following a single glucose injection (2 g glucose/kg body weight). AUC calculations ($$n = 11$$ females, $$n = 11$$ males). For B-E, grey triangles indicate the concentration of insulin or proinsulin from five islets, black circles indicate the average values per mouse. ∗ indicates $p \leq 0.05$, ∗∗ indicates $p \leq 0.01$, ∗∗∗ indicates $p \leq 0.001$; ns indicates not significant; error bars indicate SEM.Figure 4 To determine whether female islets have improved β cell function under ER stress in other contexts, we next monitored glucose-stimulated insulin secretion and glucose tolerance in mice at 20 weeks, an age where we demonstrated that insulin sensitivity was similar between the sexes (Figure 4F–H; Fig. S4). We found that fasting plasma insulin levels were higher in males (Figure 4F) and that the sexes showed similar glucose tolerance (Figure 4H). To determine the magnitude of the acute glucose-stimulated insulin secretion response in each sex, we normalized glucose-stimulated insulin secretion to basal insulin secretion levels. We found that females showed a trend toward higher acute glucose-stimulated insulin secretion response in C57BL/6J mice (Figure 4G), and significantly higher glucose-stimulated insulin secretion in 10- and 22-week-old Insrf/f:Ins1Cre−/−:nTnG+/− mice (Figs. S11A–D; data replotted from a prior Johnson lab study [53]). These findings align with data showing higher glucose-stimulated insulin secretion from prior studies in humans and rodents [5,10,16,92,93]. We also found that fasting blood glucose levels in female mice were more resilient to the near-total insulin gene knockout in Ins1−/−;Ins2fl/fl;Pdx1CreER mice given tamoxifen (Fig. S11E; replotted from published [54] and unpublished Johnson lab data). We cannot rule out all potential factors that may contribute to the sex differences in blood glucose levels following near-total loss of insulin gene function (e.g. RNA stability, translation efficiency); however, given that the burden of insulin production [54] leads to ER stress even in normal physiological conditions, our data add further support to a model in which β cells in female mice show a greater ability to maintain glucose-stimulated insulin production and secretion during ER stress.
## Sex differences in islet transcriptional and proteomic responses to ER stress in mice
To gain insight into the differential ER stress-associated phenotypes in male and female islets, we investigated global transcriptional changes after either a 6- or 12-hour Tg treatment in each sex. Principal component analysis and unsupervised clustering shows that islets clustered by sex, treatment, and treatment time (Figure 5A; Fig. S12A). The majority of the variance was explained by treatment (Figure 5B), and pathway enrichment analysis confirms the UPR as the top upregulated pathway in Tg-treated male and female islets at both 6- and 12-hours after treatment (Figs. S13A,B; Supplementary file 4). While some UPR-associated genes differentially regulated by Tg treatment were shared between the sexes (6-hour: $\frac{29}{36}$, 12-hour: $\frac{25}{31}$), biological sex explained a large proportion of variance in the gene expression response to ER stress. This suggests the transcriptional response to ER stress was not fully shared between the sexes. Indeed, after a 6-hour Tg treatment, $32.6\%$ ($\frac{2247}{4655}$) of genes that were differentially expressed between DMSO and Tg were unique to one sex (881 to females, 1376 to males). After a 12-hour Tg treatment, $29\%$ ($\frac{2259}{7785}$) were unique to one sex (1017 to males, 1242 to females).Figure 5Sex-specific transcriptomic and proteomic profiles following ER stress in mouse islets. ( A) Principal component analysis (PCA) of RNAseq data from male and female mouse islets treated with DMSO or 1 μM Tg for 6- or 12-hours. ( B) Spearman correlation depicting the variance for the first 5 principal components. ( C) Top significantly enriched Reactome pathways from the top 1000 significantly differentially expressed genes (p-adj<0.01) for females and males that were upregulated or downregulated between 6 and 12 h of Tg treatment. Gene ratio is calculated as k/n, where k is the number of genes identified in each Reactome pathway, and n is the number of genes from the submitted gene list participating in any Reactome pathway. ( D) Protein abundance from proteomics data of female and male mouse islets treated with DMSO or 1 μM Tg for 6 h. Top 45 differentially expressed proteins are shown ($p \leq 0.05$).Figure 5 To describe the transcriptional response of each sex to Tg treatment in more detail, we used a two-way ANOVA to identify genes that were upregulated, downregulated, or unchanged in male and female islets between 6- and 12-hours post-Tg (Supplementary file 5). By performing pathway enrichment analysis, we were able to determine which processes were shared, and which processes differed, between the sexes during Tg treatment. For example, we observed a significant increase in mRNA levels of genes corresponding to pathways such as cellular responses to stimuli, stress, and starvation in both male and female islets between 6- and 12-hour Tg treatments (Figure 5C; Supplementary file 4), suggesting Tg has similar effects on genes related to these pathways in both sexes. Similarly, the Tg-induced changes in mRNA levels of genes related to apoptosis were largely shared between the sexes (Figs. S14A,B). While this suggests that the sex-specific regulation of genes related to apoptosis does not fully account for the susceptibility of male islets to low-dose Tg-induced cell death, in-depth studies of β cell apoptosis will be needed to confirm this point.
In contrast to these non-sex-specific changes in gene expression, there was a male-specific increase in mRNA levels of genes associated with translation during Tg treatment (Figure 5C; Supplementary file 4). In females, there was a decrease in mRNA levels of genes associated with β cell identity, such as Pklr, Rfx6, Hnf4a, Slc2a2, Pdx1, and MafA (Fig. S15A), and in genes linked with regulation of gene expression in β cells (Figure 5C). Neither of these categories were altered between 6- and 12-hour Tg treatments in males (Figure 5C; Fig. S15B). While our data suggests some aspects of the gene expression response to ER stress were shared between the sexes, we found that many genes corresponding to important cellular processes were differentially regulated during Tg treatment in only one sex.
Beyond sex-specific transcriptional changes following Tg treatment, ER stress also had a sex-specific effect on the islet proteome. Although the majority of proteins were downregulated by Tg treatment due to the generalized repression of protein synthesis under ER stress (Figure 5D), we identified 47 proteins (35 downregulated, 12 upregulated in Tg) that were differentially expressed in female islets and 82 proteins (72 downregulated, 10 upregulated in Tg) that were differentially expressed after Tg treatment in male islets (Supplementary Table 1). Proteins downregulated only in females include proteins associated with the GO term ‘endoplasmic reticulum to Golgi vesicle-mediated transport’ (GO:0006888) (BCAP31, COG5, COG3, GOSR1), whereas proteins downregulated only in males include proteins associated with GO terms ‘insulin secretion’ (GO:0030073) (PTPRN2, CLTRN, PTPRN) and ‘lysosome pathway’ (KEGG) (NPC2, CTSZ, LAMP2, PSAP, CLTA). Importantly, only seven differentially expressed proteins were in common between the sexes (Figure 5D). This suggests that as with our phenotypic and transcriptomic data, the proteomic response to Tg treatment was largely not shared between the sexes.
To integrate our islet transcriptome and partial islet proteome data, we assessed the direction of changes to mRNA and protein levels following Tg treatment. In females, the number of differentially expressed islet proteins with concordant mRNA changes was $43\%$ ($\frac{20}{47}$), whereas the number of differentially expressed islet proteins in males with concordant mRNA changes was $49\%$ ($\frac{40}{82}$) (Fig. S16; Supplementary file 6). This data suggests that many genes with differential mRNA expression during Tg treatment show congruent changes in protein abundance. When we next asked whether the islet genes with concordant changes in mRNA and protein levels during Tg treatment were shared between the sexes, we found that only $7\%$ ($\frac{4}{56}$) of these genes were differentially expressed in both sexes during Tg treatment (Tmem27, Emb, Fkbp9, Pdcd4). The remaining $93\%$ ($\frac{51}{55}$) of islet genes with concordant changes in mRNA and protein levels during Tg treatment were differentially expressed in only one sex. Thus, when taken together, our islet transcriptome and partial proteome data suggest that male and female islets show distinct responses to ER stress.
Of the islet genes with concordant changes in mRNA and protein levels, several have been linked with β cell and/or islet function. For example, Tmem27 plays a role in enhancing GSIS [94], and Pdcd4 expression is associated with β cell death under stressed conditions [95,96]. Atp6ap2 and Lamp2 have important roles in autophagy [[97], [98], [99]], Ptprn2 is required for the accumulation of insulin granules [100], and both Ptprn2 and Chga are involved in glucose-stimulated insulin secretion [101,102]. Chga is also important for maintaining islet volume and islet cell composition [103], and studies show that Tbc1d1 influences insulin secretion and β cell mass in rodents [104,105]. Because the Tbc1d1 study [104,105], and others [[95], [96], [97], [98], [99], [100], [101], [102], [103], [104]], used single- or mixed-sex models, or cell lines, future studies will need to address whether these effects on β cell and/or islet function are shared between the sexes. Additional studies will also be needed on genes identified in our analysis that were not previously linked with β cell and/or islet function (Supplementary file 6). This will elucidate whether the sex-specific regulation of mRNA and protein levels during Tg treatment affects β cell and/or islet function in either sex.
## Discussion
Emerging evidence shows biological sex affects many aspects of β cell gene expression and function. Yet, many studies on β cells do not include both sexes, or fail to analyze male and female data separately. To address this gap in knowledge, the goal of our study was to provide detailed information on sex differences in islet and β cell gene expression and function in multiple contexts. In humans, we used a large scRNAseq dataset from ND and T2D donors to reveal significant male-female differences in the magnitude of gene expression changes, and in the identity of genes that were differentially regulated, between ND and T2D donors. While these data do not address the reasons for the sex-biased risk of T2D, our findings suggest β cell gene expression changes in T2D are not fully shared between the sexes. These data provide a useful resource to support future studies on how medication, disease progression, age, or body mass index contribute to the sex difference in β cell gene expression in T2D. Sex-based analysis of human β cell gene expression data will also clarify the mechanisms underlying our finding that β cells from female donors with T2D maintain higher insulin production than male donors with T2D.
In mice, we generated a large RNAseq dataset using islets isolated from 20-week-old males and females. We used an ITT to show that male and female mice have similar insulin sensitivity at this age. We acknowledge that the ITT may not be sensitive enough to pick up small differences, so future studies will need to use a hyperinsulinemic-euglycemic clamp to further compare insulin sensitivity between the sexes. Despite this potential limitation, our unbiased analysis of gene expression in islets from males and females revealed sex differences in genes associated with the UPR under normal physiological conditions. This differential gene expression was significant, as female islets were more resilient to phenotypes caused by ER stress and UPR activation than male islets, showed sex-specific transcriptional and proteomic changes in this context, and had a greater ability to maintain glucose-stimulated insulin production and secretion during ER stress. Collectively, these data suggest that in rodents, β cells from females are more resilient to ER stress. Considering the well-established links between ER stress and T2D [90,[107], [108], [109]], our data suggests a model in which female β cells have a greater ability to maintain glucose-stimulated insulin secretion in T2D because they are more resilient to ER stress and UPR activation. While future studies are needed to test this working model, and to assess the relative contribution of sex differences in β cells to the sex-biased risk of T2D, our findings highlight the importance of including both sexes in islet and β cell studies.
Including both sexes in our analysis of β cell gene expression in human ND and T2D allowed us to uncover genes that were differentially regulated in T2D in each sex. Because many of these genes may have been missed if the scRNAseq data was not analyzed by sex, our findings advance knowledge of β cell changes in T2D by identifying additional genes that are differentially regulated in this context. This knowledge adds to a growing number of studies that identify sex differences in β cell gene expression during aging in humans [12], and in mice fed either a normal [4,11] or a high fat diet [11]. Further, given that our RNAseq on islets from male and female mice with similar insulin sensitivity identifies genes and biological pathways that align with previous studies on sex differences in murine β cell gene expression [4,11], our data suggests that sex differences in islet and β cell gene expression cannot be explained solely by a male-female difference in peripheral insulin resistance. Instead, there is likely a basal sex difference in β cell gene expression that forms the foundation for sex-specific transcriptional responses to perturbations such as ER stress and T2D. *By* generating large islet gene expression datasets from male and female mice with similar peripheral insulin sensitivity and from islets subjected to pharmacological induction of ER stress, our studies provide a foundation of knowledge for future studies aimed at studying the causes and consequences of sex differences in islet ER stress responses and β cell function following UPR activation. This will provide deeper mechanistic insight into the sex-specific phenotypic effects reported in animal models of β cell dysfunction [[35], [36], [37], [38], [39],[110], [111], [112], [113]] and the sex-biased risk of diseases such as T2D that are associated with β cell dysfunction [12,22,113,114].
*Beyond* gene expression, our sex-based analysis of mouse islets allowed us to uncover male-female differences in ER stress-associated phenotypes (e.g. protein synthesis repression, cell death). While previous studies identify a sex difference in β cell loss in diabetic mouse models [37,39,115], and show that estrogen plays a protective role via estrogen receptor α (ERα) against ER stress to preserve β cell mass and prevent apoptosis in cell lines, mouse models, and human islets [39,115,116], we extend prior findings by showing that differences in ER stress-induced cell death were present in the context of similar insulin sensitivity between the sexes. This suggests sex differences in ER stress-associated phenotypes do not solely depend on male-female differences in peripheral insulin sensitivity. Indeed, islets isolated from males and females with similar insulin sensitivity also show a sex difference in protein synthesis repression, a classical ER stress-associated phenotype [86]. While estrogen affects insulin biosynthesis via ERα [117], future studies will need to determine whether estrogen also allows female islets to restore protein synthesis to basal levels faster than male islets following ER stress. We currently lack this knowledge, as most studies on UPR-mediated recovery from protein translation repression use single- and mixed-sex animal groups, or cultured cells [[119], [120], [121], [122], [123], [124]].
Assessing whether the recovery of protein synthesis contributes to reduced cell death in female islets following ER stress will also be an important task for future studies, as prior work suggests the inability to recover from protein synthesis repression increases ER-stress induced apoptosis [118]. Ideally, this type of study would also monitor the activity of pathways known to regulate protein synthesis repression during ER stress. For example, while we did not detect any changes in levels of phosphorylated eIF2α (also known as Eif2s1), which is known to mediate UPR-induced protein synthesis repression [86], our chosen timepoints did not overlap with the rapid changes in phospho-eIF2α following ER stress published in other studies [124,125]. A more detailed time course will therefore be necessary to assess p-eIF2α levels during ER stress in both sexes, and to test a role for phospho-eIF2α in mediating differences in protein synthesis repression. More work will also be needed to determine whether males and females differ in β cell replication [126], another ER stress-related phenotype. Ultimately, a better understanding of sex differences in ER stress-associated phenotypes in β cells will provide a mechanistic explanation for the strongly male-biased onset of diabetes-like phenotypes in mouse models of β cell ER stress (e.g. Akita, KINGS, Munich mice) [37,38,109]. Given the known relationship between ER stress, β cell death, and T2D, studies on the male-female difference in β cell ER stress-associated phenotypes may also advance our understanding of the male-biased risk of developing T2D in some population groups.
A further benefit of additional studies on the sex difference in β cell ER stress responses will be to identify mechanisms that support β cell insulin production. In rodents, we found that female islets maintained high glucose-stimulated insulin secretion and increased insulin content following ER stress, whereas male islets showed significant repression of high glucose-stimulated insulin secretion under the same conditions. In humans, while a study using a mixed-sex group of T2D donors shows β cells experience ER stress associated with β cell dysfunction [63], we found that changes to β cell insulin secretion in T2D were not the same between the sexes. Specifically, the magnitude of the reduction in insulin release by β cells from female donors with T2D was smaller than in β cells from male donors with T2D. Together with our data from rodents, this suggests female β cells maintain enhanced insulin production and/or secretion in multiple contexts, and the increased β cell function cannot be solely attributed to a sex difference in peripheral insulin sensitivity.
Clues into potential ways that female β cells maintain improved insulin production and secretion emerge from our examination of the transcriptional response to ER stress in mice of each sex. Our data shows that Tg treatment induces gene expression changes characteristic of ER stress [127], and revealed similar biological pathways that were upregulated in T2D donors. Furthermore, we identified significant differences between male and female islets in the transcriptional response to ER stress over time. One notable finding was that a greater number of β cell identity genes were downregulated between 6- and 12-hour Tg treatments in females, but not in males. Because most studies on the relationship between β cell identity and function used a mixed-sex pool of islets and β cells [79,128,129], more studies will be needed to test whether there are sex-specific changes to β cell identity during ER stress, and to determine the functional consequences of this sex-specific effect.
Overall, our data demonstrates sex differences in β cell function in multiple contexts. One potential explanation for these differences is the sex-specific regulation of β cell ER stress responses and function. Indeed, sex differences in ER stress and protein markers of apoptosis were observed in mouse kidney cells [130], suggesting that studying β cell ER stress may provide insight into this difference in other cell types. Alternatively, it is possible that there are sex differences in β cell number that account for the male-female differences in β cell function that we observe. For example, considering two published studies indicate males have fewer β cells [6,37], the burden of maintaining glucose homeostasis may fall on a smaller number of cells in males, leading to higher susceptibility to ER stress. Future studies will need to address sex differences in β cell number relative to pancreas size and body size to test this possibility. Ultimately, a better understanding of changes to β cell gene expression and function in males and females will suggest effective ways to reverse disease-associated changes to this important cell type in each sex, improving equity in health outcomes [131].
## Conclusions
Our study reports significant sex differences in islet and β cell gene expression and stress responses in both humans and mice. These differences likely contribute to sex differences in β cell resilience, allowing female β cells to show a greater ability to maintain glucose-stimulated insulin production and secretion across multiple contexts. This knowledge forms a foundation for future studies aimed at understanding how sex differences in β cell function affect physiology and the pathophysiology of diseases such as T2D.
## Author contributions
G. P. B. conceived studies, conducted experiments, interpreted experiments, wrote the manuscript Y. X. performed bioinformatic analysis and data visualization J. C. created custom R scripts (single-cell GFP tracking) S. W. analyzed data (human RNAseq)
C. C. created custom R scripts (mouse RNAseq analysis) J. A. Z. analyzed data (HPAP perifusions) S. S. conducted experiments (in vivo physiology) E. P. conducted experiments (islet western blots)
X. H. conducted experiments (dissections) J. D. J. conceived studies, interpreted experiments, edited the manuscript E. J. R. conceived studies, interpreted experiments, edited the manuscript, and is the guarantor of this work
## Funding
This study was supported by operating grants to E.J.R. from the $\frac{10.13039}{501100000245}$Michael Smith Foundation for Health Research [16876], $\frac{10.13039}{501100000024}$Canadian Institutes of Health Research (GS4-171365), the Canadian Foundation for Innovation (JELF-34879), and Diabetes Canada (OG-3-22-5646-ER), and to J.D.J. (PJT-152999) from the $\frac{10.13039}{501100000024}$Canadian Institutes of Health Research, and core support from the $\frac{10.13039}{100009881}$JDRF Centre of Excellence at $\frac{10.13039}{501100005247}$UBC (3-COE-2022-1103-M-B). J.D.J. was funded by a Diabetes Investigator Award from $\frac{10.13039}{100013528}$Diabetes Canada.
## Conflict of interest
The authors declare no competing interest.
## Supplementary data
The following are the *Supplementary data* to this article:Multimedia component 1Multimedia component 1Multimedia component 2Multimedia component 2
## Data availability
Details of all statistical tests and p-values as well as the raw data are provided in Supplementary files. Supplementary files are available as Brownrigg, George [2023], “Sex differences in islet stress responses support female β cell resilience”, Mendeley Data, V1, doi: https://doi.org/10.17632/ftcs6xj9ft.1. RNAseq data is available at PRJNA842443 and PRJNA842371.
## References
1. Parchami A., Kusha S.. **Effect of sex on histomorphometric properties of Langerhans islets in native chickens**. *Vet Res Forum* (2015) **6** 327-330. PMID: 26973769
2. Rideout E.J., Narsaiya M.S., Grewal S.S.. **The sex determination gene transformer regulates male-female differences in Drosophila body size**. *PLoS Genet* (2015 Dec 28) **11**
3. Millington J.W., Brownrigg G.P., Chao C., Sun Z., Basner-Collins P.J., Wat L.W.. **Female-biased upregulation of insulin pathway activity mediates the sex difference in Drosophila body size plasticity**. *Elife* (2021 Jan 15) **10**
4. Stancill J.S., Osipovich A.B., Cartailler J.P., Magnuson M.A.. **Transgene-associated human growth hormone expression in pancreatic β-cells impairs identification of sex-based gene expression differences**. *Am J Physiol Endocrinol Metab* (2019 Feb 1) **316** E196-E209. PMID: 30532991
5. Horie I., Abiru N., Eto M., Sako A., Akeshima J., Nakao T.. **Sex differences in insulin and glucagon responses for glucose homeostasis in young healthy Japanese adults**. *J Diabetes Investig* (2018 Nov) **9** 1283-1287
6. Marchese E., Rodeghier C., Monson R.S., McCracken B., Shi T., Schrock W.. **Enumerating β-cells in whole human islets: sex differences and associations with clinical outcomes after islet transplantation**. *Diabetes Care* (2015 Nov 1) **38** e176-e177. PMID: 26384388
7. Oliva M., Muñoz-Aguirre M., Kim-Hellmuth S., Wucher V., Gewirtz A.D.H., Cotter D.J.. **The impact of sex on gene expression across human tissues**. *Science* (2020 Sep 11) **369** eaba3066. PMID: 32913072
8. Enge M., Arda H.E., Mignardi M., Beausang J., Bottino R., Kim S.K.. **Single-cell analysis of human pancreas reveals transcriptional signatures of aging and somatic mutation patterns**. *Cell* (2017 Oct 5) **171** 321-330.e14. PMID: 28965763
9. Schaum N., Lehallier B., Hahn O., Pálovics R., Hosseinzadeh S., Lee S.E.. **Aging hallmarks exhibit organ-specific temporal signatures**. *Nature* (2020 Jul) **583** 596-602. PMID: 32669715
10. Hall E., Volkov P., Dayeh T., Esguerra J.L.S., Salö S., Eliasson L.. **Sex differences in the genome-wide DNA methylation pattern and impact on gene expression, microRNA levels and insulin secretion in human pancreatic islets**. *Genome Biol* (2014 Dec 3) **15** 522. PMID: 25517766
11. Liu G., Li Y., Zhang T., Li M., Li S., He Q.. **Single-cell RNA sequencing reveals sexually dimorphic transcriptome and type 2 diabetes genes in mouse islet β cells**. *Dev Reprod Biol* (2021 Jun 1) **19** 408-422
12. Arrojoe Drigo R., Erikson G., Tyagi S., Capitanio J., Lyon J., Spigelman A.F.. **Aging of human endocrine pancreatic cell types is heterogeneous and sex-specific**. *bioRxiv* (2019 Aug)
13. Li T., Jiao W., Li W., Li H.. **Sex effect on insulin secretion and mitochondrial function in pancreatic beta cells of elderly Wistar rats**. *Endocr Res* (2016 Jul 2) **41** 167-179. PMID: 26865180
14. Tura A., Pacini G., Moro E., Vrbíková J., Bendlová B., Kautzky-Willer A.. **Sex- and age-related differences of metabolic parameters in impaired glucose metabolism and type 2 diabetes compared to normal glucose tolerance**. *Diabetes Res Clin Pract* (2018 Dec 1) **146** 67-75. PMID: 30287229
15. Kautzky-Willer A., Brazzale A.R., Moro E., Vrbíková J., Bendlova B., Sbrignadello S.. **Influence of increasing BMI on insulin sensitivity and secretion in normotolerant men and women of a wide age span**. *Obesity* (2012) **20** 1966-1973. PMID: 22282046
16. Basu R., Man C.D., Campioni M., Basu A., Klee G., Toffolo G.. **Effects of age and sex on postprandial glucose metabolism: differences in glucose turnover, insulin secretion, insulin action, and hepatic insulin extraction**. *Diabetes* (2006 Jul 1) **55** 2001-2014. PMID: 16804069
17. Nuutila P., Knuuti M.J., Mäki M., Laine H., Ruotsalainen U., Teräs M.. **Gender and insulin sensitivity in the heart and in skeletal muscles: studies using positron emission tomography**. *Diabetes* (1995 Jan 1) **44** 31-36. PMID: 7813811
18. Lundsgaard A.M., Kiens B.. **Gender differences in skeletal muscle substrate metabolism – molecular mechanisms and insulin sensitivity**. *Front Endocrinol* (2014 Nov 13) **5** 195
19. Borissova A.M., Tankova T., Kirilov G., Koev D.. **Gender-dependent effect of ageing on peripheral insulin action**. *Int J Clin Pract* (2005) **59** 422-426. PMID: 15853858
20. Geer E.B., Shen W.. **Gender differences in insulin resistance, body composition, and energy balance**. *Gend Med* (2009 Jan 1) **6** 60-75. PMID: 19318219
21. Færch K., Borch-Johnsen K., Vaag A., Jørgensen T., Witte D.R.. **Sex differences in glucose levels: a consequence of physiology or methodological convenience? The Inter99 study**. *Diabetologia* (2010 May 1) **53** 858-865. PMID: 20182862
22. Gannon M., Kulkarni R.N., Tse H.M., Mauvais-Jarvis F.. **Sex differences underlying pancreatic islet biology and its dysfunction**. *Mol Metabol* (2018 Sep) **15** 82-91
23. Tramunt B., Smati S., Grandgeorge N., Lenfant F., Arnal J.F., Montagner A.. **Sex differences in metabolic regulation and diabetes susceptibility**. *Diabetologia* (2019 Nov 21). DOI: 10.1007/s00125-019-05040-3
24. Macotela Y., Boucher J., Tran T.T., Kahn C.R.. **Sex and depot differences in adipocyte insulin sensitivity and glucose metabolism**. *Diabetes* (2009 Apr) **58** 803-812. PMID: 19136652
25. Rudnicki M, Abdifarkosh G, Rezvan O, Nwadozi E, Roudier E, Haas TL. **Female Mice Have Higher Angiogenesis in Perigonadal Adipose Tissue Than Males in Response to High-Fat Diet**. *Front Physiol* (2018) **9** 1452. PMID: 30405427
26. Millington J.W., Biswas P., Chao C., Xia Y.H., Wat L.W., Brownrigg G.P.. **A low sugar diet enhances Drosophila body size in males and females via sex-specific mechanisms**. *Development* (2022 Feb 23)
27. Mank J.E., Rideout E.J.. **Developmental mechanisms of sex differences: from cells to organisms**. *Development* (2021 Oct 14) **148**
28. Oster R.T., Johnson J.A., Hemmelgarn B.R., King M., Balko S.U., Svenson L.W.. **Recent epidemiologic trends of diabetes mellitus among status Aboriginal adults**. *CMAJ (Can Med Assoc J)* (2011 Sep 6) **183** E803-E808. PMID: 21788417
29. Dyck R., Osgood N., Lin T.H., Gao A., Stang M.R.. **Epidemiology of diabetes mellitus among First Nations and non-First Nations adults**. *CMAJ (Can Med Assoc J)* (2010 Feb 23) **182** 249-256. PMID: 20083562
30. Zhou B., Lu Y., Hajifathalian K., Bentham J., Cesare M.D., Danaei G.. **Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4·4 million participants**. *Lancet* (2016 Apr 9) **387** 1513-1530. PMID: 27061677
31. Kautzky-Willer A., Harreiter J., Pacini G.. **Sex and gender differences in risk, pathophysiology and complications of type 2 diabetes mellitus**. *Endocr Rev* (2016 Jun) **37** 278-316. PMID: 27159875
32. Heise L., Greene M.E., Opper N., Stavropoulou M., Harper C., Nascimento M.. **Gender inequality and restrictive gender norms: framing the challenges to health**. *Lancet* (2019 Jun 15) **393** 2440-2454. PMID: 31155275
33. Kautzky-Willer A., Harreiter J.. **Sex and gender differences in therapy of type 2 diabetes**. *Diabetes Res Clin Pract* (2017 Sep) **131** 230-241. PMID: 28779681
34. Corsetti J.P., Sparks J.D., Peterson R.G., Smith R.L., Sparks C.E.. **Effect of dietary fat on the development of non-insulin dependent diabetes mellitus in obese Zucker diabetic fatty male and female rats**. *Atherosclerosis* (2000 Feb) **148** 231-241. PMID: 10657558
35. Paik S.G., Michelis M.A., Kim Y.T., Shin S.. **Induction of insulin-dependent diabetes by streptozotocin. Inhibition by estrogens and potentiation by androgens**. *Diabetes* (1982 Aug) **31** 724-729. PMID: 6219020
36. Verchere C.B., D'Alessio D.A., Palmiter R.D., Weir G.C., Bonner-Weir S., Baskin D.G.. **Islet amyloid formation associated with hyperglycemia in transgenic mice with pancreatic beta cell expression of human islet amyloid polypeptide**. *Proc Natl Acad Sci U S A* (1996 Apr 16) **93** 3492-3496. PMID: 8622964
37. Austin A.L.F., Daniels Gatward L.F., Cnop M., Santos G., Andersson D., Sharp S.. **The KINGS ins2+/G32S mouse: a novel model of β-cell endoplasmic reticulum stress and human diabetes**. *Diabetes* (2020 Dec) **69** 2667-2677. PMID: 32994272
38. Yoshioka M., Kayo T., Ikeda T., Koizuni A.. **A novel locus, Mody4, distal to D7Mit189 on chromosome 7 determines early-onset NIDDM in nonobese C57BL/6 (akita) mutant mice**. *Diabetes* (1997 May 1) **46** 887-894. PMID: 9133560
39. May C.L., Chu K., Hu M., Ortega C.S., Simpson E.R., Korach K.S.. **Estrogens protect pancreatic β-cells from apoptosis and prevent insulin-deficient diabetes mellitus in mice**. *Proc Natl Acad Sci U S A* (2006 Jun 13) **103** 9232. PMID: 16754860
40. Cohrs C.M., Panzer J.K., Drotar D.M., Enos S.J., Kipke N., Chen C.. **Dysfunction of persisting β cells is a key feature of early type 2 diabetes pathogenesis**. *Cell Rep* (2020 Apr 7) **31**
41. Saisho Y.. **β-cell dysfunction: its critical role in prevention and management of type 2 diabetes**. *World J Diabetes* (2015 Feb 15) **6** 109-124. PMID: 25685282
42. Segerstolpe Å., Palasantza A., Eliasson P., Andersson E.M., Andréasson A.C., Sun X.. **Single-cell transcriptome profiling of human pancreatic islets in health and type 2 diabetes**. *Cell Metabol* (2016 Oct 11) **24** 593-607
43. Xin Y., Kim J., Okamoto H., Ni M., Wei Y., Adler C.. **RNA sequencing of single human islet cells reveals type 2 diabetes genes**. *Cell Metabol* (2016 Oct 11) **24** 608-615
44. Avrahami D., Wang Y.J., Schug J., Feleke E., Gao L., Liu C.. **Single-cell transcriptomics of human islet ontogeny defines the molecular basis of β-cell dedifferentiation in T2D**. *Mol Metabol* (2020 Jul 30) **42**
45. Lawlor N., George J., Bolisetty M., Kursawe R., Sun L., Sivakamasundari V.. **Single-cell transcriptomes identify human islet cell signatures and reveal cell-type–specific expression changes in type 2 diabetes**. *Genome Res* (2017 Feb 1) **27** 208-222. PMID: 27864352
46. Baron M., Veres A., Wolock S.L., Faust A.L., Gaujoux R., Vetere A.. **A single-cell transcriptomic map of the human and mouse pancreas reveals inter- and intra-cell population structure**. *Cell Systems* (2016 Oct 26) **3** 346-360.e4. PMID: 27667365
47. Muraro M.J., Dharmadhikari G., Grün D., Groen N., Dielen T., Jansen E.. **A single-cell transcriptome atlas of the human pancreas**. *Cell Systems* (2016 Oct 26) **3** 385-394.e3. PMID: 27693023
48. Deng S., Vatamaniuk M., Huang X., Doliba N., Lian M.M., Frank A.. **Structural and functional abnormalities in the islets isolated from type 2 diabetic subjects**. *Diabetes* (2004 Mar 1) **53** 624-632. PMID: 14988246
49. Wu M., Lee M.Y.Y., Bahl V., Traum D., Schug J., Kusmartseva I.. **Single-cell analysis of the human pancreas in type 2 diabetes using multi-spectral imaging mass cytometry**. *Cell Rep* (2021 Nov 2) **37**
50. Salvalaggio P.R.O., Deng S., Ariyan C.E., Millet I., Zawalich W.S., Basadonna G.P.. **Islet filtration: a simple and rapid new purification procedure that avoids ficoll and improves islet mass and function**. *Transplantation* (2002 Sep 27) **74** 877-879. PMID: 12364870
51. Chu C.M.J., Modi H., Ellis C., Krentz N.A.J., Skovsø S., Zhao Y.B.. **Dynamic Ins2 gene activity defines β-cell maturity states**. *Diabetes* (2022 Dec 1) **71** 2612-2631. PMID: 36170671
52. Truchan N.A., Brar H.K., Gallagher S.J., Neuman J.C., Kimple M.E.. **A single-islet microplate assay to measure mouse and human islet insulin secretion**. *Islets* (2015 Oct 9) **7**
53. Skovsø S., Panzhinskiy E., Kolic J., Cen H.H., Dionne D.A., Dai X.Q.. **Beta-cell specific Insr deletion promotes insulin hypersecretion and improves glucose tolerance prior to global insulin resistance**. *Nat Commun* (2022 Feb 8) **13** 735. PMID: 35136059
54. Szabat M., Page M.M., Panzhinskiy E., Skovsø S., Mojibian M., Fernandez-Tajes J.. **Reduced insulin production relieves endoplasmic reticulum stress and induces β cell proliferation**. *Cell Metabol* (2016 Jan 12) **23** 179-193
55. Dobin A., Davis C.A., Schlesinger F., Drenkow J., Zaleski C., Jha S.. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013 Jan 1) **29** 15-21. PMID: 23104886
56. Love M.I., Huber W., Anders S.. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol* (2014 Dec 5) **15** 550. PMID: 25516281
57. Gillespie M., Jassal B., Stephan R., Milacic M., Rothfels K., Senff-Ribeiro A.. **The reactome pathway knowledgebase 2022**. *Nucleic Acids Res* (2022 Jan 7) **50** D687-D692. PMID: 34788843
58. Zhou G., Soufan O., Ewald J., Hancock R.E.W., Basu N., Xia J.. **NetworkAnalyst 3.0: a visual analytics platform for comprehensive gene expression profiling and meta-analysis**. *Nucleic Acids Res* (2019 Jul 2) **47** W234-W241. PMID: 30931480
59. Tyanova S., Temu T., Sinitcyn P., Carlson A., Hein M.Y., Geiger T.. **The Perseus computational platform for comprehensive analysis of (prote)omics data**. *Nat Methods* (2016 Sep) **13** 731-740. PMID: 27348712
60. Wei T., Viliam S.. (2021)
61. Wang S., Flibotte S., Camunas-Soler J., MacDonald P.E., Johnson J.D.. **A new hypothesis for type 1 diabetes risk: the at-risk allele at rs3842753 associates with increased beta-cell INS messenger RNA in a meta-analysis of single-cell RNA-sequencing data**. *Can J Diabetes* (2021 Dec 1) **45** 775-784.e2. PMID: 34052132
62. Hashemitabar M., Bahramzadeh S., Saremy S., Nejaddehbashi F.. **Glucose plus metformin compared with glucose alone on β-cell function in mouse pancreatic islets**. *Biomed Rep* (2015 Sep) **3** 721-725. PMID: 26405552
63. Marchetti P., Bugliani M., Lupi R., Marselli L., Masini M., Boggi U.. **The endoplasmic reticulum in pancreatic beta cells of type 2 diabetes patients**. *Diabetologia* (2007 Dec 1) **50** 2486-2494. PMID: 17906960
64. Lupi R., Del Guerra S., Fierabracci V., Marselli L., Novelli M., Patanè G.. **Lipotoxicity in human pancreatic islets and the protective effect of metformin**. *Diabetes* (2002 Feb) **51** S134-S137. PMID: 11815472
65. Kanda Y., Shimoda M., Hamamoto S., Tawaramoto K., Kawasaki F., Hashiramoto M.. **Molecular mechanism by which pioglitazone preserves pancreatic β-cells in obese diabetic mice: evidence for acute and chronic actions as a PPARγ agonist**. *Am J Physiol Endocrinol Metab* (2010 Feb) **298** E278-E286. PMID: 19920213
66. Murao N., Yokoi N., Takahashi H., Hayami T., Minami Y., Seino S.. **Increased glycolysis affects β-cell function and identity in aging and diabetes**. *Mol Metabol* (2022 Jan) **55**
67. Mizukami H., Takahashi K., Inaba W., Osonoi S., Kamata K., Tsuboi K.. **Age-associated changes of islet endocrine cells and the effects of body mass index in Japanese**. *J Diabetes Investig* (2014 Feb 12) **5** 38-47
68. Yinhui H.E., Haiyan X.U., Qi F.U., Tao Y.. **Effects of glycosylated hemoglobin and disease course on islet β-cell function in patients with type 2 diabetes**. *Nan Fang Yi Ke Da Xue Xue Bao* (2019 Sep 30) **39** 1003-1008. PMID: 31640950
69. Kaestner K.H., Powers A.C., Naji A., Consortium H.P.A.P., Atkinson M.A.. **NIH initiative to improve understanding of the pancreas, islet, and autoimmunity in type 1 diabetes: the human pancreas analysis Program (HPAP)**. *Diabetes* (2019 Jul) **68** 1394-1402. PMID: 31127054
70. Hellman B., Idahl L.Å., Lernmark Å., Täljedal I.B.. **The pancreatic β-cell recognition of insulin secretagogues: does cyclic AMP mediate the effect of glucose?**. *Proc Natl Acad Sci USA* (1974 Sep) **71** 3405-3409. PMID: 4372618
71. Yang X., Schadt E.E., Wang S., Wang H., Arnold A.P., Ingram-Drake L.. **Tissue-specific expression and regulation of sexually dimorphic genes in mice**. *Genome Res* (2006 Aug 1) **16** 995-1004. PMID: 16825664
72. Lu T., Mar J.C.. **Investigating transcriptome-wide sex dimorphism by multi-level analysis of single-cell RNA sequencing data in ten mouse cell types**. *Biol Sex Differ* (2020 Nov 5) **11** 61. PMID: 33153500
73. Cottet-Dumoulin D., Lavallard V., Lebreton F., Wassmer C.H., Bellofatto K., Parnaud G.. **Biosynthetic activity differs between islet cell types and in beta cells is modulated by glucose and not by secretion**. *Endocrinology* (2020 Dec 25) **162** bqaa239
74. Li N., Yang Z., Li Q., Yu Z., Chen X., Li J.C.. **Ablation of somatostatin cells leads to impaired pancreatic islet function and neonatal death in rodents**. *Cell Death Dis* (2018 Jun 7) **9** 1-12. PMID: 29298988
75. Webb G.C., Dey A., Wang J., Stein J., Milewski M., Steiner D.F.. **Altered proglucagon processing in an α-cell line derived from prohormone convertase 2 null mouse islets**. *J Biol Chem* (2004 Jul 23) **279** 31068-31075. PMID: 15143067
76. Garcia-Barrado M.J., Ravier M.A., Rolland J.F., Gilon P., Nenquin M., Henquin J.C.. **Inhibition of protein synthesis sequentially impairs distinct steps of stimulus-secretion coupling in pancreatic β cells**. *Endocrinology* (2001 Jan 1) **142** 299-307. PMID: 11145593
77. Scheuner D., Mierde D.V., Song B., Flamez D., Creemers J.W.M., Tsukamoto K.. **Control of mRNA translation preserves endoplasmic reticulum function in beta cells and maintains glucose homeostasis**. *Nat Med* (2005 Jul) **11** 757-764. PMID: 15980866
78. Izumi T., Yokota-Hashimoto H., Zhao S., Wang J., Halban P.A., Takeuchi T.. **Dominant negative pathogenesis by mutant proinsulin in the Akita diabetic mouse**. *Diabetes* (2003 Feb) **52** 409-416. PMID: 12540615
79. Xin Y., Dominguez Gutierrez G., Okamoto H., Kim J., Lee A.H., Adler C.. **Pseudotime ordering of single human β-cells reveals states of insulin production and unfolded protein response**. *Diabetes* (2018) **67** 1783-1794. PMID: 29950394
80. Lipson K.L., Fonseca S.G., Ishigaki S., Nguyen L.X., Foss E., Bortell R.. **Regulation of insulin biosynthesis in pancreatic beta cells by an endoplasmic reticulum-resident protein kinase IRE1**. *Cell Metabol* (2006 Sep) **4** 245-254
81. Teodoro T., Odisho T., Sidorova E., Volchuk A.. **Pancreatic β-cells depend on basal expression of active ATF6α-p50 for cell survival even under nonstress conditions**. *Am J Physiol Cell Physiol* (2012 Apr) **302** C992-C1003. PMID: 22189555
82. Scheuner D., Song B., McEwen E., Liu C., Laybutt R., Gillespie P.. **Translational control is required for the unfolded protein response and in vivo glucose homeostasis**. *Mol Cell* (2001 Jun) **7** 1165-1176. PMID: 11430820
83. Luciani D.S., Gwiazda K.S., Yang T.L.B., Kalynyak T.B., Bychkivska Y., Frey M.H.Z.. **Roles of IP3R and RyR Ca2+ channels in endoplasmic reticulum stress and β-cell death**. *Diabetes* (2009 Feb 1) **58** 422-432. PMID: 19033399
84. Riggs A.C., Bernal-Mizrachi E., Ohsugi M., Wasson J., Fatrai S., Welling C.. **Mice conditionally lacking the Wolfram gene in pancreatic islet beta cells exhibit diabetes as a result of enhanced endoplasmic reticulum stress and apoptosis**. *Diabetologia* (2005 Nov 1) **48** 2313-2321. PMID: 16215705
85. Li J., Lee B., Lee A.S.. **Endoplasmic reticulum stress-induced apoptosis: multiple pathways and activation of p53-up-regulated modulator of apoptosis (PUMA) and NOXA by p53**. *J Biol Chem* (2006 Mar 17) **281** 7260-7270. PMID: 16407291
86. Sharma R.B., Landa-Galván H.V., Alonso L.C.. **Living dangerously: protective and harmful ER stress responses in pancreatic β-cells**. *Diabetes* (2021 Nov) **70** 2431. PMID: 34711668
87. Gwiazda K.S., Yang T.L.B., Lin Y., Johnson J.D.. **Effects of palmitate on ER and cytosolic Ca2+ homeostasis in β-cells**. *Am J Physiol Endocrinol Metab* (2009 Apr) **296** E690-E701. PMID: 19141690
88. Fonseca S.G., Gromada J., Urano F.. **Endoplasmic reticulum stress and pancreatic beta cell death**. *Trends Endocrinol Metabol* (2011 Jul) **22** 266-274
89. Wakae-Takada N., Xuan S., Watanabe K., Meda P., Leibel R.L.. **Molecular basis for the regulation of islet beta cell mass in mice: the role of E-cadherin**. *Diabetologia* (2013 Apr) **56** 856-866. PMID: 23354125
90. Shrestha N., Franco E.D., Arvan P., Cnop M.. **Pathological β-cell endoplasmic reticulum stress in type 2 diabetes: current evidence**. *Front Endocrinol* (2021) **12**
91. Oyadomari S., Mori M.. **Roles of CHOP/GADD153 in endoplasmic reticulum stress**. *Cell Death Differ* (2004 Apr) **11** 381-389. PMID: 14685163
92. Li T., Jiao W., Li W., Li H.. **Sex effect on insulin secretion and mitochondrial function in pancreatic beta cells of elderly Wistar rats**. *Endocr Res* (2016 Aug) **41** 167-179. PMID: 26865180
93. Basu A., Dube S., Basu R., Mauvais-Jarvis F.. *Sex and gender factors affecting metabolic homeostasis, diabetes and obesity [Internet]* (2017) 53-64. DOI: 10.1007/978-3-319-70178-3_4
94. Altirriba J., Gasa R., Casas S., Ramírez-Bajo M.J., Ros S., Gutierrez-Dalmau A.. **The role of transmembrane protein 27 (TMEM27) in islet physiology and its potential use as a beta cell mass biomarker**. *Diabetologia* (2010 Jul 1) **53** 1406-1414. PMID: 20386877
95. Kumar S., Marriott C.E., Alhasawi N.F., Bone A.J., Macfarlane W.M.. **The role of tumour suppressor PDCD4 in beta cell death in hypoxia**. *PLoS One* (2017 Jul 27) **12**
96. Ruan Q., Wang T., Kameswaran V., Wei Q., Johnson D.S., Matschinsky F.. **The microRNA-21−PDCD4 axis prevents type 1 diabetes by blocking pancreatic β cell death**. *Proc Natl Acad Sci U S A* (2011 Jul 19) **108** 12030-12035. PMID: 21730150
97. Binger K.J., Neukam M., Tattikota S.G., Qadri F., Puchkov D., Willmes D.M.. **Atp6ap2 deletion causes extensive vacuolation that consumes the insulin content of pancreatic β cells**. *Proc Natl Acad Sci USA* (2019 Oct) **116** 19983-19988. PMID: 31527264
98. Masini M., Bugliani M., Lupi R., del Guerra S., Boggi U., Filipponi F.. **Autophagy in human type 2 diabetes pancreatic beta cells**. *Diabetologia* (2009 Jun 1) **52** 1083-1086. PMID: 19367387
99. Ji J., Petropavlovskaia M., Khatchadourian A., Patapas J., Makhlin J., Rosenberg L.. **Type 2 diabetes is associated with suppression of autophagy and lipid accumulation in β-cells**. *J Cell Mol Med* (2019 Apr) **23** 2890-2900. PMID: 30710421
100. Cai T., Hirai H., Zhang G., Zhang M., Takahashi N., Kasai H.. **Deletion of Ia-2 and/or Ia-2β in mice decreases insulin secretion by reducing the number of dense core vesicles**. *Diabetologia* (2011 Sep) **54** 2347-2357. PMID: 21732083
101. Torii S., Kubota C., Saito N., Kawano A., Hou N., Kobayashi M.. **The pseudophosphatase phogrin enables glucose-stimulated insulin signaling in pancreatic β cells**. *J Biol Chem* (2018 Apr 20) **293** 5920-5933. PMID: 29483197
102. Wollam J., Mahata S., Riopel M., Hernandez-Carretero A., Biswas A., Bandyopadhyay G.K.. **Chromogranin A regulates vesicle storage and mitochondrial dynamics to influence insulin secretion**. *Cell Tissue Res* (2017 Jun) **368** 487-501. PMID: 28220294
103. Portela-Gomes G.M., Gayen J.R., Grimelius L., Stridsberg M., Mahata S.K.. **The importance of chromogranin A in the development and function of endocrine pancreas**. *Regul Pept* (2008 Nov 29) **151** 19-25. PMID: 18722481
104. Paglialunga S., Simnett G., Robson H., Hoang M., Pillai R., Arkell A.M.. **The Rab-GTPase activating protein, TBC1D1, is critical for maintaining normal glucose homeostasis and β-cell mass**. *Appl Physiol Nutr Metabol* (2017 Jun) **42** 647-655
105. Stermann T., Menzel F., Weidlich C., Jeruschke K., Weiss J., Altenhofen D.. **Deletion of the RabGAP TBC1D1 leads to enhanced insulin secretion and fatty acid oxidation in islets from male mice**. *Endocrinology* (2018 Apr 1) **159** 1748-1761. PMID: 29481597
106. Laybutt D.R., Preston A.M., Åkerfeldt M.C., Kench J.G., Busch A.K., Biankin A.V.. **Endoplasmic reticulum stress contributes to beta cell apoptosis in type 2 diabetes**. *Diabetologia* (2007 Apr 1) **50** 752-763. PMID: 17268797
107. Engin F., Nguyen T., Yermalovich A., Hotamisligil G.S.. **Aberrant islet unfolded protein response in type 2 diabetes**. *Sci Rep* (2014 Feb 11) **4** 4054. PMID: 24514745
108. jiang Huang C., yu Lin C., Haataja L., Gurlo T., Butler A.E., Rizza R.A.. **High expression rates of human islet amyloid polypeptide induce endoplasmic reticulum stress mediated beta-cell apoptosis, a characteristic of humans with type 2 but not type 1 diabetes**. *Diabetes* (2007 Aug) **56** 2016-2027. PMID: 17475933
109. Herbach N., Rathkolb B., Kemter E., Pichl L., Klaften M., de Angelis M.H.. **Dominant-negative effects of a novel mutated Ins2 allele causes early-onset diabetes and severe β-cell loss in Munich Ins2C95S mutant mice**. *Diabetes* (2007 May 1) **56** 1268-1276. PMID: 17303807
110. Tiano J.P., Mauvais-Jarvis F.. **Importance of oestrogen receptors to preserve functional β-cell mass in diabetes**. *Nat Rev Endocrinol* (2012 Jun) **8** 342-351. PMID: 22330739
111. Janson J., Soeller W.C., Roche P.C., Nelson R.T., Torchia A.J., Kreutter D.K.. **Spontaneous diabetes mellitus in transgenic mice expressing human islet amyloid polypeptide**. *Proc Natl Acad Sci U S A* (1996 Jul 9) **93** 7283-7288. PMID: 8692984
112. Walker E.M., Cha J., Tong X., Guo M., Liu J.H., Yu S.. **Sex-biased islet β cell dysfunction is caused by the MODY MAFA S64F variant by inducing premature aging and senescence in males**. *Cell Rep* (2021 Oct 12) **37**
113. Mauvais-Jarvis F., Sobngwi E., Porcher R., Riveline J.P., Kevorkian J.P., Vaisse C.. **Ketosis-prone type 2 diabetes in patients of sub-saharan african origin: clinical pathophysiology and natural history of β-cell dysfunction and insulin resistance**. *Diabetes* (2004 Mar 1) **53** 645-653. PMID: 14988248
114. Iacovazzo D., Flanagan S.E., Walker E., Quezado R., de Sousa Barros F.A., Caswell R.. **MAFA missense mutation causes familial insulinomatosis and diabetes mellitus**. *Proc Natl Acad Sci U S A* (2018 Jan 30) **115** 1027-1032. PMID: 29339498
115. Xu B., Allard C., Alvarez-Mercado A.I., Fuselier T., Kim J.H., Coons L.A.. **Estrogens promote misfolded proinsulin degradation to protect insulin production and delay diabetes**. *Cell Rep* (2018 Jul 3) **24** 181. PMID: 29972779
116. Zhou Z., Ribas V., Rajbhandari P., Drew B.G., Moore T.M., Fluitt A.H.. **Estrogen receptor α protects pancreatic β-cells from apoptosis by preserving mitochondrial function and suppressing endoplasmic reticulum stress**. *J Biol Chem* (2018 Mar 30) **293** 4735. PMID: 29378845
117. Alonso-Magdalena P., Ropero A.B., Carrera M.P., Cederroth C.R., Baquié M., Gauthier B.R.. **Pancreatic insulin content regulation by the estrogen receptor ER alpha**. *PLoS One* (2008 Apr 30) **3**
118. Cnop M., Ladriere L., Hekerman P., Ortis F., Cardozo A.K., Dogusan Z.. **Selective inhibition of eukaryotic translation initiation factor 2α dephosphorylation potentiates fatty acid-induced endoplasmic reticulum stress and causes pancreatic β-cell dysfunction and apoptosis**. *J Biol Chem* (2007 Feb 9) **282** 3989-3997. PMID: 17158450
119. Oyadomari S., Koizumi A., Takeda K., Gotoh T., Akira S., Araki E.. **Targeted disruption of the Chop gene delays endoplasmic reticulum stress–mediated diabetes**. *J Clin Invest* (2002 Feb 15) **109** 525-532. PMID: 11854325
120. Han J., Back S.H., Hur J., Lin Y.H., Gildersleeve R., Shan J.. **ER-stress-induced transcriptional regulation increases protein synthesis leading to cell death**. *Nat Cell Biol* (2013 May) **15** 481-490. PMID: 23624402
121. Harding H.P., Zeng H., Zhang Y., Jungries R., Chung P., Plesken H.. **Diabetes mellitus and exocrine pancreatic dysfunction in Perk−/− mice reveals a role for translational control in secretory cell survival**. *Mol Cell* (2001 Jun 1) **7** 1153-1163. PMID: 11430819
122. Novoa I., Zhang Y., Zeng H., Jungreis R., Harding H.P., Ron D.. **Stress-induced gene expression requires programmed recovery from translational repression**. *EMBO J* (2003 Mar 3) **22** 1180. PMID: 12606582
123. Ma Y., Hendershot L.M.. **Delineation of a negative feedback regulatory loop that controls protein translation during endoplasmic reticulum stress**. *J Biol Chem* (2003 Sep 12) **278** 34864-34873. PMID: 12840028
124. Panzhinskiy E., Skovsø S., Cen H.H., Chu K.Y., MacDonald K., Soukhatcheva G.. **Eukaryotic translation initiation factor 2A protects pancreatic beta cells during endoplasmic reticulum stress while rescuing translation inhibition**. *bioRxiv* (2021 Feb 17)
125. Cnop M., Toivonen S., Igoillo-Esteve M., Salpea P.. **Endoplasmic reticulum stress and eIF2α phosphorylation: the Achilles heel of pancreatic β cells**. *Mol Metabol* (2017 Jul 12) **6** 1024-1039
126. Sharma R.B., O'Donnell A.C., Stamateris R.E., Ha B., McCloskey K.M., Reynolds P.R.. **Insulin demand regulates β cell number via the unfolded protein response**. *J Clin Invest* (2015 Oct 1) **125** 3831-3846. PMID: 26389675
127. Sharma R.B., Darko C., Alonso L.C.. **Intersection of the ATF6 and XBP1 ER stress pathways in mouse islet cells**. *J Biol Chem* (2020 Oct 9) **295** 14164-14177. PMID: 32788214
128. Nasteska D., Fine N.H.F., Ashford F.B., Cuozzo F., Viloria K., Smith G.. **PDX1LOW MAFALOW β-cells contribute to islet function and insulin release**. *Nat Commun* (2021 Jan 29) **12** 674. PMID: 33514698
129. Talchai C., Xuan S., Lin H.V., Sussel L., Accili D.. **Pancreatic β-cell dedifferentiation as mechanism of diabetic β-cell failure**. *Cell* (2012 Sep 14) **150** 1223. PMID: 22980982
130. Hodeify R., Megyesi J., Tarcsafalvi A., Mustafa H.I., Seng N.S.H.L., Price P.M.. **Pathophysiology of Acute Kidney Injury: gender differences control the susceptibility to ER stress-induced acute kidney injury**. *Am J Physiol Ren Physiol* (2013 Apr 1) **304** F875
131. Manicardi V., Russo G., Napoli A., Torlone E., Li Volsi P., Giorda C.B.. **Gender-disparities in adults with type 1 diabetes: more than a quality of care issue. A cross-sectional observational study from the AMD annals initiative**. *PLoS One* (2016 Oct 3) **11**
|
---
title: 'Pre-pandemic physical activity as a predictor of infection and mortality associated
with COVID-19: Evidence from the National Health Insurance Service'
authors:
- Saengryeol Park
- Hyeseong Kim
- So-Youn Park
- In-Hwan Oh
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9971555
doi: 10.3389/fpubh.2023.1072198
license: CC BY 4.0
---
# Pre-pandemic physical activity as a predictor of infection and mortality associated with COVID-19: Evidence from the National Health Insurance Service
## Abstract
### Introduction
During the coronavirus disease 2019 (COVID-19) pandemic, many populations have experienced reduced physical activity (PA) levels, weight gain, and increased anxiety and depression. However, according to a previous study, engaging in PA has a positive effect on damages caused by COVID-19. Therefore, this study aimed to investigate the association between PA and COVID-19 using the National Health Insurance Sharing Service Database in South Korea.
### Methods
Logistic regression analysis was used to analyze the association of PA with COVID-19 and mortality. The analysis was adjusted for body mass index, sex, age, insurance type, comorbidity, and region of residence at baseline. Disability and lifestyle (weight, smoking, and drinking status) were adjusted consecutively.
### Results
The results indicated that engaging in insufficient PA as per the WHO guidelines predicts a higher risk of COVID-19 when controlling for personal characteristics, comorbidity, lifestyle, disability, and mortality.
### Discussion
This study revealed the need to engage in PA and manage weight to reduce the risk of infection and mortality associated with COVID-19. Because engaging in PA is an important component of weight management and can help restore physical and mental health after the COVID-19 pandemic, it should be emphasized as a pillar of recovery after COVID-19.
## 1. Introduction
The novel coronavirus disease 2019 (COVID-19) has been a prominent global issue since its emergence in 2019. It is an infectious respiratory disease with mild-to-severe symptoms, which may include fever, cough, loss of taste or smell, and diarrhea [1, 2]. However, in severe cases, chest pain, loss of speech, loss of mobility, and confusion may occur. COVID-19 is rapidly spread by respiratory droplets released during coughing, sneezing, speaking, singing, or breathing by the infected individual [3]. Thus, the rapid spread of COVID-19 has caused an unprecedented number of cases and deaths. As of August 2022, there were over 593,236,266 confirmed cases and about 6,448,504 deaths worldwide [4]. In South Korea, particularly, the fatality rate of older adults over 80 years old was $2.35\%$ [5].
Recent studies show that COVID-19 has impacted mental health. There is evidence of increased severity of depression compared to those before pandemic [6, 7]. It seems that quarantine throughout the COVID-19 pandemic negatively impacted the mental health of previously unaffected individuals; for example, the anxiety and depression of people whose family, colleagues, classmates, or neighbors were affected by quarantine were increased [8]. In addition, the risks of anxiety, depression, stress, and sleep disorders in COVID-19 patients were increased [9]. Post-traumatic stress symptoms were occasionally experienced after infection, but results concerning physical health were limited [10]. However, it is unclear how factors relating to lifestyles are linked to the prognosis of COVID-19.
Engaging in physical activity (PA) has played an important role in improving psychological and physical health. It was found that engaging in regular moderate-to-vigorous PA (MVPA) is associated with reducing anxiety and negative self-perceptions, as well as improving physical health [11]. In addition, engaging in MVPA is associated with losing and managing body weight that may predispose individuals to several types of chronic diseases, such as obesity and high blood pressure [12]. However, a recent study shows that people spend more time engaging in sedentary behavior and less time engaging in PA than before the pandemic [13]. Individuals engaging in PA is found to be more associated with lower risk of COVID-19 and mortality than those who do not meet the recommended PA level (150 min of MVPA at least once a week) [14, 15], although these results did not consider other factors of health such as personal characteristics, comorbidity, and disability level.
Thus, we aimed to examine the associations of MVPA with COVID-19 and mortality. To produce more robust study results, we tried to include controlling variables beyond MVPA, using nationally representative data.
## 2.1. Database
The National Health Insurance Service (NHIS) of South *Korea is* a social insurance system for the entire nation, and registration is compulsory; approximately $97\%$ of the Korean population is currently registered [16]. The NHIS assists people with scheduling medical checkups every 2 years and records their results automatically. These data include various information such as demographic information, payment specification, consultation statement, diagnosis statements, and prescriptions. With these records they developed the National Health Insurance Sharing Service for researcher to support various studies providing sample cohort DB, customized cohort DB, health screening cohort, etc.
This study used NHIS-COVID19 DB that included 4,363 adult COVID-19 patients in South Korea between January 1, 2020, and July 14, 2020, who had medical records between 2015 and 2018, the most recent data before the pandemic. We selected 67,125 adults for the control group in NHIS DB who also had medical checkup data. We used the most recent records in our study; the final data included COVID-19 status, demographic information, comorbidity, disability status, and lifestyle, including PA and body mass index (BMI). This study conformed to the Guidelines on De-identification of Personal Data of Korea and was approved by the Kyung Hee University's Institutional Review Board (IRB No. KHSIRB20-301[EA]) as a review exemption study. Thus, the requirement for informed consent was waived.
## 2.2.1. PA
PA was measured using a self-report questionnaire from NHIS. Moderate PA (MPA) was measured with the following question: “During the last week, how many times a week and for how many hours a day did you engage in physical activity at a moderate level for more than 10 min (e.g., fast walking, doubles tennis, riding bicycle, cleaning)?” Vigorous PA (VPA) was assessed with the following question: “During the last week, how many days a week and for how many hours a day did you engage in physical activity at a vigorous level for more than 10 min (e.g., running, aerobic, fast riding a bicycle)?” PA was categorized into two groups according to PA guidelines: 150 min of MVPA at least once a week (1 min of VPA = 2 min of MPA). The items have been widely used in the literature [17].
## 2.2.2. BMI
BMI is a simple obesity indicator calculated as weight/square of height (kg/m2). In this study, BMI was categorized into four groups according to the World Health Organization (WHO) BMI classification (underweight = BMI <18.5, normal range = 18.5 ≤ BMI <25, overweight = 25 ≤ BMI <30, obese = 30 ≤ BMI).
## 2.2.3. Covariates
We adjusted for sex, age, region of residence, economic status, the number of comorbidities, disability, smoking status, drinking status, and weight, which are reportedly associated with COVID-19 [18, 19]. Residence was categorized into the five regions (Seoul, Daegu, Gyeonggi, Gyeong-buk, and other) of South Korea (from January 1, 2020 to August 14, 2020) with the most confirmed COVID-19 cases.
Economic status was measured using health insurance premiums. Health insurance premiums are categorized into five quintiles. In South Korea, every person must pay part of their income as an insurance premium. Thus, a higher quintile indicates a higher economic status. Basic livelihood security recipients were included in the medical aid.
Comorbidity refers to an underlying condition (e.g., diabetes, hypertension) that may cause and affect other diseases. In this study, the number of comorbidities was investigated with the question: “Among the following diseases, which diseases have you been diagnosed with, or have you been treated for?” with the examples of comorbidity, stroke, heart disease (e.g., myocardial infarction, angina pectoris), hypertension, diabetes, dyslipidemia, pulmonary tuberculosis, and other diseases including cancer.
Disability status included the presence, severity, and type of disability, including non-disabled, physical disability, encephalopathy, visual impairment, hearing impairment, and others. Disabled persons were registered with the Ministry of Health and Welfare of South Korea; hence, the NHIS data included their disability status. Severity and type of disability were measured according to the Act on Welfare of Persons with Disability.
Smoking was measured with the question: “In your life, have you ever smoked over five packs of cigarettes (100 pieces)?” Drinking was measured using the question, “How many times do you drink a week?”
## 2.3. Statistical analysis
The effects of PA and BMI on infection and mortality associated with COVID-19 were analyzed using logistic regression analysis, which was adjusted for sex, age, insurance type, comorbidities, and region of residence at baseline. Disability and lifestyle (e.g., need for weight management, smoking, and drinking status) were adjusted conjointly. Cases with missing data were excluded from the analysis. The $95\%$ confidence interval (CI) was estimated using the SAS PROC PHREG. Statistical significance was set at $p \leq 0.05.$ Statistical analyses were conducted using the SAS software (SAS Institute Inc., Cary, NC, USA).
## 3. Results
The analysis included 71,488 participants ($62\%$ women) over 20 years of age who had received medical checkups from 2015 to 2018. There were 4,363 COVID-19 confirmed cases ($6.1\%$), including 141 deaths ($3.3\%$; 48 women) (Table 1). Among the participants, 418 people with medical aid were developed COVID-19 ($12\%$), with the highest ratio compared to the other quintiles. According to disability status, severe disability was associated with a higher infection rate ($13.6\%$) than mild disability ($6.3\%$) and non-disability ($5.9\%$). People with comorbidities had a higher infection rate (number of comorbidities: 1 = $8\%$, 2 = $7\%$, 3 = $10.4\%$) than those without comorbidities ($5.5\%$).
**Table 1**
| Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | All | All.1 | COVID-19 infection | COVID-19 infection.1 | COVID-19 infection.2 | COVID-19 infection.3 | p-value | All.2 | All.3 | COVID-19 mortality | COVID-19 mortality.1 | COVID-19 mortality.2 | COVID-19 mortality.3 | p-value.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | | | | Infected | Infected | Non-infected | Non-infected | | | | Deaths | Deaths | Survivors | Survivors | |
| | | | n | % | n | % | n | % | | n | % | n | % | n | % | |
| Domain | Domain | Domain | 71488 | 100 | 4363 | 6.1 | 67125 | 93.9 | <0.0001 | 4363 | 100 | 141 | 3.2 | 4222 | 96.8 | <0.0001 |
| Characteristic | Sex | Male | 27044 | 100 | 1615 | 6.0 | 25429 | 94.0 | 0.2524 | 1615 | 100 | 93 | 5.8 | 1522 | 94.2 | <0.0001 |
| | | Female | 44444 | 100 | 2748 | 6.2 | 41696 | 93.8 | | 2748 | 100 | 48 | 1.7 | 2700 | 98.3 | |
| | Age | 20–29 | 6550 | 100 | 281 | 4.3 | 6269 | 95.7 | <0.0001 | 281 | 100 | 0 | 0.0 | 281 | 100.0 | <0.0001 |
| | | 30–39 | 6973 | 100 | 400 | 5.7 | 6573 | 94.3 | | 400 | 100 | 0 | 0.0 | 400 | 100.0 | |
| | | 40–49 | 11559 | 100 | 757 | 6.5 | 10802 | 93.5 | | 757 | 100 | 2 | 0.3 | 755 | 99.7 | |
| | | 50–59 | 19305 | 100 | 1257 | 6.5 | 18048 | 93.5 | | 1257 | 100 | 10 | 0.8 | 1247 | 99.2 | |
| | | 60–69 | 15676 | 100 | 991 | 6.3 | 14685 | 93.7 | | 991 | 100 | 25 | 2.5 | 966 | 97.5 | |
| | | 70–79 | 7786 | 100 | 465 | 6.0 | 7321 | 94.0 | | 465 | 100 | 44 | 9.5 | 421 | 90.5 | |
| | | 80- | 3639 | 100 | 212 | 5.8 | 3427 | 94.2 | | 212 | 100 | 60 | 28.3 | 152 | 71.7 | |
| | Region of residence | Seoul | 4174 | 100 | 267 | 6.4 | 3907 | 93.6 | 0.9053 | 267 | 100 | 3 | 1.1 | 264 | 98.9 | <0.0001 |
| | | Daegu | 47057 | 100 | 2871 | 6.1 | 44186 | 93.9 | | 2871 | 100 | 80 | 2.8 | 2791 | 97.2 | |
| | | Gyeonggi | 3811 | 100 | 224 | 5.9 | 3587 | 94.1 | | 224 | 100 | 10 | 4.5 | 214 | 95.5 | |
| | | Gyeong-buk | 9339 | 100 | 572 | 6.1 | 8767 | 93.9 | | 572 | 100 | 39 | 6.8 | 533 | 93.2 | |
| | | others | 7107 | 100 | 429 | 6.0 | 6678 | 94.0 | | 429 | 100 | 9 | 2.1 | 420 | 97.9 | |
| | Health insurance premium | Medical aid | 3497 | 100 | 418 | 12.0 | 3079 | 88.0 | <0.0001 | 418 | 100 | 18 | 4.3 | 400 | 95.7 | 0.0253 |
| | | 1st quintile | 12336 | 100 | 846 | 6.9 | 11490 | 93.1 | | 846 | 100 | 19 | 2.2 | 827 | 97.8 | |
| | | 2nd quintile | 10485 | 100 | 588 | 5.6 | 9897 | 94.4 | | 588 | 100 | 11 | 1.9 | 577 | 98.1 | |
| | | 3rd quintile | 12612 | 100 | 741 | 5.9 | 11871 | 94.1 | | 741 | 100 | 23 | 3.1 | 718 | 96.9 | |
| | | 4th quintile | 14583 | 100 | 769 | 5.3 | 13814 | 94.7 | | 769 | 100 | 25 | 3.3 | 744 | 96.7 | |
| | | 5th quintile | 17975 | 100 | 1001 | 5.6 | 16974 | 94.4 | | 1001 | 100 | 45 | 4.5 | 956 | 95.5 | |
| Comorbidity | Number of comorbidities | 0 | 53053 | 100 | 2893 | 5.5 | 50160 | 94.5 | <0.0001 | 2893 | 100 | 44 | 1.5 | 2849 | 98.5 | <0.0001 |
| | | 1 | 8524 | 100 | 684 | 8.0 | 7840 | 92.0 | | 684 | 100 | 29 | 4.2 | 655 | 95.8 | |
| | | 2 | 7169 | 100 | 502 | 7.0 | 6667 | 93.0 | | 502 | 100 | 31 | 6.2 | 471 | 93.8 | |
| | | 3+ | 2742 | 100 | 284 | 10.4 | 2458 | 89.6 | | 284 | 100 | 37 | 13.0 | 247 | 87.0 | |
| Disability | Presence | Non-disabled | 66859 | 100 | 3974 | 5.9 | 62885 | 94.1 | <0.0001 | 3974 | 100 | 107 | 2.7 | 3867 | 97.3 | <0.0001 |
| | | Disabled | 4629 | 100 | 389 | 8.4 | 4240 | 91.6 | | 389 | 100 | 34 | 8.7 | 355 | 91.3 | |
| | Severity | Non-disabled | 66859 | 100 | 3974 | 5.9 | 62885 | 94.1 | <0.0001 | 3974 | 100 | 107 | 2.7 | 3867 | 97.3 | <0.0001 |
| | | Mild | 3317 | 100 | 210 | 6.3 | 3107 | 93.7 | | 179 | 100 | 18 | 10.1 | 161 | 89.9 | |
| | | Severe | 1312 | 100 | 197 | 13.6 | 1133 | 86.4 | | 210 | 100 | 16 | 7.6 | 194 | 92.4 | |
| | Type | Non-disabled | 66859 | 100 | 3974 | 5.9 | 62885 | 94.1 | <0.0001 | 3974 | 100 | 107 | 2.7 | 3867 | 97.3 | <0.0001 |
| | | Physical disability | 2215 | 100 | 139 | 6.3 | 2076 | 93.7 | | 139 | 100 | 11 | 7.9 | 128 | 92.1 | |
| | | Encephalopathy | 372 | 100 | 34 | 9.1 | 338 | 90.9 | | 34 | 100 | 4 | 11.8 | 30 | 88.2 | |
| | | Visual impairment | 483 | 100 | 26 | 5.4 | 457 | 94.6 | | 26 | 100 | 0 | 0.0 | 26 | 100.0 | |
| | | Hearing impairment | 875 | 100 | 69 | 7.9 | 806 | 92.1 | | 69 | 100 | 6 | 8.7 | 63 | 91.3 | |
| | | Others | 684 | 100 | 121 | 17.7 | 563 | 82.3 | | 121 | 100 | 13 | 10.7 | 108 | 89.3 | |
| Lifestyle | Smoking | No | 53058 | 100 | 3523 | 6.6 | 49535 | 93.4 | <0.0001 | 4061 | 100 | 130 | 3.2 | 3931 | 96.8 | 0.6757 |
| | | Yes | 18430 | 100 | 840 | 4.6 | 17590 | 95.4 | | 302 | 100 | 11 | 3.6 | 291 | 96.4 | |
| | Drinking | No | 60961 | 100 | 4061 | 6.7 | 56900 | 93.3 | <0.0001 | 3523 | 100 | 130 | 3.7 | 3393 | 96.3 | 0.0005 |
| | | yes | 10527 | 100 | 302 | 2.9 | 10225 | 97.1 | | 840 | 100 | 11 | 1.3 | 829 | 98.7 | |
| | Weight management | Unnecessary | 36016 | 100 | 2219 | 6.2 | 33797 | 93.8 | 0.5137 | 2219 | 100 | 70 | 3.2 | 2149 | 96.8 | 0.7694 |
| | | Necessary | 35472 | 100 | 2144 | 6.0 | 33328 | 94.0 | | 2144 | 100 | 71 | 3.3 | 2073 | 96.7 | |
| | BMI | Underweight | 2684 | 100 | 148 | 5.5 | 2536 | 94.5 | <0.0001 | 148 | 100 | 4 | 2.7 | 144 | 97.3 | 0.0928 |
| | | Normal | 27927 | 100 | 1576 | 5.6 | 26351 | 94.4 | | 1576 | 100 | 45 | 2.9 | 1531 | 97.1 | |
| | | Overweight | 16784 | 100 | 1061 | 6.3 | 15723 | 93.7 | | 1061 | 100 | 27 | 2.5 | 1034 | 97.5 | |
| | | Obese | 24093 | 100 | 1578 | 6.5 | 22515 | 93.5 | | 1578 | 100 | 65 | 4.1 | 1513 | 95.9 | |
Of the confirmed cases, $73.75\%$ of the deaths were of participants aged more than 70 years, while $1.41\%$ were of those under 50 years old. Most deaths occurred in people who lived in Daegu (80 people, fatality rate of $2.8\%$) and Gyeong-Buk (39 people, fatality rate of $6.8\%$), whereas only 22 people died in other regions. Regarding comorbidities, 44 patients ($1.5\%$) who had no underlying diseases died, while 97 patients (fatality rate, 1 = $4.2\%$, 2 = $6.2\%$, over 3 = $13\%$) died of COVID-19.
## 3.1. Associations between PA and COVID-19
According to logistic regression Model 1 of COVID-19 (Table 2), which was adjusted for characteristics and comorbidity, not engaging in sufficient MVPA ($95\%$ CI: 0.989–1.119, $$p \leq 0.108$$) did not affect the risk COVID-19. According to logistic regression Model 2 of COVID-19, which was additionally adjusted for the need for weight management, smoking, and drinking status, not engaging in sufficient MVPA (OR: 1.116, $95\%$ CI: 1.046–1.191, $p \leq 0.01$) predicted a higher risk of COVID-19 than engaging in sufficient MVPA. According to logistic regression Model 3 of COVID-19, which was additionally adjusted for disability status, not engaging in sufficient MVPA (OR: 1.078, $95\%$ CI: 1.014–1.147, $p \leq 0.05$) still predicted a higher risk of COVID-19 than sufficient MVPA.
**Table 2**
| Domain | Domain.1 | Domain.2 | Model 1 | Model 1.1 | Model 1.2 | Model 1.3 | Model 2 | Model 2.1 | Model 2.2 | Model 2.3 | Model 3 | Model 3.1 | Model 3.2 | Model 3.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | | Estimate | 95% CI | 95% CI | p-value | Estimate | 95% CI | 95% CI | p-value | Estimate | 95% CI | 95% CI | p-value |
| | | | | LR | UR | | | LR | UR | | | LR | UR | |
| PA | | Sufficient MVPA | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Insufficient MVPA | 1.052 | 0.989 | 1.119 | 0.1078 | 1.116 | 1.046 | 1.191 | 0.0009 | 1.078 | 1.014 | 1.147 | 0.016 |
| BMI | | Normal | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Underweight | 0.985 | 0.827 | 1.173 | 0.8614 | 1.009 | 0.847 | 1.203 | 0.9194 | 0.976 | 0.82 | 1.161 | 0.7817 |
| | | Overweight | 1.127 | 1.039 | 1.223 | 0.0041 | 1.128 | 1.039 | 1.224 | 0.004 | 1.128 | 1.041 | 1.223 | 0.0032 |
| | | Obese | 1.144 | 1.062 | 1.232 | 0.0004 | 1.182 | 1.095 | 1.277 | < 0.0001 | 1.172 | 1.09 | 1.259 | <0.0001 |
| Characteristics | Sex | Male | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Female | 1.024 | 0.958 | 1.094 | 0.4873 | 0.754 | 0.701 | 0.81 | <0.0001 | 1.038 | 0.974 | 1.105 | 0.2543 |
| | Age | 20–29 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | 30–39 | 1.334 | 1.139 | 1.562 | 0.0004 | 1.304 | 1.113 | 1.528 | 0.001 | 1.358 | 1.161 | 1.588 | 0.0001 |
| | | 40–49 | 1.514 | 1.312 | 1.746 | <0.0001 | 1.473 | 1.276 | 1.7 | <0.0001 | 1.563 | 1.359 | 1.799 | <0.0001 |
| | | 50–59 | 1.426 | 1.245 | 1.633 | <0.0001 | 1.332 | 1.161 | 1.528 | <0.0001 | 1.554 | 1.361 | 1.774 | <0.0001 |
| | | 60–69 | 1.301 | 1.131 | 1.497 | 0.0002 | 1.164 | 1.01 | 1.342 | 0.036 | 1.506 | 1.314 | 1.724 | <0.0001 |
| | | 70–79 | 1.19 | 1.015 | 1.395 | 0.0318 | 1.008 | 0.858 | 1.185 | 0.9212 | 1.417 | 1.217 | 1.649 | <0.0001 |
| | | 80- | 1.11 | 0.917 | 1.345 | 0.2839 | 0.907 | 0.747 | 1.101 | 0.3225 | 1.38 | 1.149 | 1.657 | 0.0006 |
| | Region of residence | Seoul | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Daegu | 0.905 | 0.794 | 1.032 | 0.1367 | 0.898 | 0.788 | 1.025 | 0.1113 | 0.951 | 0.835 | 1.082 | 0.4453 |
| | | Gyeonggi | 0.906 | 0.754 | 1.089 | 0.2949 | 0.916 | 0.761 | 1.101 | 0.3489 | 0.914 | 0.761 | 1.098 | 0.335 |
| | | Gyeong-Buk | 0.905 | 0.777 | 1.054 | 0.1985 | 0.894 | 0.767 | 1.042 | 0.1511 | 0.955 | 0.822 | 1.109 | 0.5451 |
| | | others | 0.901 | 0.768 | 1.055 | 0.1953 | 0.898 | 0.766 | 1.053 | 0.1863 | 0.94 | 0.803 | 1.101 | 0.4424 |
| | Health insurance premium | Medical aid | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | 1st quintile | 0.568 | 0.502 | 0.644 | <0.0001 | 0.564 | 0.498 | 0.639 | <0.0001 | 0.542 | 0.479 | 0.614 | <0.0001 |
| | | 2nd quintile | 0.471 | 0.412 | 0.538 | <0.0001 | 0.468 | 0.409 | 0.535 | <0.0001 | 0.438 | 0.384 | 0.499 | <0.0001 |
| | | 3rd quintile | 0.494 | 0.435 | 0.562 | <0.0001 | 0.491 | 0.432 | 0.558 | <0.0001 | 0.46 | 0.405 | 0.522 | <0.0001 |
| | | 4th quintile | 0.429 | 0.378 | 0.486 | <0.0001 | 0.419 | 0.369 | 0.475 | <0.0001 | 0.41 | 0.362 | 0.465 | <0.0001 |
| | | 5th quintile | 0.447 | 0.396 | 0.505 | <0.0001 | 0.427 | 0.378 | 0.482 | <0.0001 | 0.434 | 0.385 | 0.49 | <0.0001 |
| Comorbidity | Number of comorbidities | 0 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | 1 | 1.491 | 1.365 | 1.629 | <0.0001 | 1.486 | 1.36 | 1.624 | <0.0001 | 1.513 | 1.387 | 1.65 | <0.0001 |
| | | 2 | 1.292 | 1.169 | 1.429 | <0.0001 | 1.269 | 1.148 | 1.404 | <0.0001 | 1.306 | 1.183 | 1.44 | <0.0001 |
| | | 3+ | 1.971 | 1.725 | 2.251 | <0.0001 | 1.941 | 1.698 | 2.218 | <0.0001 | 2.003 | 1.762 | 2.278 | <0.0001 |
| Lifestyle | Smoking | No | | | | | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Yes | | | | | 0.746 | 0.686 | 0.812 | <0.0001 | 0.671 | 0.622 | 0.725 | <0.0001 |
| | Drinking | No | | | | | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | yes | | | | | 0.38 | 0.334 | 0.431 | <0.0001 | 0.414 | 0.368 | 0.466 | <0.0001 |
| | Weight management | Unnecessary | | | | | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Necessary | | | | | 0.928 | 0.867 | 0.992 | 0.0289 | 0.98 | 0.922 | 1.042 | 0.5137 |
| Disability | Presence | Non-disabled | | | | | | | | | 1 | 1 | 1 | 1 |
| | | Disabled | | | | | | | | | 1.452 | 1.302 | 1.618 | <0.0001 |
| | Severity | Non-disabled | | | | | | | | | 1 | 1 | 1 | 1 |
| | | Mild | | | | | | | | | 1.07 | 0.927 | 1.234 | 0.3581 |
| | | Severe | | | | | | | | | 2.5 | 2.129 | 2.936 | <0.0001 |
| | Type | Non-disabled | | | | | | | | | 1 | 1 | 1 | 1 |
| | | Physical disability | | | | | | | | | 1.06 | 0.89 | 1.262 | 0.5166 |
| | | Encephalopathy | | | | | | | | | 1.592 | 1.117 | 2.268 | 0.0101 |
| | | Visual impairment | | | | | | | | | 0.9 | 0.606 | 1.338 | 0.6035 |
| | | Hearing impairment | | | | | | | | | 1.355 | 1.057 | 1.736 | 0.0164 |
| | | Others | | | | | | | | | 3.401 | 2.787 | 4.15 | <0.0001 |
## 3.2. Associations between PA and COVID-19-associated mortality
According to logistic regression Model 1 of COVID-19-associated mortality (Table 3), which was adjusted for characteristics and comorbidity, not engaging in sufficient PA (OR: 1.548, $95\%$ CI: 1.051–2.279, $p \leq 0.05$) predicted a higher risk of COVID-19-associated mortality than engaging in sufficient PA. According to logistic regression Model 2 of COVID-19-associated mortality, which was additionally adjusted for the need for weight management, smoking, and drinking status, not engaging in sufficient MVPA (OR: 1.623, $95\%$ CI: 1.078–2.445, $p \leq 0.05$) still predicted a higher risk of COVID-19-associated mortality than engaging in sufficient MVPA. According to logistic regression Model 3 of COVID-19-associated mortality, which was additionally adjusted for disability status, engaging in sufficient MVPA ($95\%$ CI: 0.96–1.882, $$p \leq 0.085$$) did not predict COVID-19-associated mortality. However, the presence of disability and the levels of severity of disability predicted COVID-19-associated mortality.
**Table 3**
| Domain | Domain.1 | Domain.2 | Model 1 | Model 1.1 | Model 1.2 | Model 1.3 | Model 2 | Model 2.1 | Model 2.2 | Model 2.3 | Model 3 | Model 3.1 | Model 3.2 | Model 3.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | | Estimate | 95% CI | 95% CI | p-value | Estimate | 95% CI | 95% CI | p-value | Estimate | 95% CI | 95% CI | p-value |
| | | | | LR | UR | | | LR | UR | | | LR | UR | |
| PA | | Sufficient MVPA | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Insufficient MVPA | 1.548 | 1.051 | 2.279 | 0.0269 | 1.623 | 1.078 | 2.445 | 0.0204 | 1.344 | 0.96 | 1.882 | 0.0846 |
| BMI | | Normal | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Underweight | 0.737 | 0.224 | 2.424 | 0.6159 | 0.824 | 0.246 | 2.752 | 0.7524 | 0.945 | 0.335 | 2.665 | 0.9149 |
| | | Overweight | 0.829 | 0.487 | 1.411 | 0.4894 | 0.862 | 0.504 | 1.473 | 0.5859 | 0.888 | 0.548 | 1.441 | 0.6315 |
| | | Obese | 1.189 | 0.769 | 1.839 | 0.4368 | 1.336 | 0.836 | 2.134 | 0.2258 | 1.462 | 0.993 | 2.152 | 0.0544 |
| Characteristic | Sex | Male | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Female | 0.272 | 0.182 | 0.405 | <0.0001 | 0.231 | 0.153 | 0.35 | <0.0001 | 0.291 | 0.204 | 0.414 | <0.0001 |
| | Age | 20-59 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | 60-69 | 4.48 | 2.213 | 9.067 | <0.0001 | 4.059 | 2.001 | 8.231 | 0.0001 | 5.786 | 2.896 | 11.562 | <0.0001 |
| | | 70-79 | 19.832 | 10.125 | 38.846 | <0.0001 | 18.229 | 9.24 | 35.965 | <0.0001 | 23.367 | 12.241 | 44.605 | <0.0001 |
| | | 80- | 63.703 | 31.81 | 127.573 | <0.0001 | 57.833 | 28.652 | 116.731 | <0.0001 | 88.256 | 46.492 | 167.54 | <0.0001 |
| | Region of residence | Seoul | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Daegu | 2.165 | 0.598 | 7.835 | 0.2391 | 2.079 | 0.564 | 7.668 | 0.2715 | 2.522 | 0.791 | 8.042 | 0.1178 |
| | | Gyeonggi | 4.091 | 0.954 | 17.531 | 0.0578 | 3.761 | 0.858 | 16.493 | 0.0791 | 4.112 | 1.118 | 15.13 | 0.0334 |
| | | Gyeong-buk | 3.629 | 0.967 | 13.623 | 0.0561 | 3.301 | 0.863 | 12.622 | 0.0809 | 6.439 | 1.972 | 21.029 | 0.002 |
| | | others | 1.749 | 0.411 | 7.437 | 0.4492 | 1.556 | 0.359 | 6.751 | 0.555 | 1.886 | 0.506 | 7.029 | 0.3447 |
| | Health insurance premium | Medical aid | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | 1st quintile | 0.425 | 0.203 | 0.888 | 0.0229 | 0.443 | 0.212 | 0.926 | 0.0304 | 0.511 | 0.265 | 0.983 | 0.0445 |
| | | 2nd quintile | 0.552 | 0.239 | 1.273 | 0.1633 | 0.572 | 0.248 | 1.321 | 0.1909 | 0.424 | 0.198 | 0.907 | 0.0269 |
| | | 3rd quintile | 0.824 | 0.41 | 1.66 | 0.5888 | 0.829 | 0.41 | 1.675 | 0.6012 | 0.712 | 0.38 | 1.335 | 0.2894 |
| | | 4th quintile | 0.579 | 0.289 | 1.16 | 0.1233 | 0.593 | 0.295 | 1.192 | 0.1424 | 0.747 | 0.403 | 1.385 | 0.3542 |
| | | 5th quintile | 0.554 | 0.292 | 1.054 | 0.072 | 0.563 | 0.294 | 1.079 | 0.0836 | 1.046 | 0.598 | 1.829 | 0.8746 |
| Comorbidity | Number of comorbidities | 0 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | 1 | 1.88 | 1.117 | 3.164 | 0.0174 | 1.815 | 1.075 | 3.064 | 0.0258 | 2.867 | 1.78 | 4.616 | <0.0001 |
| | | 2 | 1.893 | 1.119 | 3.202 | 0.0173 | 1.853 | 1.093 | 3.141 | 0.022 | 4.262 | 2.664 | 6.817 | <0.0001 |
| | | 3+ | 2.303 | 1.345 | 3.946 | 0.0024 | 2.347 | 1.362 | 4.045 | 0.0021 | 9.703 | 6.15 | 15.31 | <0.0001 |
| Lifestyle | Smoking | No | | | | | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Yes | | | | | 0.305 | 0.151 | 0.614 | 0.0009 | 0.346 | 0.186 | 0.644 | 0.0008 |
| | Drinking | No | | | | | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | yes | | | | | 1.5 | 0.733 | 3.068 | 0.267 | 1.143 | 0.611 | 2.139 | 0.6759 |
| | Weight management | Unnecessary | | | | | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| | | Necessary | | | | | 0.864 | 0.564 | 1.325 | 0.5035 | 1.051 | 0.752 | 1.471 | 0.7694 |
| Disability | Presence | Non-disabled | | | | | | | | | 1 | 1 | 1 | 1 |
| | | Disabled | | | | | | | | | 3.461 | 2.318 | 5.168 | <0.0001 |
| | Severity | Non-disabled | | | | | | | | | 1 | 1 | 1 | 1 |
| | | Mild | | | | | | | | | 2.981 | 1.729 | 5.139 | <0.0001 |
| | | Severe | | | | | | | | | 4.041 | 2.394 | 6.822 | <0.0001 |
| | Type | Non-disabled | | | | | | | | | 1 | 1 | 1 | 1 |
| | | Physical disability | | | | | | | | | 3.106 | 1.629 | 5.92 | 0.0006 |
| | | Encephalopathy | | | | | | | | | 4.819 | 1.668 | 13.92 | 0.0037 |
| | | Visual impairment | | | | | | | | | <0.001 | <0.001 | >999.999 | 0.9787 |
| | | Hearing impairment | | | | | | | | | 3.442 | 1.458 | 8.127 | 0.0048 |
| | | Others | | | | | | | | | 4.35 | 2.372 | 7.979 | <0.0001 |
## 4. Discussion
This study aimed to determine the effect of MVPA on COVID-19 and the association between BMI and COVID-19, considering disability status. Investigating NHIS data on COVID-19 revealed several associations. Engaging in insufficient MVPA was associated with higher risk of infection and mortality associated with COVID-19, depending on confounding variables.
The importance of engaging in an active lifestyle was found to be influential on the risk of COVID-19 over the pandemic period. The result of this study aligns with those of a recent study, which reported the decrement of PA engagement over the period of COVID-19 [20] and the lower likelihood of developing COVID-19 [14]. However, the current study has taken a step forward by utilizing the most recent data available before the pandemic. It is widely known that engaging in adequate MVPA is associated with positive health outcomes. However, the current study adds another piece of information on the role of engaging in MVPA. For example, a special focus should be paid to individuals with disabilities regardless of the type of disabilities. During the pandemic, disadvantaged populations may experience issues with accessing health information and PA programs in local communities. For example, statistics show a $4\%$ decrement in PA participation by individuals with disabilities in 2021, compared to the participation rate in 2020 [21]. Thus, it might be important to dedicate effort toward improving their participation. A recent review emphasized the role of supportive environments to stimulate one's autonomy, competence, and relatedness (i.e., social supports from close people) in PA settings [22]. Given that older individuals are more susceptible to infection, these efforts must be focused on them. For example, local communities may provide newly developed programs to educate PA leaders, who may motivate older adults in their communities.
Engaging in insufficient MVPA could play an important role in reducing mortality associated with COVID-19. Previously, insufficient MVPA was reported as a negative predictor of all-cause mortality [23]. However, the relation has not been confirmed in the patients with COVID-19. Interestingly, the estimates from the mortality results are higher than from the infection results, which suggests the importance of regular MVPA participation. Since the infection is an on-going phenomenon, engaging in MVPA should be recommended in South Korea. Particular attention should be paid to people, who are aged or physically disabled. It is noted that individuals with disability show much higher risks of mortality than those without disability. Thus, barriers to PA should be eliminated, particularly perceived barriers, such as feeling uncomfortable, a lack of time, and other priorities [24]. A technique to overcome these barriers could be providing clear intervention based on the stage of intention of individuals with disability. It is important for researchers and regional practitioners to consider levels of motivation and volition (i.e., psychological willingness to participate in PA) [25] to provide proper PA guidelines.
This study has some limitations. First, this study's sampling period (8 months) was relatively short to determine the tendency of COVID-19, as the situation continues to evolve. At the beginning of this study, in August 2020, there were 4,222 confirmed cases and only 141 people had died of COVID-19 in South Korea. However, there are now more confirmed cases and deaths in South Korea. Second, it is difficult to generalize the results of South Korea and apply them to the rest of the world. Furthermore, the fatality rate in South Korea was deferred from the rate of the world. Thus, the results of this study may not be generalized. Finally, COVID-19 variants continue to evolve and affect individuals differently [26]. As a result, the results from this study could not be applied to different variants. Thus, further studies must include various cases and variants to strengthen the associative findings between PA and COVID-19. Lastly, the current study did not examine the roles of light PA and sedentary behavior due to the limited data from the secondary source. Future research may include those variables to clarify the results of the current study.
In conclusion, this study examined the role of sufficient MVPA on both COVID-19 and associated mortality and found that engaging in sufficient MVPA may play a role in reducing the risk of COVID-19 and associated mortality. The main implication of the results is that an active lifestyle should be promoted in the community. In addition, urgent implementation is needed for people who are older or physically disabled.
## Data availability statement
Data can be accessed with the permission of NHIS. Datasets of NHIS can be found here: https://nhiss.nhis.or.kr/bd/ay/bdaya001iv.do.
## Author contributions
Conceptualization and analysis: I-HO and S-YP. Writing: SP and HK. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Klopfenstein T, Zahra H, Kadiane-Oussou N, Lepiller Q, Royer PY, Toko L. **New loss of smell and taste: uncommon symptoms in COVID-19 patients on Nord Franche-Comte cluster, France**. *Int J Infect Dis.* (2020.0) **100** 117-22. DOI: 10.1016/j.ijid.2020.08.012
2. Zhang H, Kang Z, Gong H, Xu D, Wang J, Li Z. **The digestive system is a potential route of 2019-nCov infection: a bioinformatics analysis based on single-cell transcriptomes**. *BioRxiv* (2020.0) **3** 7806. DOI: 10.1101/2020.01.30.927806
3. **Coronavirus disease (COVID-19)**
4. **COVID-19 Weekly Epidemiological Update**
5. **Cases in Korea**. (2022.0)
6. Twenge JM, Joiner TE. **Census Bureau-assessed prevalence of anxiety and depressive symptoms in 2019 and during the 2020 COVID-19 pandemic**. *Depress Anxiety.* (2020.0) **37** 954-56. DOI: 10.1002/da.23077
7. Pieh C, Budimir S, Probst T. **The effect of age, gender, income, work, physical activity, and relationship status on mental health during coronavirus disease (COVID-19) lockdown in Austria**. *J Psychosom Res.* (2020.0) **136** 110186. DOI: 10.1016/j.jpsychores.2020.110186
8. Lei L, Huang X, Zhang S, Yang J, Yang L, Xu M. **Comparison of prevalence and associated factors of anxiety and depression among people affected by versus people unaffected by quarantine during the COVID-19 epidemic in Southwestern China**. *Med Sci Monit.* (2020.0) **26** 1-12. DOI: 10.12659/MSM.924609
9. Xie Y, Xu E, Al-Aly Z. **Risks of mental health outcomes in people with COVID-19: cohort study**. *BMJ* (2022.0) **376** e068993. DOI: 10.1136/bmj-2021-068993
10. Bo HX Li W, Yang Y, Wang Y, Zhang Q, Cheung T. **Posttraumatic stress symptoms and attitude toward crisis mental health services among clinically stable patients with COVID-19 in China**. *Psychol Med.* (2021.0) **51** 1052-3. DOI: 10.1017/S0033291720000999
11. Fox KR. **The influence of physical activity on mental wellbeing**. *Public Health Nutr.* (1999.0) **2** 411-8. DOI: 10.1017/S1368980099000567
12. Chin SH, Kahathuduwa CN, Binks M. **Physical activity and obesity: what we know and what we need to know**. *Obes Rev.* (2016.0) **17** 1226-44. DOI: 10.1111/obr.12460
13. Luciano F, Cenacchi V, Vegro V, Pavei G. **COVID-19 lockdown: Physical activity, sedentary behaviour and sleep in Italian medicine students**. *Eur J Sport Sci.* (2021.0) **21** 1459-68. DOI: 10.1080/17461391.2020.1842910
14. Cho DH, Lee SJ, Jae SY, Kim WJ, Ha SJ, Gwon JG. **Physical activity and the risk of COVID-19 infection and mortality: a nationwide population-based case-control study**. *J Clin Med.* (2021.0) **10** 1539. DOI: 10.3390/jcm10071539
15. Ezzatvar Y, Ramírez-Vélez R, Izquierdo M, Garcia-Hermoso A. **Physical activity and risk of infection, severity and mortality of COVID-19: a systematic review and non-linear dose–response meta-analysis of data from 1 853 610 adults**. *Br J Sports Med.* (2022.0) **56** 1188-93. DOI: 10.1136/bjsports-2022-105733
16. Lee J, Lee JS, Park SH, Shin SA, Kim K. **Cohort profile: the national health insurance service–national sample cohort (NHIS-NSC), South Korea**. *Int J Epidemiol.* (2017.0) **46** e15. DOI: 10.1093/ije/dyv319
17. Park S, Park SY, Cheong HK, Oh IH. **Moderate-to-vigorous physical activity and risk of all-cause mortality in people with anxiety disorders in South Korea**. *Int J Sport Exerc Psychol* (2022.0) **3** 1-13. DOI: 10.1080/1612197X.2022.2066708
18. Meister T, Pisarev H, Kolde R, Kalda R, Suija K, Milani L. **Clinical characteristics and risk factors for COVID-19 infection and disease severity: a nationwide observational study in Estonia**. *PLoS ONE.* (2022.0) **17** e0270192. DOI: 10.1371/journal.pone.0270192
19. Han E, Tan J, Turk E, Sridhar D, Leung M, Shibuya K. **Lessons learnt from easing COVID-19 restrictions: an analysis of countries and regions in Asia Pacific and Europe**. *Lancet.* (2020.0) **396** 1525-34. DOI: 10.1016/S0140-6736(20)32007-9
20. Yang HJ, Park S, Yoon TY, Ryoo JH, Park SK, Jung JY. **Nationwide changes in physical activity, nutrient intake, and obesity in South Korea during the COVID-19 pandemic era**. *Front Endocrinol.* (2022.0) **13** 965842. DOI: 10.3389/fendo.2022.965842
21. **[Internet]. Sejong; Sports for the Disabled Division. Press release, the participation rate of people with disabilities in sports for the disabled in 21 years decreased by 20.2% due to the influence of COVID-19**
22. Quested E, Kritz M, Hancox JE, Ntoumanis N, Thøgersen-Ntoumani C. **Promoting self-determined motivation for physical activity: From theory to intervention work**. *In Essentials of exercise and sport psychology; An open access textbook* (2021.0) 37-61
23. Clarke AE, Janssen I. **A compositional analysis of time spent in sleep, sedentary behavior and physical activity with all-cause mortality risk**. *Int J Behav Nutr Phys Act.* (2021.0) **18** 1-12. DOI: 10.1186/s12966-021-01092-0
24. Buffart LM, Westendorp T, Van Den Berg-Emons RJ, Stam H, Roebroeck ME. **Perceived barriers to and facilitators of physical activity in young adults with childhood-onset physical disabilities**. *J Rehabil Med.* (2009.0) **41** 881-5. DOI: 10.2340/16501977-0420
25. Jaarsma EA, Smith B. **Promoting physical activity for disabled people who are ready to become physically active: a systematic review**. *Psychol Sport Exerc.* (2018.0) **37** 205-23. DOI: 10.1016/j.psychsport.2017.08.010
26. **Tracking SARS-CoV-2 Variants**. (2022.0)
|
---
title: 'The association between sex hormones and periodontitis among American adults:
A cross-sectional study'
authors:
- Xingyang Su
- Kun Jin
- Xianghong Zhou
- Zilong Zhang
- Chichen Zhang
- Yifan Li
- Mi Yang
- Xinyi Huang
- Shishi Xu
- Qiang Wei
- Xu Cheng
- Lu Yang
- Shi Qiu
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9971556
doi: 10.3389/fendo.2023.1125819
license: CC BY 4.0
---
# The association between sex hormones and periodontitis among American adults: A cross-sectional study
## Abstract
### Introduction
After adulthood, as a person grows older, the secretion of sex hormones in the body gradually decreases, and the risk of periodontitis increases. But the relationship between sex hormones and periodontitis is still controversial.
### Methods
We investigated the association between sex hormones and periodontitis among Americans over 30 years old. 4,877 participants containing 3,222 males and 1,655 postmenopausal females who had had periodontal examination and detailed available sex hormone levels, were included in our analysis from the 2009-2014 National Health and Nutrition Examination Surveys cycles. We applied multivariate linear regression models to estimate the connection between sex hormones and periodontitis after converting sex hormones into categorical variables through tertile. Additionally, to ensure the stability of the analysis results, we carried out a trend test, subgroup analysis, and interaction test.
### Results
After fully adjusting the covariates, estradiol levels were not associated with periodontitis in both males and females with a P for trend = 0.064 and 0.064, respectively. For males, we found that sex hormone-binding globulin was positively associated with periodontitis (tertile3 vs tertile1: OR=1.63, $95\%$ CI=1.17-2.28, $$p \leq 0.004$$, P for trend = 0.005). Congruously, free testosterone (tertile3 vs tertile1: OR=0.60, $95\%$ CI=0.43-0.84, $$p \leq 0.003$$), bioavailable testosterone (tertile3 vs tertile1: OR=0.51, $95\%$ CI=0.36-0.71, $p \leq 0.001$), and free androgen index (tertile3 vs tertile1: OR=0.53, $95\%$ CI=0.37-0.75, $p \leq 0.001$) was found to be negatively associated with periodontitis. Moreover, subgroup analysis of age found a closer relationship between sex hormones and periodontitis in those younger than 50 years.
### Conclusion
Our research suggested that males with lower bioavailable testosterone levels affected by sex hormone-binding globulin were at a higher risk of periodontitis. Meanwhile, estradiol levels were not associated with periodontitis in postmenopausal women.
## Introduction
By 2100, the global population of those older than 65 years is anticipated to reach 2.37 billion according to the Global Burden of Disease Study [1]. The estimated global average prevalence of severe periodontitis is about $11\%$, where the prevalence of severe periodontitis in adults and the elderly is at a minimum $16.9\%$ and possibly up to $48.0\%$ [2]. Disability-adjusted life-years (DALYs) for oral health problems due to population aging increased by $28.02\%$ ($95\%$ UI 25.87-30.09) in the 20 years from 1990 to 2010 [3]. The 2019 Global Burden of Disease study indicated that oral diseases, including severe periodontitis (chronic gum disease), cause 23.1 million DALYs all over the world [4]. Management of periodontitis is highly relevant in light of a growing aging population cohort and with the consequential disease burden increasing.
Periodontitis is a chronic, microbial-associated, host-mediated inflammatory disease of the oral cavity [5]. The inflammation initiated and driven by a dysbiosis of bacterial biofilm contributes to the progressive destruction of the dental supporting tissues [6], including the loss of periodontal ligaments and alveolar bone [2]. Several modifiable risk factors have been established, such as poor oral hygiene, cigarette smoking, and poorly controlled diabetes mellitus [7]. Studies have also found that there are many potential risk factors for periodontitis, such as metabolic syndrome, alcohol consumption, osteoporosis, a low intake of calcium and vitamin D, obesity (8–12), and hyperuricemia [13]. Interestingly, sex hormones are associated with many of the diseases or factors mentioned above.
Sex hormones are important endocrine substances in the human body that are related to many functions, such as growth, reproduction, and differentiation. Testosterone (T) is a primary male sex hormone synthesized from cholesterol in Leydig cells in males, and estradiol (E2), another of the sex hormones, is mainly derived from the ovaries in females. But E2 is also derived from the conversion of androgens by aromatase in adipose tissue in males and adrenal glands and ovaries also produce T in females. Low T and E2 are connected with a series of medical disorders including obesity, metabolic syndrome, diabetes, cardiovascular disease, and osteoporosis (14–18). Sex hormones play a powerful role in the regulation of inflammatory response and bone turnover through cytokine regulation and direct actions on osteoblast and osteoclast precursors (19–21). A study of male rats found that induced periodontitis in low and high testosterone groups resulted in increased alveolar bone loss compared to a control group [22]. Meanwhile, both low T and E2 levels and an elevated incidence of periodontitis are associated with aging (23–25). Based on the association, it is plausible that low sex hormone levels could be associated with periodontitis.
The prevalence estimate of periodontitis depends on the case definition of periodontitis and methods of periodontal surveillance [26]. There are some differences between the full mouth periodontal examination (FMPE) protocol and the partial mouth periodontal examination (PMPE) protocol for periodontal examination. Though both protocols exclude the third molars, the FMPE protocol comprehensively measures all the teeth and measures more positions per tooth compared to the PMPE protocol. It is more appropriate to detect periodontitis cases using the FMPE protocol instead of the PMPE protocol because partial-mouth protocols underestimate the prevalence of periodontitis [2, 27, 28]. As our investigation sought to assess associations between serum sex steroid levels and periodontitis in an age cohort beyond 30 years, our database necessarily included datasets from the National Health and Nutrition Examination Survey (NHANES) from 2009-2014 [28].
## Data source and participants
The data that we used for the analysis were derived from the NHANES. The Centers for Disease Control and Prevention (CDC) conducts the NHANES in the United *States via* questionnaires and examinations in a two-year cycle, with the aim of establishing a database of the health and nutritional status of the US population. As a cross-sectional survey with a complex, multistate, probabilistic sampling design, the NHANES ensures that nationally representative data are available. The National Center for Health Statistics (NCHS) institutional Review Board examines and gives permission to the study and each participant in the survey signs the appropriate written permission. More elaborate information on the methods and protocol of the NHANES is available on the CDC’s official website.
30,468 participants from the 2009-2010, 2011-2012, and 2013-2014 cycles were selected. The study population was restricted to participants with detailed sex hormone levels and a diagnosis of periodontitis. Considering the dramatic changes in hormone levels during the female physiological cycle, we chose only postmenopausal women who answered “no” to the question “Have you had at least one menstrual period in the past 12 months?” and answered “hysterectomy” or “menopause/life changes” to the question “*What is* the reason why you have not had a period in the last 12 months?” in the self-reported reproductive health questionnaire [29]. Finally, 4,877 participants were included in the analysis, containing 3,222 males and 1,655 females. The inclusion and exclusion flow of the analyzed objects was demonstrated in Figure 1.
**Figure 1:** *Flow chart of inclusion and exclusion criteria for our analysis.*
## Sex hormones
Serum total T (TT) and E2 levels were measured using an isotope dilution high-performance liquid chromatography-tandem mass spectrometry (ID-LC-MS/MS) in the NHANES, while the concentrations of sex hormone-binding globulin (SHBG) were quantified using chemo-luminescence measurements of the reaction products via a photomultiplier tube after binding to SHBG with immuno-antibodies [30]. According to previous literature, we only indirectly assessed the approximate amount of circulating free testosterone (FT) [31] and the activity of aromatase [32] through the free androgen index (FAI) that was calculated as the value of TT (ng/dL) divided by SHBG (nmol/L) and the ratio of TT to E2 (TT/E2), respectively. Bioavailable testosterone was calculated according to the Vermeulen et al. formula [33].
## Periodontitis
The FMPE that included clinical attachment loss (AL) and probing depth (PD) was used for the periodontal examination from 2009 to 2014. The dental examiner used the HU-Friedy periodontal probe to perform the assessment. AL and PD were assessed at six positions per tooth using a perioprobe for a maximum of up to 28 teeth [2, 34]. Cases of periodontitis were determined according to the suggested CDC-AAP definitions which were used for monitoring periodontitis. Periodontitis requires the patient’s periodontal measurement of ≥ 2 inter-proximal sites with ≥ 3 mm AL and ≥ 2 inter-proximal sites with ≥ 4 mm PD on different teeth or one site with ≥ 5 mm [27]. In this study, the severity of periodontitis was not differentiated. The absence of any symptoms of periodontitis described above was defined as the absence of periodontitis.
## Covariates
Covariates associated with periodontitis and T were collected, including some demographic information such as age, race, ratio of family income to poverty, educational level, and civil state. All demographic data could be obtained from questionnaire data and they were converted into the corresponding categorical variables. Smoking status was classified as never, former, and current, respectively, and alcohol intake per day was classified as never, moderate, and heavy, respectively. Time of venipuncture, BMI, diabetes, and hypertension were also potential confounding factors. White blood cells (WBC), which partly reflect inflammation in an individual’s body, were also adjusted in the analysis (35–37).
## Statistical analysis
We used the statistical software package R (http://www. Rproject.org,theRFoundation) and Empower (R) to carry out all the analysis, and a p-value of ≤ 0.05 was required to be statistically significant. In the data analysis, continuous variables and categorical variables were denoted by mean ± SD/Median (Q1-Q3) and proportions, respectively. A sample weight was assigned to each participant because of the multistate, probabilistic sampling design [38]. For categorical variables and continuous variables, we performed a weighted chi-square test and a Kruskal Wallis test to calculate characteristics of otherness between male and female groups, respectively. Our purpose was to probe the relationship between sex hormones and periodontitis, so we used weighted multivariate linear regression models in different sex groups after sex hormones tertiles. Multivariate models included the non-adjusted model, adjusted model (only age, race, BMI, ratio of family income to poverty, education level, time of venipuncture, civil state, smoking status, alcohol intake per day, and WBC level were adjusted), and adjusted model II (diabetes and hypertension were adjusted additionally). Simultaneously, we carried out a trend test to confirm the stability of the results. Finally, we used weighted stratified line regression models for subgroup analysis of age. Some categorical covariables were used to perform an interaction test. We used interaction terms between subgroup indicators to test the effect modification in subgroups, followed by a likelihood-ratio test.
## Results
Sex hormone levels and covariates of the male and female periodontitis participants in this study are characterized in Table 1. The prevalence of periodontitis in males was $59.5\%$, obviously higher than that in females ($34.2\%$). The population characteristics were not consistent in age, race, BMI, ratio of family income to poverty, education level, time of venipuncture, civil state, smoking status, alcohol intake per day, and also WBC level, diabetes, and hypertension in the two groups of participants: that is, there were statistical differences. Compared with the female groups, the male groups were more likely to be older, married, or living with a partner, a former or current smoker, and have diabetes and hypertension, as well as having a lower WBC level, a higher poverty-to-income ratio, a lower education level, a higher BMI, and a higher alcohol intake per day. It was not surprising that participants in the male groups had higher TT levels, lower E2 levels, lower SHBG levels, higher FT levels, higher FAI levels, and higher bioavailable testosterone levels. Table S1 shows the comparison of sex hormone levels between groups with and without periodontitis. There was no significant difference in testosterone and estradiol levels between the periodontitis group and the non-periodontitis group in males or females. In males, the periodontitis group had higher SHBG levels and lower bioavailable testosterone levels than the non-periodontitis group. Inversely, the periodontitis group had a lower SHBG level than the non-periodontitis group in females.
**Table 1**
| Gender | Male | Female | Standardize diff. | P-value |
| --- | --- | --- | --- | --- |
| Number | 3222 | 1655 | | |
| Age, years, mean (SD) | 51.99 (14.38) | 42.31 (10.32) | 0.77 (0.71, 0.83) | <0.001 |
| Testosterone, nmol/L, median (Q1-Q3) | 1.07 (0.81-1.42) | 0.06 (0.04-0.08) | 2.92 (2.84, 3.00) | <0.001 |
| Estradiol, pg/mL, median (Q1-Q3) | 23.40 (18.20-29.10) | 66.50 (26.20-129.00) | 0.96 (0.87, 1.04) | <0.001 |
| SHBG, nmol/L, median (Q1-Q3) | 38.59 (27.89-54.05) | 65.74 (44.88-98.10) | 0.79 (0.70, 0.88) | <0.001 |
| free testosterone, nmol/L, median (Q1-Q3) | 0.02 (0.01-0.02) | 0.00 (0.00-0.00) | 3.74 (3.60, 3.88) | <0.001 |
| Bioavailable testosterone, nmol/L, median (Q1-Q3) | 0.41 (0.33-0.50) | 0.02 (0.01-0.02) | 3.55 (3.41, 3.69) | <0.001 |
| Free androgen index | 2.81 (2.17-3.62) | 0.09 (0.06-0.14) | 2.57 (2.46, 2.69) | <0.001 |
| WBC, 10^9/L, mean (SD) | 7.01 (2.11) | 7.33 (2.24) | 0.15 (0.09, 0.21) | <0.001 |
| Race, n (%) | | | 0.10 (0.04, 0.16) | 0.02 |
| Mexican American | 411 (12.76%) | 220 (13.29%) | | |
| Other Hispanic | 285 (8.85%) | 153 (9.24%) | | |
| Non-Hispanic White | 1333 (41.37%) | 623 (37.64%) | | |
| Non-Hispanic Black | 691 (21.45%) | 346 (20.91%) | | |
| Other Race | 502 (15.58%) | 313 (18.91%) | | |
| Ratio of family income to poverty, n (%) | | | 0.11 (0.05, 0.17) | <0.001 |
| <1.3 | 862 (26.75%) | 527 (31.84%) | | |
| 1.3-3.5 | 1008 (31.28%) | 485 (29.31%) | | |
| >3.5 | 1352 (41.96%) | 643 (38.85%) | | |
| Education level, n (%) | | | 0.10 (0.04, 0.16) | 0.005 |
| Less than high school | 669 (21.27%) | 315 (19.61%) | | |
| High school or GED General educational development | 678 (21.56%) | 296 (18.43%) | | |
| Above high school | 1798 (57.17%) | 995 (61.96%) | | |
| Civil state, n (%) | | | 0.10 (0.04, 0.16) | 0.001 |
| Married or living with partner | 2301 (71.44%) | 1108 (66.99%) | | |
| Living alone | 920 (28.56%) | 546 (33.01%) | | |
| BMI, n (%) | | | 0.28 (0.22, 0.34) | <0.001 |
| <=25 | 828 (25.82%) | 543 (33.09%) | | |
| 25-30 | 1307 (40.75%) | 454 (27.67%) | | |
| >30 | 1072 (33.43%) | 644 (39.24%) | | |
| Smoking status, n (%) | | | 0.47 (0.41, 0.53) | <0.001 |
| Never | 1515 (47.05%) | 1135 (68.58%) | | |
| Former | 1134 (35.22%) | 290 (17.52%) | | |
| Current | 571 (17.73%) | 230 (13.90%) | | |
| Alcohol intake per day, n (%) | | | 0.32 (0.26, 0.38) | <0.001 |
| | 2293 (71.17%) | 1348 (81.45%) | | |
| Moderate | 390 (12.10%) | 67 (4.05%) | | |
| Heavy | 539 (16.73%) | 240 (14.50%) | | |
| Diabetes, n (%) | | | 0.23 (0.17, 0.29) | <0.001 |
| No | 2693 (86.12%) | 1505 (93.02%) | | |
| Yes | 434 (13.88%) | 113 (6.98%) | | |
| Hypertension, n (%) | | | 0.29 (0.23, 0.35) | <0.001 |
| No | 1998 (62.09%) | 1246 (75.38%) | | |
| Yes | 1220 (37.91%) | 407 (24.62%) | | |
| Time of venipuncture, n (%) | | | 0.07 (0.01, 0.13) | 0.049 |
| Morning | 1582 (49.10%) | 798 (48.22%) | | |
| Afternoon | 1172 (36.37%) | 573 (34.62%) | | |
| Evening | 468 (14.53%) | 284 (17.16%) | | |
| Periodontitis, n (%) | | | 0.52 (0.46, 0.58) | <0.001 |
| No | 1304 (40.47%) | 1089 (65.80%) | | |
| Yes | 1918 (59.53%) | 566 (34.20%) | | |
Table 2 shows the association of sex hormone levels with periodontitis in males after sorting by tertile. The results of our analysis showed no significant relationship between TT levels and periodontitis (tertile3 vs tertile1: OR=0.91, $95\%$ CI=0.73–1.13, $$p \leq 0.384$$, P for trend = 0.491). E2 levels were not associated with periodontitis (tertile2 vs tertile1: OR=0.80, $95\%$ CI=0.61-1.07, $$p \leq 0.131$$; tertile3 vs tertile1: OR=0.75, $95\%$ CI=0.56–1.01, $$p \leq 0.060$$). The trend test gave a similar result (P for trend = 0.064). For SHBG, we found that SHBG (tertile3 vs tertile1: OR=1.63, $95\%$ CI=1.17-2.28, $$p \leq 0.004$$) was positively associated with periodontitis with P for trend = 0.005. Congruously, FT (tertile3 vs tertile1: OR=0.60, $95\%$ CI=0.43-0.84, $$p \leq 0.003$$), bioavailable testosterone (tertile3 vs tertile1: OR=0.51, $95\%$ CI=0.36-0.71, $p \leq 0.001$), and FAI (tertile3 vs tertile1: OR=0.53, $95\%$ CI=0.37-0.75, $p \leq 0.001$) was found to be negatively associated with periodontitis with P for trend = 0.004, < 0.001, and < 0.001, respectively.
**Table 2**
| Exposure | Non-adjusted | Adjusted I | Adjusted II |
| --- | --- | --- | --- |
| Testosterone, ng/dL, OR (95%CI) P-value | Testosterone, ng/dL, OR (95%CI) P-value | Testosterone, ng/dL, OR (95%CI) P-value | Testosterone, ng/dL, OR (95%CI) P-value |
| T1 (1.44 - 314.08) | 1 | 1 | 1 |
| T2 (314.39 - 448.38) | 0.86 (0.72, 1.02) 0.084 | 0.83 (0.68, 1.01) 0.069 | 0.82 (0.67, 1.01) 0.061 |
| T3 (448.41 - 2543.99) | 1.04 (0.87, 1.24) 0.658 | 0.90 (0.73, 1.12) 0.345 | 0.91 (0.73, 1.13) 0.384 |
| P for trend | 1.00 (1.00, 1.00) 0.503 | 1.00 (1.00, 1.00) 0.428 | 1.00 (1.00, 1.00) 0.491 |
| Estrodiol, pg/mL, OR (95%CI) P-value | Estrodiol, pg/mL, OR (95%CI) P-value | Estrodiol, pg/mL, OR (95%CI) P-value | Estrodiol, pg/mL, OR (95%CI) P-value |
| TI (2.12 - 20.00) | 1 | 1 | 1 |
| T2 (20.10 - 27.00) | 0.77 (0.60, 0.97) 0.030 | 0.79 (0.60, 1.04) 0.091 | 0.80 (0.61, 1.07) 0.131 |
| T3 (27.10 - 95.10) | 0.91 (0.71, 1.15) 0.430 | 0.77 (0.58, 1.03) 0.078 | 0.75 (0.56, 1.01) 0.060 |
| P for trend | 0.99 (0.98, 1.01) 0.496 | 0.98 (0.97, 1.00) 0.086 | 0.98 (0.96, 1.00) 0.064 |
| SHBG, nmol/L, OR (95%CI) P-value | SHBG, nmol/L, OR (95%CI) P-value | SHBG, nmol/L, OR (95%CI) P-value | SHBG, nmol/L, OR (95%CI) P-value |
| TI (6.90 - 31.41) | 1 | 1 | 1 |
| T2 (31.42 - 48.06) | 1.59 (1.24, 2.05) 0.0003 | 1.28 (0.95, 1.73) 0.101 | 1.26 (0.93, 1.71) 0.139 |
| T3(48.08 - 196.50) | 2.47 (1.90, 3.21) <0.0001 | 1.66 (1.20, 2.31) 0.002 | 1.63 (1.17, 2.28) 0.004 |
| P for trend | 1.02 (1.02, 1.03) <0.0001 | 1.01 (1.00, 1.02) 0.003 | 1.01 (1.00, 1.02) 0.005 |
| Free testosterone, ng/dL, OR (95%CI) P-value | Free testosterone, ng/dL, OR (95%CI) P-value | Free testosterone, ng/dL, OR (95%CI) P-value | Free testosterone, ng/dL, OR (95%CI) P-value |
| TI (5.35e-05 - 0.015362954) | 1 | 1 | 1 |
| T2 (0.015384232 - 0.019664324) | 0.62 (0.48, 0.81) 0.0004 | 0.71 (0.52, 0.95) 0.023 | 0.74 (0.54, 1.00) 0.052 |
| T3 (0.01966649 - 0.103391716) | 0.46 (0.35, 0.59) <0.0001 | 0.57 (0.41, 0.80) 0.001 | 0.60 (0.43, 0.84) 0.003 |
| P for trend | 0.00 (0.00, 0.00) <0.0001 | 0.00 (0.00, 0.00) 0.001 | 0.00 (0.00, 0.00) 0.004 |
| Bioavailable testosterone, ng/dL, OR (95%CI) P-value | Bioavailable testosterone, ng/dL, OR (95%CI) P-value | Bioavailable testosterone, ng/dL, OR (95%CI) P-value | Bioavailable testosterone, ng/dL, OR (95%CI) P-value |
| TI (0.001091759 - 0.355356499) | 1 | 1 | 1 |
| T2 (0.355702816 - 0.464128567) | 0.64 (0.49, 0.83) 0.0007 | 0.74 (0.55, 1.00) 0.052 | 0.75 (0.55, 1.02) 0.068 |
| T3 (0.464900842 - 2.638546259) | 0.38 (0.29, 0.50) <0.0001 | 0.49 (0.35, 0.68) <0.0001 | 0.51 (0.36, 0.71) 0.0001 |
| P for trend | 0.02 (0.01, 0.07) <0.0001 | 0.06 (0.02, 0.22) <0.0001 | 0.07 (0.02, 0.27) <0.0001 |
| Free androgen index, OR (95%CI) P-value | Free androgen index, OR (95%CI) P-value | Free androgen index, OR (95%CI) P-value | Free androgen index, OR (95%CI) P-value |
| TI (0.006991234 - 2.4220154) | 1 | 1 | 1 |
| T2 (2.4225 - 3.27821137) | 0.60 (0.46, 0.78) 0.0001 | 0.67 (0.49, 0.90) 0.009 | 0.68 (0.50, 0.93) 0.017 |
| T3 (3.2828038 - 32.33420366) | 0.33 (0.25, 0.42) <0.0001 | 0.49 (0.35, 0.69) <0.0001 | 0.53 (0.37, 0.75) 0.0004 |
| P for trend | 0.59 (0.53, 0.67) <0.0001 | 0.72 (0.62, 0.85) <0.0001 | 0.75 (0.64, 0.88) 0.001 |
Table 3 shows the association of sex hormone levels with periodontitis in females after sorting by tertile. Similar to males, we found no significant relationship (tertile3 vs tertile1: OR=1.00, $95\%$ CI=0.82-1.22, $$p \leq 0.984$$, P for trend = 0.963) between TT level and periodontitis. E2 levels were also not associated with periodontitis in females (tertile2 vs tertile1: OR=0.67, $95\%$ CI=0.50-0.90, $$p \leq 0.009$$; tertile3 vs tertile1: OR=0.59, $95\%$ CI=0.41-0.86, $$p \leq 0.005$$; P for trend = 0.064). However, no significant relationship was observed between SHBG and periodontitis in women (tertile3 vs tertile1: OR=1.16, $95\%$ CI=0.85-1.58, $$p \leq 0.354$$, P for trend = 0.508).
**Table 3**
| Exposure | Non-adjusted | Adjusted I | Adjusted II |
| --- | --- | --- | --- |
| Testosterone, ng/dL, OR (95%CI) P-value | Testosterone, ng/dL, OR (95%CI) P-value | Testosterone, ng/dL, OR (95%CI) P-value | Testosterone, ng/dL, OR (95%CI) P-value |
| T1 (1.02 - 14.28) | 1 | 1 | 1 |
| T2 (14.3 - 22.88) | 0.86 (0.72, 1.01) 0.069 | 1.05 (0.86, 1.27) 0.649 | 1.04 (0.86, 1.26) 0.690 |
| T3 (22.9 - 575.0) | 0.78 (0.66, 0.92) 0.004 | 1.00 (0.82, 1.22) 0.997 | 1.00 (0.82, 1.22) 0.984 |
| P for trend | 0.99 (0.98, 1.00) 0.005 | 1.00 (0.99, 1.01) 0.934 | 1.00 (0.99, 1.01) 0.963 |
| Estradiol, pg/mL, OR (95%CI) P-value | Estradiol, pg/mL, OR (95%CI) P-value | Estradiol, pg/mL, OR (95%CI) P-value | Estradiol, pg/mL, OR (95%CI) P-value |
| T1 (2.117 - 7.25) | 1 | 1 | 1 |
| T2 (7.26 - 49.20) | 0.59 (0.47, 0.75) <0.0001 | 0.67 (0.50, 0.90) 0.007 | 0.67 (0.50, 0.90) 0.009 |
| T3 (49.3 - 1220) | 0.38 (0.29, 0.48) <0.0001 | 0.59 (0.41, 0.85) 0.005 | 0.59 (0.41, 0.86) 0.005 |
| P for trend | 0.99 (0.99, 0.99) <0.0001 | 1.00 (0.99, 1.00) 0.061 | 1.00 (0.99, 1.00) 0.064 |
| SHBG, nmol/L, OR (95%CI) P-value | SHBG, nmol/L, OR (95%CI) P-value | SHBG, nmol/L, OR (95%CI) P-value | SHBG, nmol/L, OR (95%CI) P-value |
| T1 (9.20 - 47.82) | 1 | 1 | 1 |
| T2 (47.86 - 79.86) | 1.15 (0.90, 1.46) 0.274 | 1.28 (0.96, 1.71) 0.090 | 1.27 (0.95, 1.70) 0.103 |
| T3 (79.90 - 758.8) | 0.89 (0.70, 1.14) 0.365 | 1.17 (0.86, 1.59) 0.322 | 1.16 (0.85, 1.58) 0.354 |
| P for trend | 1.00 (0.99, 1.00) 0.238 | 1.00 (1.00, 1.01) 0.472 | 1.00 (1.00, 1.01) 0.508 |
It is well known that the occurrence of periodontitis is closely associated with age, and the subgroup analysis of age is shown in Tables 4, 5. In adjusted model II, there was a weak association between testosterone and periodontitis in participants aged younger than 50 years in females (P for trend = 0.046). Interestingly, we found that subjects with higher SHBG levels were more likely to develop periodontitis in males younger than 50 years old (tertile2 vs tertile1: OR=1.48, $95\%$ CI=0.99-2.20, $$p \leq 0.056$$; tertile3 vs tertile1: OR=1.84, $95\%$ CI=1.11-3.07, $$p \leq 0.019$$; P for trend = 0.014), while this phenomenon was not present in males older than 50 years. At the same time, as shown in Tables S2, S3, we did not detect a significant interaction regarding the correlation between periodontitis and sex hormones.
## Discussion
The target of our study was to explore the connection between sex hormones and periodontitis. This investigation found that TT levels and E2 levels were unrelated to periodontitis, while SHBG levels were significantly positively associated with periodontitis, and FT, FAI, and bioavailable testosterone were all significantly negatively correlated with periodontitis in men.
The relationship between TT and periodontitis is still unclear. Our study found no significant relationship between TT levels and periodontitis and a positive association between SHBG with periodontitis for males. Similar to our study, Orwoll et al. reported the occurrence and progression of periodontitis and tooth loss were independent of serum sex hormone levels among older males over 65 years [35]. Consistently, a population-based longitudinal cohort study from Pomerania found no obvious connections between sex hormones with the progression of periodontal measurement or tooth loss [36]. However, some studies reported the opposite. Singh et al. reported that the T levels in participants without tooth loss were significantly higher than those in participants with tooth loss in 2011, suggesting T could predict tooth loss very well [39]. Androgen deprivation therapy (ADT) is a critical therapeutic option for treating prostate cancer, which is designed to prevent the development of prostate cancer by lowering androgen levels in a patient. Famili et al. spotted that the prevalence of parodontopathy was obviously higher among men on ADT compared with those not on ADT ($80.5\%$ vs $3.7\%$) [40]. This paradoxical result makes one wonder if SHBG is where our attention should be. Since T bound to SHBG would lose bioactivity, bioavailable testosterone, including FT and albumin-bound T, could be affected by the level of SHBG. Our conjecture is also supported by the negative correlation between FT, FAI, and bioavailable testosterone and periodontitis in our study. Our results are in concordance with another similar study in 2015 based on the NHANES III, which found that low SHBG was inversely associated with periodontitis [37]. This study found that male subjects under the age of 50 with higher SHBG levels were more likely to develop periodontitis (P for trend = 0.014), while those over the age of 50 did not (P for trend = 0.144). Consistently, a cohort study of older adults over 65 years of age found that sex steroid and SHBG levels were not associated with the baseline mean clinical attachment loss and mean PD [35]. We currently lack evidence to explain this discrepancy.
T is known to affect bone metabolism through the modulation of immunological events in existing periodontitis [41, 42]. A study has found that dihydrotestosterone (DHT) can downregulate the production of IL-6 and upregulate fibroblast proliferation simultaneously [43]. Due to lower DHT, the production of the increased proinflammatory cytokine IL-6 by osteoclasts may play a crucial role via osteoclastic activity in bone resorption [44]. As the predominant cells in periodontal connective tissues, fibroblasts play an important role under inflammatory conditions. In addition, a study showed that T replacement therapy in hypogonadal men significantly induced reductions in serum pro-inflammatory cytokines including TNF and IL-1β [45]. T may prevent and control infection of the host by bacteria by increasing expression of E-selectin (induced by TNF-α) and vascular adhesion molecule-1 in endotheliocyte, which are paramount in trans-endothelial migration of phagocytic cells [46, 47]. In male rats, testosterone therapy was found to increase the proportion of blood vessels, the extracellular matrix, and fibroblasts in the presence of periodontal inflammation, which may be related to the regulatory effects of PGE2 and IL-10 [48]. The decrease in bioavailable testosterone affected by SHBG may be one of the risk factors for the occurrence of periodontitis.
Our study was conducted utilizing the NHANES data, which is representative of the US population and has been obtained using standard protocols and measurements. This ensured that our results had good extrapolation and data reliability. However, this study had some inevitable shortcomings and limitations. Firstly, the cross-sectional nature of the NHANES limited the conclusions about gonadal hormone levels and periodontitis to possible associations, rather than causes. Secondly, serum TT levels were only measured at one time-point in the NHANES data, while American Urological Association (AUA) guidelines recommend two measurements due to intra-individual and diurnal variations of serum T [33]. Thirdly, because of the lack of relevant data, such as inflammatory biomarkers, hormone-related drugs, TSH, and free T4, these confounding factors were not adjusted.
## Conclusions
Our research suggested that males with lower bioavailable testosterone levels affected by SHBG were at a higher risk of periodontitis. But we need to carry out a large, carefully designed prospective study for clarifying the causal association between sex hormones and periodontitis due to mechanism complexity.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors.
## Author contributions
Author contributions are as follows: KJ, SQ, QW, and LY contributed equally by conceiving and designing the study. XS, KJ, and SQ organized and analyzed the data. XZ and XS wrote the paper. LY, XC, XZ, ZZ, CZ, YL, MY, XH, and SX revised the manuscript critically for important intellectual content. MY and XH helped to complete the design and evaluation of the statistical methodology in this study. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1125819/full#supplementary-material
## References
1. Vollset SE, Goren E, Yuan CW, Cao J, Smith AE, Hsiao T. **Fertility, mortality, migration, and population scenarios for 195 countries and territories from 2017 to 2100: A forecasting analysis for the global burden of disease study**. *Lancet.* (2020) **396**. DOI: 10.1016/S0140-6736(20)30677-2
2. Slots J. **Periodontitis: Facts, fallacies and the future**. *Periodontol 2000* (2017) **75** 7-23. DOI: 10.1111/prd.12221
3. Marcenes W, Kassebaum NJ, Bernabé E, Flaxman A, Naghavi M, Lopez AE. **Global burden of oral conditions in 1990-2010: a systematic analysis**. *J Dent Res* (2013) **92**. DOI: 10.1177/0022034513490168
4. Patel J, Wallace J, Doshi M, Gadanya M, Ben Yahya I, Roseman J. **Oral health for healthy ageing**. *Lancet Healthy Longev* (2021) **2**. DOI: 10.1016/S2666-7568(21)00142-2
5. Tonetti MS, Greenwell H, Kornman KS. **Staging and grading of periodontitis: Framework and proposal of a new classification and case definition**. *J Periodontol* (2018) **89 Suppl 1**. DOI: 10.1002/JPER.18-0006
6. Vollmer A, Vollmer M, Lang G, Straub A, Shavlokhova V, Kübler A. **Associations between periodontitis and COPD: An artificial intelligence-based analysis of NHANES III**. *J Clin Med* (2022) **11** 7210. DOI: 10.3390/jcm11237210
7. Kinane DF, Stathopoulou PG, Papapanou PN. **Periodontal diseases**. *Nat Rev Dis Primers* (2017) **3** 17038. DOI: 10.1038/nrdp.2017.38
8. Genco RJ, Borgnakke WS. **Diabetes as a potential risk for periodontitis: Association studies**. *Periodontol 2000* (2020) **83**. DOI: 10.1111/prd.12270
9. Bassetti MA, Bassetti RG, Sculean A, Salvi GE, Bornstein MM, Ramseier CA. **The impact of brief interventions for tobacco cessation on patients' awareness of cigarette smoking as a risk factor for chronic periodontitis**. *Oral Health Prev Dent* (2017) **15**. DOI: 10.3290/j.ohpd.a38737
10. Khan S, Barrington G, Bettiol S, Barnett T, Crocombe L. **Is overweight/obesity a risk factor for periodontitis in young adults and adolescents?: A systematic review**. *Obes Rev* (2018) **19**. DOI: 10.1111/obr.12668
11. Wang CJ, McCauley LK. **Osteoporosis and periodontitis**. *Curr Osteoporos Rep* (2016) **14**. DOI: 10.1007/s11914-016-0330-3
12. Genco RJ, Borgnakke WS. **Risk factors for periodontal disease**. *Periodontol 2000* (2013) **62** 59-94. DOI: 10.1111/j.1600-0757.2012.00457.x
13. Chen ZY, Ye LW, Zhao L, Liang ZJ, Yu T, Gao J. **Hyperuricemia as a potential plausible risk factor for periodontitis**. *Med Hypotheses* (2020) **137** 109591. DOI: 10.1016/j.mehy.2020.109591
14. Aversa A, Morgentaler A. **The practical management of testosterone deficiency in men**. *Nat Rev Urol* (2015) **12**. DOI: 10.1038/nrurol.2015.238
15. Yeap BB. **Testosterone and cardiovascular disease risk**. *Curr Opin Endocrinol Diabetes Obes* (2015) **22** 193-202. DOI: 10.1097/MED.0000000000000161
16. Carr MC. **The emergence of the metabolic syndrome with menopause**. *J Clin Endocrinol Metab* (2003) **88**. DOI: 10.1210/jc.2003-030242
17. Mauvais-Jarvis F. **Estrogen and androgen receptors: regulators of fuel homeostasis and emerging targets for diabetes and obesity**. *Trends Endocrinol Metab* (2011) **22** 24-33. DOI: 10.1016/j.tem.2010.10.002
18. Kelly DM, Jones TH. **Testosterone: a metabolic hormone in health and disease**. *J Endocrinol* (2013) **217**. DOI: 10.1530/JOE-12-0455
19. Gilliver SC. **Sex steroids as inflammatory regulators**. *J Steroid Biochem Mol Biol* (2010) **120**. DOI: 10.1016/j.jsbmb.2009.12.015
20. Clarke BL, Khosla S. **Androgens and bone**. *Steroids.* (2009) **74** 296-305. DOI: 10.1016/j.steroids.2008.10.003
21. Michael H, Härkönen PL, Väänänen HK, Hentunen TA. **Estrogen and testosterone use different cellular pathways to inhibit osteoclastogenesis and bone resorption**. *J Bone Miner Res* (2005) **20**. DOI: 10.1359/JBMR.050803
22. Steffens JP, Herrera BS, Coimbra LS, Stephens DN, Rossa C, Spolidorio LC. **Testosterone regulates bone response to inflammation**. *Horm Metab Res* (2014) **46** 193-200. DOI: 10.1055/s-0034-1367031
23. Yeap BB. **Are declining testosterone levels a major risk factor for ill-health in aging men**. *Int J Impot Res* (2009) **21** 24-36. DOI: 10.1038/ijir.2008.60
24. Harman SM, Metter EJ, Tobin JD, Pearson J, Blackman MR. **Baltimore Longitudinal study of aging. longitudinal effects of aging on serum total and free testosterone levels in healthy men. Baltimore longitudinal study of aging**. *J Clin Endocrinol Metab* (2001) **86**. DOI: 10.1210/jcem.86.2.7219
25. Sipilä S, Narici M, Kjaer M, Pöllänen E, Atkinson RA, Hansen M. **Sex hormones and skeletal muscle weakness**. *Biogerontology* (2013) **14**. DOI: 10.1007/s10522-013-9425-8
26. Holtfreter B, Albandar JM, Dietrich T, Dye BA, Eaton KA, Eke PI. **Standards for reporting chronic periodontitis prevalence and severity in epidemiologic studies: Proposed standards from the joint EU/USA periodontal epidemiology working group**. *J Clin Periodontol* (2015) **42**. DOI: 10.1111/jcpe.12392
27. Eke PI, Dye BA, Wei L, Thornton-Evans GO, Genco RJ. **CDC Periodontal disease surveillance workgroup: James beck (University of north Carolina, chapel hill, USA), Gordon Douglass (Past President, American academy of periodontology), Roy page (University of washin. prevalence of periodontitis in adults in the united states: 2009 and 2010)**. *J Dent Res* (2012) **91**. DOI: 10.1177/0022034512457373
28. Eke PI, Dye BA, Wei L, Slade GD, Thornton-Evans GO, Borgnakke WS. **Update on prevalence of periodontitis in adults in the united states: NHANES 2009 to 2012**. *J Periodontol* (2015) **86**. DOI: 10.1902/jop.2015.140520
29. Tang Y, Peng B, Liu J, Liu Z, Xia Y, Geng B. **Systemic immune-inflammation index and bone mineral density in postmenopausal women: A cross-sectional study of the national health and nutrition examination survey (NHANES) 2007-2018**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.975400
30. Luo K, Liu J, Wang Y, Aimuzi R, Luo F, Ao J. **Associations between organophosphate esters and sex hormones among 6-19-year old children and adolescents in NHANES 2013-2014**. *Environ Int* (2020) **136** 105461. DOI: 10.1016/j.envint.2020.105461
31. Wilke TJ, Utley DJ. **Total testosterone, free-androgen index, calculated free testosterone, and free testosterone by analog RIA compared in hirsute women and in otherwise-normal women with altered binding of sex-hormone-binding globulin**. *Clin Chem* (1987) **33**. DOI: 10.1093/clinchem/33.8.1372
32. Perel E, Wilkins D, Killinger DW. **The conversion of androstenedione to estrone, estradiol, and testosterone in breast tissue**. *J Steroid Biochem* (1980) **13** 89-94. DOI: 10.1016/0022-4731(80)90117-X
33. Vermeulen A, Verdonck L, Kaufman JM. **A critical evaluation of simple methods for the estimation of free testosterone in serum**. *J Clin Endocrinol Metab* (1999) **84**. DOI: 10.1210/jcem.84.10.6079
34. ALHarthi SSY, Natto ZS, Midle JB, Gyurko R, O'Neill R, Steffensen B. **Association between time since quitting smoking and periodontitis in former smokers in the national health and nutrition examination surveys (NHANES) 2009 to 2012**. *J Periodontol* (2019) **90** 16-25. DOI: 10.1002/JPER.18-0183
35. Orwoll ES, Chan BK, Lambert LC, Marshall LM, Lewis C, Phipps KR. **Sex steroids, periodontal health, and tooth loss in older men**. *J Dent Res* (2009) **88**. DOI: 10.1177/0022034509341013
36. Samietz S, Holtfreter B, Friedrich N, Mundt T, Hoffmann W, Völzke H. **Prospective association of sex steroid concentrations with periodontal progression and incident tooth loss**. *J Clin Periodontol* (2016) **43**. DOI: 10.1111/jcpe.12493
37. Steffens JP, Wang X, Starr JR, Spolidorio LC, Van Dyke TE, Kantarci A. **Associations between sex hormone levels and periodontitis in men: Results from NHANES III**. *J Periodontol* (2015) **86**. DOI: 10.1902/jop.2015.140530
38. Johnson CL, Paulose-Ram R, Ogden CL, Carroll MD, Kruszon-Moran D, Dohrmann SM. **National health and nutrition examination survey: Analytic guidelines, 1999-2010**. *Vital Health Stat 2* (2013) **161)** 1-24
39. Singh BP, Makker A, Tripathi A, Singh MM, Gupta V. **Association of testosterone and bone mineral density with tooth loss in men with chronic periodontitis**. *J Oral Sci* (2011) **53**. DOI: 10.2334/josnusd.53.333
40. Famili P, Cauley JA, Greenspan SL. **The effect of androgen deprivation therapy on periodontal disease in men with prostate cancer**. *J Urol* (2007) **177**. DOI: 10.1016/j.juro.2006.10.067
41. Taubman MA, Valverde P, Han X, Kawai T. **Immune response: the key to bone resorption in periodontal disease**. *J Periodontol* (2005) **76**. DOI: 10.1902/jop.2005.76.11-S.2033
42. Wactawski-Wende J, Hausmann E, Hovey K, Trevisan M, Grossi S, Genco RJ. **The association between osteoporosis and alveolar crestal height in postmenopausal women**. *J Periodontol* (2005) **76 Suppl 11S**. DOI: 10.1902/jop.2005.76.11-S.2116
43. Coletta RD, Reynolds MA, Martelli-Junior H, Graner E, Almeida OP, Sauk JJ. **Testosterone stimulates proliferation and inhibits interleukin-6 production of normal and hereditary gingival fibromatosis fibroblasts**. *Oral Microbiol Immunol* (2002) **17**. DOI: 10.1034/j.1399-302X.2002.170309.x
44. Kuo LC, Polson AM, Kang T. **Associations between periodontal diseases and systemic diseases: A review of the inter-relationships and interactions with diabetes, respiratory diseases, cardiovascular diseases and osteoporosis**. *Public Health* (2008) **122**. DOI: 10.1016/j.puhe.2007.07.004
45. Malkin CJ, Pugh PJ, Jones RD, Kapoor D, Channer KS, Jones TH. **The effect of testosterone replacement on endogenous inflammatory cytokines and lipid profiles in hypogonadal men**. *J Clin Endocrinol Metab* (2004) **89**. DOI: 10.1210/jc.2003-031069
46. Dennison DK, Van Dyke TE. **The acute inflammatory response and the role of phagocytic cells in periodontal health and disease**. *Periodontol* (2000) **1997**. DOI: 10.1111/j.1600-0757.1997.tb00192.x
47. Birkedal-Hansen H. **Role of cytokines and inflammatory mediators in tissue destruction**. *J Periodontal Res* (1993) **28**. DOI: 10.1111/j.1600-0765.1993.tb02113.x
48. Pelegrin ÁF, de Paiva Gonçalves V, Carvalho JS, Spolidorio DMP, Spolidorio LC. **Testosterone replacement relieves ligature-induced periodontitis by mitigating inflammation, increasing pro-resolving markers and promoting angiogenesis in rats: A preclinical study**. *Arch Oral Biol* (2023) **146** 105605. DOI: 10.1016/j.archoralbio.2022.105605
|
---
title: Effect of mango seeds as an untraditional source of energy on the productive
performance of dairy Damascus goats
authors:
- Heba A. El-Sanafawy
- Aristide Maggiolino
- Ghada S. El-Esawy
- Wasef A. Riad
- Mohamed Zeineldin
- Mohamed Abdelmegeid
- Rabiha Seboussi
- Lamiaa I. EL-Nawasany
- Mona M. M. Y. Elghandour
- Pasquale De Palo
- Abdelfattah Z. M. Salem
journal: Frontiers in Veterinary Science
year: 2023
pmcid: PMC9971558
doi: 10.3389/fvets.2023.1058915
license: CC BY 4.0
---
# Effect of mango seeds as an untraditional source of energy on the productive performance of dairy Damascus goats
## Abstract
Eighteen dairy Damascus goats weighing 38–45 kg live body weight and aged 3–4 years were divided into three groups according to their body weight, with six goats in each group. Yellow corn grain in their concentrate feed mixture was replaced with mango seeds (MS) at levels of $0\%$ MS in group 1 (G1, control), $20\%$ MS in group 2 (G2), and $40\%$ MS in group 3 (G3). The digestibility coefficients of the organic matter, dry matter, crude fiber, crude protein, ether extract, nitrogen-free extract, and total digestible nutrients increased ($P \leq 0.05$) upon feeding MS to G2 and G3. The amounts of dry matter, total digestible nutrients, and digestible crude protein required per 1 kg $3.5\%$ fat-corrected milk (FCM) were lower ($P \leq 0.05$) in G2 and G3 vs. G1. Actual milk and $3.5\%$ FCM yield increased ($P \leq 0.05$) with the increasing MS dietary level. G2 and G3 had the highest significant ($P \leq 0.05$) total solids, total protein, non-protein nitrogen, casein, ash, fat, solids not fat, lactose, and calcium contents compared with G1. Replacing yellow corn grain with MS in G2 and G3 significantly ($P \leq 0.05$) decreased the cholesterol concentration and AST activity. Feeding MS increased the concentrations of caprioc, caprylic, capric, stearic, oleic, elaidic, and linoleic acids and decreased the concentrations of butyric, laueic, tridecanoic, myristic, myristoleic, pentadecanoic, heptadecanoic, cis-10-Heptadecanoic, cis-11-eicosenoic, linolenic, arachidonic, and lignoseric acids in the milk fat. The results show that the replacement of corn grain with MS improved the digestibility, milk yield, feed conversion, and economic efficiency, with no adverse effects on the performance of Damascus goats.
## Introduction
Mango residue, which accounts for 40–$50\%$ of the total fruit weight, has potential nutritional benefits in addition to its significant levels of essential minerals [1, 2]. Recently, starch from fruit seeds, including mango, was found to possess good physicochemical properties in addition to a balanced amino acid concentration in the kernel [3]. Mango seeds kernel (MSK), pulp, and peel contents are usually considered important bioactive materials because they contain antioxidant, anti-bacterial, and antiviral immune modulatory substances as well as lipids, and have a high protein content (4–6). MSK is considered a good source of carbohydrates ($77\%$), fat ($11\%$), proteins (6–$7\%$), crude fiber ($2\%$), and ash ($2\%$) on a dry matter basis [7]. MSK contains 210 mg/100 g magnesium, 170 mg/100 g calcium, and 368 mg/100 g potassium [8] as well as vitamin E (1.30 mg/100 g db), provitamin A (15.27 IU), vitamin C (0.56 mg/100 g db), and vitamin K (0.59 mg/100 g db) [9, 10].
Mango seeds kernel, as a nonconventional concentrate mixture, can greatly lower feed costs and can be safely integrated into lambs' rations without compromising the nutrient utilization, growth rate, feed intake, or blood profile [11, 12]. The nutrient digestibility and consumption of sheep fed rice straw supplemented with mango kernel meal suggest that mango kernel meal can be used as an unusual feedstuff supplement during the lengthy dry season [13]. Mango seeds kernel can also be used to replace up to half of the yellow corn in sheep rations without affecting feed intake, water metabolism, digestion coefficients, nitrogen balance, or ruminal fermentation [14]. Replacement of corn grain with $30\%$ MS in the diets of Damascus goat bucks showed positive effects on the productive performance and increased semen production [15]. An economic evaluation of using whole mango meal in the diets of dairy goats, instead of corn, showed that the total feeding cost was reduced and the benefit-to-cost ratio improved [16]. Furthermore, adding MS to mixed diets with alfalfa hay increased the diet's fermentation rate [17]. Therefore, the objective of the present study was to investigate the effects of partial replacement of yellow corn grain with MS as a nontraditional source of energy on the productive performance of Damascus goats.
## Materials and methods
The present study was carried out at the Sakha Animal Production Research Station, which is part of the Animal Production Research Institute, Agricultural Research Center, Cairo, Egypt.
## Experimental animals
Eighteen dairy Damascus goats, weighing 38–45 kg live body weight and aged 3–4 years, were divided into three similar groups according to their body weight, with six goats in each group.
## Experimental diets
Three experimental diets were used to feed the goats during the experimental period (240 days). All diets contained a concentrate feed mixture [$50\%$ concentrate feed mixture (CFM)], fresh berseem ($40\%$), and wheat straw ($10\%$). Mango seeds were added instead of yellow corn grain in the CFM, at concentrations of $0\%$ MS (G1, control), $20\%$ (G2), and $40\%$ (G3). The ingredients of the different CFMs are shown in Table 1.
**Table 1**
| Ingredients | CFM1 | CFM2 | CFM3 |
| --- | --- | --- | --- |
| Yellow corn grains | 30 | 24 | 18 |
| Mango seeds | 0 | 6 | 12 |
| Wheat bran | 40 | 40 | 40 |
| Undecorticated cottonseed meal | 24 | 24 | 24 |
| Molasses | 3 | 3 | 3 |
| Limestone | 2 | 2 | 2 |
| Common salt | 1 | 1 | 1 |
## Mango seeds preparation
Mango seeds consist of approximately $68\%$ kernel, $29\%$ shell, and $3\%$ test [18]. The mango seeds were soaked in water for 3 days to decrease the anti-nutritional factors, air-sundried for 48 h, and then crushed in a forage machine.
## Chemical analyses
Chemical analyses of representative samples of the experimental ingredients, CFM, and diets were conducted according to [19] for OM (ID number 942.05), EE (using the Soxhlet procedure, ID number 938.06), and CP (as 6.25 × N; ID number 954.01). Neutral detergent fibers (NDF), acid detergent lignin (ADL), and acid detergent fiber (ADF) were evaluated as described previously [20]. Cellulose and hemicelluloses were also calculated. The content was calculated as follows: NFE [g/kg dry matter (DM)] = 100 – (CP + ash + EE + CF). All chemical analyses were carried out in duplicate.
## Experimental procedures
Animals in all groups were fed by group feeding, and the tested diets were offered in equal amounts for all groups twice daily at 8 a.m. and 3 p.m., while fresh water was always available. The offered amounts of CFM and roughage were adjusted every 2 weeks according to changes in the animals' body weight. Diets were formulated to meet the requirements of growing goats according to [21].
## Digestibility trials
During the feeding period, three digestibility trials were conducted on three lambs from each group to assess the feeding values and digestibility of the experimental meals. Each digestibility study had a 15-day preparatory phase and a 7-day collection period. As a natural marker, acid-insoluble ash was used [22]. During the collection period, feces samples were taken twice daily from the rectum of each animal at 12-h intervals. At the start, middle, and end of the collection period, feedstuff samples were taken. Chemical analyses of the CFM, corn silage, rice straw, and feces samples were carried out using AOAC procedures [2005]. The equation proposed by Schneider was used to calculate the nutrient digestibility as follows: where AIA is acid-insoluble ash. Digestible crude protein (DCP) and total digestible nutrients (TDN) were assessed according to the formula in [23].
## Rumen liquor samples
Rumen liquor samples were collected from goats 3 h after morning feeding using a stomach tube and the draw pulse power of an automatic milking machine. Rumen samples were strained using four layers of cheesecloth. The pH of the rumen was measured using an Orian 680 digital PH meter immediately after the samples were extracted. The amount of ammonia nitrogen (NH3-N) in the air was measured using the AOAC saturated magnesium oxide distillation method [2005]. Total volatile fatty acids in rumen liquor were measured using the steam distillation method described previously [24].
## Blood samples and biochemical analysis
At 3 h after feeding, blood samples were collected from the jugular vein in centrifuge tubes containing EDTA as an anticoagulant. The plasma was then centrifuged for 15 min at 4,000 rotations/min and stored in a deep freezer until analysis. Calorimetric determinations of blood plasma protein, albumin, globulin (by difference), urea nitrogen, AST, and ALT were performed using commercial diagnostic kits.
## Milk yield and composition
Milk production was recorded weekly using the manual milking technique. The udder was stripped completely, and then the total milk production was calculated by collecting milk during the entire experiment period and correcting for $3.5\%$ fat according to [25] as follows: FCM $3.5\%$ fat as FCM ($3.5\%$) = 0.35 M + 18.57 F, where F is the amount of fat in kilograms, and M is the quantity of milk in kilograms. Milk samples were analyzed for total solids, protein, NPN, casein, ash, and fat. The solid-not-fat portion was also determined using the formula reported previously [26].
The carbohydrate content of the milk was calculated by the difference. The Ca and P contents were determined using atomic absorption, according to the method described previously [27]. The milk fatty acid composition was determined according to [28]. The pH values of the sample were measured using a digital pH meter.
The actual milk and $3.5\%$ FCM yields increased ($P \leq 0.05$) with increasing MS concentration. G3 ($40\%$ MS) exhibited the highest actual milk and $3.5\%$ FCM yields (1.880 and 1.204 kg/day, respectively), followed by G2 (1.650 and 1.159 kg/day, respectively), with G1 presenting the lowest values (1.514 and 1.022 kg/day, respectively). Concerning the milk composition, G2 had significant ($P \leq 0.05$) TS, TP, NPN, casein, ash, fat, SNF, lactose, and Ca contents as well as fat yield G3, while G1 had the lowest (Table 8). The P content and pH values were nearly similar in all groups. Similarly, it was found [1, 44] that MSK contained considerable levels of calcium, phosphorus, sodium, magnesium, iron, zinc, and copper.
**Table 8**
| Item | Experimental diets 1 | Experimental diets 1.1 | Experimental diets 1.2 | SEM |
| --- | --- | --- | --- | --- |
| | G1 (0% MS) | G2 (20% MS) | G3 (40% MS) | |
| Milk yield (kg/day) 2 | Milk yield (kg/day) 2 | Milk yield (kg/day) 2 | Milk yield (kg/day) 2 | Milk yield (kg/day) 2 |
| Actual milk | 1.514c | 1.650b | 1.880a | 0.093 |
| 3.5% FCM | 1.022c | 1.159b | 1.204a | 0.072 |
| Milk composition 2 | Milk composition 2 | Milk composition 2 | Milk composition 2 | Milk composition 2 |
| TS % | 10.43c | 11.35a | 10.92b | 0.02 |
| TP % | 2.91b | 2.99a | 2.96ab | 0.025 |
| NPN % | 0.35b | 0.38a | 0.36b | 0.009 |
| Casein % | 2.01b | 2.06a | 2.03ab | 0.012 |
| Ash % | 0.73b | 0.82a | 0.79a | 0.014 |
| Fat % | 2.65c | 3.13a | 2.94b | 0.06 |
| SNF % | 7.77c | 8.19a | 7.96b | 0.02 |
| Lactose % | 4.14c | 4.41a | 4.25b | 0.01 |
| Ca (mg/100g) | 123.67b | 133.00a | 131.33a | 1.12 |
| P (mg/100g) | 116.00 | 117.67 | 116.33 | 1.27 |
| Fat yield (g/day) | 25.76c | 27.77a | 26.73b | 0.18 |
| pH | 6.53 | 6.54 | 6.54 | 0.007 |
## Feed conversion
The feed conversion efficiency in terms of DM, TDN, and DCP required for 1 kg $3.5\%$ FCM yield was calculated for every animal.
## Economic efficiency
The cost of feed, the feed cost/kg $4\%$ FCM, and the price of $4\%$ FCM were calculated for every goat according to 2020 prices. Additionally, the economic efficiency expressed as the ratio of the price of the $4\%$ FCM yield and the feed cost was estimated.
The feeding cost was the same for the different groups, whereas the feed cost per 1 kg $3.5\%$ FCM decreased ($P \leq 0.05$) for G2 and G3 when compared with G1 (Table 6). However, the prices of the $3.5\%$ FCM yield, net revenue, relative net revenue, economic efficiency, and relative economic efficiency were higher ($P \leq 0.05$) for G2 and G3 compared with G1. The best economic efficiency occurred with the $40\%$ MS replacement of corn compared with all treatments. This result was obtained because MS is a mango, and it is handled at a cheaper rate than yellow corn, in addition to the decreased feed consumption, which may be due to the bitter taste of MS. The relative economic efficiency of 20 and $40\%$ MS replacement of corn increased by approximately 12.70 and $17.46\%$, respectively, compared with that of the control group. Similarly, it was reported [39] that the feed cost per kilogram weight gain of broilers was lowest in the group fed $10\%$ rice polish as a replacement for corn. Additionally, it was reported [15] that the daily feed cost of growing lambs was nearly similar in all lambs, but the feed cost per kilogram weight gain decreased ($P \leq 0.05$) with increasing the level of MS. However, the price of daily gain, economic feed efficiency (EFE), and relative EFE increased ($P \leq 0.05$) with increasing MS level. In line with the present results, the economic analysis revealed a decrease in the total feeding costs and a superior benefit-to-cost ratio [16].
## Statistical analysis
Data analyses were performed using SPSS version 23.0 software (IBM, New York, NY, USA). The data were statistically analyzed using the general linear model procedure adopted by [29] for one-way ANOVA. A Duncan test was also performed to evaluate the degree of significance among the means. The significance level for all analyses was set at 0.05.
## Results and discussion
The organic matter (OM) content of mango seeds (MS) was found to be approximately similar to that of yellow corn. However, MS was superior in terms of the ether extract (EE), crude fiber (CF), and ash content in comparison with yellow corn. The ash content in yellow corn ($1.30\%$) was slightly lower than the ash content in MS ($1.75\%$), while the nitrogen free extract (NFE) and crude protein (CP) in MS were lower than those in yellow corn (Table 2). Ash, CF, and CP contents in the present study differed from the results found previously [30], where $4.73\%$ ash, $3.98\%$ CF, and $8.98\%$ CP in boiled mango seeds kernel were reported. However, the ash content was $1.40\%$ and $2.38\%$ for yellow corn and MS, respectively [14]. The chemical composition of MS in this study differed from that reported by numerous researchers, possibly because those authors used MSK. The present findings are consistent with those previously described (7, 31–33), who reported that MSK is a rich source of carbohydrates (67–$82\%$) and contains moderate amounts of protein (6–$10\%$) and fat (7–$14\%$). However, the present results were inconsistent with the results found by [34]. In line with the present study, it was reported that MSK could be a powerful energy source and would probably be a good replacement for maize [33], which is considered the main energy source in the diet of non-ruminants.
**Table 2**
| Item | DM % | Chemical analysis (%, on DM basis) a | Chemical analysis (%, on DM basis) a.1 | Chemical analysis (%, on DM basis) a.2 | Chemical analysis (%, on DM basis) a.3 | Chemical analysis (%, on DM basis) a.4 | Chemical analysis (%, on DM basis) a.5 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | OM | CP | EE | CF | NFE | Ash |
| CFM ingredients and feedstuffs | CFM ingredients and feedstuffs | CFM ingredients and feedstuffs | CFM ingredients and feedstuffs | CFM ingredients and feedstuffs | CFM ingredients and feedstuffs | CFM ingredients and feedstuffs | CFM ingredients and feedstuffs |
| Yellow corn grains | 91.58 | 98.70 | 9.29 | 1.7 | 2.1 | 85.7 | 1.3 |
| Wheat brane | 90.50 | 93.80 | 14.20 | 2.7 | 11.8 | 65.10 | 6.2 |
| Undecorticated cotton-seed meal | 90.98 | 93.30 | 26.40 | 5.64 | 24.26 | 37.00 | 6.7 |
| Mango seeds | 89.04 | 97.15 | 8.93 | 3.5 | 21.93 | 62.79 | 2.85 |
| Soaking mango seeds | 88.33 | 98.21 | 7.13 | 4.81 | 22.21 | 64.06 | 1.79 |
| Fresh Berseem | 16.31 | 88.85 | 16.62 | 1.21 | 26.30 | 44.12 | 11.15 |
| Wheat Straw | 91.75 | 87.83 | 2.96 | 1.59 | 25.8 | 57.48 | 12.17 |
| Experimental CFM | Experimental CFM | Experimental CFM | Experimental CFM | Experimental CFM | Experimental CFM | Experimental CFM | Experimental CFM |
| CFM1 (0% MS) | 91.23 | 89.52 | 14.78 | 2.94 | 11.17 | 60.63 | 10.48 |
| CFM2 (20% MS) | 91.08 | 89.49 | 14.66 | 3.13 | 12.37 | 59.33 | 10.51 |
| CFM3 (40% MS) | 90.95 | 89.47 | 14.53 | 3.30 | 13.59 | 58.04 | 10.53 |
| Experimental diets b | Experimental diets b | Experimental diets b | Experimental diets b | Experimental diets b | Experimental diets b | Experimental diets b | Experimental diets b |
| G1 (Control, 0% MS) | 32.17 | 89.07 | 14.34 | 2.11 | 18.68 | 55.51 | 10.92 |
| G2 (20% MS) | 32.16 | 89.07 | 14.26 | 2.21 | 19.29 | 53.08 | 10.94 |
| G3 (40% MS) | 32.14 | 88.99 | 14.18 | 2.28 | 19.93 | 52.41 | 10.95 |
Similar to previous reports, the highest concentration of CF was for CFM3 ($40\%$ MS), while the lowest concentration was for CFM1 ($0\%$ MS), which may be due to the high CF in MS compared with corn gain [14]. The present study also found that replacing corn with MS led to an increase in EE (2.94 to 3.30 g/100 g DM) and CF (11.17 to 13.59 g/100 g DM) in the tested feed mixtures, whereas the nitrogen-free extract content decreased from 60.63 to 58.04 g/100 g DM. These results show that MS is a good source of carbohydrates. The OM, CP, EE, and ash contents were nearly similar for the different diets, whereas CF tended to increase and NFE decrease with increasing MS levels. In line with other research, MS was shown to have greater concentrations of growth energy and CF, because the kernel is usually a significant source of fat, starch, and protein [7, 14]. Similarly, it was reported that mango kernels included a higher percentage of fat, starch, and protein [34, 35].
## Fiber fractions
The fiber fractions of the feedstuffs and experimental diets are shown in Table 3. These results show that CFM3 had the highest all-fiber fractions compared with CFM1, except for hemicellulose and non-structural carbohydrates, because MS is superior in terms of CF, NDF, ADF, ADL, and cellulose contents compared with yellow corn. The present study showed variations between the different tested diets in terms of the CF, NFE, NDF, ADF, ADL, CEL, hemicellulose, and non-fiber carbohydrate contents. The present study results suggest that NDF is positively correlated with lignin but negatively correlated with non-fiber carbohydrate. These results are in agreement with those reported previously [36].
**Table 3**
| Item | NDF | ADF | ADL | Hemi cellulose | Cellulose | NSC |
| --- | --- | --- | --- | --- | --- | --- |
| Feedstuffs | Feedstuffs | Feedstuffs | Feedstuffs | Feedstuffs | Feedstuffs | |
| CFM1 | 34.76 | 12.92 | 4.47 | 21.84 | 8.45 | 39.04 |
| CFM2 | 35.85 | 16.30 | 5.11 | 19.25 | 11.19 | 37.85 |
| CFM3 | 37.29 | 19.38 | 7.35 | 17.91 | 12.03 | 36.35 |
| Fresh berseem | 50.51 | 39.23 | 8.26 | 11.34 | 30.97 | 20.45 |
| Wheat straw | 56.98 | 38.56 | 4.15 | 18.42 | 34.41 | 26.3 |
| Experimental diets a | Experimental diets a | Experimental diets a | Experimental diets a | Experimental diets a | Experimental diets a | |
| G1 (Control, 0% MS) | 43.31 | 26.01 | 5.96 | 17.30 | 20.06 | 30.33 |
| G2 (20% MS) | 43.86 | 27.7 | 6.28 | 16.10 | 21.42 | 29.74 |
| G3 (40% MS) | 44.58 | 29.24 | 7.39 | 15.34 | 21.85 | 28.97 |
## Digestibility coefficients and feeding values
The digestibility coefficients of all nutrients (DM, OM, CP, CF, EE, and NFE) and TDN increased ($P \leq 0.05$) in the diets that included CFMs containing MS in G2 and G3, with the highest observed in G3. The DCP content was nearly similar for the different groups and not significantly affected by feeding of MS (Table 4). This result may imply that a $40\%$ concentration of MS supplementation would support rumen microbial activity fermentation, which, in turn, facilitates improved digestibility. Similarly, it was revealed [37] that a specific minimum daily dry matter intake is required to fulfill an animal's hunger and to allow the digestive tract to function properly. This indicates a healthy source of roughage that could improve rumination and prevent rumen digestive disorders [38]. Similarly, it was found [15] that the digestibility coefficients of DM, CP, EE, and CF increased ($P \leq 0.05$) as the level of MS increased. The digestibility of OM and NFE, and feeding values, as well as the TDN and DCP were higher ($P \leq 0.05$) in groups fed MS, at $20\%$ in G2 and $40\%$ in G3, compared with G1. Similarly, it was reported [13] that the nutrient digestibility of growing West African dwarf sheep fed a diet containing $75\%$ mango kernel meal was the best in terms of CF, EE, and NFE and differed significantly ($P \leq 0.05$) from those containing 0, 50, and $100\%$ mango kernel meal. The addition of $25\%$ or $50\%$ yellow corn instead of MSK considerably increased the nutritional digestibility coefficients of DM, OM, CP, CF, and NFE ($P \leq 0.05$) [14]. When MSK was substituted with $25\%$ or $50\%$ yellow corn content in the control ration, the TDN and DCP values improved ($P \leq 0.005$).
**Table 4**
| Item | Experimental diets 1 | Experimental diets 1.1 | Experimental diets 1.2 | SEM |
| --- | --- | --- | --- | --- |
| | G1 (0%MS) | G2 (20% MS) | G3 (40% MS) | |
| Digestibility coefficients (%) | Digestibility coefficients (%) | Digestibility coefficients (%) | Digestibility coefficients (%) | Digestibility coefficients (%) |
| Dry matter | 68.83b | 71.39a | 72.82a | 0.63 |
| Organic matter | 70.03c | 72.03b | 73.51a | 0.55 |
| Crude protein | 71.17b | 73.08a | 73.97a | 0.49 |
| Ether extract | 67.41c | 70.83b | 73.23a | 0.88 |
| Crude fiber | 65.69b | 69.32a | 70.51a | 0.75 |
| Nitrogen free extract | 71.26b | 72.78ab | 74.57a | 0.61 |
| Feeding value (%) | Feeding value (%) | Feeding value (%) | Feeding value (%) | Feeding value (%) |
| Total digestible nutrients | 63.47b | 65.45a | 66.92a | 0.55 |
| Digestible crude protein | 10.28 | 10.44 | 10.52 | 0.05 |
## Feed intake and feed conversion
The average daily feed intake of goats fed different experimental rations is presented in Table 5. Because the goats were group fed, comparisons regarding the feed intakes of all nutrients are made on a relative basis rather than on a statistical basis. The CFM, fresh berseem, wheat straw, DM, TDN, and DCP intakes tended to increase for goats fed CFMs containing MS and tended to increase as the MS concentration increased from 20 to $40\%$. These results agreed with previous findings [15], where the average daily feed intake from feedstuffs (CFM, fresh berseem, and wheat straw) or DM, TDN, and DCP was slightly higher in treatment groups containing MS than in the control group. Similarly, significant differences ($P \leq 0.05$) were observed in the digestible nutrient intake in growing West African dwarf sheep fed different levels of mango kernel meal-based diets [13].
**Table 5**
| Item | Experimental diets 1 | Experimental diets 1.1 | Experimental diets 1.2 | SEM |
| --- | --- | --- | --- | --- |
| | G1 (0% MS) | G2 (20% MS) | G3 (40% MS) | |
| As fed basis (kg/day) | As fed basis (kg/day) | As fed basis (kg/day) | As fed basis (kg/day) | As fed basis (kg/day) |
| CFM (kg) | 0.702 | 0.722 | 0.737 | |
| Fresh berseem (kg) | 3.140 | 3.227 | 3.286 | |
| Wheat straw (kg) | 0.140 | 0.143 | 0.146 | |
| Total | 3.982 | 4.092 | 4.169 | |
| On DM basis (kg/day) 2 | On DM basis (kg/day) 2 | On DM basis (kg/day) 2 | On DM basis (kg/day) 2 | On DM basis (kg/day) 2 |
| DM (kg) | 1.281 | 1.316 | 1.340 | 0.027 |
| TDN (kg) | 0.813c | 0.861b | 0.897a | 0.012 |
| DCP (kg) | 0.132 | 0.137 | 0.141 | 0.002 |
The results of feed conversion reveal that introducing MS in the CFMs of dairy goats improved feed conversion, and the amounts of DM, TDN, and DCP required per kilogram of $3.5\%$ FCM were lower ($P \leq 0.05$) in G2 and G3 when compared with G1 (Table 6). These results may be attributed to the improved FCM yield upon feeding with MS. Similar results were previously obtained [15], where the feed conversion in growing lambs improved in lambs fed CFMs containing MS.
**Table 6**
| Item | Experimental diets 1 | Experimental diets 1.1 | Experimental diets 1.2 | SEM |
| --- | --- | --- | --- | --- |
| | G1 (0% MS) | G2 (20% MS) | G3 (40% MS) | |
| Feed conversion (kg/kg 3.5% FCM) | Feed conversion (kg/kg 3.5% FCM) | Feed conversion (kg/kg 3.5% FCM) | Feed conversion (kg/kg 3.5% FCM) | Feed conversion (kg/kg 3.5% FCM) |
| Dry matter | 1.253a | 1.135b | 1.113b | 0.23 |
| Total digestible nutrients | 0.795a | 0.743b | 0.745b | 0.16 |
| Digestible crude protein | 0.129a | 0.118b | 0.117b | 0.045 |
| Economic efficiency | Economic efficiency | Economic efficiency | Economic efficiency | Economic efficiency |
| Feed cost (L.E./day) | 4.85 | 4.89 | 4.88 | 0.18 |
| Feed cost (L.E./kg 3.5% FCM) | 4.75a | 4.22b | 4.05b | 0.21 |
| Price of 3.5% FCM (L.E./day) | 6.13b | 6.95a | 7.22a | 0.26 |
| Net revenue (LE) | 1.28b | 2.06a | 2.34a | 0.12 |
| Relative net revenue (%) | 100.00b | 160.94a | 182.81a | 2.32 |
| Economic efficiency | 1.26b | 1.42a | 1.48a | 0.03 |
| Relative economic efficiency % | 100.00 | 112.70 | 117.46 | 1.17 |
## Rumen liquor parameters
The ruminal pH value and the total volatile fatty acids (TVFA) and NH3-N concentrations were not affected by incorporating MS into the diets (Table 7). Similarly, MSK replaced 25 or $50\%$ of the yellow maize in the control ratio which significantly increased the ruminal pH ($P \leq 0.05$); however, it had no significant ($P \leq 0.05$) effect on the NH3-N and total volatile fatty acid (TVFA) concentrations [14]. The high fermentation of mango waste likely stimulated microbial growth, resulting in high NH3-N capture by microorganisms [40]. The greater proportion of minor VFA is consistent with the higher NH3-N concentrations, as minor VFA and NH3-N are the final products of protein degradation in the rumen [41].
**Table 7**
| Item | Experimental diets 1 | Experimental diets 1.1 | Experimental diets 1.2 | SEM |
| --- | --- | --- | --- | --- |
| | G1 (0% MS) | G2 (20% MS) | G3 (40% MS) | |
| Rumen liquor parameters | Rumen liquor parameters | Rumen liquor parameters | Rumen liquor parameters | Rumen liquor parameters |
| pH value | 6.19 | 6.25 | 6.16 | 0.05 |
| TVFA's (Mm/100 ml) | 12.41 | 12.00 | 12.67 | 0.35 |
| NH3-N(mg/100 ml) | 7.84 | 7.84 | 7.65 | 0.57 |
| Blood plasma biochemical | Blood plasma biochemical | Blood plasma biochemical | Blood plasma biochemical | Blood plasma biochemical |
| Total protein (g/100 ml) | 7.08 | 6.92 | 7.15 | 0.11 |
| Albumin (g/100 ml) | 3.42 | 3.30 | 3.37 | 0.08 |
| Globulin (g/100 ml) | 3.66 | 3.62 | 3.78 | 0.12 |
| Albumin: Globulin ration | 0.93 | 0.91 | 0.89 | 0.04 |
| Cholesterol (mg/dl) | 77.00a | 65.17b | 63.72b | 1.89 |
| Creatinine (mg/100 ml) | 0.58 | 0.61 | 0.59 | 0.03 |
| AST (IU/l) | 75.17a | 67.33b | 68.83b | 2.23 |
| ALT (IU/l) | 11.83 | 11.67 | 12.33 | 0.29 |
| Urea-N (mg/100 ml) | 56.50 | 57.83 | 55.67 | 4.19 |
## Blood parameters
Only the cholesterol concentration and AST activity were decreased ($P \leq 0.05$) when yellow corn grain was replaced with MS in G2 and G3. Other plasma biochemical parameters, including total protein, albumin, globulin, albumin to globulin ratio, creatinine, ALT, and urea–N were not affected by MS (Table 7). This result agrees with that found previously [42], where $0.28\%$ dietary mango saponin supplementation decreased the plasma total cholesterol content in cockerels. This could be explained by the presence of mangiferin in the mango saponin. The hypocholesterolemia effect of MSK may be related to flavonoid components that may prevent lipid peroxidation, which regulates cholesterol synthesis. Serum creatinine kinase (CK) activity for Gimmizah cockerels decreased ($P \leq 0.05$) in the treated groups. At $10\%$ MSK substitution with corn, the lowest creatinine kinase (CK) concentration was reported. According to a previous study [43], elevated serum CK levels are linked to cell damage and muscle cell disintegration. The enhancement of muscle cells in that study, by replacing 10 and $15\%$ of the maize in the cockerels' diet with MSK, was linked to phenolic chemicals, which may reduce oxidation reactions in the cell.
## Milk fatty acid profile
Feeding the goats with MS increased the concentrations of caprioc, caprylic, capric, stearic, oleic, elaidic, and linoleic acids (Table 9). In contrast, feeding with MS decreased the concentrations of butyric, laueic, tridecanoic, myristic, myristoleic, pentadecanoic, heptadecanoic, cis-10-Heptadecanoic, cis-11-Eicosenoic, linolenic, arachidonic, and lignoseric acids in the milk fat (Table 9). The apparent change in the milk's stearic acid content when the goats were fed with MS may be due to the high percentage of that acid in MS, as explained previously [4]. In line with other reports, the assessment of the milk fatty acid profile of goats revealed no difference among treatments [16]. However, the level of myristoleic fatty acid C14:1 cis-9 decreased linearly by 0.15 g/100 g of fatty acids for every $1\%$ increase in whole mango meal levels ($P \leq 0.05$).
**Table 9**
| Fatty acids concentration, % | Experimental diets 1 | Experimental diets 1.1 | Experimental diets 1.2 | SEM |
| --- | --- | --- | --- | --- |
| | G1 (0% MS) | G2 (20% MS) | G3 (40% MS) | |
| Butyric acid (C4:0) | 0.29a | 0.12c | 0.21b | 0.07 |
| Caproic acid (C6:0) | 0.87c | 1.07b | 1.32a | 0.19 |
| Caprylic acid (C8:0) | 1.78c | 2.40b | 2.60a | 0.37 |
| Capric acid (C10:0) | 9.44a | 10.06b | 10.36a | 0.41 |
| Laueic acid (C12:0) | 5.47a | 4.52b | 3.92c | 0.68 |
| Tridecanoic acid (C13:0) | 0.19a | 0.06b | 0.05b | 0.07 |
| Myristic acid (C14:0) | 11.52a | 9.74c | 10.06b | 0.83 |
| Myristoleic acid methyl ester (14.1) | 0.39a | 0.29b | 0.20c | 0.83 |
| Pentadecanoic acid (C15:0) | 1.39a | 0.96b | 0.68b | 0.31 |
| Palmitic acid (C16:0) | 29.44a | 25.74c | 29.02b | 1.75 |
| Palmitoleic acid (C16:1n7) | 0.52a | 0.49b | 0.39c | 0.06 |
| Heptadecanoic acid (C17:0) | 0.57a | 0.50b | 0.41c | 0.07 |
| Cis-10-Heptadecanoic acid (C17:1) | 0.84a | 0.73b | 0.56c | 0.12 |
| Stearic acid (C18:0) | 10.02c | 13.69a | 11.73b | 1.59 |
| Oleic acid (C18:1n9c) | 21.10b | 23.55a | 20.72c | 1.33 |
| Elaidic acid (C18:1n9t) | 0.71c | 0.98b | 1.24a | 0.23 |
| Linoleic acid (C18:2n6c) | 2.43c | 2.89b | 3.25a | 0.36 |
| Linolelaidic acid (C18:2n6t) | 0.24b | 0.37a | 0.24b | 0.07 |
| Arachidic acid (20.0) | 0.20b | 0.22a | 0.15c | 0.03 |
| γ- Linolenic acid (C18:3n6) | —c | 0.07a | 0.04b | 0.02 |
| Cis-11- Eicosenoic acid(20.1) | 0.52a | 0.46b | 0.34c | 0.08 |
| Linolenic acid (C18:3n3) | 0.68a | 0.60c | 0.65b | 0.04 |
| Cis−11,14- Eicosadienoic acid (C 20.2) | —b | 0.06a | 0.06a | 0.01 |
| Arachidonic acid (C20: 4n6) | 0.23a | 0.19b | 0.19b | 0.02 |
| Eicosapentaenoic acid (C20:5n3) | —a | 0.04b | —a | 0.02 |
| Lignoseric acid (C24:0) | 1.18a | 0.12c | 0.98b | 0.49 |
| Docosahexaenoic acid (C22:6n3) | —a | 0.05b | —a | 0.02 |
## Conclusions
The results revealed that replacement of 20–$40\%$ of yellow corn grain with mango seeds resulted in no adverse effects on the performance of Damascus goats. In addition, the inclusion of mango seeds at a concentration of $40\%$ improved the digestibility, milk yield, feed conversion, economic efficiency, and milk composition.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding authors.
## Ethics statement
This study was conducted according to the research ethics approved by the committee on research of the Animal Production Research Institute, Agricultural Research Center. Written informed consent was obtained from the owners for the participation of their animals in this study.
## Author contributions
HE-S, GE-E, WR, MZ, and MA designed the experiment, carried out the research, and laboratory analysis. AM, AS, ME, LE-N, and PD did the data analysis, wrote the manuscript, and revised the manuscript. All authors reviewed and agreed on the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Fowomola M. **Some nutrients and antinutrients contents of mango (Magnifera indica) seed**. *Afr J Food Sci.* (2010) **4** 472-6
2. Akter R, Afrose A, Rahman MR, Chowdhury R, Nirzhor SSR, Khan RI. **A comprehensive analysis into the therapeutic application of natural products as SIRT6 modulators in Alzheimer's disease, aging, cancer, inflammation, and diabetes**. *Int J Mol Sci.* (2021) **22** 4180. DOI: 10.3390/ijms22084180
3. Hassan L, Muhammad A, Aliyu R, Idris Z, Izuagie T, Umar K. **Extraction and characterisation of starches from four varieties of Mangifera indica seeds**. *IOSR J Appl Chem.* (2013) **3** 16-23. DOI: 10.9790/5736-0361623
4. Torres-León C, Rojas R, Contreras-Esquivel JC, Serna-Cock L, Belmares-Cerda RE, Aguilar CN. **Mango seed: Functional and nutritional properties**. *Trends Food Sci Technol.* (2016) **55** 109-17. DOI: 10.1016/j.tifs.2016.06.009
5. Ben-Othman S, Jõudu I, Bhat R. **Bioactives from agri-food wastes: present insights and future challenges**. *Molecules.* (2020) **25** 510. DOI: 10.3390/molecules25030510
6. Mohanasundaram S, Ramirez-Asis E, Quispe-Talla A, Bhatt MW, Shabaz M. **Experimental replacement of hops by mango in beer: production and comparison of total phenolics, flavonoids, minerals, carbohydrates, proteins and toxic substances**. *Int J Syst Assur Eng Manag.* (2022) **13** 132-45. DOI: 10.1007/s13198-021-01308-3
7. Diarra SS. **Potential of mango (Mangifera indica L.) seed kernel as a feed ingredient for poultry: a review**. *World's Poult Sci J.* (2014) **70** 279-88. DOI: 10.1017/S0043933914000294
8. Yatnatti S, Vijayalakshmi D, Chandru R. **Processing and nutritive value of mango seed kernel flour**. *Curr Res Nutr Food Sci J.* (2014) **2** 170-5. DOI: 10.12944/CRNFSJ.2.3.10
9. Diarra S, Usman B, Igwebuike J. **Replacement value of boiled mango kernel meal for maize in broiler finisher diets**. *J Agric Biol Sci.* (2010) **5** 47-52
10. Ravindran V, Sivakanesan R. **The nutritive value of mango seed kernels for starting chicks**. *J Sci Food Agric* (1996) **71** 245-50
11. Saiyed L, Parnerkar S, Wadhwani K, Pandya P, Patel A. **Performance of weaner lambs on conventional feeds or supplemented with Mango seed kernel (Mangifera indica) and Babul Pods chuni (Acacia nilotica) under intensive production system**. *Asian-Aust J Anim Sci.* (2003) **16** 1469-74. DOI: 10.5713/ajas.2003.1469
12. De Evan T, Carro M, Yepes JF, Haro A, Arbesú L, Romero-Huelva M. **Feeding mango wastes to dairy goats: Effects on diet digestibility, ruminal fermentation, and milk yield and composition**. *Anim Feed Sci Technol.* (2022) **286** 115252. DOI: 10.1016/j.anifeedsci.2022.115252
13. Abel I, Winifred P, John A. **Digestibility and nutrient intake of mango (Mangifera indica) kernel meal based diets by growing West African dwarf sheep feed rice straw**. *J Res Rep Genet* (2018) **2** 1-3
14. Omer HA, Tawila MA, Gad SM, Abdel-Magid SS. **Mango (Mangifera indica) seed kernels as untraditional source of energy in Rahmani sheep rations**. *Bull Natl Res Cent.* (2019) **43** 1-10. DOI: 10.1186/s42269-019-0241-4
15. Shehabeldin AM, El-Esawy GS, El-Sanafawy HA. **Effect of dietary partial replacement of corn grains by mango seeds on productive and reproductive characterization of damascus goat bucks at prepubertal stage**. *J Anim Poult Prod.* (2021) **12** 61-9. DOI: 10.21608/jappmu.2021.153289
16. de Carvalho FF, Mattos CW, Menezes DR, Coelho MCS, Garcia DA, Pereira Neto JD. *Replacement of corn with mango meal for dairy goats. Revista Colomb Cienc Pecu.* (2016) **29** 178-87. DOI: 10.17533/udea.rccp.v29n3a03
17. Marcos CN, Carro MD, Fernández-Yepes JE, Arbesu L, Molina-Alcaide E. **Utilization of avocado and mango fruit wastes in multi-nutrient blocks for goats feeding: In vitro evaluation**. *Animals.* (2020) **10** 2279. DOI: 10.3390/ani10122279
18. Odunsi A. **Response of laying hens and growing broilers to the dietary inclusion of mango (Mangifera indica L.) seed kernel meal**. *Trop Anim Health Prod.* (2005) **37** 139-50. DOI: 10.1023/B:TROP.0000048455.96694.85
19. Thiex N, Novotny L, Crawford A. **Determination of ash in animal feed: AOAC official method 94205 revisited**. *J AOAC Int.* (2012) **95** 1392-7. DOI: 10.5740/jaoacint.12-129
20. Slansky F. **Insect nutritional ecology as a basis for studying host plant resistance**. *Fla Entomol* (1990) 359-78. DOI: 10.2307/3495455
21. **Committee on Nutrient Requirements of Small Ruminants, National Research Council, Committee on the Nutrient Requirements of Small Ruminants, Board on Agriculture, Division on Earth and Life Studies**. *Nutrient Requirements of Small Ruminants: Sheep, Goats, Cervids, and New World Camelids* (2007)
22. Van Keulen J, Young B. **Evaluation of acid-insoluble ash as a natural marker in ruminant digestibility studies**. *J Anim Sci.* (1977) **44** 282-7. DOI: 10.2527/jas1977.442282x
23. McDonald P, Edwards RA, Greenhalgh JFD. *Animal Nutrition* (1973)
24. Warner A. **Production of volatile fatty acids in the rumen: methods of measurement**. *Nutr. Abstr. Rev.* (1964) **34** 339-52. PMID: 14144939
25. Parekh H. **A new formula for FCM (3.5%) fat [Fat-corrected milk]. Short communication**. *Indian J Anim Sci* (1986)
26. Harding F. *Milk Quality* (1995) 165
27. Sparks DL, Page AL, Helmke PA, Loeppert RH. *Methods of Soil Analysis, Part 3: Chemical Methods* (2020)
28. Zahran HA, Tawfeuk HZ. **Physicochemical properties of new peanut (Arachis hypogaea L) varieties**. *OCL.* (2019) **26** 19. DOI: 10.1051/ocl/2019018
29. Weinberg SL, Abramowitz SK. *Statistics Using SPSS: An Integrative Approach.* (2008)
30. Admasu S, Wondifraw Z, Gash M. **Effects of replacing maize with boiled mango (Mangifera indica) seed kernel on feed intake, body weight gain and feed conversion ratio of cobb 500 broiler chicken**. *Poult Fish Wildl Sci.* (2020) **8** 211. DOI: 10.35248/2375-446X.20.8.211
31. Hutagalung R. **3.4 The use of tree crops and their by-products for intensive animal production**. *BSAP Occas Publ.* (1981) **4** 151-88. DOI: 10.1017/S0263967X00000525
32. Badifu G, Ilochi J, Dutse J, Akpapunam M. **Use of mango mesocarp flour to enrich the provitamin A content of a complementary food blend of maize and soya bean flours for porridge**. *Food Nutr Bull.* (2000) **21** 316-22. DOI: 10.1177/156482650002100312
33. Dhingra S, Kapoor AC. **Nutritive value of mango seed kernel**. *J Sci Food Agric.* (1985) **36** 752-6. DOI: 10.1002/jsfa.2740360817
34. Sruamsiri S, Silman P. **Nutritive value and nutrient digestibility of ensiled mango by-products**. *Maejo Int J Sci Technol.* (2009) **3** 371-8
35. Elegbede J, Achoba I, Richard H. **Nutrient composition of mango (Mangnifera indica) seed kernel from Nigeria**. *J Food Biochem.* (1995) **19** 391-8. DOI: 10.1111/j.1745-4514.1995.tb00543.x
36. Jayanegara A, Leiber F, Kreuzer M. **Meta-analysis of the relationship between dietary tannin level and methane formation in ruminants from**. *J Anim Physiol Anim Nutr.* (2012) **96** 365-75. DOI: 10.1111/j.1439-0396.2011.01172.x
37. Ogundipe SO. **Ration formulation and least cost rations for small ruminants. Manual for Small Ruminant Production in Nigeria**. *Nat Prod Res Inst.* (2002) **1** 49-62
38. Demment MW, Van Soest PJ. **A nutritional explanation for body-size patterns of ruminant and nonruminant herbivores**. *Am Nat.* (1985) **125** 641-72. DOI: 10.1086/284369
39. Khalique A, Lone K, Pasha T, Khan A. **Chemical composition and nutritional evaluation of variously treated defatted rice polishing for broiler feeding**. *Asian Aust J Anim Sci.* (2003) **16** 873-9. DOI: 10.5713/ajas.2003.873
40. Satter L, Slyter L. **Effect of ammonia concentration on rumen microbial protein production in vitro**. *Br J Nutr.* (1974) **32** 199-208. DOI: 10.1079/BJN19740073
41. Pereira L, Barreiros D, Oliveira L, Ferreira A, Mauricio R, Azevedo J. **Chemical composition and ruminal fermentation kinetics of fruit by-products in south Bahia–Brazil**. *Livest Res Rural Dev.* (2008) **20** 1-13
42. Zhang Y, Wang J, Qi B, Wu S, Chen H, Luo H. **Evaluation of mango saponin in broilers: effects on growth performance, carcass characteristics, meat quality and plasma biochemical indices**. *Asian Aust J Anim Sci.* (2017) **30** 1143. DOI: 10.5713/ajas.16.0847
43. Baird MF, Graham SM, Baker JS, Bickerstaff GF. **Creatine-kinase-and exercise-related muscle damage implications for muscle performance and recovery**. *J Nutr Metabol* (2012) **2012** 203-5. DOI: 10.1155/2012/960363
44. Kiflewahid B, Potts GR. *By-Product Utilization for Animal Production: Proceeding of a Workshop on Applied Research.* (1983)
|
---
title: 'Impact of educational attainment and economic globalization on obesity in
adult females and males: Empirical evidence from BRICS economies'
authors:
- Gamze Sart
- Yilmaz Bayar
- Marina Danilina
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9971565
doi: 10.3389/fpubh.2023.1102359
license: CC BY 4.0
---
# Impact of educational attainment and economic globalization on obesity in adult females and males: Empirical evidence from BRICS economies
## Abstract
Obesity has considerably increased since 1980 and become a global epidemic. Obesity-related health problems and the negative social and economic implications of obesity have led international institutions and countries to combat it. This study investigates the role of educational attainment and economic globalization in the global prevalence of obesity in samples of adult females and males in BRICS economies for 1990–2016 through causality and cointegration tests. The results of the causality tests reveal that educational attainment and economic globalization have a significant influence on obesity in both adult females and males in the short run. Furthermore, cointegration analysis indicates a negative effect of educational attainment on obesity in all BRICS economies in the long run, but the influence of economic globalization on obesity differs among the BRICS economies. Furthermore, the negative influence of educational attainment on obesity is revealed to be relatively higher in females than males.
## 1. Introduction
Obesity has become a serious public health and economic problem in the globalized world, and the World Health Organization (WHO) accepted obesity as a global epidemic in 1997 [1]. On the one hand, obesity can be a significant threat to public health in terms of life expectancy, life quality, and being the source of many non-communicable diseases (NCDs) such as cancer, type 2 diabetes, coronary heart disease, and stroke [2]. On the other hand, obesity can negatively influence economies by decreasing life expectancy and productivity and increasing health care expenditures and disability [3]. However, the prevalence rates of obesity are continuing to increase in all countries. The World Obesity Federation (WOF) predicts that one in five women and one in seven men will become obese [in other words, their body mass index (BMI) will be greater than or equal to 30 kg/m2], and, in turn, over a billion people worldwide will be obese [4]. So, obesity is more prevalent in women than men and this trend is predicted not to change in the near future. Furthermore, most obese people have been living in low- and middle-income countries (LMICs), and the number of obese people in LMICs and low-income countries has doubled and more than tripled, respectively, as of 2010 [4].
Countries and international institutions, such as the WHO and the United Nations, have tried to combat obesity given its negative health and economic effects. In this context, the WHO's Global Strategy on Diet, Physical Activity and Health suggests actions to support healthy diets and regular physical activity and calls for stakeholders to take action at local, regional, and global levels to improve the diets and physical activity patterns of individuals [5]. Furthermore, the “Global Action Plan on Physical Activity 2018–2030: More Active People for a Healthier World” by the WHO makes policy suggestions to raise physical activity [6]. Obesity is not explicitly mentioned in the 17 Sustainable Development Goals (SDGs), but it is implicitly targeted in the context of SDG-2 (zero hunger), SDG-3 (good health and wellbeing), and SDG-12 (responsible consumption and production) [7].
The unveiling of the factors underlying obesity are critical for policy-making to combat it. In this context, economic development, technological progress, dietary factors, physical activity, sleep duration, genetics, demographics, social and lifestyle factors, stress levels, environment, and built environment have been documented as the major factors underlying obesity (8–14). However, the determinants of obesity vary considerably between countries based on their economic and social development levels. In this research, the influence of educational attainment and economic globalization on obesity is separately investigated in adult females and males because educational attainment can also influence most of the factors underlying obesity, and the economic globalization can also affect obesity through facilitating the flows of capital, goods and services among the countries. Furthermore, obesity is more prevalent in females than males in accordance with global obesity distribution by gender [4]. In this context, the means of obesity in adult females and males in the BRICS economies are respectively $16.562\%$ and 7.768 during the 1990–2016 period and consistent with World Obesity Federation [4].
Educational attainment is expected to influence obesity through the following channels: (a) education is a significant factor underlying economic growth and development; (b) education is a significant factor underlying personal income and life quality; (c) individuals with higher education are more aware of the determinants of obesity and the associated health risks; (d) individuals with higher education have greater access to information about healthy living and healthcare services [15]. Hence, Cutler and Lleras-Muney [16] discovered that individuals with higher education levels are less likely to be obese, smoke, drink a lot, or use illegal drugs. Therefore, a negative influence of education on obesity is expected, depending on countries' economic development levels.
On the other hand, the world has experienced a significant globalization process as of 1980, and, in turn, the mobility of goods, services, and individuals has considerably grown, and economies and societies have integrated to a great extent. As a result, economic globalization has led to many economic and non-economic changes in the world. In this context, economic globalization can influence obesity in different ways through diverse channels: (a) economic globalization can affect obesity through economic growth and development; (b) economic globalization can ease the entry of food manufacturers and supermarket and fast-food chains into countries and, in turn, foster obesity by increasing accessibility to obesogenic products; (c) economic globalization can influence obesity through the dissemination of the modern workplace, technology use, and motorized transportation; (d) economic globalization can affect obesity through urbanization and cultural changes [17, 18]. Therefore, the influence of economic globalization on obesity can change depending on which factors are dominant in the relationship between economic globalization and obesity.
Furthermore, there is a close interaction between educational attainment and globalization. Education is also internationalized and new concepts such as knowledge economy and lifelong learning are integrated with education policies [19]. The countries have increased their education investments and updated their education curriculum and teaching methods to survive in the highly competitive global economy. The globalized world has also experienced the significant technological progress during the past four decades and in turn the need for a highly skilled workforce is increased in the global labor markets. As a result, educational attainment is going to increase in the world through demand and supply side causes such as higher income and the need for a highly skilled workforce [20] and thus economic globalization can also affect the obesity through the channel of education attainment.
Extensive empirical studies have been conducted on the determinants of obesity in different samples from various countries. This research aims to make a contribution to the literature about the determinants of obesity in three ways. First, the study is one of the first studies to investigate the interaction among educational attainment, economic globalization, and obesity in samples of the economies of Brazil, Russia, India, China, and South Africa (BRICS). BRICS economies are the drivers of global economic expansion and account for $40\%$ of world population, $25\%$ of nominal global GDP, and $30\%$ of world land coverage, and $18\%$ of international trade [21]. Second, the influence of educational attainment and economic globalization on obesity has been relatively less explored, and studies have generally utilized the regression approach and the regression analysis enables us to see the common effect of a variable on dependent variable for all countries. Therefore, another novelty of the study is the utilization of causality and cointegration tests to determine the short- and long-term influence of educational attainment and economic globalization on obesity for each country in the sample. Finally, the study investigates the interaction of educational attainment, economic globalization, and obesity through macro-data, unlike many empirical studies, and its findings may be useful for policy-making to combat obesity. The next part of the paper evaluates and summarizes the empirical studies in the relevant literature, and the data and methods are explained in Section 3. The econometric applications are conducted and their findings are evaluated in view of the related literature in Section 4. The paper comes to its conclusion in Section 5.
## 2. Literature review
Obesity is a global epidemic and the source of many diseases and social and economic problems. Therefore, the determinants of obesity have been explored in a widespread manner; economic development, technological progress, dietary factors, physical activity, sleep duration, genetics, demographics, social and lifestyle factors, stress levels, environment, and built environment have been documented as the major factors underlying obesity (8–14). However, the influence of educational attainment and economic globalization, which also affect all these factors, on obesity has not been explored sufficiently. The influence of educational attainment, including nutritional, physical, and virtual education, on obesity has been investigated relatively more when compared with economic globalization.
The literature summary on the various education indicators–obesity nexus in Table 1 shows that researchers have generally utilized regression analysis and reached the conclusion that education proxied by different indicators has generally had a negative influence on obesity in countries with diverse economic development levels (15, 23–30). However, the interaction between lower education and obesity was generally weaker in men than women [22, 29]. Monteiro et al. [ 22] also found that education did not have a significant impact on obesity risk in men in the less-developed region of Brazil. Furthermore, Curry [12] discovered an insignificant influence of educational attainment on risk of being obese among Black women in the United States. The literature research has uncovered that the influence of educational attainment and economic globalization on obesity has not been analyzed in sample of BRICS economies yet. Therefore, this study investigated the interaction among obesity, educational attainment, and economic globalization in sample of the BRICS economies, the drivers of the global economy during the past a few decades.
**Table 1**
| Study | Sample | Method | Impact of education on obesity |
| --- | --- | --- | --- |
| Monteiro et al. (22) | Brazil | Logistic regression | Education did not have a significant impact on obesity risk for men in the less-developed region. But education had a negative impact on obesity risk for men in the more-developed region. There existed a negative interaction between education and obesity risk for women in both regions |
| Anyanwu et al. (23) | Nigeria (325 males and 254 females with Ibo ethnicity) | Descriptive statistics and correlation analysis | Obesity was the highest in the group with the lowest education |
| Devaux et al. (15) | Australia, Canada, England, and Korea | Regression analysis | Negative |
| Faeh et al. (24) | Switzerland (53,588 adult individuals) | Logistic regression | Negative |
| Brunello et al. (25) | Austria, Denmark, Germany, Greece, Italy, Portugal, Spain, Sweden, and the United Kingdom | Regression analysis | Negative |
| Chung et al. (26) | Republic of Korea | Logistic regression | Negative in both adult men and women |
| Chung and Lim (27) | Republic of Korea (14,577 women) | Extended Oaxaca–Blinder method | Obesity was much more prevalent in women with less education, and lifestyle factors contributed most to obesity |
| Curry (12) | Black women in the United States | Regression analysis | Insignificant |
| Hsieh et al. (28) | Taiwan (28,092 old men and 31,835 old women) | Logistic regression | Negative |
| Witkam et al. (29) | Meta-analysis about the studies on the nexus of education and obesity | | Negative, but the interaction between lower education and obesity was weaker in men than women |
| Iriyani et al. (30) | Indonesia (38 adult participants from Samarinda City Junior High School) | Quasi-experiment approach | Nutritional education and physical activity contributed to weight loss |
| Milla et al. (31) | Indonesia (19 participants from Surabaya) | Quasi-experimental approach | Virtual education had a positive influence on obesity awareness |
The influence of globalization and its main dimensions on obesity has been investigated by very few researchers in Table 2, and they have generally utilized regression analysis to find a positive influence of globalization on obesity (17, 18, 32–34, 36–38). However, Ghosh [35] found that the influence of globalization on obesity changed depending on countries' income levels.
**Table 2**
| Study | Sample | Method | Impact of globalization on obesity |
| --- | --- | --- | --- |
| De Vogli et al. (32) | 127 countries | Regression analysis | Economic globalization had a positive influence on obesity |
| Goryakin et al. (33) | 56 countries (887,000 women) | Regression analysis | Positive |
| Costa-Font and Mas (34) | 26 countries | Regression | Positive |
| Ghosh (35) | Asian countries | Westerlund cointegration test | Economic and social globalization had a positive influence on obesity in low- and low-middle income countries, but globalization had a negative influence on obesity in relatively richer economies |
| Lopez et al. (36) | 44 low- and middle-income economies | Regression analysis | Trade liberalization made a positive contribution to the spread of sugar-sweetened beverages |
| Lin et al. (37) | 172 countries | Regression analysis | Sugar and processed food imports had a positive influence on obesity |
| Fox et al. (38) | 190 countries | Regression analysis | Economic globalization had a positive influence on obesity, but the influence of cultural globalization on obesity varied among the countries |
| García (17) | 10 Latin American and Caribbean economies (320,873 non-pregnant women) | Multiple logistic regression | Positive |
| An et al. (18) | Review of 16 studies about the globalization–obesity nexus | | 14 studies revealed a positive interaction between economic globalization and obesity, one study revealed a negative interaction between two variables, and one discovered an insignificant interaction between two variables |
## 3. Data and method
This article studies the effects of educational attainment and economic globalization on obesity in females and males in BRICS economies for 1990–2016 through cointegration and causality tests. In the econometric analyses, adult obesity (OBS) is proxied by males or females with BMIs of 30 kg/m2 or higher as a percentage of the male/female population aged 18+ and is obtained from the World Bank database [39]. The BMI is calculated through weight (kilograms) divided by squares of the height (meters). Educational attainment (EDU) is substituted by the mean years of schooling of males/females by UNDP [40], and economic globalization (EG) is substituted with the economic globalization index calculated by the KOF Swiss Economic Institute [41] and measures the trade and financial globalization and gets value between 1 and 100 (higher values reflect higher economic globalization level). All series are yearly, and the study period is specified as 1990–2016 because adult obesity data is available for this period.
The main characteristics of the obesity, educational attainment, and economic globalization reported in Table 3 indicate that the means of obesity in adult females and males are, respectively, $16.562\%$ and 7.768, so obesity is more prevalent in females than males in the BRICS economies. Furthermore, South Africa, Russia, and Brazil had a larger obesity rate than China and India, and females also had considerably larger obesity rates in these countries than males. On the other hand, the mean years of schooling are 6.93 years in females and 7.63 years in men, and the gap in schooling years by gender is relatively very low. However, females had relatively larger schooling years in Brazil and Russia, but males had relatively larger schooling years in China, India, and South Africa. The mean economic globalization level is 41.203 in BRICS economies during 1990–2016 and Russia and South Africa had relatively higher economic globalization level. Furthermore, variations in obesity and economic globalization levels in these countries are larger than those in education.
**Table 3**
| Characteristics | Characteristics.1 | Females | Females.1 | Females.2 | Males | Males.1 | Males.2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | OBS * | EDU ** | EG *** | OBS * | EDU ** | EG *** |
| Mean | Mean | 16.562 | 6.933 | 41.203 | 7.768 | 7.631 | 41.203 |
| Maximum | Maximum | 39.600 | 12.565 | 58.374 | 18.500 | 12.562 | 58.374 |
| Minimum | Minimum | 1.300 | 1.793 | 14.507 | 0.500 | 3.529 | 14.507 |
| Std. Dev. | Std. Dev. | 11.953 | 3.112 | 10.446 | 5.570 | 2.562 | 10.446 |
| Brazil | Mean | 19.174 | 5.898 | 38.074 | 12.452 | 5.581 | 38.074 |
| | Maximum | 25.400 | 7.908 | 47.000 | 18.500 | 7.528 | 47.000 |
| | Minimum | 13.100 | 3.796 | 25.000 | 7.000 | 3.529 | 25.000 |
| | Std. Dev. | 3.762 | 1.266 | 6.805 | 3.526 | 1.261 | 6.805 |
| China | Mean | 3.703 | 5.404 | 40.815 | 2.648 | 6.543 | 40.815 |
| | Maximum | 6.500 | 6.992 | 52.000 | 5.900 | 7.594 | 52.000 |
| | Minimum | 1.700 | 3.421 | 43.000 | 0.800 | 4.867 | 43.000 |
| | Std. Dev. | 1.481 | 1.049 | 6.873 | 1.557 | 0.806 | 6.873 |
| India | Mean | 2.874 | 3.243 | 32.778 | 1.326 | 5.576 | 32.778 |
| | Maximum | 5.100 | 5.534 | 46.000 | 2.700 | 7.250 | 46.000 |
| | Minimum | 1.300 | 1.793 | 15.000 | 0.500 | 3.620 | 15.000 |
| | Std. Dev. | 1.127 | 1.058 | 11.453 | 0.679 | 1.083 | 11.453 |
| Russia | Mean | 25.085 | 11.731 | 45.926 | 13.233 | 11.668 | 45.926 |
| | Maximum | 26.900 | 12.565 | 55.000 | 18.100 | 12.562 | 55.000 |
| | Minimum | 23.500 | 9.401 | 24.000 | 9.200 | 9.974 | 24.000 |
| | Std. Dev. | 1.013 | 1.037 | 8.176 | 2.714 | 0.865 | 8.176 |
| South Africa | Mean | 31.974 | 8.391 | 48.556 | 9.181 | 8.787 | 48.556 |
| | Maximum | 39.600 | 10.037 | 58.000 | 15.400 | 10.451 | 58.000 |
| | Minimum | 24.200 | 6.378 | 29.000 | 4.500 | 6.601 | 29.000 |
| | Std. Dev. | 4.723 | 1.102 | 10.146 | 3.318 | 1.175 | 10.146 |
The causal and cointegration interactions of educational attainment, economic globalization, and obesity are, respectively, investigated with the Dumitrescu and Hurlin [42] causality test and Westerlund and Edgerton [43] LM (Lagrange Multiplier) bootstrap cointegration test in view of the fact that there exists heterogeneity and cross-sectional dependence among education, globalization, and obesity.
Cointegration tests investigate whether the long-run linear relationship among two or more series stationary even if there is not the linear relationship in the short-run [44]. Therefore, cointegration test is employed to analyze the cointegration among educational attainment, economic globalization, and obesity, because increasing the educational attainment generally is a long-term phenomenon. However, causality analysis is also utilized to see the short run interaction among educational attainment, economic globalization, and obesity.
The LM bootstrap cointegration test permits autocorrelation and heteroscedasticity in the cointegration equation and also produces relatively more robust results for small sample sizes. The test is based on the LM test of McCoskey and Kao [45], and bootstrap critical values are taken into account in case there exists cross-sectional dependence [46]. The cointegration test is generated from Equation [1]: $t = 1$….,T and $i = 1$….,N respectively specifies the time series and cross-sections and zit (zit=μit+vit∑$j = 1$tnij)is the error term. nij is an error term with a zero mean and σi2 variance.
The null hypothesis of the cointegration test suggests a significant cointegration among education, globalization, and obesity in all countries and is tested by the LM test statistic in Equation [2].
sit2 is partial sum of zit, and ω^i-2 is long-term variance of μit.
The causality analysis investigates a bidirectional interaction among educational attainment, economic globalization, and obesity. In other words, it tests whether educational attainment has a significant effect on obesity or obesity has a significant effect on educational attainment. The Dumitrescu and Hurlin [42] causality test can be utilized in case of unbalanced panels, existence of cross-sectional dependence, N > T, and T > N and the test employs the following equation: In [3] numbered equation, k is lag length, γ and β are respectively dependent and independent variables lags' coefficients. All variables used in the causality analysis should be stationary. The null hypothesis of the test suggests an insignificant causality between two series and the null hypothesis is tested by Wald (WN,THnc (Homogeneous non causality)) and ZN,THnc test statistics as following [see Dumitrescu and Hurlin [42] for detailed information about calculation of test statistics]: Zhnc(ZNTHnc) test statistic in Equation [4] with asymptotic distribution is taken into account if N<T, but Ztild (ZNHnc) test statistic in Equation [5] with semi- asymptotic distribution is taken into account if T<N.
## 4. Results and discussion
The interaction of educational attainment, economic globalization, and obesity is analyzed by cointegration and causality tests. In this context, pretests of cross-sectional dependence and heterogeneity are, respectively, investigated by LM and delta tilde tests at first. The existence of cross-sectional dependence among countries is examined with LMadj., LM CD, and LM tests and their results are depicted in Table 4. The alternative hypothesis of three tests (“there exists cross-sectional dependence”) is accepted because the probability values of these tests are lower than 0.05. Then, the existence of slope coefficients' homogeneity is controlled by delta tilde tests, and their results are depicted in Table 4. The alternative hypothesis of two tests (“there exists heterogeneity”) is accepted because the probability values of these tests are lower than 0.05. So, the effect of educational attainment and economic globalization on obesity in adult females and males differs among the countries.
**Table 4**
| Test | Model 1 (Females) | Model 1 (Females).1 | Model 2 (Males) | Model 2 (Males).1 |
| --- | --- | --- | --- | --- |
| | Test statistic | P value | Test statistic | P value |
| Cross-sectional dependence tests | Cross-sectional dependence tests | Cross-sectional dependence tests | Cross-sectional dependence tests | Cross-sectional dependence tests |
| LM adj (47) | 22.61 | 0.0000 | 27.4 | 0.000 |
| LM CD (48) | 3.929 | 0.0001 | 6.286 | 0.000 |
| LM (49) | 47.96 | 0.0000 | 55.81 | 0.000 |
| Heterogeneity tests | Heterogeneity tests | Heterogeneity tests | Heterogeneity tests | Heterogeneity tests |
| Delta tilde (50) | 21.041 | 0.000 | 12.634 | 0.000 |
| Adjusted delta tilde (50) | 22.797 | 0.000 | 13.689 | 0.008 |
The stationarity analysis of OBS, EDU, and EG in Model 1 and Model 2 is implemented by Pesaran [46] cross-sectional augmented Dickey–Fuller (CADF) unit root test, and the results are depicted in Table 5. All series are not stationary at their level values, but the series have become stationary at first-differenced values.
**Table 5**
| Variables | Model 1 (Females) | Model 1 (Females).1 | Model 2 (Males) | Model 2 (Males).1 |
| --- | --- | --- | --- | --- |
| | Constant | Constant + Trend | Constant | Constant + Trend |
| OBS | –1.178 | –1.325 | –0.983 | –1.023 |
| D(OBS) | –8.312*** | −8.793 *** | –6.462*** | −7.063 *** |
| EDU | –1.205 | –1.467 | –1.142 | –1.156 |
| D(EDU) | –9.735*** | −10.042 *** | –8.606*** | −8.993 *** |
| EG | –0.821 | 0.820 | –0.821 | 0.820 |
| D(EG) | –2.532*** | –4.562*** | –2.532*** | −4.562 *** |
The long-term interaction of educational attainment, economic globalization, and obesity in adult females and males in BRICS economies is investigated by the LM bootstrap cointegration test in deference to small sample sizes and subsistence of cross-sectional dependence. The results of the LM bootstrap cointegration test are depicted in Table 6. As a result, the null hypothesis (“there exists a significant cointegration interaction among educational attainment, economic globalization, and obesity for females and females”) is accepted, and a significant long-term relationship between the three variables is reached.
**Table 6**
| Models | Constant | Constant.1 | Constant.2 | Constant + Trend | Constant + Trend.1 | Constant + Trend.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | Test statistic | Asymptotic P value | Bootstrap P value | Test statistic | Asymptotic P value | Bootstrap P value |
| Model 1 (Females) | 8.914 | 0.214 | 0.289 | 9.127 | 0.302 | 0.326 |
| Model 2 (Males) | 7.265 | 0.145 | 0.168 | 8.902 | 0.210 | 0.311 |
The cointegration coefficients are predicted by AMG estimator [51, 52], and the coefficients are denoted in Table 7. The estimated coefficients reveal that educational attainment has a negative impact on obesity in females and males in all BRICS economies, but the negative influence of educational attainment on obesity is found to be relatively higher in females than males. Furthermore, the negative impact of educational attainment on obesity in both females and males is relatively higher in China, India, and Russia.
**Table 7**
| Countries | Model 1 (Females) | Model 1 (Females).1 | Model 2 (Males) | Model 2 (Males).1 |
| --- | --- | --- | --- | --- |
| | EDU | EG | EDU | EG |
| Brazil | −0.121 *** | 0.007 | −0.081 *** | 0.005 |
| China | −0.196 *** | −0.023 *** | −0.142 *** | −0.021 *** |
| India | −0.162 * | −0.031 * | −0.151 ** | −0.011 ** |
| Russia | −0.147 *** | 0.005 *** | −0.144 *** | −0.002 |
| South Africa | −0.113 * | 0.027 *** | −0.089 ** | 0.012 *** |
| Panel | −0.012 | −0.003 | −0.0621 | −0.008 |
On the other hand, economic globalization has a negative influence on obesity in China and India, but a positive influence on obesity in females in Russia and South Africa. Economic globalization also has a negative influence on obesity in males in China and India, but a positive influence on obesity in males in South Africa.
The findings of the study are compatible with theoretical expectations and related empirical literature about the education–obesity nexus. The cointegration analysis indicates that improvements in educational attainment contribute to decreases in obesity in the long run. Educational attainment proxied by different indicators can make a direct contribution to decreases in obesity by raising awareness of obesity-related health problems and encouraging healthy eating and regular physical activity. Higher educational attainment can also cause individuals to earn higher income and, in turn, foster healthy nutrition and lifestyles. Furthermore, educational attainment can contribute to decreases in obesity by enhancing economic growth and development because educational attainment is a critical factor for human capital, which is a significant determinant of economic growth and development. In the related empirical literature, Devaux et al. [ 15], Anyanwu et al. [ 23], Faeh et al. [ 24], Brunello et al. [ 25], Chung et al. [ 26], Chung and Lim [27], Hsieh et al. [ 28], Witkam et al. [ 29], Iriyani et al. [ 30], and Monteiro et al. [ 22] also discovered a negative influence of various education indicators on obesity in different countries with different income levels in a similar way.
Our findings also reveal that the influence of educational attainment on obesity is relatively higher in females than males in the BRICS economies. Witkam et al. [ 29] and Monteiro et al. [ 22] similarly reached the conclusion that education is more effective for obesity in women than men. Furthermore, the influence of educational attainment on obesity in both genders varies among the BRICS economies. China and India achieved significant progress in educational attainment, GDP per capita and human development during the 1990–2016 period and Russia was the leading country among the BRICS economies in terms of socio-economic development as seen in Table 8. Therefore, we evaluate that the variations about the influence of educational attainment on obesity can be resulted from the differences in human and economic development of the BRICS economies.
**Table 8**
| Countries | GDP per capita (constant 2015 US $) | GDP per capita (constant 2015 US $).1 | Human development index | Human development index.1 |
| --- | --- | --- | --- | --- |
| | 1990.0 | 2016.0 | 1990.0 | 2016.0 |
| Brazil | 6155.645 | 8455.312 | 0.61 | 0.755 |
| China | 905.031 | 8516.514 | 0.484 | 0.74 |
| India | 532.755 | 1719.318 | 0.434 | 0.639 |
| Russian Federation | 7849.512 | 9313.967 | 0.743 | 0.828 |
| South Africa | 5031.464 | 6209.365 | 0.632 | 0.719 |
Economic globalization can influence obesity through fostering economic growth and development, increasing accessibility to obesogenic products, disseminating the modern workplace, technology use, and motorized transportation, and urbanization [17, 18]. Therefore, which of these factors is dominant determines the effect of economic globalization on obesity. In the related empirical literature, García [17], De Vogli et al. [ 32], Goryakin et al. [ 33], Costa-Font and Mas [34], Lopez et al. [ 36], Fox et al. [ 38], and Lin et al. [ 37] discovered a positive influence of various globalization components on obesity. Only Ghosh [35] reached the conclusion that the influence of globalization on obesity varies based on countries' income levels. In the study, the influence of economic globalization on obesity differs among the BRICS economies and the coefficients are revealed to much lower when compared with educational attainment. The positive influence of economic globalization on obesity in females in Russia and South Africa and in males in South *Africa is* consistent with the related empirical literature to a great extent. However, economic globalization has a very small negative influence on obesity in both genders in China and India in compatible with findings by Ghosh [35]. The weak positive or negative influence of economic globalization on obesity can be probably resulted from low economic globalization levels of BRICS economies, because the mean of economic globalization in the BRICS economic is only 41.203 over the 1990–2016 period. Furthermore, the mean of economic globalization in South Africa, Russia, China, Brazil, and India for the 1990–2016 period are respectively 48.556, 45.926, 40.815, 38.074, and 32.778.
The causal interaction of educational attainment, economic globalization, and obesity for females and males is investigated by the Dumitrescu and Hurlin [42] causality test, and the results of this test are depicted in Table 9. The causality analysis uncovered a unidirectional causality from educational attainment and economic globalization to obesity for both genders. In other words, economic globalization and educational attainment have a significant effect on obesity in the short term. A significant interaction between educational attainment, economic globalization, and obesity is theoretically expected, but the studies have mainly conducted one-way analyses from educational attainment and economic globalization to obesity and discovered a significant influence of both variables on obesity. In this research, mutual interaction of educational attainment, economic globalization, and obesity is analyzed, but an insignificant influence from obesity on educational attainment and economic globalization is discovered. Therefore, the causality findings are revealed to be compatible with the related empirical literature to a great extent.
**Table 9**
| Unnamed: 0 | Unnamed: 1 | Females | Females.1 | Males | Males.1 |
| --- | --- | --- | --- | --- | --- |
| Null hypothesis | Test | Test statistics | P value | Test statistics | P value |
| DEDU ↛DOBS | Whnc | 5.721 | 0.000 | 8.453 | 0.000 |
| | Zhnc | 6.909 | 0.000 | 8.912 | 0.000 |
| | Ztild | 7.112 | 0.000 | 9.067 | 0.000 |
| DOBS ↛DEDU | Whnc | 1.256 | 0.109 | 1.563 | 0.109 |
| | Zhnc | 1.310 | 0.115 | 1.879 | 0.127 |
| | Ztild | 1.378 | 0.128 | 2.112 | 0.156 |
| DEG ↛DOBS | Whnc | 6.808 | 0.006 | 5.302 | 0.000 |
| | Zhnc | 7.114 | 0.000 | 6.478 | 0.000 |
| | Ztild | 7.913 | 0.001 | 6.801 | 0.012 |
| DOBS ↛DEG | Whnc | 1.195 | 0.132 | 0.982 | 0.145 |
| | Zhnc | 1.624 | 0.149 | 1.314 | 0.166 |
| | Ztild | 2.089 | 0.161 | 1.722 | 0.210 |
## 5. Conclusion
Worldwide obesity has increased considerably, and obesity is accepted as a global epidemic and one of the most important threats to public health. Obesity is not only a significant source of many NCDs but also leads many negative economic and social implications for societies. Therefore, international institutions and national governments have tried to control and decrease abnormal increases in obesity.
In this study, the influence of educational attainment and economic globalization on obesity in adult females and males are separately investigated in sample of BRICS economies through causality and cointegration analyses by paying attention to significant differences in obesity rate between females and males. The causality analysis reveals that education and economic globalization have significant influence on female and male obesity in the short term. On the other hand, the cointegration analysis shows that educational attainment has a negative influence on obesity in both adult females and males, but the influence of educational attainment on obesity is generally revealed to be higher in females than males. Furthermore, the influence of economic globalization on obesity varies among the BRICS economies.
Our findings and the related literature indicate that educational attainment has a negative influence on obesity in countries with different income levels and also suggest that educational attainment is one of the most effective instruments for decreasing obesity. Therefore, increasing educational attainment should be used as a policy instrument to decrease obesity. In addition, the related literature has widely revealed a positive influence of economic globalization on obesity because economic globalization can increase access to obesogenic products and disseminate the modern workplace, technology use, motorized transportation, urbanization, and cultural changes. However, a positive influence of economic globalization on obesity is discovered for females in Russia and South Africa and for males in South Africa and a very small negative influence of economic globalization on obesity is revealed for both genders in China and India. Both positive and negative influence of economic globalization on obesity is very small when compared with that of educational attainment. We evaluate that the small influence of economic globalization on obesity can be resulted low economic globalization levels of the BRICS economies.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
## Author contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Haththotuwa RN, Wijeyaratne CN, Senarath U, Tahir A.. **Chapter 1 - worldwide epidemic of obesity**. *Mahmood, Sabaratnam Arulkumaran, Frank A. Chervenak, Obesity and Obstetrics (Second Edition)* (2020) 3-8. DOI: 10.1016/B978-0-12-817921-5.00001-1
2. Tiwari A,., Balasundaram P.. (2022)
3. Shekar M, Popkin B. *Obesity: Health and Economic Consequences of an Impending Global Challenge* (2020). DOI: 10.1596/978-1-4648-1491-4
4. **World Obesity Atlas,**. (2022)
5. **Obesity and overweight,**. (2022)
6. **Global Action Plan on Physical Activity 2018-2030: More Active People for a Healthier World, Geneva: World Health Organization**. (2018)
7. United Nations. **Do you know all 17 SDGs?**. (2022)
8. Finkelstein EA, Ruhm CJ, Kosa KM. **Economic causes and consequences of obesity**. *Annu Rev Public Health.* (2005) **26** 239-57. DOI: 10.1146/annurev.publhealth.26.021304.144628
9. Hruby A, Manson JE, Qi L, Malik VS, Rimm EB, Sun Q. **Determinants and consequences of obesity**. *Am J Public Health* (2016) **106** 1656-1662. DOI: 10.2105/AJPH.2016.303326
10. Kerkadi A, Sadig AH, Bawadi H, Al Thani AAM, Al Chetachi W, Akram H. **The relationship between lifestyle factors and obesity indices among adolescents in Qatar**. *Int J Environ Res Public Health.* (2019) **16** 4428. DOI: 10.3390/ijerph16224428
11. Bu T, Popovic S, Huang H, Fu T, Gardasevic J. **Relationship between national economic development and body mass index in Chinese children and adolescents aged 5–19 from 1986 to 2019**. *Front Pediatr.* (2021) **9** 671504. DOI: 10.3389/fped.2021.671504
12. Curry GD. **The impact of educational attainment on black women's obesity rate in the United States**. *J Racial Ethnic Health Dispar.* (2020) **7** 345-54. DOI: 10.1007/s40615-019-00663-z
13. Lam TM, Vaartjes I, Grobbee DE. **Associations between the built environment and obesity: an umbrella review**. *Int J Health Geogr.* (2021) **20** 7. DOI: 10.1186/s12942-021-00260-6
14. Tanveer M, Hohmann A, Roy N, Zeba A, Tanveer U, Siener M. **The current prevalence of underweight, overweight, and obesity associated with demographic factors among Pakistan school-aged children and adolescents—An empirical cross-sectional study**. *Int J Environ Res Public Health.* (2022) **19** 11619. DOI: 10.3390/ijerph191811619
15. Devaux M, Sassi F, Church J, Cecchini M, Borgonovi F. **Exploring the relationship between education and obesity**. *OECD J Econ Stud.* (2011) **2011** 1-40. DOI: 10.1787/eco_studies-2011-5kg5825v1k23
16. Cutler D, Lleras-Muney A. *Education and health: Evaluating theories and evidence* (2006)
17. García CM. **Association of globalization in its different dimensions with overweight and obesity: An analysis in 10 Latin American and Caribbean countries**. *Salud Pública de México.* (2019) **61** 174-83. DOI: 10.21149/8886
18. An R, Shen J, Bullard T, Han Y, Qiu D, Wang S. **A scoping review on economic globalization in relation to the obesity epidemic**. *Obes Rev.* (2020) **21** e12969. DOI: 10.1111/obr.12969
19. Zajda J, Fan G, Popkewitz T. **Globalisation, Education and Policy Reforms**. *Handbook of Education Policy Studies.* (2020). DOI: 10.1007/978-94-024-1743-2
20. 20.Liu MY. How does globalization affect educational attainment? Evidence from China, 2019. (2019). Available online: https://www.freit.org/WorkingPapers/Papers/Development/FREIT1585.pdf (accessed January 20, 2023).. *How does globalization affect educational attainment? Evidence from China, 2019.* (2019)
21. Iqbal BA. **BRICS as a driver of global economic growth and development**. *Global J Emer Market Econ.* (2021) **14** 7-8. DOI: 10.1177/09749101211067096
22. Monteiro CA, Conde WL, Popkin BM. **Independent effects of income and education on the risk of obesity in the Brazilian adult population**. *J Nutr.* (2001) **131** 881S-6S. DOI: 10.1093/jn/131.3.881S
23. Anyanwu GE, Ekezie J, Danborno B, Ugochukwu AI. **Impact of education on obesity and blood pressure in developing countries: A study on the Ibos of Nigeria**. *N Am J Med Sci.* (2010) **2** 320-4. DOI: 10.4297/najms.2010.2320
24. Faeh D, Braun J, Bopp M. **Prevalence of obesity in Switzerland 1992-2007: The impact of education, income and occupational class**. *Obes Rev.* (2011) **12** 151-66. DOI: 10.1111/j.1467-789X.2010.00793.x
25. Brunello G, Fabbri D, Fort M. **The causal effect of education on body mass: Evidence from Europe**. *J Labor Econ.* (2013) **31** 195-223. DOI: 10.1086/667236
26. Chung W, Lim SJ, Lee S, Kim R, Kim J. **Gender-specific interactions between education and income in relation to obesity: A Cross-sectional analysis of the fifth Korea National Health and Nutrition Examination Survey (KNHANES V)**. *BMJ Open.* (2017) **7** e014276. DOI: 10.1136/bmjopen-2016-014276
27. Chung W, Lim S. **Factors contributing to educational differences in obesity among women: Evidence from South Korea**. *BMC Public Health.* (2020) **20** 1136. DOI: 10.1186/s12889-020-09221-3
28. Hsieh TH, Lee JJ, Yu EWR, Hu HY, Lin SY, Ho CY. **Association between obesity and education level among the elderly in Taipei, Taiwan between 2013 and 2015: A cross-sectional study**. *Sci Rep.* (2020) **10** 20285. DOI: 10.1038/s41598-020-77306-5
29. Witkam R, Gwinnutt JM, Humphreys J, Gandrup J, Cooper R, Verstappen SMM. **Do associations between education and obesity vary depending on the measure of obesity used? A systematic literature review and meta-analysis**. *SSM - Population Health.* (2021) **15** 100884. DOI: 10.1016/j.ssmph.2021.100884
30. Iriyani K, Anwar F, Kustiyah L, Riyadi H. **The impact of nutrition education and physical activity on changes in nutritional status in adolescents overweight and obesity**. *Teikyo Med J.* (2022) **45** 5651-62
31. Milla C, Dwipayana MA, Novita SM. **Impact of virtual education on obesity knowledge, attitude, and practice**. *Int J Res Public.* (2022) **107** 124-34. DOI: 10.47119/IJRP1001071820223791
32. De Vogli R, Kouvonen A, Elovainio M, Marmot M. **Economic globalization, inequality and body mass index: A cross-national analysis of 127 countries**. *Crit Public Health.* (2014) **24** 7-21. DOI: 10.1080/09581596.2013.768331
33. Goryakin Y, Lobstein T, James W, Suhrcke M. **The impact of economic, political and social globalization on overweight and obesity in the 56 low and middle income countries**. *Soc Sci Med.* (2015) **133** 67-76. DOI: 10.1016/j.socscimed.2015.03.030
34. Costa-Font J, Mas N. **‘Globesity'? The effects of globalization on obesity and caloric intake**. *Food Policy.* (2016) **64** 121-32. DOI: 10.1016/j.foodpol.2016.10.001
35. 35.Ghosh S,. Globalization obesity: Asian experiences of ‘Globesity'. MPRA Paper 2017, No. 94601, (2017). Available online at: https://mpra.ub.uni-muenchen.de/94601/ (accessed July 5, 2022). (2017)
36. Lopez AM, Loopstra R, McKee M, Stuckler D. **Is Trade liberalisation a vector for the spread of sugar-sweetened beverages? A cross-national longitudinal analysis of 44 low- and middle-income**. *Soc Sci Med* (2017) **172** 21-27. DOI: 10.1016/j.socscimed.2016.11.001
37. Lin TK, Teymourian Y, Tursini MS. **The effect of sugar and processed food imports on the prevalence of overweight and obesity in 172 countries**. *Global Health.* (2018) **14** 35. DOI: 10.1186/s12992-018-0344-y
38. Fox A, Feng W, Asal V. **What is driving global obesity trends? Globalization or “Modernization”?**. *Global Health.* (2019) **15** 32. DOI: 10.1186/s12992-019-0457-y
39. **Prevalence of obesity (% of population ages 18+)**. (2022)
40. 40.UNDP. Human Development Reports. (2022). Available online at: https://hdr.undp.org/data-center/documentation-and-downloads (accessed July 10, 2022).. (2022)
41. 41.KOF Swiss Economic Institute. KOF Globalisation Index. (2022). Available online at: https://kof.ethz.ch/en/forecasts-and-indicators/indicators/kof-globalisation-index.html (accessed July 10, 2022).. *KOF Globalisation Index.* (2022)
42. Dumitrescu E-I, Hurlin C. **Testing for Granger non-causality in heterogeneous panels**. *Econ Model.* (2012) **29** 1450-60. DOI: 10.1016/j.econmod.2012.02.014
43. Westerlund J, Edgerton DL. **A panel bootstrap cointegration test**. *Econ Lett.* (2007) **97** 185-90. DOI: 10.1016/j.econlet.2007.03.003
44. Tu C, Fan Y, Fan J. **Universal cointegration and its applications**. *IScience.* (2019) **9** 986-95. DOI: 10.1016/j.isci.2019.08.048
45. McCoskey S, Kao C. **A residual-based test of the null of cointegration in panel data**. *Econometr Rev.* (1998) **17** 57-84. DOI: 10.1080/07474939808800403
46. Pesaran MH. **A simple panel unit root test in the presence of cross-section dependence**. *J Appl Econom.* (2007) **22** 265-312. DOI: 10.1002/jae.951
47. Pesaran MH, Ullah A, Yamagata T. **A bias-adjusted LM test of error cross-section independence**. *Econom J.* (2008) **11** 105-27. DOI: 10.1111/j.1368-423X.2007.00227.x
48. Pesaran MH. *General diagnostic tests for cross-section dependence in panels.* (2004). DOI: 10.2139/ssrn.572504
49. Breusch TS, Pagan AR. **The lagrange multiplier test and its applications to model specification tests in econometrics**. *Rev Econ Stud.* (1980) **47** 239-53. DOI: 10.2307/2297111
50. Pesaran H, Yamagata T. **Testing slope homogeneity in large panels**. *J Econom.* (2008) **142** 50-93. DOI: 10.1016/j.jeconom.2007.05.010
51. Eberhardt M, Bond S. *Cross-section dependence in nonstationary panel models: A novel estimator.* (2009)
52. Teal F, Eberhardt M. *Productivity analysis in global manufacturing production.* (2010)
53. **GDP per capita (constant 2015 US$),**. (2022)
|
---
title: 'Incidence of Parkinson’s disease and modifiable risk factors in Korean population:
A longitudinal follow-up study of a nationwide cohort'
authors:
- Sung Hoon Kang
- Seok-Joo Moon
- Minwoong Kang
- Su Jin Chung
- Geum Joon Cho
- Seong-Beom Koh
journal: Frontiers in Aging Neuroscience
year: 2023
pmcid: PMC9971569
doi: 10.3389/fnagi.2023.1094778
license: CC BY 4.0
---
# Incidence of Parkinson’s disease and modifiable risk factors in Korean population: A longitudinal follow-up study of a nationwide cohort
## Abstract
### Introduction
We aimed to investigate the incidence of Parkinson’s disease (PD) by age and year for each sex as well as the modifiable risk factors for PD. Using data from the Korean National Health Insurance Service, 938,635 PD and dementia-free participants aged ≥40 years who underwent general health examinations were followed to December 2019.
### Methods
We analyzed the PD incidence rates according to age, year and sex. To investigate the modifiable risk factors for PD, we used the Cox regression model. Additionally, we calculated the population-attributable fraction to measure the impact of the risk factors on PD.
### Results
During follow-up, 9,924 of the 938,635 ($1.1\%$) participants developed PD. The incidence of PD increased continuously from 2007 to 2018, reaching 1.34 per 1,000 person-years in 2018. The incidence of PD also increases with age, up to 80 y. Presence of hypertension (SHR = 1.09, $95\%$ CI 1.05 to 1.14), diabetes (SHR = 1.24, $95\%$ CI 1.17 to 1.31), dyslipidemia (SHR = 1.12, $95\%$ CI 1.07 to 1.18), ischemic stroke (SHR = 1.26, $95\%$ CI 1.17 to 1.36), hemorrhagic stroke (SHR = 1.26, $95\%$ CI 1.08 to 1.47), ischemic heart disease (SHR = 1.09, $95\%$ CI 1.02 to 1.17), depression (SHR = 1.61, $95\%$ CI 1.53 to 1.69), osteoporosis (SHR = 1.24, $95\%$ CI 1.18 to 1.30), and obesity (SHR = 1.06, $95\%$ CI 1.01 to 1.10) were independently associated with a higher risk for PD.
### Discussion
Our results highlight the effect of modifiable risk factors for PD in the Korean population, which will help establish health care policies to prevent the development of PD.
## 1. Introduction
As the population ages, the number of patients with neurodegenerative diseases also rapidly increase, along with an increased socioeconomic burden (Bach et al., 2011). Parkinson’s disease (PD) is the second most common neurodegenerative disease and is characterized by progressive non-motor symptoms and motor deficits, including bradykinesia, tremors, and rigidity. Given that PD currently has no disease-modifying treatment, it is important to identify the incidence and modifiable risk factors of PD and to find effective strategies for preventing PD in public health care policies.
The incidence of PD varies across countries, ranging from 80.4 to 678 per 100,000 person-years (Baldereschi et al., 2000; Benito-León et al., 2004; de Lau et al., 2004; Taylor et al., 2006; Alves et al., 2009; Driver et al., 2009; Linder et al., 2010; Winter et al., 2010; Caslake et al., 2013). In the Korean population, the prevalence of PD has been steadily increasing, and the prevalence in people aged ≥50 y is approximately $0.4\%$ (Park J. H. et al., 2019). However, a detailed information regarding age-specific PD incidence is lacking.
A variety of known modifiable risk factors exist for PD with varying degrees of impact on PD (Ascherio and Schwarzschild, 2016). However, the extent to which modifiable risk factors affect PD remains controversial. The prevention of PD has been the focus of research owing to the absence of disease-modifying medications. Risk factors that have a higher relative risk (RR) for PD and a higher prevalence in the elderly population may contribute more to PD incidence. Therefore, the RR and prevalence of each risk factor in the elderly population should be considered to establish public health care measures for PD prevention. In addition, among the risk factors, the effect of cardiometabolic syndrome on PD risk has not been established, although a growing body of evidence has shown that cardiometabolic syndrome is closely related to Alzheimer’s disease, included in neurodegenerative disease with PD.
Using the Korean National Health Insurance Service (KNHIS) data, the first goal of our study was to investigate PD incidence by age and year for each sex. The second goal was to explore the hazard ratio (HR) of each modifiable risk factor for PD. The third goal was to evaluate the RR of each modifiable risk factor for PD and estimate the attributable fraction of the risk factors in elderly Koreans.
## 2. Materials and methods
This study was approved by the Institutional Review Board of the Korea University Guro Hospital and adhered to the principles of the Declaration of Helsinki.
## 2.1. Data source
We used a customized dataset from the KNHIS, which includes more than $99\%$ of the Korean population (approximately 50 million).1 The KNHIS database includes personal information; health insurance claim codes (procedures and prescriptions); diagnostic codes from the Korean Standard Classification of Diseases, 7th Revision, which is based on the International Classification of Diseases, 10th Revision (ICD-10); death records from the Korean National Statistical Office; and general medical examination data for each participant from 2002 to 2019. Data on body mass index (BMI) and behavioral characteristics, including frequency of physical activity, smoking, and alcohol consumption, were obtained from the general health examinations in the KNHIS database.
## 2.2. PD and dementia-free cohort
To exclude participants with PD and dementia, PD was defined according to the ICD-10 code (G20) and prescriptions of PD medication. Dementia was defined according to the ICD-10 codes (F00, F01, F02, F03, F05, G30, or G31) and dementia medication prescription.
In the NHIS dataset, 6,257,567 PD and dementia-free participants aged 45 y or older who underwent general health examinations were identified. We randomly selected $15\%$ [938,635] of the participants and enrolled them in the present study.
## 2.3. Definition of modifiable risk factors
With respect to the modifiable risk factors for PD, we considered hypertension, diabetes, hyperlipidemia, ischemic stroke, hemorrhagic stroke, ischemic heart disease, depression, osteoporosis, obesity, physical inactivity, smoking status, and heavy alcohol consumption. The presence of hypertension was defined according to ICD-10 code (I10-15) and prescription of antihypertensive medication. The presence of diabetes was defined according to the ICD-10 code (E8-14) and prescription of antidiabetic medication. The presence of hyperlipidemia was defined according to the ICD-10 code (E78) and the prescription of lipid-lowering medication. The presence of ischemic stroke was defined according to the ICD-10 code (I63-66) and prescription of antiplatelet or anticoagulation agents. Hemorrhagic stroke was defined according to the ICD-10 code (I60-62). The presence of ischemic heart disease was defined according to the ICD-10 code (I20-25) and the prescription of antiplatelet or anticoagulation agents. Depression was defined according to the ICD-10 code (F32-34). Osteoporosis was defined according to the ICD-10 code (M80-82). Obesity was defined as a BMI ≥ 25 kg/m2. Physical inactivity was defined as the absence of physical activity, even once a week. Smoking status was grouped into three levels: never smoked, ex-smoker, and current smoker. Heavy alcohol consumption was defined as alcohol consumption more than three times per week.
## 2.4. Definition of outcome and follow-Up
The outcome of the study was the development of PD, which was defined according to the ICD-10 code (G20) and prescription of PD medication for ≥3 months. Furthermore, to exclude secondary parkinsonism and atypical parkinsonism, such as progressive supranuclear palsy and multiple system atrophy, we excluded participants who additionally had ICD-10 code (G21-23) after diagnosis of PD from the outcome. Participants without PD during follow-up were considered to have completed the study on the date of death or at the end of follow-up. The patients were followed up from the date of the general health examinations (baseline) to the date of PD diagnosis, date of death, or until December 2019.
## 2.5. Statistical analyses
Baseline characteristics are presented as mean ± standard deviation or median (interquartile range) and frequency (%). First, in the PD-free cohort, we calculated the PD incidence rates and confidence intervals of the incidence rates under the assumption that the number of outcomes follows a Poisson distribution. Second, to investigate the modifiable risk factors for PD, we calculated HR using the Cox regression model, including each modifiable risk factor as a separate predictor after controlling for age and sex (model 1). We further performed the Cox regression model including modifiable risk factors as predictors that showed statistical significance in model 1 after controlling for age and sex (model 2). Third, we calculated the RR using log-binomial regression to adjust for age, sex, and modifiable risk factors (McNutt et al., 2003). These models included the risk factors that showed statistical significance in Cox regression model 1 and were controlled for age and sex. Finally, the population-attributable fraction (PAF) was calculated using Levin’s formula: With respect to the prevalence of modifiable risk factors, we considered the prevalence in our cohort as presented in Table 1. To identify the combined effects of risk factors, we obtained the overall PAFs using the following formula: Sensitivity analyses were used to exclude participants with stroke to eliminate the mediating or confounding effects of stroke on the relationship between the modifiable risk factors and the development of PD.
**Table 1**
| Characteristics at initial visit | Total (n = 938,635) | PD-free at follow-up (n = 928,711) | Incident PD at follow-up (n = 9,924) |
| --- | --- | --- | --- |
| Sex | | | |
| Male (%) | 475609.0 | 471,207 (50.7%) | 4,402 (44.4%) |
| Female (%) | 463026.0 | 457,504 (49.3%) | 5,522 (55.6%) |
| Age groups, y | | | |
| ≤50 | 155240.0 | 154,895 (16.7%) | 345 (3.5%) |
| 51–55 | 216621.0 | 215,862 (23.2%) | 759 (7.7%) |
| 56–60 | 169606.0 | 168,432 (18.1%) | 1,174 (11.8%) |
| 61–65 | 138092.0 | 136,329 (14.7%) | 1,763 (17.8%) |
| 66–70 | 121046.0 | 118,542 (12.8%) | 2,504 (25.2%) |
| 71–75 | 77919.0 | 75,890 (8.2%) | 2,029 (20.5%) |
| 76–80 | 38910.0 | 37,890 (4.1%) | 1,020 (10.3%) |
| 81–85 | 15687.0 | 15,407 (1.7%) | 280 (2.8%) |
| ≥86 | 5514.0 | 5,464 (0.6%) | 50 (0.5%) |
| Hypertension (%) | 344839.0 | 339,625 (36.6%) | 5,214 (52.5%) |
| Diabetes (%) | 97162.0 | 95,514 (10.3%) | 1,648 (16.6%) |
| Dyslipidemia (%) | 139135.0 | 137,049 (14.8%) | 2,086 (21.0%) |
| Ischemic stroke (%) | 33408.0 | 32,576 (3.5%) | 832 (8.4%) |
| Hemorrhagic stroke (%) | 8538.0 | 8,362 (0.9%) | 176 (1.8%) |
| Ischemic heart disease (%) | 92466.0 | 53,126 (5.7%) | 1,039 (10.5%) |
| Depression (%) | 107609.0 | 105,493 (11.36%) | 2,116 (21.3%) |
| Osteoporosis (%) | 199675.0 | 196,332 (21.1%) | 3,343 (33.7%) |
| Obesity (%) | 340780.0 | 337,065 (36.3%) | 3,715 (37.4%) |
| Physical inactivity (%) | 513526.0 | 507,495 (54.7%) | 6,031 (60.8%) |
| Heavy alcohol consumption (%) | 100300.0 | 99,374 (10.7%) | 926 (9.3%) |
| Smoking | | | |
| Never | 672659.0 | 664,801 (71.6%) | 7,858 (79.2%) |
| Ex-smoker | 86394.0 | 85,671 (9.2%) | 723 (7.3%) |
| Current smoker | 179582.0 | 178,239 (19.2%) | 1,342 (13.5%) |
All reported p-values were two-sided and the significance level was set at 0.05. All analyses were performed using SAS (version 9.3; SAS Institute Inc., Cary, NC, United States).
## 3.1. Clinical characteristics of the study participants at baseline
Among the 938,635 participants in the PD and dementia-free cohorts, 463,026 ($49.3\%$) were women. The most prevalent modifiable risk factor was physical inactivity ($54.7\%$), followed by hypertension ($36.7\%$), obesity ($36.3\%$), osteoporosis ($21.3\%$), and current smoking ($19.1\%$, Table 1). Subjects who developed PD were more likely to be and have hypertension, diabetes, dyslipidemia, ischemic heart disease, ischemic stroke, hemorrhagic stroke, depression, and osteoporosis than those who were PD-free at follow-up.
## 3.2. PD incidence by year, age, and sex
During follow-up, 9,924 of the 938,635 ($1.1\%$) participants developed PD. The incidence rate of PD showed annual growth, increasing from 0.56 per 1,000 person-years in 2006 to 1.34 per 1,000 person-years in 2018 (Table 2). The incidence rate of PD also increases with age, up to 80 y. Specifically, the incidence rate of PD was only 0.19 per 1,000 person-years among participants who were 50 y and less, while that of PD increased to 2.91 per 1,000 person-years among participants who were 76 to 80 y old (Table 3). In terms of sex, women (9,447; $54.4\%$) were more likely to develop PD than men (7,910; $45.57\%$). The incidence rate of PD in women (1.05 per 1,000 person-years) was higher than that in men (0.84 per 1,000 person-years, Table 3).
## 3.3. Modifiable risk factors for PD
In model 1, presence of hypertension (subdistribution hazard ratio [SHR] = 1.27, $95\%$ confidence interval [CI] 1.22 to 1.33), diabetes (SHR = 1.39, $95\%$ CI 1.32 to 1.47), dyslipidemia (SHR = 1.34, $95\%$ CI 1.28 to 1.41), ischemic stroke (SHR = 1.58, $95\%$ CI 1.47 to 1.70), hemorrhagic stroke (SHR = 1.54, $95\%$ CI 1.32 to 1.78), ischemic heart disease (SHR = 1.38, $95\%$ CI 1.30 to 1.47), depression (SHR = 1.76, $95\%$ CI 1.68 to 1.85), osteoporosis (SHR = 1.34, $95\%$ CI 1.28 to 1.41), and obesity (SHR = 1.12, $95\%$ CI 1.07 to 1.16) increased the risk of PD (Table 4). In model 2, presence of hypertension (SHR = 1.09, $95\%$ CI 1.05 to 1.14), diabetes (SHR = 1.24, $95\%$ CI 1.17 to 1.31), dyslipidemia (SHR = 1.12, $95\%$ CI 1.07 to 1.18), ischemic stroke (SHR = 1.26, $95\%$ CI 1.17 to 1.36), hemorrhagic stroke (SHR = 1.26, $95\%$ CI 1.08 to 1.47), ischemic heart disease (SHR = 1.09, $95\%$ CI 1.02 to 1.17), depression (SHR = 1.61, $95\%$ CI 1.53 to 1.69), osteoporosis (SHR = 1.24, $95\%$ CI 1.18 to 1.30), and obesity (SHR = 1.06, $95\%$ CI 1.01 to 1.10) remained independently associated with a higher risk for PD (Table 4).
**Table 4**
| Unnamed: 0 | Model 1* | Model 1*.1 | Model 2# | Model 2#.1 |
| --- | --- | --- | --- | --- |
| | HR (95% CI) | p | HR# (95% CI) | p |
| Hypertension | 1.27 (1.22–1.33) | <0.001 | 1.09 (1.05–1.14) | <0.001 |
| Diabetes | 1.39 (1.32–1.47) | <0.001 | 1.24 (1.17–1.31) | <0.001 |
| Dyslipidemia | 1.34 (1.28–1.41) | <0.001 | 1.12 (1.07–1.18) | <0.001 |
| Ischemic stroke | 1.58 (1.47–1.70) | <0.001 | 1.26 (1.17–1.36) | <0.001 |
| Hemorrhagic stroke | 1.54 (1.32–1.78) | <0.001 | 1.26 (1.08–1.47) | 0.003 |
| Ischemic heart disease | 1.38 (1.30–1.47) | <0.001 | 1.09 (1.02–1.17) | 0.012 |
| Depression | 1.76 (1.68–1.85) | <0.001 | 1.61 (1.53–1.69) | <0.001 |
| Osteoporosis | 1.34 (1.28–1.41) | <0.001 | 1.24 (1.18–1.30) | <0.001 |
| Obesity | 1.12 (1.07–1.16) | <0.001 | 1.06 (1.01–1.10) | 0.011 |
| Physical inactivity | 0.99 (0.95–1.03) | 0.544 | | |
| Heavy alcohol consumption | 0.96 (0.89–1.03) | 0.247 | | |
| Smoking | | | | |
| Never | Reference | | Reference | |
| Ex-smoker | 0.82 (0.76–0.89) | <0.001 | 0.82 (0.76–0.89) | <0.001 |
| Current smoker | 0.76 (0.74–0.83) | <0.001 | 0.81 (0.76–0.86) | <0.001 |
## 3.4. Population-attributable fraction for PD
As presented in Table 5, among the modifiable risk factors, depression had the greatest impact on PD (PAF, $6.5\%$), followed by osteoporosis (PAF, $4.8\%$), and hypertension (PAF, $3.3\%$). The overall PAF of the modifiable risk factors was $20.4\%$.
**Table 5**
| Unnamed: 0 | Risk factor prevalence | Parkinson’s disease | Parkinson’s disease.1 |
| --- | --- | --- | --- |
| | Risk factor prevalence | Relative risk* (95% CI) | PAF (95% CI) |
| Hypertension | 36.7% | 1.09 (1.05–1.14) | 3.3% (1.6–5.0) |
| Diabetes | 10.4% | 1.24 (1.17–1.31) | 2.4% (1.8–3.1) |
| Dyslipidemia | 14.8% | 1.12 (1.07–1.18) | 1.8% (1.0–2.6) |
| Ischemic stroke | 3.6% | 1.26 (1.17–1.36) | 0.9% (0.6–1.3) |
| Hemorrhagic stroke | 0.9% | 1.26 (1.08–1.47) | 0.2% (0.1–0.4) |
| Ischemic heart disease | 5.8% | 1.09 (1.02–1.17) | 0.5% (0.1–1.0) |
| Depression | 11.5% | 1.61 (1.53–1.69) | 6.5% (5.7–7.3) |
| Osteoporosis | 21.3% | 1.24 (1.18–1.30) | 4.8% (3.7–6.0) |
| Obesity | 36.3% | 1.06 (1.01–1.10) | 2.0% (0.5–3.5) |
| Overall PAF | | | 20.4% |
## 3.5. Sensitivity analyses
Among the participants without stroke, hypertension (SHR = 1.09, $95\%$ CI 1.03 to 1.13), diabetes (SHR = 1.24, $95\%$ CI 1.17 to 1.32), dyslipidemia (SHR = 1.14, $95\%$ CI 1.08 to 1.21), ischemic heart disease (SHR = 1.13, $95\%$ CI 1.04 to 1.22), depression (SHR = 1.68, $95\%$ CI 1.59 to 1.77), osteoporosis (SHR = 1.25, $95\%$ CI 1.19 to 1.31), and obesity (SHR = 1.05, $95\%$ CI 1.00 to 1.10) were independently associated with a higher risk for PD (Supplementary Table 1).
## 4. Discussion
In the present study, we identified the incidence and modifiable risk factors of PD using the Korean nationwide cohort data. The major findings of this study are as follows. First, the incidence of PD increased continuously from 2007 to 2018, reaching 1.34 per 1,000 person-years in 2018. Second, the incidence of PD increases with age, up to 80 y. Third, cardiometabolic syndromes, depression, and osteoporosis are associated with a higher incidence of PD, independent of stroke. Overall, our results will help in the design of public health policies for PD prevention.
Our first major finding was the increasing trend in the incidence of PD in South Korea from 2007 to 2018. Trends in PD incidence vary depending on the study design, population, and period. Stable or slightly decreasing trends have been reported in Western countries, such as the United States, the United Kingdom, France, and the Netherlands during the 2010s (Akushevich et al., 2013; Horsfall et al., 2013; Blin et al., 2015; Darweesh et al., 2016; Evans et al., 2016). Conversely, several studies have reported an annual increase in PD incidence (Liu et al., 2016; Savica et al., 2016). In the Minnesota population aged ≥70 years old and older, the incidence of PD increased from 0.80 per 1,000 person-years to 1.37 per 1,000 person-years over 30 y (Savica et al., 2016). An increasing trend has also been identified in Taiwan, which is included in the far-eastern Asian countries along with Korea (Liu et al., 2016). This increasing trend may be attributed to better recognition of PD in older patients with comorbidities. In recent years, physicians have begun to diagnose elderly individuals with cancer, cardiovascular diseases, or other conditions as having PD, because parkinsonism symptoms have become important in the overall clinical outcome and are considered one of the major causes of disability and mortality. In addition, the KNHIS started to cover dopamine transporter images in 2016, and consequently, the diagnosis of PD became relatively simplified, which could contribute to an increasing point of PD incidence in 2016. The increasing incidence of PD may also be explained by the increase in the prevalence of modifiable risk factors for PD, such as hypertension and dyslipidemia, and the dramatic decrease in the rate of smoking, a protective factor for PD, in Korea (Korea Health Statistics 2019, Korea National Health and Nutrition Examination Survey,2).
Our second major finding was that the incidence of PD increased with age, up to 80 y. It is well known that PD prevalence is low (0.13–$1.6\%$) in populations aged less than 60 y, after which there is a sharp increase in incidence (Kis et al., 2002; Benito-León et al., 2003; Chan et al., 2005; Blin et al., 2015). These results are consistent with our findings that PD incidence was only 0.36 per 1,000 person-years among participants aged 60 y or less, whereas PD incidence increased to 2.91 per 1,000 person-years among those aged 76–80 y. Although age may be an important risk factor for PD, peak age-specific incidence varies among studies. Several studies have reported that PD incidence uniformly increases up to the ninth decade (Allyson Jones et al., 2012; Caslake et al., 2013; Blin et al., 2015), whereas others have found a decline in PD incidence among the oldest old population. We also found that the age-specific incidence peaked in the population aged 76–80 y and declined beyond this age. However, a direct comparison between the studies is difficult because of the small sample sizes in the oldest old group and varying definitions (von Campenhausen et al., 2005). The decline in the oldest old population may be due to the following reasons. First, the high burden of comorbidities, such as dementia and musculoskeletal disease, in the oldest old group increases the diagnostic uncertainty for PD (Meara et al., 1999). Second, individuals with PD at the oldest old age may not use a medical institution (Bowling et al., 1991). Third, mortality selection may cause unobserved heterogeneity within the oldest old group, which in turn determines the ratio of individuals with and without PD in favor of the latter group (Vaupel et al., 1979).
Contrary to our expectations, we observed that PD incidence in women was significantly higher than that in men. A growing body of evidence shows a predominance of PD incidence in the male population (Baldereschi et al., 2000; Clavería et al., 2002; Benito-León et al., 2003; Alves et al., 2009; Nerius et al., 2017) or no sex difference in PD incidence (Linder et al., 2010; Winter et al., 2010). However, a few Asian studies have reported female predominance in PD (Kimura et al., 2002; Park J. H. et al., 2019). Although the underlying mechanisms for this discrepancy remain unclear, genetic, hormonal, cultural, and environmental factors may mediate such outcomes. First, Asian women may have different genetic susceptibilities to PD. Second, substantial differences in cultural and environmental factors during childhood in the elderly population (in the mid 1900s) in Korea may cause sex-related disparities in educational levels and literacy rates associated with brain reserves. Furthermore, Korean women may encounter more risk factors, including dietary deficiencies, agricultural occupations, pesticide use, and head trauma. Finally, the greater average longevity of women in Korea may lead to a quicker increase in the elderly female population, which in turn causes female predominance. In fact, the sex-specific difference in life expectancy was higher in South Korea (men, 79.7 y and women, 85.7 y) than in European countries (men, 79.7 y and women, 82.8 y) according to the Office for National Statistics.
Our third major finding was that cardiometabolic syndrome, depression, and osteoporosis were associated with a higher incidence of PD, independent of stroke. In particular, among cardiometabolic syndromes, diabetes contributes the most to the development of PD, which is consistent with previous findings that diabetes is associated with a higher PD risk (Hu et al., 2007; Driver et al., 2008; Schernhammer et al., 2011; Xu et al., 2011; Sun et al., 2012). However, the association between hypertension, dyslipidemia, obesity, and PD remains controversial. Many previous studies in Western countries have shown that hypertension, dyslipidemia, and obesity are not associated with a higher risk of PD (Abbott et al., 2002; Logroscino et al., 2007; Simon et al., 2007; Kyrozis et al., 2013; Ascherio and Schwarzschild, 2016), whereas a few studies have reported that dyslipidemia and obesity may be risk factors for PD (Hu et al., 2006, 2008). These discordant results may be related to the modest effect of cardiometabolic syndrome on incident PD, or undiscerned confounding or modifying factors that modulate the relationship between cardiometabolic syndrome and PD risk. Another explanation for this discrepancy is ethnic differences in the effects of cardiometabolic syndromes on PD. Compared to European populations, Asian populations have a higher incidence of cardiometabolic syndrome (Yoon et al., 2006) and associated complications, including coronary artery disease (McKeigue et al., 1989), stroke (Eastwood et al., 2015), dementia (Niu et al., 2017; Park J. E. et al., 2019; Jang et al., 2021), and high mortality rates (Wild et al., 2007). Additionally, studies have shown that Asian populations tend to have higher visceral fat and lower subcutaneous fat than European populations with similar BMI (Nazare et al., 2012). Such unequal fat distribution may be associated with more severe cardiometabolic complications commonly seen in Asian populations, given that visceral fat is associated with arteriosclerosis and compromised brain health (Debette et al., 2010; Isaac et al., 2011; Kato et al., 2011; Widya et al., 2015).
We also found that depression and osteoporosis increased PD risk, suggesting that depression and osteoporosis preceded the diagnosis of PD. Although depression is a well-known premotor symptom (Tolosa et al., 2007), the temporal relationship between osteoporosis and PD remains poorly understood. Contrary to previous findings of a greater risk of osteoporosis in patients with PD (Torsney et al., 2014), the present study revealed that patients with osteoporosis were at a high risk of PD. Although further studies should be necessary to identify the mechanism, our findings suggested that osteoporosis and PD might have an interactive relationship or that osteoporosis might be a veiled premotor symptom of PD.
Our study has several limitations that should be addressed. First, the discordance between the diagnosis of PD in clinical practice and that recorded in the KNHIS may have led to inaccurate results. However, these issues could be mitigated by the fact that the diagnostic code of PD was classified into the registration code in the program for rare intractable diseases to increase diagnostic accuracy. Additionally, we added the prescription of PD medication for ≥3 months as the outcome definition. Second, we could not define cardiometabolic syndrome using blood pressure measurements and laboratory data such as fasting glucose and total cholesterol levels. Third, we could not assess the exposure time and changes in the risk factors and risk factors that occurred after 2006. Fourth, we could not consider the potential effect of antihypertensive, antidiabetic, and lipid-lowering medication. Fifth, due to a lack of information on a positive family history, traumatic brain injury, exposure to pesticides, dietary patterns, air pollution, and social isolation, we did not identify the PAF of these factors.
Despite the aforementioned limitations, our study aimed to identify age-and sex-specific PD incidence based on a nationwide cohort that included a larger number of elderly participants. In addition, our results highlight the effect of modifiable risk factors for PD in the Korean population, which will help establish health care policies to prevent the development of PD.
## Data availability statement
The datasets presented in this article are not readily available because The Korean NHIS database is confidential and approved for use by researchers who meet the criteria for access through the Korea National Health Insurance Sharing Service (NHISS) Institutional Data Access Committee (https://nhiss.nhis.or.kr/bd/ay/bdaya001iv.do). If data are requested for additional analysis, the corresponding author would consider it deliberately to offer after passing the review process of the Korea NHISS Institutional Data Access Committee and after payment of the data access fee charged to the requester. Requests to access the datasets should be directed to S-BK, [email protected].
## Ethics statement
The studies involving human participants were reviewed and approved by this study was approved by the Institutional Review Board of the Korea University Guro Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
SK analyzed and interpreted the data and drafted the manuscript for intellectual content. S-JM and MK analyzed and interpreted the data. SC and GC played major roles in data acquisition. S-BK acquired the data, designed and conceptualized the study, and revised the manuscript for intellectual content. All authors contributed to the article and approved the submitted version.
## Funding
This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (grant number: 2022R1I1A1A01056956) and Korea University Guro Hospital (KOREA RESEARCH-DRIVEN HOSPITAL; grant number: O2208241).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnagi.2023.1094778/full#supplementary-material
## References
1. Abbott R. D., Ross G. W., White L. R., Nelson J. S., Masaki K. H., Tanner C. M.. **Midlife adiposity and the future risk of Parkinson's disease**. *Neurology* (2002) **59** 1051-1057. DOI: 10.1212/wnl.59.7.1051
2. Akushevich I., Kravchenko J., Ukraintseva S., Arbeev K., Yashin A. I.. **Time trends of incidence of age-associated diseases in the US elderly population: medicare-based analysis**. *Age Ageing* (2013) **42** 494-500. DOI: 10.1093/ageing/aft032
3. Allyson Jones C., Wayne Martin W. R., Wieler M., King-Jesso P., Voaklander D. C.. **Incidence and mortality of Parkinson's disease in older Canadians**. *Parkinsonism Relat. Disord.* (2012) **18** 327-331. DOI: 10.1016/j.parkreldis.2011.11.018
4. Alves G., Müller B., Herlofson K., HogenEsch I., Telstad W., Aarsland D.. **Incidence of Parkinson's disease in Norway: the Norwegian ParkWest study**. *J. Neurol. Neurosurg. Psychiatry* (2009) **80** 851-857. DOI: 10.1136/jnnp.2008.168211
5. Ascherio A., Schwarzschild M. A.. **The epidemiology of Parkinson's disease: risk factors and prevention**. *Lancet Neurol.* (2016) **15** 1257-1272. DOI: 10.1016/s1474-4422(16)30230-7
6. Bach J. P., Ziegler U., Deuschl G., Dodel R., Doblhammer-Reiter G.. **Projected numbers of people with movement disorders in the years 2030 and 2050**. *Move. disorders Off. J. Move. Disord. Soc.* (2011) **26** 2286-2290. DOI: 10.1002/mds.23878
7. Baldereschi M., Di Carlo A., Rocca W. A., Vanni P., Maggi S., Perissinotto E.. **Parkinson's disease and parkinsonism in a longitudinal study: two-fold higher incidence in men**. (2000) **55** 1358-1363. DOI: 10.1212/wnl.55.9.1358
8. Benito-León J., Bermejo-Pareja F., Morales-González J. M., Porta-Etessam J., Trincado R., Vega S.. **Incidence of Parkinson disease and parkinsonism in three elderly populations of Central Spain**. *Neurology* (2004) **62** 734-741. DOI: 10.1212/01.wnl.0000113727.73153.68
9. Benito-León J., Bermejo-Pareja F., Rodríguez J., Molina J. A., Gabriel R., Morales J. M.. **Prevalence of PD and other types of parkinsonism in three elderly populations of Central Spain**. *Move. Disord. Off. J. Move. Disord. Soc.* (2003) **18** 267-274. DOI: 10.1002/mds.10362
10. Blin P., Dureau-Pournin C., Foubert-Samier A., Grolleau A., Corbillon E., Jové J.. **Parkinson's disease incidence and prevalence assessment in France using the national healthcare insurance database**. *Eur. J. Neurol.* (2015) **22** 464-471. DOI: 10.1111/ene.12592
11. Bowling A., Farquhar M., Browne P.. **Use of services in old age: data from three surveys of elderly people**. *Soc. Sci. Med.* (1991) **33** 689-700. DOI: 10.1016/0277-9536(91)90023-6
12. Caslake R., Taylor K., Scott N., Gordon J., Harris C., Wilde K.. **Age-, gender-, and socioeconomic status-specific incidence of Parkinson's disease and parkinsonism in Northeast Scotland: the PINE study**. *Parkinsonism Relat. Disord.* (2013) **19** 515-521. DOI: 10.1016/j.parkreldis.2013.01.014
13. Chan D. K., Cordato D., Karr M., Ong B., Lei H., Liu J.. **Prevalence of Parkinson's disease in Sydney**. *Acta Neurol. Scand.* (2005) **111** 7-11. DOI: 10.1111/j.1600-0404.2004.00348.x
14. Clavería L. E., Duarte J., Sevillano M. D., Pérez-Sempere A., Cabezas C., Rodríguez F.. **Prevalence of Parkinson's disease in Cantalejo, Spain: a door-to-door survey**. *Move. Disord. Off. J. Move. Disord. Soc.* (2002) **17** 242-249. DOI: 10.1002/mds.10087
15. Darweesh S. K., Koudstaal P. J., Stricker B. H., Hofman A., Ikram M. A.. **Trends in the incidence of Parkinson disease in the general population: the Rotterdam study**. *Am. J. Epidemiol.* (2016) **183** 1018-1026. DOI: 10.1093/aje/kwv271
16. de Lau L. M., Giesbergen P. C., de Rijk M. C., Hofman A., Koudstaal P. J., Breteler M. M.. **Incidence of Parkinsonism and Parkinson disease in a general population: the Rotterdam study**. *Neurology* (2004) **63** 1240-1244. DOI: 10.1212/01.wnl.0000140706.52798.be
17. Debette S., Beiser A., Hoffmann U., DeCarli C., O'Donnell C. J., Massaro J. M.. **Visceral fat is associated with lower brain volume in healthy middle-aged adults**. *Ann. Neurol.* (2010) **68** 136-144. DOI: 10.1002/ana.22062
18. Driver J. A., Logroscino G., Gaziano J. M., Kurth T.. **Incidence and remaining lifetime risk of Parkinson disease in advanced age**. *Neurology* (2009) **72** 432-438. DOI: 10.1212/01.wnl.0000341769.50075.bb
19. Driver J. A., Smith A., Buring J. E., Gaziano J. M., Kurth T., Logroscino G.. **Prospective cohort study of type 2 diabetes and the risk of Parkinson's disease**. *Diabetes Care* (2008) **31** 2003-2005. DOI: 10.2337/dc08-0688
20. Eastwood S. V., Tillin T., Chaturvedi N., Hughes A. D.. **Ethnic differences in associations between blood pressure and stroke in south Asian and European men**. *Hypertension* (2015) **66** 481-488. DOI: 10.1161/hypertensionaha.115.05672
21. Evans J. R., Cummins G., Breen D. P., Foltynie T., Mason S. L., Brayne C. E.. **Comparative epidemiology of incident Parkinson's disease in Cambridgeshire, UK**. *J. Neurol. Neurosurg. Psychiatry* (2016) **87** 1034-1036. DOI: 10.1136/jnnp-2015-312581
22. Horsfall L., Petersen I., Walters K., Schrag A.. **Time trends in incidence of Parkinson's disease diagnosis in UK primary care**. *J. Neurol.* (2013) **260** 1351-1357. DOI: 10.1007/s00415-012-6804-z
23. Hu G., Antikainen R., Jousilahti P., Kivipelto M., Tuomilehto J.. **Total cholesterol and the risk of Parkinson disease**. *Neurology* (2008) **70** 1972-1979. DOI: 10.1212/01.wnl.0000312511.62699.a8
24. Hu G., Jousilahti P., Bidel S., Antikainen R., Tuomilehto J.. **Type 2 diabetes and the risk of Parkinson's disease**. *Diabetes Care* (2007) **30** 842-847. DOI: 10.2337/dc06-2011
25. Hu G., Jousilahti P., Nissinen A., Antikainen R., Kivipelto M., Tuomilehto J.. **Body mass index and the risk of Parkinson disease**. *Neurology* (2006) **67** 1955-1959. DOI: 10.1212/01.wnl.0000247052.18422.e5
26. Isaac V., Sim S., Zheng H., Zagorodnov V., Tai E. S., Chee M.. **Adverse associations between visceral adiposity, brain structure, and cognitive performance in healthy elderly**. *Front. Aging Neurosci.* (2011) **3** 12. DOI: 10.3389/fnagi.2011.00012
27. Jang J.-W., Park J. H., Kim S., Lee S.-H., Lee S.-H., Kim Y.-J.. **Prevalence and incidence of dementia in South Korea: a Nationwide analysis of the National Health Insurance Service senior cohort**. *J. Clin. Neurol.* (2021) **17** 249-256. DOI: 10.3988/jcn.2021.17.2.249
28. Kato A., Ishida J., Endo Y., Takita T., Furuhashi M., Maruyama Y.. **Association of abdominal visceral adiposity and thigh sarcopenia with changes of arteriosclerosis in haemodialysis patients**. *Nephrol. Dialysis Transpl.* (2011) **26** 1967-1976. DOI: 10.1093/ndt/gfq652
29. Kimura H., Kurimura M., Wada M., Kawanami T., Kurita K., Suzuki Y.. **Female preponderance of Parkinson's disease in Japan**. *Neuroepidemiology* (2002) **21** 292-296. DOI: 10.1159/000065527
30. Kis B., Schrag A., Ben-Shlomo Y., Klein C., Gasperi A., Spoegler F.. **Novel three-stage ascertainment method: prevalence of PD and Parkinsonism in South Tyrol, Italy**. *Neurology* (2002) **58** 1820-1825. DOI: 10.1212/wnl.58.12.1820
31. Kyrozis A., Ghika A., Stathopoulos P., Vassilopoulos D., Trichopoulos D., Trichopoulou A.. **Dietary and lifestyle variables in relation to incidence of Parkinson's disease in Greece**. *Eur. J. Epidemiol.* (2013) **28** 67-77. DOI: 10.1007/s10654-012-9760-0
32. Linder J., Stenlund H., Forsgren L.. **Incidence of Parkinson's disease and parkinsonism in northern Sweden: a population-based study**. *Move. Disord. Off. J. Move. Disord. Soc.* (2010) **25** 341-348. DOI: 10.1002/mds.22987
33. Liu C. C., Li C. Y., Lee P. C., Sun Y.. **Variations in incidence and prevalence of Parkinson's disease in Taiwan: a population-based Nationwide study**. *Park. Dis.* (2016) **2016** 8756359. DOI: 10.1155/2016/8756359
34. Logroscino G., Sesso H. D., Paffenbarger R. S., Lee I. M.. **Body mass index and risk of Parkinson's disease: a prospective cohort study**. *Am. J. Epidemiol.* (2007) **166** 1186-1190. DOI: 10.1093/aje/kwm211
35. McKeigue P. M., Miller G. J., Marmot M. G.. **Coronary heart disease in south Asians overseas: a review**. *J. Clin. Epidemiol.* (1989) **42** 597-609. DOI: 10.1016/0895-4356(89)90002-4
36. McNutt L. A., Wu C., Xue X., Hafner J. P.. **Estimating the relative risk in cohort studies and clinical trials of common outcomes**. *Am. J. Epidemiol.* (2003) **157** 940-943. DOI: 10.1093/aje/kwg074
37. Meara J., Bhowmick B. K., Hobson P.. **Accuracy of diagnosis in patients with presumed Parkinson's disease**. *Age Ageing* (1999) **28** 99-102. DOI: 10.1093/ageing/28.2.99
38. Nazare J. A., Smith J. D., Borel A. L., Haffner S. M., Balkau B., Ross R.. **Ethnic influences on the relations between abdominal subcutaneous and visceral adiposity, liver fat, and cardiometabolic risk profile: the international study of prediction of intra-abdominal adiposity and its relationship with Cardiometabolic risk/intra-abdominal adiposity**. *Am. J. Clin. Nutr.* (2012) **96** 714-726. DOI: 10.3945/ajcn.112.035758
39. Nerius M., Fink A., Doblhammer G.. **Parkinson's disease in Germany: prevalence and incidence based on health claims data**. *Acta Neurol. Scand.* (2017) **136** 386-392. DOI: 10.1111/ane.12694
40. Niu H., Álvarez-Álvarez I., Guillén-Grima F., Aguinaga-Ontoso I.. **Prevalence and incidence of Alzheimer's disease in Europe: a meta-analysis**. *Neurologia (Barcelona, Spain)* (2017) **32** 523-532. DOI: 10.1016/j.nrl.2016.02.016
41. Park J. E., Kim B. S., Kim K. W., Hahm B. J., Sohn J. H., Suk H. W.. **Decline in the incidence of all-cause and Alzheimer's disease dementia: a 12-year-later rural cohort study in Korea**. *J. Korean Med. Sci.* (2019) **34** e293. DOI: 10.3346/jkms.2019.34.e293
42. Park J. H., Kim D. H., Kwon D. Y., Choi M., Kim S., Jung J. H.. **Trends in the incidence and prevalence of Parkinson's disease in Korea: a nationwide, population-based study**. *BMC Geriatr.* (2019) **19** 320. DOI: 10.1186/s12877-019-1332-7
43. Savica R., Grossardt B. R., Bower J. H., Ahlskog J. E., Rocca W. A.. **Time trends in the incidence of Parkinson disease**. *JAMA Neurol.* (2016) **73** 981-989. DOI: 10.1001/jamaneurol.2016.0947
44. Schernhammer E., Hansen J., Rugbjerg K., Wermuth L., Ritz B.. **Diabetes and the risk of developing Parkinson's disease in Denmark**. *Diabetes Care* (2011) **34** 1102-1108. DOI: 10.2337/dc10-1333
45. Simon K. C., Chen H., Schwarzschild M., Ascherio A.. **Hypertension, hypercholesterolemia, diabetes, and risk of Parkinson disease**. *Neurology* (2007) **69** 1688-1695. DOI: 10.1212/01.wnl.0000271883.45010.8a
46. Sun Y., Chang Y. H., Chen H. F., Su Y. H., Su H. F., Li C. Y.. **Risk of Parkinson disease onset in patients with diabetes: a 9-year population-based cohort study with age and sex stratifications**. *Diabetes Care* (2012) **35** 1047-1049. DOI: 10.2337/dc11-1511
47. Taylor K. S., Counsell C. E., Harris C. E., Gordon J. C., Smith W. C.. **Pilot study of the incidence and prognosis of degenerative parkinsonian disorders in Aberdeen, United Kingdom: methods and preliminary results**. *Move. Disord. Off. J. Move. Disord. Soc.* (2006) **21** 976-982. DOI: 10.1002/mds.20866
48. Tolosa E., Compta Y., Gaig C.. **The premotor phase of Parkinson's disease**. *Parkinsonism Relat. Disord.* (2007) **13** S2-S7. DOI: 10.1016/j.parkreldis.2007.06.007
49. Torsney K. M., Noyce A. J., Doherty K. M., Bestwick J. P., Dobson R., Lees A. J.. **Bone health in Parkinson's disease: a systematic review and meta-analysis**. *J. Neurol. Neurosurg. Psychiatry* (2014) **85** 1159-1166. DOI: 10.1136/jnnp-2013-307307
50. Vaupel J. W., Manton K. G., Stallard E.. **The impact of heterogeneity in individual frailty on the dynamics of mortality**. *Demography* (1979) **16** 439-454. DOI: 10.2307/2061224
51. von Campenhausen S., Bornschein B., Wick R., Bötzel K., Sampaio C., Poewe W.. **Prevalence and incidence of Parkinson's disease in Europe**. *Euro. Neuropsychopharmacol. J. Euro. Coll. Neuropsychopharmacol.* (2005) **15** 473-490. DOI: 10.1016/j.euroneuro.2005.04.007
52. Widya R. L., Kroft L. J., Altmann-Schneider I., van den Berg-Huysmans A. A., van der Bijl N., de Roos A.. **Visceral adipose tissue is associated with microstructural brain tissue damage**. *Obesity* (2015) **23** 1092-1096. DOI: 10.1002/oby.21048
53. Wild S. H., Fischbacher C., Brock A., Griffiths C., Bhopal R.. **Mortality from all causes and circulatory disease by country of birth in England and Wales 2001-2003**. *J. Public Health* (2007) **29** 191-198. DOI: 10.1093/pubmed/fdm010
54. Winter Y., Bezdolnyy Y., Katunina E., Avakjan G., Reese J. P., Klotsche J.. **Incidence of Parkinson's disease and atypical Parkinsonism: Russian population-based study**. *Move. Disord. Off. J. Move. Disord. Soc.* (2010) **25** 349-356. DOI: 10.1002/mds.22966
55. Xu Q., Park Y., Huang X., Hollenbeck A., Blair A., Schatzkin A.. **Diabetes and risk of Parkinson's disease**. *Diabetes Care* (2011) **34** 910-915. DOI: 10.2337/dc10-1922
56. Yoon K. H., Lee J. H., Kim J. W., Cho J. H., Choi Y. H., Ko S. H.. **Epidemic obesity and type 2 diabetes in Asia**. *Lancet (London, England)* (2006) **368** 1681-1688. DOI: 10.1016/s0140-6736(06)69703-1
|
---
title: Analyzing risky behaviors among different minority and majority race in teenagers
in the USA using latent classes
authors:
- Zeeshan Aslam
- Muhammad Asim
- Iqra Javaid
- Faisal Rasheed
- Muhammad Naveed Akhter
journal: Frontiers in Behavioral Neuroscience
year: 2023
pmcid: PMC9971590
doi: 10.3389/fnbeh.2023.1089434
license: CC BY 4.0
---
# Analyzing risky behaviors among different minority and majority race in teenagers in the USA using latent classes
## Abstract
Objective: This study is to ascertain any inconsistencies in the trend of co-occurrence by sex of teenage health risk behavior patterns such as smoking, behaviors contributing to deliberate and unintentional injuries, risky sexual behavior, and sedentary lifestyle.
Methods: The study’s purpose was accomplished using Youth Risk Behavior Surveillance System (YRBSS) 2013 data. A Latent Class Analysis (LCA) was conducted for the entire sample of teenagers as well as separately for each sex.
Results: *In this* subset of youths, marijuana use was acknowledged by more than half of them, and smoking cigarettes was far more likely. More than half of the individuals in this subset engaged in risky sexual practices, like not using a condom during their most recent encounter. Males were split into three categories based on their involvement in risky behavior, while females were split into four subgroups.
Conclusion: Regardless of gender, various risk behaviors among teenagers are connected. However, gender variations in the higher risk of particular trends, such as mood disorders and depression among females, underline the significance of creating treatments that take adolescent demographics into account.
## Introduction
Adolescence and young adults in the United States are typically deemed to be in good health according to traditional indicators, including mortality, chronic illness, and hospitalization rates (Mulye et al., 2009; Shaw et al., 2015). According to research on the health condition of teenagers in the United States, the primary health dangers to adolescents are based on their behavioral decisions (Oh et al., 2019, 2022). Teenagers who are involved in unsupervised sexual activities have an increased risk of contracting STIs (Sexually Transmitted Infections) like HIV (Human Immunodeficiency Virus) and becoming pregnant unintentionally (Hamilton et al., 2010; Cavazos-Rehg et al., 2013). Adolescent risk behavior includes smoking, drinking alcohol, using illegal drugs, and engaging in risky sexual behavior, which can have negative physiological condition effects like unwanted pregnancy and sexually transmitted illnesses. This class also includes activities that expose participants to aggression and unintended harm (Bozzini et al., 2021). The individual’s personal well-being as well as his or her interactions with loved ones may suffer major effects if a continuous trend of harmful activity is not identified early on and adequately managed, even when these actions are often inconsistent (Tsevat et al., 2017).
Understanding teenage health-risk behaviors is crucial for implementing policies and creating successful preventative initiatives. Adolescent fatalities are most frequently brought on by accidental accidents ($48\%$) and substance abuse (cigarettes, liquor, and other drug use). Teenagers who engage in unsupervised sexual activity have an increased risk of contracting STIs like HIV and becoming pregnant unintentionally. According to recent statistics, children between the ages of 15 and 19 experience an estimated 329,772 births, 548,032 instances of chlamydia, gonorrhea, and syphilis, and 2,240 cases of HIV per year (Chesson et al., 2010). Infertility, ectopic pregnancy, premature births, inflammation of the pelvis, and fetal abnormalities are among the harmful health effects of STIs in adolescents. Unwanted adolescent pregnancies may not only exacerbate socioeconomic limitations such as lower socioeconomic position and educational achievement but also raise the likelihood of adverse baby and mother health outcomes.
According to research, teenagers who are overweight or obese are more vulnerable to engaging in dangerous activities than their peers who are of a healthy weight (Farhat et al., 2010). In the US (United States), $32\%$ of children ages 2–19 are overweight or obese. Poor eating practices, unhealthful weight control, and physical inactivity are examples of harmful behaviors that are established throughout adolescence and frequently followed into adulthood. Early sexual activity, having several partners, and unprotected sexual contact are all linked to liquor consumption (Fergusson and Lynskey, 1996). For instance, concurrent marijuana or drug use has been linked to the initiation of smoking (Wills et al., 2004), and drug use raises the chance of contemplating suicide (Park et al., 2006). Drug use and delinquent behavior have been linked to an increased risk of depression (Costello et al., 2008). The six high-risk behaviors that are the main causes of illness and death among children and young people in the US have been identified by the CDC (Centers for Disease Control). Some of the more prevalent ones are smoking, liquor and other drug use, sexual practices that increase the risk of unwanted pregnancy and STIs, bad eating habits, and inactivity.
Cigarette usage is highlighted as a significant public health challenge, particularly cigarette smoking (Costello et al., 2008; Owen and Halford, 2007). Smoking products accelerate the deterioration of lung function in children and teenagers. Smoking is linked to slowed lung development, persistent coughing, and wheezing. Although variations depending on other demographics are unknown, research findings have provided some indication that the trends or dangerous behaviors differ by gender (Centers for Disease Control and Prevention, 2014). $9\%$ of American students in grades 9 through 12 reported using smokeless cigarettes, according to the 2013 Youth Risk Behavior Surveillance System (YRBSS). Smokeless cigarette use is linked to a number of harmful health effects, such as gingival depression, nicotine obsession, and oral, laryngeal, and pharyngeal malignancies (MacArthur et al., 2012). Cigarette makers have launched a plethora of colored brands with tastes that resemble candies. In an effort to attract young people, cigarette companies have more recently launched a plethora of vibrant brands with candy-like flavors, as well as smokeless and spitless products. Adolescent usage of alcoholic drinks, marijuana, and other illegal substances has been linked to morbidity and death. These chemicals have been related to a number of negative social and economic outcomes, such as family breakdown, crime, school leavers, and joblessness (Olajide et al., 2022).
The majority of Hispanic students have consumed alcohol at some point in their lives. According to the CDC, $35\%$ of high school students in the United States currently use alcohol. Females are more prone than males to have consumed booze at some point in their lives ($68\%$ vs. $64\%$). Other high-risk behaviors (HRBs) in teenagers have been linked to alcohol consumption, including risky sexual behaviors (Khan et al., 2012; de la Haye et al., 2014), mental health disorders such as melancholy and suicidal behavior (Strachman et al., 2009; Fisher et al., 2018), and behaviors that Bleakley et al. [ 2017] mention. Liquor abuse, for example, has been associated with depression, suicidal thoughts and attempts, risky sexual behavior, aggressive behavior, poor academic or professional performance, and unexpected damage. Cannabis usage among high school students varies by race and ethnicity, with black and Hispanic students using it more frequently than white students (Owen and Halford, 2007). Usage of cannabis has been linked to aggressive conduct, delinquency, sadness, and violent behavior.
The incidence of other illegal drug usage among teenagers across the nation is as follows: $5\%$ of teenagers say they have ever used cocaine in any form (powder, crack), $7\%$ have used hallucinogens, $9\%$ have used inhalants, $7\%$ have used ecstasy, $2\%$ have used heroin, $3\%$ have used methamphetamines, and $18\%$ have taken a prescription. $7\%$ of people report using hallucinogens. $18\%$ of people have used prescription medications without a prescription from a doctor. In the United States, accidental accidents are a significant cause of illness and death in adolescents and young people. Teenagers are more likely to drink liquor while driving, smoking, and engaging in risky sexual behavior (Parks et al., 2014). According to the most current statistics, one in four high school students said they had engaged in at least one violent altercation in the previous year (Khan et al., 2012). According to research, drug abuse is linked to aggressive conduct and actions that might cause harm (Weden and Zabin, 2005).
Gender refers to the array of socially constructed roles and relationships, personality traits, attitudes, behaviors, values, and relative power and influences that society ascribes to the two sexes on a differential basis. Simply put, sex refers to biological differences, whereas gender refers to social differences (Tsevat et al., 2017). Gender analysis in health has primarily been undertaken by social scientists who have discovered that biological differences alone cannot adequately explain health behavior. Health outcomes also depend on social and economic factors that, in turn, are influenced by cultural and political conditions in society (Khan et al., 2012).
The aim purpose of this study is to ascertain any inconsistencies in the trend of co-occurrence among different minority and majority races by sex of teenage health risk behavior patterns such as smoking, behaviors contributing to deliberate and unintentional injuries, risky sexual behavior, and sedentary lifestyle.
## Methods
The protocol for the YRBS was approved by the Institutional Review Board of the CDC. In order to preserve students’ privacy, survey protocols were created to enable voluntary, anonymous participation. Local parental approval procedures were followed prior to survey administration. The self-administered survey was completed by students during one class hour, and they wrote their answers directly on a computer-scannable booklet. Depression, a factor that influences hazardous and health-risk behaviors, was the main outcome factor. The performance measurement factors are examined as categorical variables in particular: [1] consensual sex with four or more people; [2] consensual sex with at least one person within the previous 3 months; [3] use of condoms; and [4] prior consumption of alcohol or drug use.
The YRBSS question “Sexuality?” served as the basis for our demographic variable, which was self-reported and quantified as a binary variable in the Latent class analysis (LCA). The descriptive statistics also contained data on age, academic level, and race/ethnic group. The individuals’ self-reported ages at the time the YRBSS was administered were used to compute chronological age. Based on two questions in the YRBSS, are you Hispanic? The racial group was assessed as a self-reported minimal variable. “ Yes/No” and “What race are you?” Asian; Black or African American; Native Hawaiian or another Pacific Islander; White, American Indian, or Alaska Native the survey question “At what academic level are you?” serves as a gauge of the respondents’ academic level at the time of the survey.
The YRBSS, a nationally representative sample of American teenagers in grades 9 through 12, provided the data for the study. The Centers for Disease Control and Prevention created YRBSS to track six types of high-risk behavior for health. Using cigarettes, liquor, or other drugs, engaging in sexual conduct, or making poor food choices fall under these categories. Asthma and obesity prevalence are also monitored by YRBSS. The YRBSS data is collected every other year, in odd-numbered years, and represents all high school students enrolled that year. To collect data, anonymous self-administered questionnaires are employed. An LCA was performed to look for potential sex-specific patterns of hazardous behavior. The latent class analysis employed a total of 40 manifest variables, or items, that covered several issue behavior categories. The goal was to include as many potential issue behaviors as feasible in this list, including risk-taking activities like chatting while operating a motor vehicle. Each record includes a weight factor based on student gender, racial group, and school grade to adjust for student non-response and oversampling of black and Hispanic children. The final overall weights are modified such that the total sample size equals the weighted count of students, and the weighted proportions of students in each grade correspond to the country’s demographic estimations for each survey year. The findings are based on weighted proportions of students in each grade and national population forecasts for each survey year. Students who declined to participate in the state surveys or the national YRBS were not replaced in the sampled courses, schools, or students. The response rate for schools was $77.3\%$, while the response rate for students was $87.8\%$, for a response rate of $67.6\%$. All variables in this study were dichotomized in accordance with the 2013 YRBSS Data User’s Manual (Youth Risk Behavior Survey, 2013). The percentage of students reporting participation or not engaging in a behavior was indicated by dichotomous variables. Based on a favorable reaction to engaging in a behavior once or more during the course of the previous 12 months or a recent month, a behavior was considered to be dangerous (Nylund et al., 2007).
The survey comprised 13,583 teens from US high schools. Figure 1 show that more than half of the respondents were Caucasian ($55.4\%$), and the majority were between the ages of 15 and 18 (Figure 2).
**Figure 1:** *Bar chart showing Race Ethnicity of Youth Risk Behavior Surveillance System (YRBSS) 2013 survey.* **Figure 2:** *Bar chart showing the Age of YRBSS 2013 survey.*
Latent class analysis (LCA) is a statistical technique used in this research to identify latent classes within a population based on their patterns of responses to multiple categorical or binary variables. The idea behind LCA is that individuals within a population may be grouped into different classes based on the similarities in their responses to multiple items or questions.
The basic assumptions of LCA is that the respondents in each class have a certain probability of giving a certain answer to each item and that these probabilities are class specific.
LCA can be considered a type of unsupervised machine learning algorithm, because it does not require prior knowledge about which individuals belong to which class, and the classes themselves are not directly observable. Instead, the classes are inferred from the patterns of responses to multiple items.
The likelihood ratio test (LRT) is a statistical test that compares a model with a more complex hypothesis to a model with a simpler hypothesis. The likelihood ratio test statistic (also known as the TECH 11 statistic) is a measure of the goodness of fit of a model and can be used to determine whether the more complex model provides a significantly better fit to the data than the simpler model.
## Statistical analysis
For analysis of the YRBSS data, SAS v. 9.4 was used to generate descriptive statistics so as to identify the socioeconomic and behavioral characteristics of the sample.
A structural main model was found by fitting an unconstrained latent class model to 40 risky behavioral signals. With respect to previous research, the model started with a two-class model and gradually added more classes till the best model class was matched with the data (Nylund et al., 2007). This strategy is used when there are no preconceived preconceptions about the grouping of manifest variables, as there were in this study (Chandler et al., 2021). The fitness was the ratio of probability, Akaike information criteria, revised Akaike information criterion, and the Bayesian information criterion. The latent class probabilities and conditional probabilities are derived using the population dispersion across latent classes, latent class characteristics, and entropy (Schreiber, 2022). Analyses were updated employing sample weights from the YRBSS to make the outcome relevant to high school students in the United States.
## Results
The prevalence of specific risk behaviors among the youth in our sample is shown in Figure 3. According to our research, many youths participate in risky behavior. Drug usage is common in this demographic, similar to reports by Thorsen [2018]. Overall, $31.8\%$ of young people had drunk alcohol on more than a day in the preceding month, and nearly a quarter had used marijuana on more than a day in the previous month. Almost $15.8\%$ of teenagers reported being current smokers, defined as having smoked more than one cigarette in the preceding month. *This* generation also used prescription medications often, with around $18.3\%$ of youngsters having done so without a doctor’s prescription in the preceding year. Teenagers in our sample demonstrated behaviors that resulted in unintentional harm and aggression, as seen in Figure 3. More than a quarter of the teenagers admitted to chatting while driving at least once during the month before.
**Figure 3:** *Sunburst of specific risk behaviors of YRBSS 2013 survey.*
A large number of students engage in risky sexual behavior. According to the findings, $32.3\%$ of the young people in our survey are now interested in sex (i.e., had sexual relations with one or more people in the three months preceding the survey). $13.4\%$ of sexually active teens at the time had not used a condom for their most recent experience.
The fitness for various classes taken into consideration in this study is shown in Table 1. From LCA, starting with a two-class solution and gradually adding more classes until the fit indices reached a plateau in terms of overall data fit for risk behaviors among teenagers in our sample, a five-class model gave the best results. The p-value for LRT, the value of entropy, and the review of BIC, were used to make this determination. On a scale of [0, 1], entropy is a measure of categorization precision, with values close to one denoting high classification confidence and values close to zero denoting poor classification confidence.
**Table 1**
| Fit indices | 2 | 3 | 4 | 5 |
| --- | --- | --- | --- | --- |
| AIC | 367453.6 | 376590.6 | 373116.9 | 373426.2 |
| ABIC | 361785.1 | 762519.9 | 341604.1 | 365441.3 |
| BIC | 390062.5 | 381307.6 | 376622.1 | 380089.6 |
| TECH 11 LRT | 38795.5 | 7546.5 | 6529.1 | 2606.5 |
| p-Value | 0.0 | 0.35 | 0.47 | 0.63 |
| Entropy | 0.92 | 0.88 | 0.86 | 0.8 |
Five types (subsets) of risky behavior were found in the sample of teenagers as a whole, as shown in Table 2. The majority of teenagers ($52.6\%$) were in class 1. $14.9\%$ of teenagers were in class 2. $14.3\%$ of the teenagers were in class 3, $9.4\%$ were in class 4, and $8.5\%$ were in class 5, respectively. We included teenagers from Classes 1 and 2 who abstained from hazardous conduct. As a result, we classified the teenagers in these classrooms as low-risk. Compared to classes 1, 2, or 4, the adolescents in class 3 were more likely to be sad and suicidal. In one class, more than $79.8\%$ of the teenagers reported having depression, $63.8\%$ had experienced suicidal thoughts in the previous year, and more than $50.4\%$ intended to commit suicide. Adolescents in class 4 who use cigarettes and liquor are shown. In comparison to the other classes, these teenagers had a higher likelihood of being current drinkers (more than one drink in the previous month) and current smokers (more than one cigar in the previous month). Class 5 was made up of high-risk youth who were more likely to engage in the most risky behaviors, including depression. These young people were classified as high-risk polydrug users. These teenagers had a high probability of taking prescription medications such as Codeine, OxyContin, Xanax, or Ritalin, in addition to a high probability of using nicotine, marijuana, and liquor. For the teenagers in this grouping, the likelihood of violent altercations over the previous 12 months was moderate.
**Table 2**
| Unnamed: 0 | Class 1 | Class 2 | Class 3 | Class 4 | Class 5 |
| --- | --- | --- | --- | --- | --- |
| Currently Smoking Cigarette | 0.0 | 0.0 | 0.004 | 0.97 | 0.972 |
| Regular Cigarette Smoker | 0.003 | 0.0029 | 0.0027 | 0.354 | 0.543 |
| Currently Using Marijuana | 0.044 | 0.462 | 0.168 | 0.543 | 0.827 |
| Use of alcohol recently | 0.064 | 0.876 | 0.305 | 0.746 | 0.954 |
| Cocaine | 0.002 | 0.064 | 0.043 | 0.054 | 0.543 |
| Had sexual relations with at least four people | 0.043 | 0.261 | 0.076 | 0.262 | 0.611 |
| Engaged in sexual activity | 0.164 | 0.614 | 0.271 | 0.573 | 0.834 |
| Didn’t use condom | 0.327 | 0.382 | 0.457 | 0.384 | 0.571 |
| Took drugs or alcohol prior to previous sexual encounter | 0.005 | 0.251 | 0.072 | 0.205 | 0.572 |
A 4-class model for females in our sample and a 3-class model for males based on fit statistics was identified. In both groups of teenagers, $50.3\%$ of females and $62.3\%$ of males were classified as low-risk, also known as class 1.
More than a quarter of the males in Class 2 were teenagers. Due to the significant likelihood that they now consume liquor, the kids in this category were labeled as “liquor drinkers”. Males who were in Class 3 made up $12\%$ of the sample and were classified as “High-risk Polysubstance Users”. Compared to sets 1 and 2, all members of this class showed a greater likelihood of using cigarettes, liquor, binge drinking, and cigarettes. In addition, compared to classes 1 and 2, males in this category were more likely to have used prescription drugs at some point in their lives. In Class 3, around two-thirds of the males admitted to carrying a weapon in the previous month, such as a knife, rifle, or club. Males in this set also had a high likelihood of getting into a fight physically at least once in the previous year.
Over $50\%$ of the females in our sample belonged to Class 1, which is a low-risk subset. $20\%$ of ladies in Class 2 supported the use of alcoholic beverages. Approximately half of the females in this cohort reported excessive drinking (more than five drinks of liquor in a row, within a couple of hours, more than a day in the previous month). Class 3 consisted of slightly more than $19\%$ of the females in our sample who were classified as “liquor drinkers.” More than half of the females in this set acknowledged creating a suicide plan in the preceding year, and the majority of them experienced depressive symptoms and suicidal tendencies. We found that the females in this category were suicidal and sad. Almost $12\%$ of the females were in Class 4. Females in this category were classified as "high risk" because they supported the majority of the risk behaviors that were investigated in our study. $80.3\%$ of females reported smoking cigarettes, $74.5\%$ reported using marijuana, and $88.3\%$ reported drinking alcohol, with $71.4\%$ reporting binge drinking. Additionally, $68.3\%$ of these females supported using drugs, particularly prescription drugs. This subset of females also expressed suicidal ideation and depressive symptoms. Also, $79.3\%$ of the females in this subset claimed to be sexually active; of those, $61.4\%$ admitted to not wearing a condom during their most recent sex, and $44.4\%$ admitted to using liquor or drugs just before their most recent sex.
## Key findings
The LCA identifies five subsets of youth interacting in risky behavior for the entire sample of adolescents in this study.
Class 1 and 2 (low risk): those who do not engage in risky behavior.
Those in Class 3 have depressive symptoms and suicidal thoughts.
Class 4: Those who have a high likelihood of using cigarettes, liquor, Those classified as polysubstance users fall into Class 5.
In keeping with previous research, we observed that the majority of high school students ($62\%$ of males and $50\%$ of females) were in the low-risk category. Adolescents with the highest risk profiles made up the smallest fraction of the population ($12.2\%$ for males and $11.6\%$ for females). We discovered that young teenagers in the greatest subsets were involved in numerous risk behaviors, including substance usage, unsafe sex, and behavior patterns that resulted in accidental injuries and violence, lending credence to past studies that connected two or three risk behaviors (Finch, 2015; Thorsen, 2018). Because males and females had similar risky behavior categories, there were gender disparities in the volume and types of hazardous behavior among adolescents overall.
According to gender studies, females were more likely than males to have depressive symptoms, suicidal thoughts, and suicidal plans, as well as to be sedentary. Males in the polysubstance use subset exhibited considerably lower rates of depressive symptoms and suicidal thoughts and attempts than did females, who reported higher rates of both of these risk behaviors along with the majority of other risk behaviors. This result deviates from other studies that identified a specific risk profile among males, including high rates of marijuana use and suicidal behavior (Finch, 2015; Cheng et al., 2016). In our analysis, there were no discernible variations in the prevalence of drug use between males and females. Males reported using weapons and engaging in physical altercations at greater rates than females, though.
These findings support the issue behavior hypothesis by showing that risk behaviors co-occur and may have significant effects on efforts at prevention, intervention, and health promotion. These behavior modification initiatives must specifically acknowledge that health risk behaviors in this demographic are interconnected rather than discrete, unconnected activities. The majority of currently available interventions or programs concentrate on one or two main correlates that are thought to cause or contribute to problem behavior in adolescents, and these programs frequently offer a straightforward or one-dimensional approach to addressing and preventing risk behavior in adolescents. The Scared Straight program, for instance, targets young people who are delinquent or at-risk of becoming delinquent and aims to prevent them from engaging in criminal activity in the future. It is based on the idea that young people need to have a better understanding of the negative effects of their behavior as well as a change in attitude. The program’s elements include visits to prisons and talks by offenders to provide children an understanding of the harsh reality of life behind bars and the repercussions of their aggressive or illegal behavior (Ball and Weisberg, 2014; Cheng et al., 2016). This program posits that a specific undesirable behavior is caused by a single determinant, and that youngsters just need to be “Scared Straight” by witnessing the consequences of such negative behavior. Despite research showing that people who engage in initiatives comparable to Mortified are more inclined to perpetrate felonies in the future (Cheng et al., 2016), the program is nevertheless extensively employed across the country due to public safety concerns.
Specifically, the data from our research suggests that marijuana use is common among this group, with more than half of the individuals acknowledging use. Similarly, the data suggest that smoking cigarettes is also common, with a high likelihood of occurrence. In addition, more than half of the individuals in this group engage in risky sexual practices, such as not using condoms during their most recent encounter.
The LCA results also suggest that there are differences in the patterns of risky behavior among males and females in this group. Specifically, the LCA identified three categories of males based on their involvement in risky behavior, while four subgroups of females were identified. This suggests that there may be gender-specific patterns of risky behavior in this group of youths. It’s also worth noting that the high levels of risky behavior among this group are concerning, and it would be valuable to conduct further research to understand the factors that contribute to these behaviors in this group of youths. Additionally, identifying these subgroups could inform the design and delivery of public health interventions to target these specific at-risk subgroups and help to reduce the prevalence of these risky behaviors among them.
## Conclusion
Using LCA to examine each risk activity as a different but linked area of potentially dangerous adolescent behavior allows researchers to evaluate the co-occurrence of risky behaviors. This method allowed us to incorporate sex as a covariate and evaluate how patterns of certain behaviors affect participation in other hazardous behaviors.
LCA has been used to examine each risk activity as a different but linked area of potentially dangerous adolescent behavior and has helped to evaluate the co-occurrence of risky behaviors.
This research has helped in the understanding of how different risky behaviors may be related to one another and how they may interact to influence overall risk-taking behavior. Additionally, by incorporating sex as a covariate, research has evaluated how patterns of certain behaviors are affected by the participant’s sex. This has provided insights into how gender may influence the relationship between different risky behaviors and how the patterns of co-occurrence may differ between males and females. The LCA has provided a more nuanced understanding of risk-taking behaviors by identifying latent classes within the larger population based on their patterns of risky behaviors. This research has helped identify complex patterns of co-occurrence among multiple variables.
Integrated risk behavior prevention models must not only give correct information and impart useful life lessons for handling stress and conflict but they also must be strategically given while taking demographic aspects into consideration and be scheduled appropriately for adolescent growth. Interventions may be more successful and have a greater impact if they take into consideration the psychological and environmental factors that influence adolescent decision-making. In research looking at the impact of physical exercise on youth smoking cessation, implementing a multi-behavior program that addressed both smoking and physical activity in high school teenagers was shown to be much more successful than a short intervention in research. High schools were randomly allocated to one of three research conditions by Horn and colleagues: brief intervention, Not on Cigarettes, a tried-and-true program for teen smoking cessation, which also contained a physical activity module. The outcomes demonstrated that including a physical exercise component in cigarettes was very effective in helping males quit smoking.
There is a need for further research on health risk behaviors that makes use of representative sample observational studies with situational features. Our findings imply that future research should concentrate on sex variations in youths’ risky health behaviors. In our study, females were more often and significantly associated with polysubstance use and depression. Drug usage and either a low or high frequency of multi-partner sexual activity were significantly more correlated in females. Females are less likely than males to be classed as having several partners or as having a high multi-partner sexual risk, but when they use various drugs often, their relative risk for engaging in these behaviors increases. It may be essential to consider how one’s social and physical circumstances may affect behavioral decision-making while developing effective prevention and intervention techniques, as well as legislation relevant to adolescent health.
## Data availability statement
The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.
## Author contributions
ZA and FR: main manuscript writing and draft. MA, IJ, and MNA: editing, proofreading. All authors contributed to the article and approved the submitted version.
## Conflict of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s Note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Ball W., Weisberg R.. **The new normal? Prosecutorial charging in california after public safety realignment**. *SSRN Electronic J.* (2014). DOI: 10.2139/ssrn.2403040
2. Bleakley A., Ellithorpe M. E., Hennessy M., Jamieson P. E., Khurana A., Weitz I.. **Risky movies, risky behaviors and ethnic identity among Black adolescents**. *Soc. Sci. Med.* (2017) **195** 131-137. DOI: 10.1016/j.socscimed.2017.10.024
3. Bozzini A. B., Bauer A., Maruyama J., Simões R., Matijasevich A.. **Factors associated with risk behaviors in adolescence: a systematic review**. *Braz. J. Psychiatr.* (2021) **43** 210-221. DOI: 10.1590/1516-4446-2019-0835
4. Cavazos-Rehg P. A., Krauss M. J., Spitznagel E. L., Schootman M., Cottler L. B., Bierut L. J.. **Characteristics of sexually active teenage girls who would be pleased with becoming pregnant**. *Matern. Child Health J.* (2013) **17** 470-476. DOI: 10.1007/s10995-012-1020-0
5. Centers for Disease Control and Prevention (2014). Youth Risk Behavior Survey Data. Available online at: www.cdc.gov/yrbs.. *Youth Risk Behavior Survey Data* (2014)
6. Chandler L., Abdujawad A. W., Mitra S., Mceligot A. J.. **Marijuana use and high-risk health behaviors among diverse college students post- legalization of recreational marijuana use**. *Public Health Pract. (Oxf)* (2021) **2** 100195. DOI: 10.1016/j.puhip.2021.100195
7. Cheng T. J., Johnston C., Kerr T., Nguyen P., Wood E. K., DeBeck K. K.. **Substance use patterns and unprotected sex among street-involved youth in a Canadian setting: a prospective cohort study**. *BMC Public Health* (2016) **16** 4. DOI: 10.1186/s12889-015-2627-z
8. Chesson H., Sternberg M., Leichliter J., Aral S.. **The distribution of chlamydia, gonorrhoea and syphilis cases across states and counties in the USA, 2007**. *Sex. Transm. Infect.* (2010) **86** 52-57. DOI: 10.1136/sti.2009.040873
9. Costello D. M., Swendsen J., Rose J. S., Dierker L. C.. **Risk and protective factors associated with trajectories of depressed mood from adolescence to early adulthood**. *J. Consult. Clin. Psychol.* (2008) **76** 173-183. DOI: 10.1037/0022-006X.76.2.173
10. de la Haye K., D’Amico E. J., Miles J. N. V., Ewing B., Tucker J. S.. **Covariance among multiple health risk behaviors in adolescents**. *PLoS One* (2014) **9** e98141. DOI: 10.1371/journal.pone.0098141
11. Farhat T., Iannotti R. J., Simons-Morton B. G.. **Overweight, obesity, youth and health-risk behaviors**. *Am. J. Prev. Med.* (2010) **38** 258-267. DOI: 10.1016/j.amepre.2009.10.038
12. Fergusson D. M., Lynskey M. T.. **Alcohol misuse and adolescent sexual behaviors and risk taking**. *Pediatrics* (1996) **98** 91-96. PMID: 8668418
13. Finch H.. **A comparison of statistics for assessing model invariance in latent class analysis**. *Open J. Stat.* (2015) **5** 191-210. DOI: 10.4236/ojs.2015.53022
14. Fisher J. A., McManus L., Cottingham M. D., Kalbaugh J. M., Wood M. M., Monahan T.. **Healthy volunteers’ perceptions of risk in US Phase I clinical trials: a mixed-methods study**. *PLoS Med.* (2018) **15** e1002698. DOI: 10.1371/journal.pmed.1002698
15. Hamilton B. E., Martin J. A., Ventura S. J.. **Births: preliminary data for 2008**. *National Vital Stat. Rep.* (2010) **58** 1-18
16. Khan M. R., Berger A. T., Wells B. E., Cleland C. M.. **Longitudinal associations between adolescent alcohol use and adulthood sexual risk behavior and sexually transmitted infection in the United States: assessment of differences by race**. *Am. J. Public Health* (2012) **102** 867-876. DOI: 10.2105/AJPH.2011.300373
17. MacArthur G. J., Smith M. C., Melotti R., Heron J., Macleod J., Hickman M.. **Patterns of alcohol use and multiple risk behaviour by gender during early and late adolescence: the ALSPAC cohort**. *J. Public Health (Oxf)* (2012) **34** i20-i30. DOI: 10.1093/pubmed/fds006
18. Mulye T. P., Park M. J., Nelson C. D., Adams S. H., Irwin C. E., Brindis C. D.. **Trends in adolescent and young adult health in the United States**. *J. Adolesc. Health* (2009) **45** 8-24. DOI: 10.1016/j.jadohealth.2009.03.013
19. Nylund K. L., Asparouhov T., Muthén B. O.. **Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study**. *Struct. Equ. Modeling* (2007) **14** 535-569. DOI: 10.1080/10705510701575396
20. Oh H., Banawa R., Zhou S., Devylder J., Koyanagi A.. **The mental and physical health correlates of psychotic experiences among US college students: findings from the Healthy Mind Study 2020**. *J. Am. Coll. Health* (2022) 1-7. DOI: 10.1080/07448481.2022.2058879
21. Oh H., Waldman K., Stickley A., Devylder J., Koyanagi A.. **Psychotic experiences and physical health conditions in the United States**. *Compr. Psychiatry* (2019) **90** 1-6. DOI: 10.1016/j.comppsych.2018.12.007
22. Olajide D., Eberth B., Ludbrook A.. **Analysis of multiple health risky behaviours and associated disease outcomes using scottish linked hospitalisation data**. *Front. Public Health* (2022) **10** 847938. DOI: 10.3389/fpubh.2022.847938
23. Owen N., Halford K.. **Psychology, public health and cigarette smoking**. *Aust. Psychol.* (2007) **23** 137-152. DOI: 10.1080/00050068808255600
24. Park H. S., Schepp K. G., Jang E. H., Koo H. Y.. **Predictors of suicidal ideation among high school students by gender in South Korea**. *J. Sch. Health* (2006) **76** 181-188. DOI: 10.1111/j.1746-1561.2006.00092.x
25. Parks S. E., Johnson L. L., McDaniel D. D., Gladden M.. **Surveillance for violent deaths—national violent death reporting system, 16 states, 2010**. *MMWR Surveill. Summ.* (2014) **63** 1-33. DOI: 10.1016/j.jadohealth.2009.03.025
26. Schreiber J.. **Latent class and latent class regression**. *Contemp. Res. Methods Pharm. Health Services* (2022) **55** 541-552. DOI: 10.1016/B978-0-323-91888-6.00044-2
27. Shaw P. H., Reed D. R., Yeager N., Zebrack B., Castellino S. M., Bleyer A.. **Adolescent and young adult (AYA) oncology in the United States: a specialty in its late adolescence**. *J. Pediatr. Hematol. Oncol.* (2015) **37** 161-169. DOI: 10.1097/MPH.0000000000000318
28. Strachman A., Impett E. A., Henson J. M., Pentz M. A.. **Early adolescent alcohol use and sexual experience by emerging adulthood: a 10-year longitudinal investigation**. *J. Adolesc. Health* (2009) **45** 478-482. DOI: 10.1016/j.jadohealth.2009.03.025
29. Thorsen M. L.. **A latent class analysis of behavioral and psychosocial dimensions of adolescent sexuality: exploring race differences**. *J. Sex Res.* (2018) **55** 45-59. DOI: 10.1080/00224499.2016.1254143
30. Tsevat D. G., Wiesenfeld H. C., Parks C., Peipert J. F.. **Sexually transmitted diseases and infertility**. *Am. J. Obstet. Gynecol.* (2017) **216** 1-9. DOI: 10.1016/j.ajog.2016.08.008
31. Weden M. M., Zabin L. S.. **Gender and ethnic differences in the co-occurrence of adolescent risk behaviors**. *Ethn. Health* (2005) **10** 213-234. DOI: 10.1080/13557850500115744
32. Wills T. A., Resko J. A., Ainette M. G., Mendoza D.. **Smoking onset in adolescence: a person-centered analysis with time-varying predictors**. *Health Psychol.* (2004) **23** 158-167. DOI: 10.1037/0278-6133.23.2.158
33. Youth Risk Behavior Survey (2013). Division of Adolescent and School Health, National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention. Available online at: www.cdc.gov/yrbs.. *Division of Adolescent and School Health, National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention* (2013)
|
---
title: Sex differences in factors influencing hospital-acquired pneumonia in schizophrenia
patients receiving modified electroconvulsive therapy
authors:
- Mi Yang
- Yan Yang
- Liju Liu
- Di Kong
- Min Xu
- Xincheng Huang
- Cheng Luo
- Guocheng Zhao
- Xiangyang Zhang
- Yan Huang
- Yunzhong Tu
- Zezhi Li
journal: Frontiers in Psychiatry
year: 2023
pmcid: PMC9971594
doi: 10.3389/fpsyt.2023.1127262
license: CC BY 4.0
---
# Sex differences in factors influencing hospital-acquired pneumonia in schizophrenia patients receiving modified electroconvulsive therapy
## Abstract
### Background
Sex differences may be presented in the clinical features or symptoms of schizophrenia patients but also affect the occurrence of hospital-acquired pneumonia (HAP). Modified electroconvulsive therapy (mECT) is a common treatment method for schizophrenia, used in combination with antipsychotics. This retrospective research explores the sex difference in HAP affecting patients with schizophrenia who have received mECT treatment during hospitalization.
### Methods
We included schizophrenia inpatients treated with mECT and antipsychotics between January 2015 and April 2022. Blood-related and demographic data collected on admission were analyzed. Influencing factors of HAP in male and female groups were assessed separately.
### Results
A total of 951 schizophrenia patients treated with mECT were enrolled in the study, including 375 males and 576 females, of which 62 patients experienced HAP during hospitalization. The risk period of HAP in these patients was found to be the first day after each mECT treatment and the first three sessions of mECT treatment. Statistically significant differences in the incidence of HAP were identified in male vs. female groups, with an incidence in men about 2.3 times higher than that in women ($P \leq 0.001$). Lower total cholesterol (Z = −2.147, $$P \leq 0.032$$) and the use of anti-parkinsonian drugs (χ2 = 17.973, $P \leq 0.001$) were found to be independent risk factors of HAP in male patients, while lower lymphocyte count (Z = −2.408, $$P \leq 0.016$$), hypertension (χ2 = 9.096, $$P \leq 0.003$$), and use of sedative-hypnotic drugs (χ2 = 13.636, $P \leq 0.001$) were identified in female patients.
### Conclusion
Influencing factors of HAP in schizophrenia patients treated with mECT have gender differences. The first day after each mECT treatment and the first three sessions of mECT treatment were identified to have the greatest risk for HAP development. Therefore, it would be imperative to monitor clinical management and medications during this period according to these gender differences.
## 1. Introduction
Schizophrenia (SCZ) is a common severe psychiatric disorder and usually manifests clinically with psychotic symptoms such as hallucinations, delusions, emotional indifference, and cognitive dysfunction [1, 2]. According to a 2018 World Health Organization report, more than 20 million people are already living with SCZ worldwide, and China accounts for about half of this population [3]. SCZ is characterized by high incidence, high disability, and a low cure rate, with a lifetime incidence of about $1\%$ [4]. As such, the disease brings heavy psychological and economic burdens to patient families and broader society [5, 6] and has become a major societal challenge [4].
Gender differences may exist in the clinical symptoms of schizophrenia patients; for example, the age of onset for men may be a 3-year younger than that for women [7, 8]. The correlation between non-social functioning and objective social cognition in men may be much stronger than in women [9], whereas hostile bias correlates with verbal fluency found in women [10]. Women are good at processing speed and verbal situational memory, but men are good at visual working memory [11]. There are also sex differences in the cognitive correlates of first-episode schizophrenia [12], i.e., working memory and executive function were correlated to onset age, negative symptoms were associated with memory or working memory in women, whereas processing speed was correlated with antipsychotic dosage in men. Antipsychotics may induce gender differences in extrapyramidal and anticholinergic responses, sexual problems, and subjective tolerance [13]. However, no sex differences in the efficacy of amisulpride or risperidone medication were found in elderly schizophrenia patients [14]. Although women may be more prone to gain weight from the antipsychotic medication [15], women may have a better prognosis than men [16]. According to the current pathology of schizophrenia, these gender differences may be related to gene expression [17], the mechanism of the microbiota-brain-gut axis [18], differences in brain structure and function [19], or even sociocultural [20]. However, these hypotheses are inconsistent to a large extent, so the mechanisms underlying the clinical characterization induced by gender differences need further investigation.
Electroconvulsive therapy (ECT) is widely used, particularly in patients with refractory schizophrenia (21–24). To avoid the generalized convulsions triggered by ECT treatment and the related fear of convulsions [25], modified ECT (mECT) was developed. Due to the use of anesthetic and muscle relaxants before the mECT treatment [26, 27], the risk of post-treatment infections may be increased [28]. Prior research on ECT has shown that the incidence of pneumonia infection is 3.8 per 10,000 ECT treatments [29]. Hospital-acquired pneumonia (HAP) is a leading cause of morbidity and mortality, and the incidence of HAP remains high, but effective treatment is usually lacking [30]. Sex differences in the incidence of HAP exist [31], with men being a risk factor for HAP [32, 33], but a lower incidence of non-ventilator HAP in women [34]. Some pathologies may be explained by these differences induced by genders, such as immune response to the virus [24], diabetes [35], or chronic obstructive pulmonary disease (COPD) [36]. However, there are few studies on gender differences in risk factors related to hospital-acquired pneumonia (HAP) in SCZ patients with mECT around the world. Therefore, this paper aims to analyze the risk factors associated with the development of HAP in SCZ patients who have received mECT in recent years, explore the possible pathogenesis of HAP caused by gender differences, and provide a basis for guiding clinical management and improving the quality of patient treatment.
## 2.1. Patients
This retrospective study included inpatients with schizophrenia admitted between January 2015 and April 2022. Patients met the diagnosis criteria of schizophrenia according to the International Classification of Diseases-10 (ICD-10) and received mECT treatment during their hospitalization. The diagnosis of HAP required all the following criteria: new lung infiltrates on chest imaging, respiratory decline, fever, and productive cough [37]. Patients with infections within 48 h of hospitalization were excluded. This study was approved by the Ethics Committee of the Fourth People's Hospital of Chengdu.
Patient information collected included name, age, gender, as well as the status of diabetes mellitus, hypertension, epilepsy, or substance dependence (smoking or drinking), and excluded patients with comorbid cardiovascular disease. Blood samples were collected on admission for routine biochemical testing (white blood cells, red blood cells, platelets, lipids, glucose, blood proteins, etc.). Other medications of patients receiving mECT at the time of hospitalization were recorded, such as sedative-hypnotic drugs (SHD), antidepressant drugs (ADD), anti-anxiety drugs (AAD), antimanic drugs (AMD), anti-epileptic drugs (AED), anti-parkinsonian drugs (APD), and other neurological drugs. mECT-related conditions only for patients with HAP, such as the days from the first mECT to HAP (dfE2H) occurrence, the numbers of mECT treatments before HAP (nE2H) occurrence, and the days from the last mECT treatment to HAP (dlE2H) occurrence.
## 2.2. mECT parameters
The ECT Instrument was Thymatron System IV (Somatics, LLC, 149 Amityville Street Islip Terrace, NY 11752 USA). Electrode placement was a bilateral temporal energization mode. The electrical stimulation procedure was LOW 0.5: Frequency: 10–70 Hz; Pulse width: 0.5 ms; Duration: ≤ 8s; Waveform: bipolar, brief pulsed, square wave. Stimulus intensity (power): “Age mode” is used. For the first treatment session, the power is set to age × $80\%$ for younger than 30 and age × $100\%$ for older than 50; The power of follow-up treatment increased by 1–$5\%$ depending on the seizure index. Convulsions quality assessments: EEG Endpoint was 20–60 s; Average Seizure Energy Index was over 5,000; Postictal Suppression Index was over $80\%$.
## 2.3. Statistical analysis
Software SPSS 26 (IBM Corporation, New Orchard Road, Armonk, NY 10504, USA) was used for statistical calculations. These patients were divided into two groups by gender, with subgroups of HAP and non-HAP. Then the statistical analysis was conducted as follows: First, a general linear model (univariate model) was used for assessing the interaction effect of gender vs. other factors on HAP. Second, influencing factors for HAP men and women were analyzed separately. The χ2 test was used for categorical variables. Continuous variables were first tested for normality by Kolmogorov-Smirnov; if variables conformed to a normal distribution, t-testing was used, while non-normality variables were tested by non-parametric (Mann-Whitney) test. Since the included data were almost all non-normal variables, Spearman correlation analysis was used. Third, binary logistic regression modeling was later used for risk factors analysis. Finally, statistical calibration was performed using the Bonferroni method ($P \leq 0.05$/31 ≈ 0.0016). Continuous variables were expressed as mean ± standard deviation (x̄ ± std.), and $P \leq 0.05$ was considered to be statistically significant.
## 3.1. General characteristics of included SCZ patients
A total of 951 inpatients, aged between 14 and 71 years old, were included. Of the 951 inpatients, 375 were male, and 576 were female, with mean ages of 32.57 ± 11.93 years and 36.85 ± 13.96 years, respectively, and males being significantly younger than females ($t = 5.053$, $P \leq 0.001$). 62 inpatients were HAP, with 37 in men and 25 in women. Covariance analysis results indicate interaction between sex and epilepsy, TC, anti-depressant drugs, anti-epileptic drugs, anti-parkinsonian drugs, and other neuron drugs (all Ps < 0.05), as shown in Table 1. Only epilepsy passed the Bonferroni test.
**Table 1**
| Unnamed: 0 | Factors | Factors.1 | Sex | Sex.1 | Factor × Sex | Factor × Sex.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | F | P | F | P | F | P |
| Demographic data | Demographic data | Demographic data | Demographic data | Demographic data | Demographic data | Demographic data |
| Age (years) | 0.106 | 0.745 | 2.716 | 0.100 | 0.222 | 0.637 |
| Days in hospital | 4.071 | 0.044 | 0.688 | 0.407 | 1.478 | 0.224 |
| Hypertension | 7.561 | 0.006 | 11.763 | 0.001 | 0.107 | 0.744 |
| Diabetes | 5.940 | 0.015 | 11.589 | 0.001 | 0.007 | 0.935 |
| Epilepsy | 8.247 | 0.004 | 12.557 | <0.001 | 12.496 | <0.001 |
| Substance dependence | 0.016 | 0.899 | 10.840 | 0.001 | 0.313 | 0.576 |
| Blood data | Blood data | Blood data | Blood data | Blood data | Blood data | Blood data |
| TB (μmol/L) | 1.145 | 0.285 | 0.842 | 0.359 | 0.628 | 0.428 |
| DB (μmol/L) | 0.120 | 0.914 | 3.336 | 0.068 | 0.09 | 0.891 |
| IB (μmol/L) | 0.035 | 0.851 | 0.808 | 0.369 | 0.711 | 0.399 |
| Glucose (mmol/L) | 0.963 | 0.327 | 0.263 | 0.608 | 0.185 | 0.667 |
| TC (mmol/L) | 4.983 | 0.026 | 7.460 | 0.006 | 4.164 | 0.042 |
| TG (mmol/L) | 0.395 | 0.530 | 2.959 | 0.086 | 0.001 | 0.973 |
| HDL (mmol/L) | 0.458 | 0.499 | 0.028 | 0.688 | 0.451 | 0.502 |
| LDL (mmol/L) | 0.351 | 0.553 | 8.051 | 0.005 | 3.683 | 0.055 |
| Albumin (g/L) | 1.719 | 0.190 | 0.287 | 0.593 | 0.672 | 0.413 |
| Uric acid (μmol/L) | 0.920 | 0.338 | 1.211 | 0.271 | 0.002 | 0.963 |
| Leukocyte (109/L) | 2.207 | 0.138 | 0.069 | 0.793 | 0.632 | 0.427 |
| Monocyte (109/L) | 1.770 | 0.184 | 0.288 | 0.591 | 3.212 | 0.073 |
| Lymphocyte (109/L) | 2.450 | 0.118 | 0.106 | 0.745 | 0.800 | 0.371 |
| Eosinophils (109/L) | 2.439 | 0.119 | 7.823 | 0.005 | 0.166 | 0.683 |
| Basophils (109/L) | 2.153 | 0.143 | 2.721 | 0.099 | 0.013 | 0.909 |
| RBC (1012/L) | 2.343 | 0.126 | 0.244 | 0.622 | 0.003 | 0.960 |
| Hemoglobin (g/L) | 0.555 | 0.456 | 0.068 | 0.794 | 0.014 | 0.907 |
| Thrombocyte (109/L) | 2.552 | 0.110 | 3.416 | 0.065 | 0.951 | 0.330 |
| Affiliated medication | Affiliated medication | Affiliated medication | Affiliated medication | Affiliated medication | Affiliated medication | Affiliated medication |
| Sedative-hypnotics | 26.827 | <0.001 | 2.559 | 0.110 | 1.641 | 0.201 |
| Anti-anxiety drugs | 0.323 | 0.570 | 10.858 | 0.001 | 0.046 | 0.831 |
| Anti-manic drugs | 4.374 | 0.037 | 10.525 | 0.001 | 0.051 | 0.822 |
| Anti-depressant drugs | 2.984 | 0.084 | 6.244 | 0.013 | 6.422 | 0.011 |
| Anti-epileptic drugs | 7.032 | 0.008 | 4.255 | 0.040 | 4.191 | 0.041 |
| Anti-parkinsonian drugs | 31.591 | <0.001 | 2.364 | 0.124 | 6.444 | 0.011 |
| Other neuron drugs | 6.169 | 0.013 | 9.439 | 0.002 | 9.552 | 0.002 |
## 3.2. HAP occurrence in patients receiving mECT
HAP occurred in 62 of the 951 study subjects, with an incidence of $6.52\%$. Patients developed HAP 20.10 ± 18.29 days after admission, received about 3.47 ± 2.57 sessions of mECT treatment before HAP, and HAP occurred approximately 2.78 ± 4.70 days after the last MECT treatment. The prevalence of HAP was significantly higher in men than in women ($\frac{37}{375}$:$\frac{25}{576}$ = $9.87\%$:$4.34\%$ ≈ 2.3, χ2 = 11.382, $P \leq 0.001$). There was no statistically significant difference between men and women in dfE2H, nE2H, or dlE2H, as shown in Table 2.
**Table 2**
| Unnamed: 0 | Woman (n = 25) | Woman (n = 25).1 | Man (n = 37) | Man (n = 37).1 | Z | P |
| --- | --- | --- | --- | --- | --- | --- |
| | x̄ ± std | Median (Q1, Q3) | x̄ ± std | Median (Q1, Q3) | | |
| dfE2H | 19.39 ± 16.90 | 13.20 (5.91, 26.38) | 20.59 ± 19.38 | 13.81 (4.79, 32.11) | −0.194 | 0.846 |
| nE2H | 3.45 ± 4.98 | 1.00 (0.41, 5.52) | 2.33 ± 4.51 | 1.00 (0.39, 2.03) | −0.854 | 0.393 |
| dlE2H | 3.44 ± 2.66 | 2.00 (2.00, 4.50) | 3.49 ± 2.53 | 3.00 (1.00, 5.50) | −0.037 | 0.971 |
## 3.3. Risk factors for HAP in male SCZ patients receiving mECT
In Table 3, hospitalized days were significantly higher in the HAP group, as compared to the non-HAP group (Z = −2.760, $$P \leq 0.006$$). Total cholesterol (mmol/L) was significantly lower in the HAP group as compared to the non-HAP group (Z = −2.147, $$P \leq 0.032$$), while monocyte count was significantly higher (Z = −2.001, $$P \leq 0.045$$). There was a statistically significant difference in basophils between the HAP and non-HAP group (Z = −2.065, $$P \leq 0.039$$). There were also statistically significant increases in HAP incidence found in patients who had taken any medications of SHD (χ2 = 11.396, $P \leq 0.001$), ADD (χ2 = 5.237, $$P \leq 0.002$$), AED (χ2 = 7.993, $$P \leq 0.005$$), APD (χ2 = 17.973, $P \leq 0.001$), or other neurological drugs (χ2 = 7.323, $$P \leq 0.007$$), though only SHD and APD passed the Bonferroni test.
**Table 3**
| Unnamed: 0 | Non-HAP (n = 338) | HAP (n = 37) | t, Z, χ2 | P |
| --- | --- | --- | --- | --- |
| Demographic data | Demographic data | Demographic data | Demographic data | Demographic data |
| Age (year) | 32.58 ± 11.62 | 32.43 ± 14.66 | −0.803 | 0.422 |
| Diabetes | Diabetes | Diabetes | Diabetes | Diabetes |
| No | 331 | 35 | 1.583 | 0.208 |
| Yes | 7 | 2 | | |
| Hypertension | Hypertension | Hypertension | Hypertension | Hypertension |
| No | 331 | 35 | 1.583 | 0.208 |
| Yes | 7 | 2 | | |
| Epilepsy | Epilepsy | Epilepsy | Epilepsy | Epilepsy |
| No | 336 | 37 | 0.220 | 0.639 |
| Yes | 1 | 0 | | |
| Substance addiction | Substance addiction | Substance addiction | Substance addiction | Substance addiction |
| No | 333 | 36 | 0.317 | 0.573 |
| Yes | 5 | 1 | | |
| Days in hospital (day) | 46.51 ± 30.59 | 55.18 ± 23.85 | −2.760 | 0.006 |
| Blood parameters | Blood parameters | Blood parameters | Blood parameters | Blood parameters |
| TB (μmol/L) | 17.00 ± 9.00 | 16.76 ± 7.95 | −0.212 | 0.832 |
| DB (μmol/L) | 4.24 ± 2.66 | 4.17 ± 2.42 | −0.133 | 0.894 |
| IB (μmol/L) | 12.47 ± 6.9 | 12.91 ± 6.19 | −0.736 | 0.461 |
| Glucose (mmol/L) | 4.95 ± 1.27 | 5.1 ± 1.57 | −0.998 | 0.318 |
| TC (mmol/L) | 4.41 ± 1.05 | 4.00 ± 1.14 | −2.147 | 0.032 |
| TG (mmol/L) | 1.22 ± 0.73 | 1.17 ± 0.66 | −0.545 | 0.586 |
| HDL (mmol/L) | 1.38 ± 0.33 | 1.42 ± 0.38 | −0.695 | 0.554 |
| LDL (mmol/L) | 2.27 ± 0.72 | 2.11 ± 0.54 | −1.223 | 0.221 |
| Albumin (g/L) | 42.97 ± 3.71 | 43.67 ± 3.33 | −1.089 | 0.276 |
| UA (μmol/L) | 414.31 ± 131.1 | 401.62 ± 136.6 | −0.443 | 0.658 |
| Leukocyte (109/L) | 7.69 ± 2.47 | 8.23 ± 3.05 | −0.814 | 0.416 |
| Monocyte (109/L) | 0.49 ± 0.18 | 0.54 ± 0.16 | −2.001 | 0.045 |
| Lymphocyte (109/L) | 1.77 ± 0.65 | 1.73 ± 0.54 | −0.093 | 0.926 |
| Eosinophils (109/L) | 0.12 ± 0.11 | 0.10 ± 0.10 | −1.012 | 0.311 |
| Basophils (109/L) | 0.03 ± 0.02 | 0.03 ± 0.03 | −2.065 | 0.039 |
| RBC (1012/L) | 4.88 ± 0.57 | 4.79 ± 0.36 | −0.660 | 0.509 |
| Hemoglobin (g/L) | 146.37 ± 13.71 | 145.57 ± 12.87 | −0.218 | 0.827 |
| Thrombocyte (109/L) | 209.41 ± 62.92 | 194.81 ± 70.01 | −1.187 | 0.235 |
| Medications Sedative-hypnotics | Medications Sedative-hypnotics | Medications Sedative-hypnotics | Medications Sedative-hypnotics | Medications Sedative-hypnotics |
| No | 181 | 9 | 11.396 | <0.001 |
| Yes | 157 | 28 | | |
| Anti-anxiety | Anti-anxiety | Anti-anxiety | Anti-anxiety | Anti-anxiety |
| No | 337 | 37 | 0.110 | 0.740 |
| Yes | 1 | 0 | | |
| Anti-manic | Anti-manic | Anti-manic | Anti-manic | Anti-manic |
| No | 332 | 35 | 2.105 | 0.147 |
| Yes | 6 | 2 | | |
| Antidepressants | Antidepressants | Antidepressants | Antidepressants | Antidepressants |
| No | 312 | 30 | 5.237 | 0.002 |
| Yes | 26 | 7 | | |
| Anti-epileptics | Anti-epileptics | Anti-epileptics | Anti-epileptics | Anti-epileptics |
| No | 283 | 24 | 7.993 | 0.005 |
| Yes | 55 | 13 | | |
| Anti-Parkinsonians | Anti-Parkinsonians | Anti-Parkinsonians | Anti-Parkinsonians | Anti-Parkinsonians |
| No | 250 | 15 | 17.973 | <0.001 |
| Yes | 88 | 22 | | |
| Other neuron drugs | Other neuron drugs | Other neuron drugs | Other neuron drugs | Other neuron drugs |
| No | 336 | 35 | 7.323 | 0.007 |
| Yes | 2 | 2 | | |
There were correlations between HAP and days in hospital (rs = 0.143, $$P \leq 0.006$$), total cholesterol (rs = −0.111, $$P \leq 0.032$$), monocyte count (rs = 0.103, $$P \leq 0.045$$), SHD (rs = 0.174, $$P \leq 0.001$$), ADD (rs = 0.118, $$P \leq 0.022$$), AED (rs = 0.146, $$P \leq 0.005$$), APD (rs = 0.219, $P \leq 0.001$), and other neurological drugs (rs = 0.140, $$P \leq 0.007$$) by Spearman test. After input of these variables into the logistic regression equation, only total cholesterol [Beta = −0.373, Wald = 3.920, $$P \leq 0.048$$, Exp(B) = 0.688 (0.476, 0.996)] and APD [Beta = 1.366, Wald = 14.84, $P \leq 0.001$, Exp(B) = 3.920 (1.925, 7.981)] survived.
## 3.4. Risk factors for HAP in female SCZ patients receiving mECT
In Table 4, statistically significant higher incidences of diabetes (χ2 = 6.203, $$P \leq 0.013$$), hypertension (χ2 = 9.096, $$P \leq 0.003$$), epilepsy (χ2 = 5.643, $P \leq 0.001$), lymphocyte count (Z = −2.408, $$P \leq 0.016$$), eosinophil count (Z = −2.141, $$P \leq 0.032$$), SHD (χ2 = 13.636, $P \leq 0.001$), and APD (χ2 = 8.526, $$P \leq 0.004$$) were identified in the HAP group, as compared to the non-HAP group, while lower levels of total bilirubin (Z = −2.024, $$P \leq 0.043$$) on admission were found in the HAP group. Only variables of epilepsy and SHD passed the Bonferroni test.
**Table 4**
| Unnamed: 0 | Non-HAP (n = 551) | HAP (n = 25) | t, Z, χ2 | P |
| --- | --- | --- | --- | --- |
| Demographic data | Demographic data | Demographic data | Demographic data | Demographic data |
| Age (year) | 36.74 ± 13.80 | 39.16 ± 17.22 | −0.508 | 0.611 |
| Diabetes | Diabetes | Diabetes | Diabetes | Diabetes |
| No | 535 | 22 | 6.203 | 0.013 |
| Yes | 16 | 3 | | |
| Hypertension | Hypertension | Hypertension | Hypertension | Hypertension |
| No | 539 | 22 | 9.096 | 0.003 |
| Yes | 12 | 3 | | |
| Epilepsy | Epilepsy | Epilepsy | Epilepsy | Epilepsy |
| No | 551 | 24 | 22.078 | <0.001 |
| Yes | 0 | 1 | | |
| Substance addiction | Substance addiction | Substance addiction | Substance addiction | Substance addiction |
| No | 549 | 25 | 0.091 | 0.763 |
| Yes | 2 | 0 | | |
| Days in hospital (day) | 46.43 ± 32.16 | 51.67 ± 30.44 | −1.301 | 0.193 |
| Blood parameters | Blood parameters | Blood parameters | Blood parameters | Blood parameters |
| TB (μmol/L) | 15.39 ± 7.92 | 12.70 ± 7.18 | −2.024 | 0.043 |
| DB (μmol/L) | 3.24 ± 1.68 | 3.25 ± 2.13 | −0.774 | 0.439 |
| IB (μmol/L) | 10.69 ± 5.80 | 9.62 ± 5.77 | −0.923 | 0.356 |
| Glucose (mmol/L) | 5.25 ± 1.64 | 5.45 ± 1.31 | −1.110 | 0.267 |
| TC (mmol/L) | 4.58 ± 1.06 | 4.54 ± 1.00 | −0.176 | 0.861 |
| TG (mmol/L) | 1.11 ± 0.58 | 1.05 ± 0.54 | −0.621 | 0.535 |
| HDL (mmol/L) | 1.57 ± 0.37 | 1.57 ± 0.40 | −0.088 | 0.930 |
| LDL (mmol/L) | 2.26 ± 0.69 | 2.45 ± 0.78 | −1.079 | 0.281 |
| Albumin (g/L) | 41.54 ± 3.96 | 41.94 ± 3.91 | −0.573 | 0.567 |
| UA (μmol/L) | 312.70 ± 97.62 | 296.48 ± 74.93 | −0.481 | 0.630 |
| Leukocyte (109/L) | 7.29 ± 2.39 | 7.61 ± 2.54 | −1.103 | 0.270 |
| Monocyte (109/L) | 0.43 ± 0.18 | 0.42 ± 0.15 | −0.299 | 0.765 |
| Lymphocyte (109/L) | 1.76 ± 0.63 | 1.46 ± 0.56 | −2.408 | 0.016 |
| Eosinophils (109/L) | 0.11 ± 0.16 | 0.06 ± 0.06 | −2.141 | 0.032 |
| Basophils (109/L) | 0.03 ± 0.02 | 0.02 ± 0.02 | −1.640 | 0.101 |
| RBC (1012/L) | 4.32 ± 0.47 | 4.22 ± 0.47 | −1.030 | 0.303 |
| Hemoglobin (g/L) | 127.02 ± 12.76 | 124.92 ± 15.12 | −0.567 | 0.571 |
| Thrombocyte (109/L) | 230.49 ± 67.93 | 221.8 ± 75.92 | −0.464 | 0.643 |
| Medications Sedative-hypnotics | Medications Sedative-hypnotics | Medications Sedative-hypnotics | Medications Sedative-hypnotics | Medications Sedative-hypnotics |
| No | 274 | 3 | 13.636 | <0.001 |
| Yes | 277 | 22 | | |
| Anti-anxiety | Anti-anxiety | Anti-anxiety | Anti-anxiety | Anti-anxiety |
| No | 533 | 25 | 0.843 | 0.359 |
| Yes | 18 | 0 | | |
| Anti-manics | Anti-manics | Anti-manics | Anti-manics | Anti-manics |
| No | 546 | 24 | 2.219 | 0.136 |
| Yes | 5 | 1 | | |
| Anti-depressants | Anti-depressants | Anti-depressants | Anti-depressants | Anti-depressants |
| No | 506 | 24 | 0.565 | 0.452 |
| Yes | 45 | 1 | | |
| Anti-epileptics | Anti-epileptics | Anti-epileptics | Anti-epileptics | Anti-epileptics |
| No | 501 | 22 | 0.245 | 0.621 |
| Yes | 50 | 3 | | |
| Anti-Parkinsonians | Anti-Parkinsonians | Anti-Parkinsonians | Anti-Parkinsonians | Anti-Parkinsonians |
| No | 393 | 11 | 8.526 | 0.004 |
| Yes | 158 | 14 | | |
| Other Neuron Drugs | Other Neuron Drugs | Other Neuron Drugs | Other Neuron Drugs | Other Neuron Drugs |
| No | 541 | 25 | 0.462 | 0.497 |
| Yes | 10 | 0 | | |
Spearman testing showed that HAP is correlated with diabetes (rs = 0.104, $$P \leq 0.013$$), hypertension (rs = 0.126, $$P \leq 0.003$$), epilepsy (rs = 0.196, $P \leq 0.001$), total bilirubin (rs = −0.084, $$P \leq 0.043$$), lymphocyte count (rs = −0.100, $$P \leq 0.016$$), eosinophil count (rs = −0.089, $$P \leq 0.032$$), SHD (rs = 0.154, $P \leq 0.001$), and APD (rs = 0.122, $$P \leq 0.003$$). After input of these variables into the logistic regression equation, only lymphocyte count [Beta = 1.702, Wald = 5.432, $$P \leq 0.020$$, Exp(B) = 5.483 (1.311, 22.937)], hypertension [Beta = −0.835, Wald = 4.764, $$P \leq 0.029$$, Exp(B) = 0.434 (0.205, 0.918)], and SHD [Beta = 2.287, Wald = 9.387, $$P \leq 0.002$$, Exp(B) = 9.847 (2.280, 42.536)] survived.
## 4. Discussion
To the best of our knowledge, this is the first study to investigate the risk factors of HAP in patients with mECT treatment. We found that the incidence of HAP in mECT patients was $6.52\%$, greater than the $1.80\%$ previously reported by Han et al. in patients with schizophrenia spectrum disorder [38], but slightly less than the $7.8\%$ reported in a study of elderly SCZ patients (age >50) by Yang et al. [ 39].
ECT treatment improves the structure and function of the hippocampus and insula in SCZ patients and regulates the function of the prefrontal and thalamic striatum of the default network [40, 41]. Animal experiments have also shown that ECT attenuates microglia and astrocyte proliferation, thus improving schizophrenic behavior [42]. However, ECT treatment may also induce acute immunoinflammatory responses, such as elevated plasma cortisol and levels of interleukin-1 or−6; lowered levels of blood tumor necrosis factor alpha (TNF-α) and interleukin-6 in long-term treatment [43]; or even detrimentally alter blood parameters [44], leading to a decrease in immunity. Schizophrenia itself is also a risk factor for the development of pneumonia [45], thus ECT may increase the risk of HAP in patients, and our prior work [46] had found that mECT may be a risk factor influencing the occurrence of HAP. Further, we wanted to find whether there were gender differences in HAP, and the discussion of the results was divided into the following four parts.
## 4.1. mECT induced increased incidence of HAP
Our results showed that among HAP patients who underwent mECT, there was no statistical difference between men and women in the three indicators: days from first mECT treatment to HAP occurrence (dfE2H), the numbers of mECT treatments before HAP occurrence (nE2H), and days from last mECT treatment to HAP occurrence (dlE2H). However, the results showed that the median number of days from the last mECT treatment to HAP was 1 day in both men and women, indicating that all patients had a very high risk of developing HAP within 1 day after receiving mECT treatment. In addition, the risk of HAP within the first three mECT treatments was quite high in men and women, respectively, which may be due to the fact that patients need to gradually adapt to the clinical symptoms that may occur after mECT and ignore the possible risk of HAP. Therefore, healthcare providers should be on alert the first day after each mECT treatment, and special attention should be paid to clinical care after the first 3 sessions of mECT treatment.
Interestingly, we found that male patients with SCZ are more prone to HAP than women, and the risk factors predisposing both groups to HAP also differ from each other.
## 4.2. Risk factors for male mECT patients
The prevalence of HAP in male mECT patients in this study was significantly higher than that in women (male incidence ~2.3 times that of females). This increased prevalence may be related to lifestyle habits of male patients, such as smoking, alcohol abuse, low weight, frequent contact with children, and poor oral hygiene, which are all risk factors for HAP [47]. In addition, this study found that lower total cholesterol prior to hospital admission and APD medication in hospitalization may be risk factors for the development of HAP in men.
Our results showed that male SCZ patients treated with mECT with lower total cholesterol levels at admission were more likely to develop HAP, and logistic regression analysis showed that low levels of total cholesterol might be an independent risk factor for HAP in SCZ patients treated with mECT. Cholesterol is widely distributed in many tissues of the body, especially the brain and neural tissues, and may be associated with a variety of diseases [48], and even mental status or personality changes [49]. Importantly, cholesterol has an important role in coronavirus entry, membrane fusion, and pathological syncytium formation. 25-hydroxycholesterol (25HC) is one of the metabolites of cholesterol, and 25HC inhibits coronavirus infection by blocking membrane fusion [50], so lower total cholesterol levels may lead to lower 25HC levels, increasing the chance of coronavirus infection in SCZ patients and potentially explaining the correlation between viral infection and lower cholesterol levels [51, 52]. Lower total cholesterol may also be a factor of increased short-term mortality in elderly patients with community-acquired pneumonia (CAP) [53]. In addition, men may be more sensitive to low levels of cholesterol [54], which accounts for the fact that men with low total cholesterol levels were more likely to develop HAP in our study, whereas women showed no statistical difference in HAP based on the level of total cholesterol.
APDs are used to improve Parkinsonian-like symptoms and extrapyramidal effects in schizophrenic patients. Our results show that male mECT SCZ patients using APDs during hospitalization were more likely to develop HAP, and logistic regression analysis showed that APD use is an independent risk factor for the incidence of HAP. Relatedly, it has been shown that anticholinergic drugs may increase the risk of HAP [55]; despite some studies showing that amantadine can be used to prevent pneumonia [56], the presence of drowsiness, falls, and skin problems [57] may in turn lead to an increased risk of developing HAP in patients.
## 4.3. Risk factors for female mECT patients
Risk factors for HAP in women are hypertension, low lymphocyte count on admission, and use of SHDs during hospitalization. There was no statistically significant difference in the prevalence of hypertension between men and women in patients treated with mECT ($\frac{9}{375}$:$\frac{15}{576}$, χ2 = 0.038, $$P \leq 0.844$$), however, SCZ patients with comorbid hypertension in women had a higher risk of HAP, compared to men. ECT treatment may not result in significant changes in blood pressure in hypertensive patients [58], however, patients with comorbidities such as chronic renal insufficiency or diabetes mellitus are more likely to be affected by COVID-19, potentially fatally (59–61). Therefore, for female SCZ patients with comorbid diseases such as hypertension, clinical management should be particularly strengthened during hospitalization to reduce the risk of HAP and improve the quality of patient survival.
Lymphocytes have an important role in immune regulation and can be involved in the pathogenesis of respiratory diseases such as pneumonia, infections, asthma, and acute respiratory distress syndrome [62]. The lower lymphocyte count in female mECT patients suggests that immunity may be reduced, potentially increasing the incidence of HAP.
Benzodiazepines drugs (BZD) are often used in sedation-hypnosis and can modulate peripheral γ-aminobutyric acid (GABA) type A on macrophages and increase the incidence of infection by inhibiting proinflammatory cytokines [63]. Experiments performed on mice have further shown that benzodiazepines enhance GABA signaling, leading to increased mortality from pneumonia [64]. In addition, certain meta-analyses have shown an increased risk of pneumonia with recent or current exposure to benzodiazepines [65, 66], which is consistent with our findings. Therefore, BZDs should be used with caution in SCZ patients receiving mECT.
## 4.4. Limitations
Although mECT is used as a routine treatment for patients with schizophrenia, the number of patients receiving it remains low, which may stem from the fear of electrical stimulation. Since the outbreak of COVID-19, more standardized clinical management and increased awareness of personal protection implemented by medical institutions have led to a decrease in the number of patients with mECT experiencing HAP; therefore, random errors induced by small sample size may have a greater impact on the statistical analysis. Incomplete demographic indicators for some patients, such as height, weight, and education, as well as the lack of cognitive assessment of patients with schizophrenia during data collection, prevented our results from fully reflecting the full picture of patients. Data from an individual psychiatric hospital is another limitation, as regional differences, local culture, economic conditions, and ethnic groups of patients may also influence our results. Therefore, these risk factors deserve to be investigated as a prospective study with a larger patient sample size and collaboration of multiple clinical centers.
## 5. Conclusions
Among schizophrenia patients treated with mECT, men were more likely to develop HAP than women. Schizophrenia patients were at very high risk of developing HAP within the first day after each mECT treatment or in the first three sessions of mECT treatment. Lower levels of total cholesterol and use of anti-parkinsonian drugs were identified as independent risk factors of HAP in male patients, while hypertension, lower lymphocyte count on admission, and use of sedative-hypnotic drugs in hospitalization were identified as independent risk factors in female patients. These gender-based differences may be due to differences in physiological immune function, lifestyle habits, and the complicated nature of schizophrenia itself. Our results may help guide future clinical management and care of patients with SCZ, and help elucidate the potential direction of follow-up studies.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethics Committee of the Fourth People's Hospital of Chengdu. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.
## Author contributions
MY: conception, writing—original draft preparation, and funding acquisition. YY, DK and MX: data curation. LL: validation. XH: resources. GZ and XZ: writing—review. YH: writing—review and editing. YT: project administration and funding acquisition. ZL: writing—review, supervision, and funding acquisition. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Keeley JW, Gaebel W. **Symptom rating scales for schizophrenia and other primary psychotic disorders in ICD-11**. *Epidemiol Psychiatr Sci.* (2018) **27** 219-24. DOI: 10.1017/S2045796017000270
2. Tandon R, Gaebel W, Barch DM, Bustillo J, Gur RE, Heckers S. **Definition and description of schizophrenia in the DSM-5**. *Schizophr Res.* (2013) **150** 3-10. DOI: 10.1016/j.schres.2013.05.028
3. James SL, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N. **Global regional and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic analysis for the global burden of disease study 2017**. *Lancet.* (2018) **392** 1789-858. DOI: 10.1016/S0140-6736(18)32279-7
4. Charlson FJ, Ferrari AJ, Santomauro DF, Diminic S, Stockings E, Scott JG. **Global epidemiology and burden of schizophrenia: findings from the global burden of disease study 2016**. *Schizophr Bull.* (2018) **44** 1195-203. DOI: 10.1093/schbul/sby058
5. Peng MM, Xing J, Tang X, Wu Q, Wei D, Ran MS. **Disease-related risk factors for caregiver burden among family caregivers of persons with schizophrenia: a systematic review and meta-analysis**. *Int J Environ Res Public Health.* (2022) **19** 1862. DOI: 10.3390/ijerph19031862
6. Weber S, Scott JG, Chatterton ML. **Healthcare costs and resource use associated with negative symptoms of schizophrenia: a systematic literature review**. *Schizophr Res.* (2022) **241** 251-9. DOI: 10.1016/j.schres.2022.01.051
7. Hafner Han der Heiden W. **Epidemiology of schizophrenia**. *Can J Psychiatry.* (1997) **42** 139-51. DOI: 10.1177/070674379704200204
8. Adachi N, Hara T, Oana Y, Matsuura M, Okubo Y, Akanuma N. **Difference in age of onset of psychosis between epilepsy and schizophrenia**. *Epilepsy Res.* (2008) **78** 201-6. DOI: 10.1016/j.eplepsyres.2007.12.001
9. Ferrer-Quintero M, Green MF, Horan WP, Penn DL, Kern RS, Lee J. **The effect of sex on social cognition and functioning in schizophrenia**. *NPJ Schizophr.* (2021) **7** 57. DOI: 10.1038/s41537-021-00188-7
10. Kubota R, Okubo R, Ikezawa S, Matsui M, Adachi L, Wada A. **Sex differences in social cognition and association of social cognition and neurocognition in early course schizophrenia**. *Front Psychol.* (2022) **13** 867468. DOI: 10.3389/fpsyg.2022.867468
11. Torniainen M, Suvisaari J, Partonen T, Castaneda AE, Kuha A, Perala J. **Sex differences in cognition among persons with schizophrenia and healthy first-degree relatives**. *Psychiatry Res.* (2011) **188** 7-12. DOI: 10.1016/j.psychres.2010.11.009
12. Li AWY, Hui CLM, Lee EHM, Chang WC, Chan SKW, Chen EYH. **Gender differences in correlates of cognition in first-episode psychosis**. *Psychiatry Res.* (2019) **271** 412-20. DOI: 10.1016/j.psychres.2018.12.011
13. Barbui C, Nose M, Bindman J, Schene A, Becker T, Mazzi MA. **Sex differences in the subjective tolerability of antipsychotic drugs**. *J Clin Psychopharmacol.* (2005) **25** 521-6. DOI: 10.1097/01.jcp.0000185423.15891.02
14. Riedel M, Eich FX, Moller HJ. **A pilot study of the safety and efficacy of amisulpride and risperidone in elderly psychotic patients**. *Eur Psychiatry.* (2009) **24** 149-53. DOI: 10.1016/j.eurpsy.2008.10.005
15. Seeman MV. **Schizophrenia: women bear a disproportionate toll of antipsychotic side effects**. *J Am Psychiatr Nurses Assoc.* (2010) **16** 21-9. DOI: 10.1177/1078390309350918
16. Grossman LS, Harrow M, Rosen C, Faull R, Strauss GP. **Sex differences in schizophrenia and other psychotic disorders: a 20-year longitudinal study of psychosis and recovery**. *Compr Psychiatry.* (2008) **49** 523-9. DOI: 10.1016/j.comppsych.2008.03.004
17. Hoffman GE, Ma Y, Montgomery KS, Bendl J, Jaiswal MK, Kozlenkov A. **Sex differences in the human brain transcriptome of cases with schizophrenia**. *Biol Psychiatry.* (2022) **91** 92-101. DOI: 10.1016/j.biopsych.2021.03.020
18. Shobeiri P, Kalantari A, Teixeira AL, Rezaei N. **Shedding light on biological sex differences and microbiota-gut-brain axis: a comprehensive review of its roles in neuropsychiatric disorders**. *Biol Sex Differ.* (2022) **13** 12. DOI: 10.1186/s13293-022-00422-6
19. Rivera-Garcia MT, McCane AM, Chowdhury TG, Wallin-Miller KG, Moghaddam B. **Sex and strain differences in dynamic and static properties of the mesolimbic dopamine system**. *Neuropsychopharmacology.* (2020) **45** 2079-86. DOI: 10.1038/s41386-020-0765-1
20. Lewine R. **At issue: sex and gender in schizophrenia**. *Schizophr Bull.* (2004) **30** 755-62. DOI: 10.1093/oxfordjournals.schbul.a007128
21. Sinclair DJM, Zhao S, Qi F, Nyakyoma K, Kwong JSW, Adams CE. **Electroconvulsive therapy for treatment-resistant schizophrenia**. *Schizophr Bull.* (2019) **45** 730-2. DOI: 10.1093/schbul/sbz037
22. Chan CYW, Abdin E, Seow E, Subramaniam M, Liu J, Peh CX. **Clinical effectiveness and speed of response of electroconvulsive therapy in treatment-resistant schizophrenia**. *Psychiatry Clin Neurosci.* (2019) **73** 416-22. DOI: 10.1111/pcn.12855
23. Purohith AN, Chatorikar SA, Praharaj SK, Bhandary RP, Sharma P. **Efficacy and safety of maintenance electroconvulsive therapy (M-ECT) in treatment-resistant schizophrenia: a case series**. *Asian J Psychiatr.* (2022) **73** 103132. DOI: 10.1016/j.ajp.2022.103132
24. Kellner CH, Obbels J, Sienaert P. **When to consider electroconvulsive therapy (ECT)**. *Acta Psychiatr Scand.* (2020) **141** 304-15. DOI: 10.1111/acps.13134
25. Lava-Parmele S, Lava C, Parmele JB. **The historical struggles of modified electroconvulsive therapy: how anesthesia came to the rescue**. *J Anesth Hist.* (2021) **7** 17-25. DOI: 10.1016/j.janh.2021.03.001
26. Soehle M, Bochem J, Kayser S, Weyerhauser J, Valero R. **Challenges and pitfalls in anesthesia for electroconvulsive therapy**. *Best Pract Res Clin Anaesthesiol.* (2021) **35** 181-9. DOI: 10.1016/j.bpa.2020.12.012
27. Bryson EO, Aloysi AS, Farber KG, Kellner CH. **Individualized anesthetic management for patients undergoing electroconvulsive therapy: a review of current practice**. *Anesth Analg.* (2017) **124** 1943-56. DOI: 10.1213/ANE.0000000000001873
28. Kuehnert MJ, Webb RM, Jochimsen EM, Hancock GA, Arduino MJ, Hand S. **Staphylococcus aureus bloodstream infections among patients undergoing electroconvulsive therapy traced to breaks in infection control and possible extrinsic contamination by propofol**. *Anesth Analg.* (1997) **85** 420-5. DOI: 10.1097/00000539-199708000-00031
29. Blumberger DM, Seitz DP, Herrmann N, Kirkham JG, Ng R, Reimer C. **Low medical morbidity and mortality after acute courses of electroconvulsive therapy in a population-based sample**. *Acta Psychiatr Scand.* (2017) **136** 583-93. DOI: 10.1111/acps.12815
30. Roquilly A, Torres A, Villadangos JA, Netea MG, Dickson R, Becher B. **Pathophysiological role of respiratory dysbiosis in hospital-acquired pneumonia**. *Lancet Respir Med.* (2019) **7** 710-20. DOI: 10.1016/S2213-2600(19)30140-7
31. Jiao J, Yang XY Li Z, Zhao YW, Cao J, Li FF. **Incidence and related factors for hospital-acquired pneumonia among older bedridden patients in china: a hospital-based multicenter registry data based study**. *Front Public Health.* (2019) **7** 221. DOI: 10.3389/fpubh.2019.00221
32. Kim BG, Kang M, Lim J, Lee J, Kang D, Kim M. **Comprehensive risk assessment for hospital-acquired pneumonia: sociodemographic, clinical, and hospital environmental factors associated with the incidence of hospital-acquired pneumonia**. *BMC Pulm Med.* (2022) **22** 21. DOI: 10.1186/s12890-021-01816-9
33. Goncalves-Pereira J, Mergulhao P, Nunes B, Froes F. **Incidence and impact of hospital-acquired pneumonia: a portuguese nationwide four-year study**. *J Hosp Infect.* (2021) **112** 1-5. DOI: 10.1016/j.jhin.2021.03.012
34. Lopez-de-Andres A, Albaladejo-Vicente R, de Miguel-Diez J, Hernandez-Barrera V, Ji Z, Zamorano-Leon JJ. **Gender differences in incidence and in-hospital outcomes of community-acquired, ventilator-associated and nonventilator hospital-acquired pneumonia in Spain**. *Int J Clin Pract.* (2021) **75** e13762. DOI: 10.1111/ijcp.13762
35. Lopez-de-Andres A, Lopez-Herranz M, Hernandez-Barrera V, de-Miguel-Diez J, de-Miguel-Yanes JM, Carabantes-Alarcon D. **Sex differences in hospital-acquired pneumonia among patients with type 2 diabetes mellitus patients: retrospective cohort study using hospital discharge data in Spain (2016-2019)**. *Int J Environ Res Public Health.* (2021) **18** 12645. DOI: 10.3390/ijerph182312645
36. de-Miguel-Diez J, Jimenez-Garcia R, Hernandez-Barrera V, de-Miguel-Yanes JM, Carabantes-Alarcon D, Lopez-de-Andres A. **Assessing the impact of gender and COPD on the incidence and mortality of hospital-acquired pneumonia. A retrospective cohort study using the Spanish National Discharge Database (2016–2019)**. *J Clin Med.* (2021) **10** e5453. DOI: 10.3390/jcm10225453
37. Modi AR, Kovacs CS. **Hospital-acquired and ventilator-associated pneumonia: diagnosis, management, and prevention**. *Cleve Clin J Med.* (2020) **87** 633-9. DOI: 10.3949/ccjm.87a.19117
38. Han J, Lv Z, Shen M, Wan Q, Xiao L, Wang G. **Risk factors for hospital-acquired pneumonia among inpatients with mental disorders in a large mental health center within a tertiary general hospital**. *Am J Infect Control.* (2022). DOI: 10.1016/j.ajic.2022.06.014
39. Yang M, Li Q, Wang C, Li L, Xu M, Yan F. **Influencing factors of hospital-acquired pneumonia infection in the middle-aged and elderly patients with schizophrenia**. *Front Psychiatry.* (2021) **12** 746791. DOI: 10.3389/fpsyt.2021.746791
40. Moon SY, Kim M, Lho SK, Oh S, Kim SH, Kwon JS. **Systematic review of the neural effect of electroconvulsive therapy in patients with schizophrenia: hippocampus and insula as the key regions of modulation**. *Psychiatry Investig.* (2021) **18** 486-99. DOI: 10.30773/pi.2020.0438
41. Thomann PA, Wolf RC, Nolte HM, Hirjak D, Hofer S, Seidl U. **Neuromodulation in response to electroconvulsive therapy in schizophrenia and major depression**. *Brain Stimul.* (2017) **10** 637-44. DOI: 10.1016/j.brs.2017.01.578
42. Limoa E, Hashioka S, Miyaoka T, Tsuchie K, Arauchi R, Azis IA. **Electroconvulsive shock attenuated microgliosis and astrogliosis in the hippocampus and ameliorated schizophrenia-like behavior of gunn rat**. *J Neuroinflammation.* (2016) **13** 230. DOI: 10.1186/s12974-016-0688-2
43. Yrondi A, Sporer M, Peran P, Schmitt L, Arbus C, Sauvaget A. **Electroconvulsive therapy, depression, the immune system and inflammation: a systematic review**. *Brain Stimul.* (2018) **11** 29-51. DOI: 10.1016/j.brs.2017.10.013
44. Chaturvedi S, Chadda RK, Rusia U, Jain N. **Effect of electroconvulsive therapy on hematological parameters**. *Psychiatry Res.* (2001) **104** 265-8. DOI: 10.1016/s0165-1781(01)00303-1
45. Karaoulanis SE, Christodoulou NG. **Do patients with schizophrenia have higher infection and mortality rates due to COVID-19? A systematic review**. *Psychiatriki.* (2021) **32** 219-23. DOI: 10.22365/jpsych.2021.027
46. Yang Y, Mi Y, Di K, Wei C, Guocheng Z, Xi T. **Non-antipsychotic medicines and modified electroconvulsive therapy are risk factors for hospital-acquired pneumonia in schizophrenia patients**. *Front Psychiatry.* (2023). DOI: 10.3389/fpsyt.2022.1071079
47. Torres A, Peetermans WE, Viegi G, Blasi F. **Risk factors for community-acquired pneumonia in adults in Europe: a literature review**. *Thorax.* (2013) **68** 1057-65. DOI: 10.1136/thoraxjnl-2013-204282
48. Song Y, Liu J, Zhao K, Gao L, Zhao J. **Cholesterol-induced toxicity: an integrated view of the role of cholesterol in multiple diseases**. *Cell Metab.* (2021) **33** 1911-25. DOI: 10.1016/j.cmet.2021.09.001
49. Boston PF, Dursun SM, Reveley MA. **Cholesterol and mental disorder**. *Br J Psychiatry.* (1996) **169** 682-9. DOI: 10.1192/bjp.169.6.682
50. Zang R, Case JB, Yutuc E, Ma X, Shen S, Gomez Castro MF. **Cholesterol 25-hydroxylase suppresses SARS-CoV-2 replication by blocking membrane fusion**. *Proc Natl Acad Sci USA.* (2020) **117** 32105-13. DOI: 10.1073/pnas.2012197117
51. Dai J, Wang H, Liao Y, Tan L, Sun Y, Song C. **Coronavirus infection and cholesterol metabolism**. *Front Immunol.* (2022) **13** 791267. DOI: 10.3389/fimmu.2022.791267
52. Zinellu A, Paliogiannis P, Fois AG, Solidoro P, Carru C, Mangoni AA. **Cholesterol and triglyceride concentrations, COVID-19 severity, and mortality: a systematic review and meta-analysis with meta-regression**. *Front Public Health.* (2021) **9** 705916. DOI: 10.3389/fpubh.2021.705916
53. Ko SH, Lee JS, Kim SK, Jeong KY. **Serum cholesterol as a predictor of mortality among the elderly patients with pneumonia in the emergency department**. *Am J Emerg Med.* (2021) **45** 404-9. DOI: 10.1016/j.ajem.2020.09.012
54. Tomson-Johanson K, Harro J. **Low cholesterol, impulsivity and violence revisited**. *Curr Opin Endocrinol Diabetes Obes.* (2018) **25** 103-7. DOI: 10.1097/MED.0000000000000395
55. Lee CY, Cheng YD, Cheng WY, Tsai TH, Huang KH. **The prevalence of anticholinergic drugs and correlation with pneumonia in elderly patients: a population-based study in Taiwan**. *Int J Environ Res Public Health.* (2020) **17** 6260. DOI: 10.3390/ijerph17176260
56. Lehnert R, Pletz M, Reuss A, Schaberg T. **Antiviral medications in seasonal and pandemic influenza**. *Dtsch Arztebl Int.* (2016) **113** 799-807. DOI: 10.3238/arztebl.2016.0799
57. Mohammad Zadeh N, Mashinchi Asl NS, Forouharnejad K, Ghadimi K, Parsa S, Mohammadi S. **Mechanism and adverse effects of COVID-19 drugs: a basic review**. *Int J Physiol Pathophysiol Pharmacol.* (2021) **13** 102-9. PMID: 34540130
58. Albin SM, Stevens SR, Rasmussen KG. **Blood pressure before and after electroconvulsive therapy in hypertensive and nonhypertensive patients**. *J ECT.* (2007) **23** 9-10. DOI: 10.1097/01.yct.0000263252.75926.98
59. Huang S, Wang J, Liu F, Liu J, Cao G, Yang C. **COVID-19 patients with hypertension have more severe disease: a multicenter retrospective observational study**. *Hypertens Res.* (2020) **43** 824-31. DOI: 10.1038/s41440-020-0485-2
60. Petrakis V, Panagopoulos P, Papazoglou D, Papanas N. **Diabetes mellitus and hypertension as major risk factors of mortality from COVID-19 pneumonia**. *Exp Clin Endocrinol Diabetes.* (2022) **130** 205-6. DOI: 10.1055/a-1325-0381
61. Denova-Gutierrez E, Lopez-Gatell H, Alomia-Zegarra JL, Lopez-Ridaura R, Zaragoza-Jimenez CA, Dyer-Leal DD. **The association of obesity, type 2 diabetes, and hypertension with severe coronavirus disease 2019 on admission among Mexican patients**. *Obesity (Silver Spring).* (2020) **28** 1826-32. DOI: 10.1002/oby.22946
62. Deng Z, Zheng Y, Cai P, Zheng Z. **The role of B and T lymphocyte attenuator in respiratory system diseases**. *Front Immunol.* (2021) **12** 635623. DOI: 10.3389/fimmu.2021.635623
63. Skicki E, Morgan M, Brown C, Bradburn E, Rogers F. **Benzodiazepines increase the likelihood of both infectious and thrombotic complications**. *J Trauma Acute Care Surg.* (2021) **91** 206-11. DOI: 10.1097/TA.0000000000003134
64. Sanders RD, Godlee A, Fujimori T, Goulding J, Xin G, Salek-Ardakani S. **Benzodiazepine augmented gamma-amino-butyric acid signaling increases mortality from pneumonia in mice**. *Crit Care Med.* (2013) **41** 1627-36. DOI: 10.1097/CCM.0b013e31827c0c8d
65. Rajamaki B, Hartikainen S, Tolppanen AM. **Psychotropic drug-associated pneumonia in older adults**. *Drugs Aging.* (2020) **37** 241-61. DOI: 10.1007/s40266-020-00754-1
66. Sun GQ, Zhang L, Zhang LN, Wu Z, Hu DF. **Benzodiazepines or related drugs and risk of pneumonia: a systematic review and meta-analysis**. *Int J Geriatr Psychiatry.* (2019) **34** 513-21. DOI: 10.1002/gps.5048
|
---
title: Evaluation of food and nutrient intake in a population of subjects affected
by periodontal disease with different levels of bone mineral density
authors:
- Leonardo Guasti
- Luisella Cianferotti
- Barbara Pampaloni
- Francesco Tonelli
- Francesco Martelli
- Teresa Iantomasi
- Maria Luisa Brandi
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9971598
doi: 10.3389/fendo.2023.1098366
license: CC BY 4.0
---
# Evaluation of food and nutrient intake in a population of subjects affected by periodontal disease with different levels of bone mineral density
## Abstract
### Introduction
Both osteoporosis and periodontitis are pathologies characterized by an imbalance in the bone tissue. Vitamin C is an important factor involved in maintaining the health of the periodontium; its deficiency causes characteristic lesions to periodontal tissues such as bleeding and redness of the gums. Among the essential minerals for the health of the periodontium we find instead calcium.
### Objectives of the study
The objectives of the proposed study are to study the association between the presence of osteoporosis and periodontal disease. We tried to identify the possible connections between particular dietary patterns and therefore the etiopathogenesis of periodontal disease and secondarily of osteoporosis.
### Materials and methods
110 subjects were recruited in a single-center observational cross-sectional study carried through the collaboration between the University of Florence and the private institute of dentistry Excellence Dental Network based in Florence, suffering of periodontitis, 71 osteoporotic/osteopenic and 39 non-osteoporotic/osteopenic. Anamnestic data and information on eating habits were collected.
### Results
The population showed eating habits that do not meet the intake levels recommended by the L.A.R.N. Regarding the relationship between nutrient intake and plaque index, it appears that in the population, the higher the intake of vitamin C through food, the lower the plaque index value is. This result could reinforce the scientific evidence that there is a protective factor in the onset of periodontal disease by the consumption of vitamin C which to date is still the subject of investigation. In addition, the same type of trend would also have been observed for calcium intake, but a larger sample size would be required to make this effect significant.
### Conclusions
The relationship between osteoporosis and periodontitis and the role of nutrition in influencing the evolution of these pathologies still seems to be deeply explored. However, the results obtained seem to consolidate the idea that there is a relationship between these two diseases and that eating habits play an important role in their prevention.
## Introduction
Periodontitis is a chronic inflammatory disease in which bacterial infection of periodontal tissues is necessary and sufficient for the onset and progression of the oral pathology. Yet, numerous other factors negatively affect the course of the disease, like smoking, hormonal changes, endocrine or systemic comorbidities, and poor oral hygiene. It is a condition with poly-microbial etiology which specifically affects the supporting tissues of the teeth. According to the Global Burden of Disease 2010, the global prevalence of severe periodontitis between 1990 and 2010, standardized by age, was $11.2\%$, i.e. the sixth most widespread disease in the world [1]. Age-standardized incidence of the more severe forms in 2010 was similar to that of 1990, with 701 cases per 100,000 individuals per year. Prevalence gradually increased with age, showing a large increase between the third and fourth decades of life, with a peak at about 38 years old.
Osteoporosis is a systemic skeletal disease characterized by a reduction in bone mass and qualitative changes in the material properties of the macro- and micro-architecture of the bones, this leading to an increased risk of fracture even in case of minor traumas [2]. Nowadays, densitometry allows to accurately measure the bone mass and BMD (bone mineral density) in g/cm2 of bone surface, a property responsible for 60-$80\%$ of the mechanical resistance of a bone. According to the WHO (World Health Organization), diagnosis of osteoporosis made by densitometry relies on the assessment of bone mineral density through dual-energy x-ray absorptiometry (DXA) and on its comparison with the average value in healthy adult subjects. Standard deviation from the average peak of bone mass (T-score) is used as unit of measurement. This procedure represents the diagnostic test for osteoporosis and risk of fracture [2]. To this regard, it has been observed that risk of fracture starts to exponentially increase with densitometry values of T-score <-2.5 SD, which indeed represents the threshold for diagnosing the presence of osteoporosis.
QUS systems (calcaneus Ultrasonography) are techniques used more and more commonly for densitometry investigations within population groups. Through the measurement of ultrasound variables applied to the calcaneus, they provide a clinical parameter known as Stiffness Index. The Stiffness Index describes the risk of osteoporotic fracture for post-menopausal women comparable to the BMD measured by DXA of the spine or hip.
After reviewing several studies performed using the QUS systems, the FDA approved the use of this procedure for the prevention from the risk of hip fracture, comparable to the DEXA study of the hip/spine.
Based on these diagnostic criteria, approximately $6\%$ of men and $22\%$ of women between 50 and 84 years old in the EU have osteoporosis, i.e. 27.6 million people [3].
Both osteoporosis and periodontitis are characterized by an imbalance between bone tissue resorption and bone tissue neo-genesis, which finally results in bone tissue loss. Furthermore, recent studies suggested the possibility that these two pathologies are inter-connected, playing a role as reciprocal risk factors [4] [5].
Evaluating the possible implications that different diets could have on the oral bacterial population is an aspect still poorly investigated.
A large number of nutrients impact periodontal health, among these macro and micro-elements must be distinguished.
With reference to periodontal disease, the most important macronutrients (carbohydrates, proteins and lipids) are carbohydrates, which are involved in the progressing of periodontal disease associated with dental caries. In fact, high carbohydrate intake seems to promote oral dysbiosis, while its reduction decreases gingival inflammation. Furthermore, diets rich in saturated fats, known to increase oxidative stress, should be avoided in order to prevent the onset of periodontitis [6]. To this regard, a higher body fat content has been associated with an increased gingival bleeding in older patients. On the other hand, polyunsaturated fats (such as omega 3) have shown a positive effect on periodontium conditions [7]. Other studies conducted instead on protein deprived animals have resulted in the rupture of periodontal ligaments, degeneration of gingival tissues, and resorption of the alveolar bone [8]. Another study has finally suggested an inverse relationship between high protein intake and periodontitis [9].
Strong association has also been reported between periodontal disease and obesity [10], a problem that has long been underestimated. Obesity has been indeed identified as risk factor for a number of systemic diseases with inflammatory background. A longitudinal study identified a positive correlation between BMI and the incidence of periodontal disease [6].
The correlation between nutrition and periodontal disease is for various reasons rather uncertain. However, although dental bacterial plaque is accepted as the main causative agent, incorrect dietary can affect both onset and course of the disease.
Several studies have demonstrated the influence of dental plaque as the main etiological factor for gingival inflammation, noting worsening of gingivitis when study participants stopped oral hygiene procedures [11] [12] [13]. The authors therefore concluded that the gingivitis experimental protocol is not applicable if the diet does not include refined carbohydrates. Diet seems then to have a strong influence on the gum and on the inflammatory reaction of the periodontal. These effects could affect both tissue repair and immune system mechanisms, which are affected by: However, although dietary imbalance is not sufficient to induce periodontal disease without the simultaneous presence of the bacterial plaque, it can likewise affect its severity and extension, altering the resistance of the host organism and the regenerative capability of the tissues.
Nutritional recommendations to keep the periodontium in health include reduction of carbohydrate consumption and supplementation of omega 3 fatty acids, vitamin C, vitamin D, antioxidants and fibers [14]. In vitro studies have shown that vitamin C and D intake may play an important role in the prevention of gingivitis and periodontal inflammation [15].
Evidence in the recent literature reinforces the concept of how important is vitamin D for periodontal health. Periodontal disease seems to be correlated with low vitamin D serum levels; in a cohort of pregnant women over 20 weeks of gestation from the University Hospital “*Maggiore della* Carità”, Novara, Italy, the authors assessed serum levels of vitamin D and oral health status through the following indexes: Oral Hygiene Index (OHI), Plaque Control Record (PCR), Gingival Bleeding Index (GBI), and Community Periodontal Index of Treatment Needs (CPTIN). They finally found strong correlation between low serum levels of vitamin D and the idexes that identify periodontal disease [16].
Another relevant study indicates that Breast cancer (BC) survivors treated with aromatase inhibitors (AIs) commonly show several pathological issues, including poor oral health, bone health impairment, and vitamin D deficiency. This study evaluates the correlation between oral health and vitamin D status in BC survivors undergoing treatment with AIs through a machine learning approach. It’s showed a significant correlation between DMFT and vitamin D levels; the regression machine learning model showed that vitamin D status and the use of dental floss were the most relevant variables in terms of correlation with Filled Permanent Teeth Index (DMFT). Vitamin D deficiency, inadequate use of dental floss have a negative impact on oral health in BC women [17].
Vitamin C is an important factor for maintaining the health of the periodontium. Its deficiency causes characteristic lesions of the periodontal tissues, such as bleeding and redness of the gums. In vitro studies suggest that local applications of vitamin C and magnesium salt decrease inflammation at the level of the gingival fibroblasts [18].
Among the essential minerals for the health of the periodontium there is calcium, which is essential for calcified tissues such as bones and teeth. Deficiencies of this mineral can affect the health of the periodontium. A Danish study demonstrated that a higher dairy intake decreases the severity of periodontitis in adulthood [19].
The aim of this work is therefore to study the association between the presence of osteoporosis and periodontal pathology. We did it by identifying possible connections between specific diets and the etio-pathogenesis of periodontal disease, primarily, and osteoporosis, secondarily.
## Materials and methods
In the scope of a monocentric cross-sectional observational study carried out from November 2019 to December 2021 through the collaboration between the University of Florence and the private dentistry institute Excellence Dental Network based in Florence, were recruited 110 subjects (36 males and 73 females) affected by periodontitis, 71 of which also affected by osteoporosis or osteopenia. Detailed protocol was submitted to and approved by the Ethics Committee.
Criteria for the inclusion in the study were [1] written and signed declaration of informed consent, [2] age ≥ 18 years old, for both sexes, and [3] the diagnosis of periodontitis.
Exclusion criteria were instead [1] to have done antibiotic therapies or [2] steroid therapies in the past 3 months prior to the beginning of the study, [3] presence of parathyroid diseases or [4] diseases related to bone metabolism (except osteoporosis), and [5] development of neoplastic pathologies in the last five years. Subjects presenting eating disorders or pregnancy were also excluded.
Participants, all attending the abovementioned dental institute and all respecting the above listed inclusion criteria, were introduced to the project and provided of the relative documentation (patient information, informed consent, privacy policy). Data was collected anonymously, through the assignment of a specific alphanumeric code to each participant, organized in an electronic database and stored for the purposes expressed in the following scientific research. Two questionnaires for the collection of anamnestic information were administered to the participants, one for information related to the existence of previous pathologies and risk factors for osteoporosis, one for information on eating habits, with particular attention to vitamins and minerals diet intake. An attendance questionnaire already validated by Montomoli and collaborators was also administered [20]. Information about potential intake of supplements has also been recorded. The questionnaire consisted of sixteen questions related to nutrient intake. Food selection included in the questionnaire was based on data obtained by the Italian Institute of Nutrition, related to the composition of the Italian diet, to the frequency with which foods are consumed, and their relative importance as sources of calcium. Regarding cheese, more questions were asked to better identify the types of cheese consumed. Main food classes were included in the questionnaire: cereals (pasta, rice, bread and similar and potatoes), fish-meat, eggs, legumes, vegetables, and fruit. Two questions related to the consumption of sweets rich in calcium that are commonly consumed by the Italian population, such as milk-based ice cream and milk chocolate, were also present. Finally, contribution to calcium intake from drinking water was also carefully evaluated, as it can represent an important source of this mineral. A list of the most consumed calcium-rich mineral waters commercialized in Italy has been attached to the Food frequency Questionnaire (FFQ). For each question, participants were also asked to indicate the amount of product consumed, selecting between small, medium, or large portion. In addition to calcium intake, the questionnaire also aimed to estimate the participants’ intake of other macro and micro nutrients important for the health of bones and teeth, i.e. carbohydrates, proteins, lipids, phosphorus, sodium, iron, magnesium, potassium, selenium, zinc, vitamin C, vitamin D, and vitamin B12.
All data collected by the qualified staff of the IRF institute in Microdentistry and the Biomolecular Diagnostic laboratory were transcribed and archived in a database suitably prepared based on the purposes of this study: *Statistical analysis* was carried out using SPSS software and Microsoft Excel. Quantitative data derived from the analysis of parameters relating to osteoporosis, periodontal disease and dietary habits (t-score, bacterial concentrations, frequency of food consumption, etc.) were described through the use of statistical indices of central tendency mean and variability (minimum, maximum, range, standard deviation). For qualitative data, most suitable descriptive statistics (frequency distributions, relative frequencies, etc.) were presented.
After classifying the patients on the basis of the presence or absence of osteoporosis/ostepenia, groups were compared in relation to a series of quantitative variables (bacterial concentrations, daily calcium intake, etc.) using the most suitable statistical tests for independent samples (or an equivalent non-parametric test in the case of non-normal distributions). In addition, the association between osteoporosis/osteopenia and qualitative variables (presence of periodontitis, eating habits) was evaluated.
After having classified patients according to the severity of periodontal disease (mild, moderate, severe), groups were compared in relation to a series of quantitative variables (t-score, blood parameters of bone metabolism) by ANOVA and consequent post-hoc corrected with the Bonferroni method (or an equivalent non-parametric test in the case of non-normal distributions). In addition, the association between the severity of periodontal disease and qualitative variables (presence of osteoporosis/osteopenia, eating habits, etc.) was evaluated.
## Descriptive statistics of the study population
Over the three-year project, 181 individuals were asked to take part in the study, 110 of which agreed to participate. The others refused to participate because not interested.
All 110 participants reported data about the mineralometric and metabolic condition of their bones, genetic analyses and microbiological features of their periodontal.
Only 44 people were finally able to provide anamnestic information and eating habits, this likely because of the unfortunate ongoing pandemic situation, which prevented direct contact with the subjects. Participants were thus contacted by telephone in order to have answers to the questions. Many subjects, at the time of the call, were either not found or no longer interested in participating to the project.
The descriptive data collected are shown below. Overall, we obtained data from 110 subjects, 36 males and 74 females.
## Average age and anthropometry
The average age across the whole sample was 55 years old, while in regard to the anthropometric characteristics the average weight was 67 kg, the average height 1.68 m, and BMI 23.65, i.e. normal weight (Table 1).
**Table 1**
| Age and anthropometry | N | Minimum | Maximum | Average | std. deviation | Asymmetry |
| --- | --- | --- | --- | --- | --- | --- |
| Age and anthropometry | Statistics | Statistics | Statistics | Statistics | Statistics | Statistics |
| Age | 110 | 27 | 101 | 5522 | 13208 | ,478 |
| Weight (Kg) | 107 | 410 | 1200 | 67250 | 150090 | ,835 |
| Height (m) | 107 | 153 | 189 | 16832 | ,08513 | ,271 |
| BMI | 107 | 156226 | 419143 | 23652578 | 46736212 | 1449 |
## Bone turnover marker and bone mineralometry
Mineralometric data were collected from the entire sample of subjects, obtained through computerized bone mineralometry (MOC) or calcaneus ultrasound (QUS). When possible, data on blood concentration of 25OHD3, PTH, calcium, phosphatemia, alkaline phosphatase, and bone alkaline phosphatase, were also collected (Table 2).
**Table 2**
| Mineralometric data | N | Average | std. deviation |
| --- | --- | --- | --- |
| Mineralometric data | Statistics | Statistics | Statistics |
| Tscore lumbar | 15 | -2093 | 15917 |
| Tscore femur tot | 13 | -1823 | ,9671 |
| Tscore femur neck | 14 | -2157 | ,9788 |
| Tscore right foot | 99 | -,961 | 12538 |
| Tscore left foot | 99 | -,948 | 12966 |
On average, the value detected for 25OHD3 is insufficient if compared to the reference values.
The values of PTH, calcium, phosphatemia, alkaline phosphatase and bone alkaline phosphatase are instead in the average when compared to the reference values.
The average t-scores detected in the lumbar and femoral area through MOC examination fall below the desirable values, while the values detected by ultrasonography at the heel are on average in an optimal range, albeit at the limit.
$35.5\%$ of the sample had normal mineralometric values, $48.2\%$ had values tending to osteopenia and $16.4\%$ had values testifying a condition of osteoporosis (Table 3).
**Table 3**
| Exam result | Frequency | Valid percentage |
| --- | --- | --- |
| normal | 39 | 355 |
| osteopenia | 53 | 482 |
| osteoporosis | 18 | 164 |
| Total | 110 | 1000 |
## Classification, pocket depth (PPD), gingival recession (REC), plaque index
The classification of the severity of periodontal disease was obtained using the classification proposed by Amitage GC [21].
The analysis shows that patients always have generalized forms of periodontitis, in which chronic forms prevail over youthful/aggressive forms; in particular, $16.4\%$ have a mild chronic form, $41.8\%$ a moderate form and $32.7\%$ a severe form (Table 4).
**Table 4**
| Periodontitis (Amitage CG 1999) | Frequency | Valid percentage |
| --- | --- | --- |
| Chronic, mild, generalized | 18 | 164 |
| Chronic, moderate, generalized | 46 | 418 |
| Chronic, severe, generalized | 36 | 327 |
| Aggressive, mild, generalized | 1 | ,9 |
| Aggressive, moderate, generalized | 1 | ,9 |
| Aggressive, severe, generalized | 8 | 73 |
| Total | 110 | 1000 |
The mean pocket depth (PPD) detected by the test was 5.54 mm, therefore beyond the values considered physiological, while the mean gingival recession was 0.59 mm. The average plaque index was found to be $31.16\%$ (Table 5).
**Table 5**
| Unnamed: 0 | N | Minimum | Maximum | Average | Std.Deviation |
| --- | --- | --- | --- | --- | --- |
| | Statistics | Statistics | Statistics | Statistics | Statistics |
| PPD average (mm) | 110 | 220 | 1500 | 55414 | 142438 |
| REC average (mm) | 110 | ,00 | 240 | ,5877 | ,59648 |
| Plaque Index (%) | 97 | 0 | 9048 | 311609 | 2510672 |
## Consumption of the main foods important for the health of bones and teeth
To the subgroup of people interviewed by telephone was asked about eating habits for estimating calcium intake [20] and other nutrients important for bone and tooth health: Tables 6 and 7 show the average weekly intake frequencies and the average portions of the various foods taken into consideration.
In subjects who have declared their intake, milk is consumed almost every day, yogurt more occasionally, cheese only 2-3 times a week on average.
Pasta and bread have an average consumption of 3-4 times a week.
Meat and fish are eaten on average 4 times a week.
Legumes are consumed on average 2 times a week and vegetables at least once a day.
About fruit consumption, the question asked how many fruits per week were consumed; 37 out of 44 subjects replied to consume at least one fruit a week, the rest said they did not consume fruit. The average consumption of fresh fruit in the population was found to be 11.89 fruits per week (standard deviation ± 1479.56), or 1.69 fruits/day.
## Daily intake of the main nutrients useful for the health of bones and teeth
Table 8 shows the daily intake values of the main nutrients analyzed, which are important for the prevention of bone and dental health.
**Table 8**
| Unnamed: 0 | N | Minimum | Maximum | Average | Std. Deviation |
| --- | --- | --- | --- | --- | --- |
| | Statistics | Statistics | Statistics | Statistics | Statistics |
| Ca/die mg | 44 | 2189843 | 15075456 | 760717584 | 3216166047 |
| P/die mg | 44 | 3856909 | 14826531 | 931571488 | 2373435570 |
| Na/die mg | 44 | 4399885 | 19212101 | 973602366 | 3461657990 |
| Fe/die mg | 44 | 20775 | 127557 | 7893709 | 25206753 |
| Mg/die mg | 44 | 608884 | 2547129 | 171823180 | 532108933 |
| K/die mg | 44 | 5500149 | 33860009 | 2105757209 | 6676191528 |
| Se/die mcg | 44 | 114049 | 434157 | 27981607 | 80220100 |
| Zn/die mg | 44 | 36157 | 145086 | 8410503 | 22191760 |
| VitC mg | 44 | 48416 | 2303100 | 132337146 | 604172013 |
| VitD mcg | 44 | ,4771 | 31066 | 1710894 | ,7033637 |
| VitB12 mcg | 44 | 11937 | 72377 | 3843221 | 14569935 |
Considering the average age of the population (55 years), our data show that the intake of some nutrients in the population included in this study is below the values recommended by the Reference levels of nutrient and energy intake (L.A.R.N.) [22], in particular: result insufficient compared to the recommended daily intake.
## Study of the differences in food consumption and nutrient intake between subjects with normal bone density, osteopenic and osteoporotic
By carrying out a univariate ANOVA test and related post hoc tests, an attempt was made to investigate whether there are significant differences in the intake of certain foods and nutrients between the different study groups, i.e. subjects with normal, osteopenic and osteoporotic t-score values.
From the analyses shown in Table 9, in which only the analyses carried out that have produced significant results are shown, we have obtained:
**Table 9**
| Dependent Variable | Result | Result.1 | Sig. |
| --- | --- | --- | --- |
| FrequencyLEGUME | Normal | Osteopenia | 32 |
| | | Osteoporosis | 332 |
| | Osteopenia | Normal | 32 |
| | | Osteoporosis | 998 |
| | Osteoporosis | Normal | 332 |
| | | Osteopenia | 998 |
| PortionVEGETABLES | Normal | Osteopenia | 1 |
| | | Osteoporosis | 38 |
| | Osteopenia | Normal | 1 |
| | | Osteoporosis | 852 |
| | Osteoporosis | Normal | 38 |
| | | Osteopenia | 852 |
| Frequency VEGETABLES | Normal | Osteopenia | 559 |
| | | Osteoporosis | 122 |
| | Osteopenia | Normal | 559 |
| | | Osteoporosis | 379 |
| | Osteoporosis | Normal | 122 |
| | | Osteopenia | 379 |
• Significant difference in the portion and frequency of consumption of fresh vegetables between normal and osteopenic subjects (higher consumption); • Significant difference in the portion and frequency of vegetable consumption between normal and osteoporotic subjects (higher consumption); • Significant difference in the frequency of consumption of legumes between normal and osteopenic subjects (higher consumption); • Significant difference in the frequency of consumption of legumes between normal and osteoporotic subjects (higher consumption). The other analyzes did not give significant results.
As regards the intake of the different nutrients in the three different categories of subjects, the following was found: The p values are shown in Table 10.
**Table 10**
| Dependent variable | Result mineralometry | Resultmineralometry | Error std. | Sig. |
| --- | --- | --- | --- | --- |
| Ca/die_mg | Normal | Osteopenia | 1035798 | 847 |
| | | Osteoporosis | 1616607 | 477 |
| | Osteopenia | Normal | 1035798 | 847 |
| | | Osteoporosis | 1625093 | 697 |
| | Osteoporosis | Normal | 1616607 | 477 |
| | | Osteopenia | 1625093 | 697 |
| P/die_mg | Normal | Osteopenia | 7630852 | 436 |
| | | Osteoporosis | 1190975 | 914 |
| | Osteopenia | Normal | 7630852 | 436 |
| | | Osteoporosis | 1197227 | 921 |
| | Osteoporosis | Normal | 1190975 | 914 |
| | | Osteopenia | 1197227 | 921 |
| Na/die_mg | Normal | Osteopenia | 1114268 | 472 |
| | | Osteoporosis | 1739078 | 824 |
| | Osteopenia | Normal | 1114268 | 472 |
| | | Osteoporosis | 1748207 | 986 |
| | Osteoporosis | Normal | 1739078 | 824 |
| | | Osteopenia | 1748207 | 986 |
| Fe/die_mg | Normal | Osteopenia | 717718 | 4 |
| | | Osteoporosis | 1120168 | 233 |
| | Osteopenia | Normal | 717718 | 4 |
| | | Osteoporosis | 1126048 | 849 |
| | Osteoporosis | Normal | 1120168 | 233 |
| | | Osteopenia | 1126048 | 849 |
| Mg/die_mg | Normal | Osteopenia | 1555401 | 13 |
| | | Osteoporosis | 2427571 | 167 |
| | Osteopenia | Normal | 1555401 | 13 |
| | | Osteoporosis | 2440314 | 999 |
| | Osteoporosis | Normal | 2427571 | 167 |
| | | Osteopenia | 2440314 | 999 |
| K/die_mg | Normal | Osteopenia | 1971462 | 15 |
| | Normal | Osteoporosis | 3076932 | 381 |
| | Osteopenia | Normal | 1971462 | 15 |
| | | Osteoporosis | 3093084 | 858 |
| | Osteoporosis | Normal | 3076932 | 381 |
| | | Osteopenia | 3093084 | 858 |
| Se/die_mcg | Normal | Osteopenia | 2515653 | 158 |
| | | Osteoporosis | 392627 | 756 |
| | Osteopenia | Normal | 2515653 | 158 |
| | | Osteoporosis | 3946881 | 881 |
| | Osteoporosis | Normal | 392627 | 756 |
| | | Osteopenia | 3946881 | 881 |
| Zn/die_mg | Normal | Osteopenia | 69102 | 178 |
| | | Osteoporosis | 10785 | 333 |
| | Osteopenia | Normal | 69102 | 178 |
| | | Osteoporosis | 1084161 | 96 |
| | Osteoporosis | Normal | 10785 | 333 |
| | | Osteopenia | 1084161 | 96 |
| VitC_mg | Normal | Osteopenia | 1735689 | 13 |
| | | Osteoporosis | 2708952 | 46 |
| | Osteopenia | Normal | 1735689 | 13 |
| | | Osteoporosis | 2723173 | 844 |
| | Osteoporosis | Normal | 2708952 | 46 |
| | | Osteopenia | 2723173 | 844 |
| VitD_mcg | Normal | Osteopenia | 225023 | 5 |
| | | Osteoporosis | 351201 | 912 |
| | Osteopenia | Normal | 225023 | 5 |
| | | Osteoporosis | 353045 | 502 |
| | Osteoporosis | Normal | 351201 | 912 |
| | | Osteopenia | 353045 | 502 |
| VitB12_mcg | Normal | Osteopenia | 472399 | 81 |
| | | Osteoporosis | 737289 | 88 |
| | Osteopenia | Normal | 472399 | 81 |
| | | Osteoporosis | 74116 | 659 |
| | Osteoporosis | Normal | 737289 | 88 |
| | | Osteopenia | 74116 | 659 |
The other analyses did not give significant results.
## Evaluation of the correlations between the plaque index and nutrient intake (calcium, carbohydrates and vitamin C)
Using the Spearman rank correlation coefficient, it was investigated whether there was a correlation between the intake values of certain nutrients important for bone and tooth health (calcium, carbohydrates and vitamin C) and the plaque index values (Table 11):
**Table 11**
| Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Plaque Index | Plaque Index.1 |
| --- | --- | --- | --- | --- |
| Rho di Spearman | Plaque Index | Correlation coefficientSig. (2-code) N | 1000 | 1000 |
| | Plaque Index | Correlation coefficientSig. (2-code) N | | |
| | Plaque Index | Correlation coefficientSig. (2-code) N | 97 | 97 |
| | VitC_mg | Correlation coefficientSig. (2-code) N | -,385 | -,385 |
| | VitC_mg | Correlation coefficientSig. (2-code) N | ,025 | ,025 |
| | VitC_mg | Correlation coefficientSig. (2-code) N | 34 | 34 |
| | Ca/die_mg | Correlation coefficientSig. (2-code) N | -,199 | -,199 |
| | Ca/die_mg | Correlation coefficientSig. (2-code) N | ,259 | ,259 |
| | Ca/die_mg | Correlation coefficientSig. (2-code) N | 34 | 34 |
## Descriptive evaluations
The data collected show us that the population involved in this study is made up of a group of subjects tending to belong to the over 50 age group, with BMI values within the recommended range, for both the male and female population. With regard to the subgroup to which it was possible to administer the anamnestic questionnaire, a population that tends to be “healthy” emerged. The data relating to bone turnover and mineralometric values describe a population with blood values of PTH, calcemia and phosphatemia tendentially in line with the desirable values, but with values of 25OHD3 below the optimal value. Furthermore, the t-score values of both QUS and MOC, when available, identify a population mainly made up of fragile subjects from the point of view of the risk of fracture and osteoporosis, with a smaller number of subjects with values tending to normal, which they were then compared for the variables under study with the more frail subjects.
## Periodontal and bone evaluations
The assumption on which subjects were recruited for participation to the study was that of being affected by periodontal disease. On this regard, sample analysis shows that the most represented forms have a chronic rather than an aggressive nature, especially of moderate or advanced type, with high gingival recession values and periodontal pocket depths in the tested sites.
## Nutritional evaluations
The population that provided answer to the food questionnaire showed eating habits that do not meet the intake levels recommended by the L.A.R.N. Nutrients important for bone health and relevant for the prevention of periodontal disease and mouth health in general, such as calcium, iron, magnesium, potassium and fiber, were found to be below the recommended values. It is no coincidence that these nutrients are mainly taken from foods of plant origin, such as fruit, vegetables and legumes. Foods that have proved to be deficient in the consumption habits of the population, both in terms of frequency of consumption and in amount consumed per portion. About the calcium intake, it is interesting to note that in our population this nutrient mainly derives from the intake of cheese, which also bring nutrients that are dangerous for health when consumed in excess (cholesterol and triglycerides), rather than from leaner foods, (milk and yogurt) which seem to be consumed much less frequently on average. This situation brings the attention on the need for a more efficient dietary education of the patients. To this regard, the dental staff could play an important intermediary role between patients and nutrition specialists, like nutritionists and dieticians, in order to limit and prevent nutritional imbalances predisposing to a whole series of pathologies like the periodontitis itself, osteoporosis, dismetabolisms and hypertension. Quite relevant results emerged from the analysis of significance and correlation; Specifically, it seems that subjects with lower t-score values, and therefore predisposed to a greater risk of bone fragility, show unsatisfactory but still higher intakes on average for many nutrients important for bone health among those that we have already mentioned (Fe, Mg, K, vit C, fibers). This result could be explained by the fact that a subject already defined at risk of developing pathology has already been sensitized, at least in part, to the importance of nutrition in the prevention of these pathologies by their generic physician or likewise by health personnel previously encountered. Real primary prevention, the one intended for healthy subjects, would therefore be an element of greater criticality, as people who still do not perceive a problem would tend to be less attentive to their eating habits. For this reason, it would be essential in the future to increase awareness of food choices even in healthy subjects, in order to prevent any nutrient shortcomings and future pathological developments.
Another result of great interest was obtained about the relationship between nutrient intake and plaque index, a factor that describes the severity of periodontal disease. From the correlation analyzes carried out, it appears that across the population, the higher the intake of vitamin C through food, the lower the plaque index value is. This result could reinforce the scientific evidence that there is a protective factor in the onset of periodontal disease from the consumption of vitamin C, which is currently subject of investigation and debate). Systematic reviews on this aspect were performed in latest years; in 2019 Akio et al. selected 14 articles corresponding to inclusion criteria after a full revision of 716 articles. The vitamin C intake and blood levels were negatively related to periodontal disease in all seven cross-sectional studies. The subjects who suffer from periodontitis presented a lower vitamin C intake and lower blood-vitamin C levels than the subjects without periodontal disease in the two case-control studies. The patients with a lower dietary intake or lower blood level of vitamin C showed a greater progression of periodontal disease than the controls. The intervention using vitamin C administration improved gingival bleeding in gingivitis, but not in periodontitis. Alveolar bone absorption was also not improved. The present systematic review suggested that vitamin C contributes to a reduced risk of periodontal disease [23]. In 2021 Hytham N. et al. performed another systematic review in which they found six studies fulfilled the inclusion criteria. Vitamin C supplementation helped improve bleeding indices in gingivitis but did not significantly lead to reduction of probing depths or clinical attachment gain for periodontitis. In this case, administration of vitamin C as an adjunct to non-surgical periodontal therapy did not result in clinically significant improvements in pocket probing depths at 3 months in periodontitis patients. With the limited evidence available, no recommendation can be made for supplementation of vitamin C in conjunction with initial periodontal therapy for subjects with periodontitis to improve primary treatment outcome measures. [ 24].
Once this aspect will be consolidated, we could then focus on the role that this vitamin plays in the prevention of the osteoporotic disease [25], another aspect currently under investigation, and define its importance in the prevention from the two diseases. Furthermore, a similar trend would have also been observed for calcium intake, but a larger sample size would be required to make this effect significant. In the future, we hope to be able to deepen the relationship of this nutrient with the prevention of periodontal disease, which is already fundamental in the prevention of bone health.
## Conclusions
The relationship between osteoporosis and periodontitis and the role of nutrition in influencing the course of these pathologies seems still to be extensively explored. However, our results consolidate the idea that there is a relationship between these two diseases, and that eating habits play an important role in their prevention. The analyses presented here may be of great interest for the development of future studies aiming to expand the sample size and to reduce the confounding factors present at the level of the studied populations.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by Comitato etico regionale per la sperimentazione clinica della regione Toscana. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
LG: Main author and performer of the study. LC: Supervisor of the project. BP: Revisor of the main article. FT: organized the collaboration and the cooperation between F.I.R.M.O. and EDN. FM: Main Chief of EDN. TI: organized the collaboration and the cooperation between F.I.R.M.O. and University of Florence. MB: Main Chief of F.I.R.M.O. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Kassebaum NJ, Bernabé E, Dahiya M, Bhandari B, Murray CJ, Marcenes W. **Global burden of severe periodontitis in 1990-2010: a**. *J Dent Res* (2014) **93**
2. Adami G, Fassio A, Rossini M, Caimmi C, Giollo A, Orsolini G. **Osteoporosis in rheumatic diseases**. *Int J Mol Sci* (2019) **20** 5867. PMID: 31766755
3. Hernlund E, Svedbom A, Ivergård M, Compston J, Cooper C, Stenmark J, McCloskey EV. **Osteoporosis in the European union: medical management, epidemiology and economic burden**. *Arch Osteoporosis* (2013)
4. Tezal M, Wactawski-Wende J, Grossi SG, Dmochowski J, Genco RJ. **Periodontal disease and the incidence of tooth loss in postmenopausal woman**. *J Periodontol* (2005) **76**. DOI: 10.1902/jop.2005.76.7.1123
5. Straka M, Straka-Trapezanlidis M, Deglovic J, Varga I. **Periodontitis and osteoporosis**. *Neuroendocrinol Lett* (2015) **36**
6. Varela-López A, Giampieri F, Bullón P, Battino M, Quiles JL. **Role of lipids in the onset, progression and treatment of periodontal disease. a systematic review of studies in humans**. *Int J Mol Sci* (2016) **17** 1-13. DOI: 10.3390/ijms17081202.
7. Schifferle RE. **Periodontal disease and nutrition: separating the evidence from current fads**. *Periodontol 2000* (2009) **50** 78-89. DOI: 10.1111/j.1600-0757.2008.00297.x
8. Chawla TN, Glickman I. **Protein deprivation and the periodontal structures of the albino rat**. *Oral Surg Oral Med Oral Pathol* (1951) **4** 578-602. DOI: 10.1016/0030-4220(51)90052-7
9. Stahl SS, Sandler HC, Cahn L. **The effect of protein deprivation upon the oral tissues of the rat and particularly upon periodontal structures under irritation**. *Oral Surg Oral Med Oral Pathol* (1955) **8**. DOI: 10.1016/0030-4220(55)90040-2
10. Park JB, Nam G. **Obesity in relation to oral health behaviors: an analysis of the Korea national health and nutrition examination survey**. *Exp Ter Med* (2016) **12**. DOI: 10.3892/etm.2016.3697
11. Loe H, Theilade E, Jensen SB. **Experimental gingivitis in man**. *J Periodontol* (1965). DOI: 10.1902/jop.1965.36.3.177
12. Brecx MC, Fröhlicher I, Gehr P, Lang NP. **Stereological observations on long-term experimental gingivitis in man**. *J Clin Periodontol* (1988) **15**. DOI: 10.1111/j.1600-051X.1988.tb02262.x
13. Baumgartner S, Imfeld T, Schicht O, Rath C, Persson RE, Persson GR. **The impact of the stone age diet on gingival conditions in the absence of oral hygiene**. *Periodontol* (2009) **80**. DOI: 10.1902/jop.2009.080376
14. Nielsen SJ, Trak-Fellermeier MA, Joshipura K, Dye BA. **Dietary fiber intake is inverselly associated with periodontal disease among US adults**. *J Nutr* (2016) **146**. DOI: 10.3945/jn.116.237065
15. Najeeb S, Zafar MS, Khurshid Z, Zohaib S, Almas K. **The role of nutrition in periodontal health: an update**. *Nutrients* (2016) **8** 530. DOI: 10.3390/nu8090530
16. Ferrillo M, Migliario M, Roccuzzo A, Molinero-Mourelle P, Falcicchio G, Umano GR. **Periodontal disease and vitamin d deficiency in pregnant**. *J Clin Med* (2021) **10** 4578. DOI: 10.3390/jcm10194578
17. Ferrillo M, Migliario M, Marotta N, Lippi L, Antonelli A, Calafiore D. **Oral health in breast cancer women with vitamin D deficiency: A machine learning study**. *J Clin Med* (2022) **11** 4662. DOI: 10.3390/jcm11164662
18. Rekha C, Poornima G, Manasa M, Abhipsa V, Pavithra Devi J, Vijay Kumar HT. **Ascorbic acid, total phenol content and antioxidant activity of fresh juices of four ripe and unripe citrus fruits**. *Chem Sci Trans* (2012) **1**. DOI: 10.7598/cst2012.182
19. Adegboye ARA, Chrisrtensen LB, Heitmann BL. **Intake of dairy products in relation to periodontitis in older danish adults**. *Nutrients* (2012) **4**. DOI: 10.3390/nu4091219
20. Montomoli M, Gonnelli S, Giacchi M, Mattei R, Cuda C, Rossi S, Gennari C. **Validation of a food frequency questionnaire for nutritional calcium intake assessment in Italian women**. *Eur J Clin Nutr* (2002) **56** 21-30. DOI: 10.1038/sj.ejcn.1601278
21. Armitage GC, Jeffcoat MK, Chadwick DE, Taggart EJ, Numabe Y, Landis JR. **Longitudinal evaluation of elastase as a marker for the progression of periodontitis**. *J Periodontol* (1994) **65**. DOI: 10.1902/jop.1994.65.2.120
22. 22
SINU. Livelli di assunzione di riferimento di nutrienti ed energia per una sana popolazione italiana. (2014), Melzo (MI): SICS Editore Srl.. *Livelli di assunzione di riferimento di nutrienti ed energia per una sana popolazione italiana* (2014)
23. Miura H, Tada A. **The relationship between vitamin c and periodontal diseases: A systematic review**. *Int J Environ Res Public Health* (2019) **16** 2472. PMID: 31336735
24. Hytham N, Fageeh HI. **Efficacy of vitamin c supplementation as an adjunct in the non-surgical management of periodontitis: a systematic review**. *Syst Rev* (2021) **10** 5. DOI: 10.1186/s13643-020-01554-9
25. Olga Brzezińska ZŁ. **Role of vitamin c in osteoporosis development and treatment-a literature review**. *Nutrients* (2020) **12** 2394. PMID: 32785080
|
---
title: The relationship between shifts in the rhizosphere microbial community and
root rot disease in a continuous cropping American ginseng system
authors:
- Yan-Meng Bi
- Xi-Mei Zhang
- Xiao-Lin Jiao
- Jun-Fei Li
- Na Peng
- Gei-Lin Tian
- Yi Wang
- Wei-Wei Gao
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC9971623
doi: 10.3389/fmicb.2023.1097742
license: CC BY 4.0
---
# The relationship between shifts in the rhizosphere microbial community and root rot disease in a continuous cropping American ginseng system
## Abstract
The root rot disease causes a great economic loss, and the disease severity usually increases as ginseng ages. However, it is still unclear whether the disease severity is related to changes in microorganisms during the entire growing stage of American ginseng. The present study examined the microbial community in the rhizosphere and the chemical properties of the soil in 1–4-year-old ginseng plants grown in different seasons at two different sites. Additionally, the study investigated ginseng plants' root rot disease index (DI). The results showed that the DI of ginseng increased 2.2 times in one sampling site and 4.7 times in another during the 4 years. With respect to the microbial community, the bacterial diversity increased with the seasons in the first, third, and fourth years but remained steady in the second year. The seasonal changing of relative abundances of bacteria and fungi showed the same trend in the first, third, and fourth years but not in the second year. Linear models revealed that the relative abundances of Blastococcus, Symbiobacterium, Goffeauzyma, Entoloma, Staphylotrichum, Gymnomyces, Hirsutella, Penicillium and Suillus spp. were negatively correlated with DI, while the relative abundance of Pandoraea, Rhizomicrobium, Hebeloma, Elaphomyces, Pseudeurotium, Fusarium, Geomyces, Polyscytalum, Remersonia, Rhizopus, Acremonium, Paraphaeosphaeria, Mortierella, and Metarhizium spp. were positively correlated with DI ($P \leq 0.05$). The Mantel test showed that soil chemical properties, including available nitrogen, phosphorus, potassium, calcium, magnesium, organic matter, and pH, were significantly correlated to microbial composition. The contents of available potassium and nitrogen were positively correlated with DI, while pH and organic matter were negatively correlated with DI. In summary, we can deduce that the second year is the key period for the shift of the American ginseng rhizosphere microbial community. Disease aggravation after the third year is related to the deterioration of the rhizosphere microecosystem.
## 1. Introduction
Plant-associated microorganisms play important roles in plant growth, nutrition, and resistance to biotic and abiotic stresses (Vandenkoornhuyse et al., 2015; Hu et al., 2018). Plant rhizospheres provide a rich environment in which diverse microbial communities, including plant-beneficial microbes and pathogenic microbes, coexist (Berendsen et al., 2012; Trivedi et al., 2020; Xia et al., 2021). It is believed that the plant selects its microbial partners through the influence of its rhizodeposits (Sasse et al., 2018). Thus, different plant species support host-specific microbial communities when grown on the same soil (Garbeva et al., 2008; Berg and Smalla, 2009; Xia et al., 2022).
For the past several years, the use of molecular approaches based on high-throughput sequencing has dramatically extended our knowledge of the plant rhizosphere microbial community and revealed the relationship between plant disease and its microbiome. A previous study on the microbiome of *Arabidopsis thaliana* indicated that plants could specifically recruit a group of resistance-inducing and growth-promoting beneficial microbes upon pathogen infection (Berendsen et al., 2018). Shen et al. [ 2019] demonstrated that biofertilizer application and fumigation could reduce banana Panama disease by establishing a beneficial soil microbiome. Wang et al. [ 2020] reported that no-tillage and residue management influenced the composition of the soil microbial community and increased the risk of maize root rot. Chen et al. [ 2020] studied the rhizosphere soil of a 12-year cropping strawberry and found that physicochemical properties, the abundance of key microorganisms, and some phenolic acids accumulated significantly, which might lead to the disease under a continuous cropping system. However, how the specific microbial community of perennial crops forms little by little is still not fully understood. Perennial crops have their particularities, and many perennial plants are valuable economic crops, but perennial plants' diseases usually worsen over years of cultivation (Li et al., 2020; Moore et al., 2022). It is unclear whether the aggravation of plant disease during the development stages was associated with the succession of its rhizosphere microbial community, though. The upshot of this better understanding will substantially impact various research and applications about soil microbial ecology and plant disease. American ginseng (*Panax quinquefolius* L.) is a perennial plant that is well-known globally for its eutherapeutic effect on some diseases (Sen et al., 2012; Singh et al., 2017). In agricultural practice, American ginseng is continuously cultivated for 3 or 4 years before harvest. During the growth of American ginseng, the root rot diseases caused by pathogens reduced the products and quality severely (Yang et al., 2009; Farh et al., 2018), and these diseases became more and more severe over the years of cultivation.
This study aimed to investigate the succession of bacterial and fungal communities and their relationships with root rot disease in American ginseng over years of cultivation. In addition, chemical properties were detected in a correspondence soil sample to analyze their correlation with microbial communities and disease. The results obtained from this study will be valuable for gaining insight into the impact of different cropping systems on soil micro-ecology, which can aid in the cultivation of perennial crops and enhance sustainable development in the medicinal industry.
## 2.1. Sample processing
The sample sites were in Wendeng Dist., the Wehai City of Shandong Province, one of China's main American ginseng-producing regions. In the first year of the study, 7.5 metric tons of organic fertilizer (with an organic matter content >$80\%$) were applied per hectare. In the following years, increasing amounts of compound fertilizer were applied per hectare: 150–200 kg in the second year, 300–400 kg in the third year, and 450–600 kg in the fourth year. The compound fertilizer used had a nitrogen, phosphorus, and potassium content >$16\%$. This region has a northern temperate marine monsoon climate and receives an annual precipitation of ~762 mm. The plow layer in the plantation consists of gray-brown soil. In 2017, it was decided to directly sow American ginseng for 1 to 4 years in adjacent charmilles of Xishuipo Village, Dashuipo Town (122°14′07″E, 37°10′48″N, marked as Site I). For each ginseng age, four sampling points (replicates) of about 10 m2 were chosen and marked for subsequent sampling.
During the spring (late May), summer (late July), and autumn (late September), at each sampling location, half-row ginsengs (6–10 individuals) were collected, and the rhizosphere soil of these ginsengs was thoroughly mixed. Some amounts of soil was stored at −80°C for DNA extraction, while the rest was air-dried for chemical analysis. The ginseng from each site collected in the summer and autumn was cleaned to calculate the disease index (DI). In the summer (July 20th) and autumn (September 20th) of 2018, the experiment was repeated in Liujiatuan Village of Zetou Town (121°51′44″E, 37°03′14″N, referred to as Site II), another town of Wendeng District. In total, 80 = 4 replications × 4 ages × [3 seasons (of 2017, Site I) + 2 seasons (of 2018, Site II)] soil samples were obtained (Supplementary Table S1).
## 2.2. Disease index calculations
The root rot disease index (DI) was estimated by dividing the number of diseased ginseng roots by the total number of plants investigated. The severity of root disease observed in the pot experiment was determined by the presence of surface lesions, which were quantified on a scale from 0 to 4, with 0 representing no lesions and 1, 2, 3, and 4 standing for lesions that are <$10\%$, $10\%$−$33\%$, $33\%$−$67\%$, and larger than $67\%$ of the total area of the root. The severity of the disease at one sampling site was recorded as the DI, which was calculated as follows: *Si is* the severity rating, *Xi is* the number of roots with the corresponding severity rating, and N is the total number of roots in one sampling site (Jiao et al., 2019).
## 2.3. Soil chemical properties
The soil chemical properties, including available nitrogen (AN), available phosphorus (AP), available potassium (AK), organic matter (OM), pH, exchangeable calcium, and magnesium (E-Ca and E-Mg), were measured with the alkali hydrolysis diffusion method, the NaHCO3 extraction method, the NH4OAC extraction method, the potassium dichromate volumetric method, the pH meter, and the atomic absorption spectrophotometry, respectively, as described by Bao [2000].
## 2.4. DNA extraction, PCR, and high-throughput sequencing
Soil microbial DNA was extracted from 0.4 g soil with the PowerSoil DNA Isolation Kit (Mobio Laboratories Inc., Carlsbad, CA, USA) according to the manufacturer's instructions. The bacterial universal V3-V4 region of the 16S rRNA gene was amplified with amplicon PCR forward primer 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and reverse primer 806R (5′-GGACTACCAGGGTATCTAAT-3′) (Zheng et al., 2017). The fungal universal ITS1 region was amplified with the amplicon PCR forward primer (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and reverse primer (5′-TGCGTTCTTCATCGATGC-3′) (Mukherjee et al., 2014). Three PCR products per sample were pooled, purified, and quantified by real-time PCR. Parallel-tagged sequencing was performed on an Illumina MiSeq platform (Allwegene, Beijing, China) according to standard protocols (Edgar, 2013; Zhang et al., 2018). Specifically, split reads were merged using FLASH (Magoc and Salzberg, 2011), where forward and reverse reads had overlapping base lengths ≥10 bp and were sorted into each sample by the unique barcodes with QIIME (Caporaso et al., 2010). The sequences with a quality score below 20 contained ambiguous bases or did not exactly match the primer sequences, and barcode tags were removed to become raw data (Wang et al., 2019). Chimeras were removed with USEARCH against the Gold and UNITE reference databases, and sequences shorter than 200 bp were removed to become clean data. The high-quality sequences were clustered into operational taxonomic units (OTUs) at a threshold of $97\%$ similarity using the UPARSE pipeline (Edgar, 2013). Singletons that occurred only one time in the entire data set were removed from subsequent analyses to reduce the overprediction of rare OTUs (Jiao et al., 2019). The representative OTU sequences were aligned and annotated using the Ribosomal Database Project (RDP) for 16S and Unite for ITS (Shen et al., 2019). The datasets generated for this study can be found in NCBI, and the BioProject accession numbers are PRJNA890432 for bacteria and PRJNA890915 for fungi.
## 2.5. Statistical analyses
DI was expressed as means and standard deviation of four replicates. ANOVA was performed with SPSS 19.0 (SPSS Inc., Chicago, IL, USA), and significant differences among groups were determined at the $P \leq 0.05$ level according to the Duncan multiple range test. Other statistical analyses were based on R programs (v3.2.2; http://www-r-project.org/). The alpha (α) diversity indexes Chao1 and Shannon were calculated to assess the microbial abundance and diversity, and the differences among different groups were tested by Duncan's multiple range test. Microbial beta-diversity was quantified with two axes of a non-metric multidimensional scaling (NMDS) analysis of Bray–Curtis dissimilarities in the OTUs community matrix using the “vegan” package in R. The alluvial figures over time for bacterial phylum and fungal class were based on the “ggalluvial“ package in R, and Proteobacteria was divided into Alpha-, Beta-, Delta-, and Gamma-Proteobacteria classes due to the high relative abundance of Protecobacteria. To obtain the biomarkers of microbial taxa across American ginseng residence time in the field, we used the Random Forests approach provided by Zhang J. et al. [ 2018] to regress the relative abundances of bacterial and fungal taxa at the genus level against American ginseng residence time in the field. We showed the obtained biomarkers with the “phetamap” package in R. Linear models for the relationships of microbial indicators with DI and the relative importance of each of the predictors in this model was also tested after stepwise model selection using stepAIC in R to select the best explanatory power. The Mantel test was used to identify correlations between microbial composition and soil chemical properties, and perMANOVA was used to test the effects of American ginseng over years of cultivation and seasons on soil chemical properties (Shen et al., 2019). Bray–Curtis distance matrices and Euclid distance matrices were used for microbial composition and soil environmental factors, respectively (Jiao et al., 2016). The ordination diagram was done based on original data of soil chemical properties, DIs of American ginseng, and relative abundances of disease-related microorganisms, using Canoco 4.5 software (Microcomputer Power, Ithaca, NY, USA) with the method described by Bi et al. [ 2018].
## 3.1. Root disease indices of American ginseng
During the four years of the growing phase of American ginseng, the roots gradually became larger (Supplementary Figure S1A), and the root weight of ginseng increased from 0.36 ± 0.14 g to 17.81 ± 7.46 g (Supplementary Figure S1B). Meanwhile, the root rot disease of American ginseng exhibited an aggravating trend with ginseng ages in both sites (Figures 1A, B). The disease indices of ginseng were higher in autumn than in summer for the corresponding ages, though no significant difference was found. In Site I, the disease indices of 3- and 4-year ginsengs were significantly higher ($P \leq 0.05$) than those of 1- and 2-year ginsengs in the corresponding seasons, while in Site II, 4-year ginsengs exhibited a significantly higher DI than 1-, 2-, and 3-year ginsengs. Specifically, the disease indices ranged between 7.3 and 42.0 in Site I and between 11.3 and 36.7 in Site II.
**Figure 1:** *Root disease index of American ginseng of different ages in summer and autumn of Dashuipo in 2017 (A) and Zetou in 2018 (B). Error bars represent standard errors of four replicates, and different letters indicate a significant difference among different ages of ginseng (P < 0.05, according to Duncan's multiple range test).*
## 3.2. Dynamic of rhizosphere microbiota overtime during the 1-year to 4-year growth of American ginseng
Across all the samples, we obtained a total of 3,817,186 and 4,471,469 high-quality 16S and ITS sequences, which were respectively grouped into 9,704 and 4,415 OTUs when using the $97\%$ sequence similarity cutoff. The most abundant bacterial phylum were Proteobacteria ($50.8\%$), Acidobacteria ($15.5\%$), Actinobacteria ($11.4\%$), and Chloroflexi ($6.3\%$), while fungal sequences were primarily composed of the phylum Ascomycota ($62.1\%$), Mortierellomycota ($16.7\%$), and Basidiomycota ($11.3\%$). According to the rarefaction and species accumulation curves, it can be inferred that the sequencing depth and sample amount are enough for subsequent analysis (Supplementary Figure S2).
For bacteria at the two sites, the Shannon indices increased with season changes in the first year, peaked in the autumn, and then maintained a high level in the second year. In the third and fourth years, the indices exhibited a similar trend to the first year (Figures 2A, B). Besides, in spring and summer, the Shannon indices for 2-year American ginseng cultivated soil were compared to the other 3 years. However, the indices for autumn soil over the 4 years were similar. The fungi Shannon indices did not change significantly between years, and the trend was similar with bacteria (Figures 2C, D).
**Figure 2:** *Alpha-diversity of microorganisms during the growing stages of American ginseng. Chao1 of bacteria (A), Shannon of bacteria (B), Chao1 of fungi (C), and Shannon of fungi (D). Thick horizontal bars show the median. The upper and lower “hinges” correspond to the 25th and 75th percentiles, and whiskers extend from the hinge to the highest (or lowest) value that is within 1.5 × interquartile range (IQR) of the hinge. Boxes that do not share the same letter indicate a significant difference (P < 0.05, according to Duncan's multiple range test).*
The NMDS revealed that the soil microbial community of American ginseng rhizosphere soil exhibited a gradient change among seasons during 4 years of growth (Figure 3), with significant differences being found at taxonomic levels (ANOSIM test). The differences in bacterial communities among years were larger than those between seasons at both sites (Figures 3A, B), indicating that soil bacterial communities were more influenced by years of cultivation than seasons. With respect to the fungal community, the differences among years were larger than seasons at Site I (Figure 3B) but were smaller than seasons at Site II (Figure 3D). Moreover, we found that the microbial communities of 3- and 4-year American ginseng rhizosphere soil at both sites were so similar that they could not be separated in the NMDS plots. However, the microbial communities of 1-year ginseng rhizosphere soil were far from those of 3- and 4-year ginseng, while the microbial communities of 2-year ginseng rhizosphere soil sat between those of 1-, 3-, and 4-year ginseng rhizosphere soil.
**Figure 3:** *The general pattern of microbial beta-diversity in the soil of three seasons during four years. NMDS showed the structure of the bacterial community of Site I (A) and Site II (B), and the fungal community of Site I (C) and Site II (D). Similarity values among the samples of different seasons (“Seasons”) during 1- to 4-year (“Years”) were examined via the ANOSIM test, which are shown in each plot. 60% confidence ellipses were shown around the samples grouped based on different ages of American ginseng (years).*
The relative abundances of bacterial phyla exhibited a distinct cyclic variation with seasons during the 4 years except for the second year (Figure 4A). From spring to autumn, the relative abundance of Acidobacteria increased while that of Gammaproteobacteria decreased dramatically in 1-, 3-, and 4-year soils. For 2-year-old soil, among different seasons, the bacterial composition kept steady. Compared with the bacterial phylum, the fungal composition fluctuated slightly in class level among different seasons and years, except for the samples of spring in the 3-year soil (Figure 4B). In spring in the 3-year soil, the relative abundance of Dothideomicetes was higher, while the relative abundance of Mortierellomycetes was lower than the other groups. Furthermore, we discovered that Dothideomicetes was most abundant in the spring of all 4 years of soil. Besides, the relative abundance of Tremellomycetes in the three seasons' soil during 1 and 2 years the soil was higher than that in 3 and 4 years.
**Figure 4:** *Average relative abundances change over time of bacterial phylum (A) and fungal class (B). Alluvial figures were plotted based on 80 samples of the two sites.*
## 3.3. Specific taxa of the root microbiota are associated with residence time
Based on the cross-validation result of the rfcv() function in the R package “randomForest,” nine bacterial genera were screened as biomarkers (Figure 5A). The heatmap showed that the relative abundance of Sulfuriferula, Pseudarthrobacter, Oryzihumus, Blastococcus, Nakamurella, Variibacter, and Symbiobacterium decreased over residence time, while that of Pandoraea and Burkholderia-Paraburkholderia increased. Based on how these bacteria are classified at the phylum level, the number of biomarkers that belong to Actinobacteria and Firmicutes has decreased over time, while the number of biomarkers that belong to Proteobacteria has both increased and decreased (Figure 5B).
**Figure 5:** *Bacterial taxonomic biomarkers of American ginseng cultivated time in fields. (A) The top nine biomarker bacterial genera were identified by applying Random Forests regression of their relative abundances in soil against American ginseng years of cultivation and seasons in the field. Biomarker taxa are ranked in descending order of importance to the accuracy of the model. (B) A heat map showing the relative abundances of the top nine predictive biomarker bacterial genera.*
With the same method, 22 fungal genera were screened as biomarkers (Supplementary Figure S3A). The relative abundances of Staphylotrichum, Gymnomyces, Solicoccozyma, Cystodendron, Dioszegia, Lipomyces, Byssocorticium, Stilbella, Pseudaleuria, and Trechispora of the first year were higher than those of the last 3 years and exhibited a decreasing trend over the years; those of Hebeloma, Remersonia, Chaetomidium, Chaetomium, Polyscytalum, and Suillus increased and reached a peak in the second year and then decreased; while those of Fusarium, Pseudeurotium, Geomyces, Elaphomyces, Mortierella, and Sordaria were higher in the last 2 years (Supplementary Figure S3B) when the root rot was severe, their abundances decreased. Mortierella belongs to Mortierellomycota, while the rest of the biomarkers belong to Ascomycota and Basidiomycota.
In total, 14 fungi genera were isolated and identified in the diseased root, with Fusarium accounting for the highest proportion ($63.9\%$) and Trichoderma accounting for the lowest ($12.0\%$), and the proportion of Fusarium increased while that of Trichoderma decreased with ginseng age. In addition, Rhizopus, Plectosphaerella, Mortierella, Alternaria, Zalerion, Rosellinia, Rhizoctonia, Pythium, Penicillium, Paraphaeosphaeria, Mucor, and Chaetomium were also identified, and a few isolated fungi were not identified (Supplementary Figure S4).
## 3.4. Relationship between microbial indicators and DI
To study the relationship between DI and microbial indicators, two linear models, bacterial indicators and fungal indicators, were constructed, respectively. Bacterial indicators included Chao1, Shannon, NMDS1, and NMDS2, the screened biomarkers of time (Figure 5). The retained indicators after stepwise selection are shown in Table 1, and the proportion of variance explained by the model was $46.2\%$. Of the retained indicators, the relative abundances of Pandoraea, Rhizomicrobium, and bacterial Chao1 were positively correlated with DI, while those of Blastococcus and Symbiobacterium were negatively correlated with DI ($P \leq 0.05$, Table 1). Fungal indicators included Chao1, Shannon, NMDS1, and NMDS2, the screened biomarkers of time (Supplementary Figure S4), and the fungi isolated from the rotted root of American ginseng (Supplementary Figure S4) were initially selected. The retained indicators after stepwise selection are shown in Table 2, and the proportion of variance explained by the model was $87.5\%$, which was higher than the model constructed with bacterial indicators and DI, indicating that fungi may play a more important role in the occurrence of root rot disease in American ginseng.
## 3.5. Relationship between soil chemical properties and community composition
Soil chemical properties differed significantly among different samples. According to the multi-factor analysis of the general linear model, years of cultivation influenced pH, OM, AP, AK, and E-Mg, and sampling seasons influenced AN, AK, E-Ca, and E-mg. Sampling sites influenced pH, OM, AN, AP, AK, and E-Ca (Table 3). In both two sites, the soil is slightly acidic, with the pH at a range of 4.5–5.4. The contents of AN, AK, E-Ca, and E-Mg were higher, while AP was lower in Site I than in Site II. Regarding the over 2 years of cultivation, the pH and AP of the 2-year soil were the highest of the 4 years' soil. No clear trend was found, considering the changes in chemical properties among seasons.
**Table 3**
| Sample | pH | OM (%) | AN (mg kg−1) | AP (mg kg−1) | AK (mg kg−1) | E-Ca (g kg−1) | E-Mg (mg kg−1) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| D-Spr1 | 5.04 ± 0.25ab | 1.18 ± 0.16bcd | 107.91 ± 16.03abc | 19.89 ± 12.79b | 164.97 ± 93.92c | 2.20 ± 0.54ab | 198.63 ± 63.61ab |
| D-Sum1 | 5.02 ± 0.15ab | 1.22 ± 0.28bcd | 96.88 ± 3.94bc | 21.95 ± 9.04b | 180.64 ± 3.78c | 1.66 ± 0.50abc | 168.38 ± 48.18abc |
| D-Aut1 | 5.21 ± 0.14ab | 1.33 ± 0.16bcd | 86.89 ± 3.42c | 25.61 ± 10.49b | 157.15 ± 30.85c | 0.96 ± 0.42cd | 151.27 ± 80.04abc |
| D-Spr2 | 5.13 ± 0.21ab | 1.84 ± 0.52a | 115.40 ± 4.63ab | 62.43 ± 19.46a | 223.09 ± 34.06bc | 2.36 ± 0.08a | 190.21 ± 19.50ab |
| D-Sum2 | 5.42 ± 0.40a | 1.70 ± 0.31ab | 93.7 ± 9.29bc | 66.32 ± 16.81a | 212.68 ± 40.75bc | 1.73 ± 0.68abc | 155.49 ± 41.62abc |
| D-Aut2 | 5.19 ± 0.01ab | 1.63 ± 0.19abc | 108.80 ± 5.32abc | 61.80 ± 7.93a | 250.53 ± 1.61bc | 0.97 ± 0.12cd | 85.01 ± 3.07bc |
| D-Spr3 | 4.84 ± 0.24b | 0.92 ± 0.15d | 89.23 ± 7.24c | 32.87 ± 9.60b | 211.24 ± 24.37bc | 1.92 ± 0.05abc | 193.36 ± 13.70ab |
| D-Sum3 | 5.01 ± 0.15ab | 1.12 ± 0.23bcd | 93.84 ± 13.17bc | 33.49 ± 10.9b | 283.84 ± 45.07b | 1.32 ± 0.50bcd | 140.12 ± 47.54abc |
| D-Aut3 | 4.99 ± 0.13ab | 1.13 ± 0.28bcd | 126.16 ± 6.87a | 31.80 ± 9.48b | 374.34 ± 64.95a | 0.67 ± 0.22d | 106.09 ± 19.65bc |
| D-Spr4 | 5.07 ± 0.13ab | 1.14 ± 0.22bcd | 102.39 ± 11.49bc | 48.07 ± 10.62ab | 190.02 ± 19.04bc | 2.31 ± 0.27a | 222.29 ± 40.80a |
| D-Sum4 | 5.20 ± 0.27ab | 1.08 ± 0.15cd | 98.21 ± 12.33bc | 49.09 ± 12.9ab | 246.60 ± 44.08bc | 2.12 ± 0.83ab | 203.26 ± 28.62ab |
| D-Aut4 | 5.19 ± 0.19ab | 1.01 ± 0.18d | 114.01 ± 14.22ab | 39.82 ± 18.20ab | 282.08 ± 41.74b | 1.05 ± 0.29cd | 132.44 ± 22.59abc |
| Z-Sum1 | 4.94 ± 0.19cd | 1.17 ± 0.16a | 72.92 ± 10.29ab | 82.27 ± 36.17a | 211.61 ± 53.11ab | 0.85 ± 0.07abc | 206.90 ± 38.34a |
| Z-Aut1 | 5.16 ± 0.06bcd | 1.29 ± 0.12a | 102.13 ± 17.85a | 58.24 ± 7.58ab | 137.66 ± 12.93b | 1.05 ± 0.18ab | 239.24 ± 20.84a |
| Z-Sum2 | 5.46 ± 0.12a | 0.99 ± 0.15a | 78.20 ± 18.37ab | 86.90 ± 9.70a | 155.15 ± 14.79b | 0.78 ± 0.05abc | 113.28 ± 8.26b |
| Z-Aut2 | 5.19 ± 0.11bcd | 1.12 ± 0.07a | 81.72 ± 17.39ab | 79.89 ± 10.34a | 150.71 ± 4.04b | 0.62 ± 0.04bc | 111.31 ± 6.20b |
| Z-Sum3 | 4.81 ± 0.03d | 1.13 ± 0.06a | 93.96 ± 20.59ab | 52.56 ± 6.42ab | 253.74 ± 42.12a | 0.54 ± 0.19c | 122.20 ± 22.66b |
| Z-Aut3 | 5.02 ± 0.02bcd | 1.00 ± 0.11a | 78.22 ± 11.29ab | 46.97 ± 5.21b | 161.36 ± 17.09b | 1.17 ± 0.25a | 217.71 ± 36.20a |
| Z-Sum4 | 4.97 ± 0.21bcd | 1.00 ± 0.22a | 65.48 ± 9.42b | 44.90 ± 13.12b | 176.88 ± 64.03b | 0.89 ± 0.42abc | 166.61 ± 93.11ab |
| Z-Aut4 | 4.86 ± 0.03d | 1.01 ± 0.11a | 72.71 ± 4.02ab | 54.29 ± 13.51ab | 179.70 ± 35.46b | 0.63 ± 0.20bc | 115.35 ± 15.05b |
| Significant due to | | | | | | | |
| Years | * | * | NS | * | * | NS | * |
| Seasons | NS | NS | * | NS | * | * | * |
| Sites | * | * | * | * | * | * | NS |
Soil chemical properties were significantly correlated to both the bacterial (Mantel: $R = 0.2516$, $P \leq 0.001$) and fungal (Mantel: $R = 0.3030$, $P \leq 0.001$) (Supplementary Table S2). To disentangle the relationship between soil chemical properties, DI, and bacterial and fungal genera related to DI screened by linear models, RDA was done based on the above indicators (Figure 6). The ordination diagram showed that AK and AN were positively correlated with Pandoraea and Rhizomicrobium, pH, E-Ca, and E-Mg were positively correlated with Suillus, Blastococcus, Mortierella, and Symbiobacterium, organic matter was positively correlated with Symbiobacterium and Elaphomyces, and AP was positively correlated with Laccaria, Hebeloma, Geomyces, Gymnomyces, Hirsutella, Polyscytalum, Remersonia, Acremonium, Goffeauzyma, Pseudeurotium, Entoloma, Inocybe, Rhizopus and Paraphaeosphaeria. Moreover, the concentrations of AK and AN were positively correlated with DI, while those of AP, E-Mg, E-Ca, organic matter, and pH were negatively correlated with DI. In addition, AK and AN were higher in 3- and 4-year ginseng residence soil, while pH, AP, E-Mg, E-Ca, and organic matter were relatively high in 1- and 2-year ginseng residence soil, indicating that the soil chemical properties affected the occurrence of root rot disease in ginseng.
**Figure 6:** *Redundancy analysis (RDA) between soil chemical properties, Disease index (DI), and the relative abundance higher than 1% of microbial genera related with DI. The environmental parameters are represented by red lines, different treatments are solid triangles, and microbial genera are blue lines. Fus, Fusarium; Rhi, Rhizomicrobium; Pan, Pandoraea; Mor, Mortierella; Pen, Penicillium; Geo, Geomyces; Acr, Acremonium; Pse, Pseudeurotium; Rhi, Rhizopus; Par: Paraphaeosphaeria.*
## 4. Discussion
In our study, the DI became larger with the American ginseng age, which is similar to the study by Dong et al. [ 2016] on Panax notoginseng, which also belongs to Panax. The isolated Fusarium increased as the American ginseng aged, whereas Trichoderma decreased. Fusarium is a genus that contains root rot disease causal pathogens (Punja et al., 2005, 2007), and the *Thichoderma genus* contains the most common pathogen antagonists in soil (Vinale et al., 2008; Chen et al., 2011; Gao et al., 2019). The shifts of the two genera in ginseng roots reflect that pathogens become stronger while their antagonism becomes weaker during the growth of American ginseng. In addition, the relative abundance of Fusarium spp. in the soil was much higher in the third and fourth years compared with the first 2 years, indicating that the soil environment became less favorable with the increase in the planting years of American ginseng.
Besides, our study found that the occurrences of root rot disease were affected by seasons. As the soil moisture and temperature are quite different among different seasons, the pathogenicity of fungi is different (Wong et al., 1984; Hudec and Muchova, 2010; Guo et al., 2022). In this study, the diversity of bacteria and fungi showed similar seasonal (annual) periodicity during the 4 years of growth of ginseng except for the second year (Figures 2, 4). Similar to our study, Cregger et al. [ 2012] also found that the variation of microbial communities is highly dependent on seasonal dynamics. Because the microbial community is regulated by climate conditions (e.g., precipitation and temperature), which vary largely between seasons, it is not surprising that the microbial diversity presents a gradient change in different seasons (Guo et al., 2022). Our results revealed that the variation of American ginseng rhizosphere microbial diversity in the second year was not in accordance with the seasonal periodicity. As previously reported, plants can affect the composition of the soil microbial community through plant-soil feedback during growth (Barbara et al., 2015). The soil microbial community should change during rotation. In our study, because the biomass of the maize plant is much larger than that of 1-year-old American ginseng, the effect of maize on the soil microbial community may last until the biomass of American ginseng is large enough to shape the rhizosphere microbial community after 1 year's growth. As a result, the transformation of the maize-shaped microbial community into the American ginseng-shaped microbial community broke the seasonal periodicity in the second year and formed an ecotone of the two communities in the spring of the second year but not the first year, which made the microbial diversity of the 2-year-old ginseng rhizosphere much higher than that in the spring of the other 3 years (Figure 2) because temporal heterogeneity of plant inputs increases soil biodiversity (Eisenhauer, 2016). After the second year, the microbial community finished transforming, thus following the seasonal pattern again in the third and fourth years of American growth. Therefore, the second year is the turning point of new microbial community formation during the 4-year growth of American ginseng. The results of the NMDS analysis also verify the above viewpoint. From the NMDS map, we observed the driving force of the bacterial and fungal community of the American ginseng rhizosphere changing gradually over 1–4 years. The composition of the microbial community in the third and fourth years is close, while that in the second year is between the first year and the last two years (Figure 3). NMDS shows the process of continuous and gradual change in the microbial community under the effect of plants, which is consistent with the previous research on rice (Zhang J. et al., 2018) and cotton (Li D. et al., 2022). Studies have shown that rotation can improve microbial diversity, but the mechanisms are not fully elaborated (Venter et al., 2016; Xie et al., 2022). A potential reason may be that an ecotone of two microbial communities forms when the effects on microbial communities transform from one plant species to another, according to our study.
Proteobacteria, Acidobacteria, and Actinobacteria were the most abundant phyla across all samples, which is consistent with previous studies where results were derived from soils cultivated with maize (Li Y. et al., 2022), American ginseng (Jiang et al., 2019), and even other crops (Shen et al., 2019), indicating that there is no clear distinction among different species of the plant rhizosphere bacteria at the phylum level. At the genus level, Blastococcus and Symbiobacterium spp. were negatively correlated with DI (Table 2) and decreased over time (Figure 5). Blastococcus and Symbiobacterium spp. belong to Actinobacteria and Firmicutes, respectively, which were reported to be involved in disease suppression due to the production of biocontrol agents that exhibit antimicrobial effects (Mendes et al., 2011; Cha et al., 2016; Shen et al., 2019). With our findings, we can speculate that Blastococcus and Symbiobacterium spp. may be biocontrol agents of American ginseng, though more research is needed to confirm this. Similarly, Pandoraea spp. decreased over time and were negatively correlated with DI in our study, which could suppress pathogens (Jin et al., 2007) and be a potential biocontrol agent (Kotan et al., 2013), as previously reported.
With respect to fungi, our study showed that Ascomycota and Basidiomycota dominated at the phylum level across all samples, which is consistent with a previous study (Shen et al., 2019). Nevertheless, the relative abundance of Ascomycota was positively correlated with the index of disease caused by Fusarium in the study of Shen et al. [ 2019], while some of the fugal genera belonging to Ascomycota were positively correlated with DI, and some were negatively correlated with DI in our study. This discrepancy between the two studies may result from the fact that the taxon is quite different in Ascomycota among different soils, which indicates that microorganisms at the phylum level cannot be used as biomarkers to judge whether their resident soil is conducive to disease or not. Among biomarkers of American ginseng resident time, Laccaria, Goffeauzyma, Entoloma, Staphylotrichum, Gymnomys, Inocybe, Hirsutella, Penicillium, Tomentella, and Suillus spp. had a significant negative correlation with DI, while Hebeloma, Elaphomyces, Pseudeurotium, Fusarium, Geomyces, Polyscytalum, Remersonia, Rhizopus, Paraphaeosphaeria, Mortierella, and Metarhizium spp. were significantly positively correlated with the DI of American ginseng. Previous studies about the microbial community and root rot disease of plants of Panax, Staphylotrichum, Penicillium, Fusarium, and Mortierella spp. were also found to be associated with plant disease (Tan et al., 2017; Jiang et al., 2019; Wei et al., 2020). We also revealed their relationship with resident time to find correlations between DI and fungi. We found that fungi negatively correlated with DI in 1- and 2-year samples, while fungi positively correlated with DI in 3- and 4-year samples. These phenomena can partly explain why the disease in 3- and 4-year-old American ginseng was more severe than that in 1- and 2-year-old American ginseng. Unfortunately, the factors leading to the change in the rhizosphere microbial community are complex and diverse, such as plants (Liu et al., 2022) and nutrients (Rosinger et al., 2022). Thus, to establish a causal relationship between rhizosphere microbial community shift and disease index, more experiments, as described in the study of Zhou et al. [ 2022], need to be performed in future studies.
When discussing microbial diversity, many researchers believe that a high microbial diversity index signifies healthy soil and can suppress plant disease (Janvier et al., 2007; Larkin, 2015; Vukicevich et al., 2016). In our study, however, the bacterial and fungal Shannon had no significant correlation with DI, while the bacterial and fungal Chao1 were positively correlated with DI. Although it is not common, previous studies also reported a similar phenomenon: the bacterial Chao1 index of disease-suppressive soil was lower than that of disease-conducive soil (Xiong et al., 2017), and the fungal Chao1 showed increasing trends in soils used to cultivate American ginseng compared with those of traditional cropping systems (Dong et al., 2017). Unfortunately, neither of the two studies explained the possible reasons. According to previous studies, when infected by pathogens in soil, plants may recruit many other rhizosphere microorganisms to fight against pathogens (Xiong et al., 2017; Huang et al., 2019), which sometimes contributes to the increase in microbial diversity. To our knowledge, however, the exact mechanism still needs to be elucidated. Microbial diversity may not be a robust index for evaluating whether the soil is disease-suppressive or conducive, especially for agricultural soils, which are disturbed most by human activity.
Moreover, the composition of soil microorganisms is complex, and the diversity index is not competent to reflect the composition of microorganisms. For example, microbial communities with the same diversity may have different richness and evenness (Kennedy, 1999; Hu et al., 2022). In this study, the microbial community structure changed severely during the four years, especially in the second year of American ginseng growth, but the diversity index did not change significantly among the 1-, 3-, and 4-year soils in the same seasons, indicating that a relatively steady microbial structure formed gradually, which can be supported by the study of Jiao et al. [ 2019] who demonstrated that at least three years' rotation is needed to replant American ginseng.
Soil chemical properties gradually change with plant growth and fertilization in the field (Jiao et al., 2019). In our study, organic matter and pH were higher in 1- and 2-year soil, while AK and AN were higher in 3- and 4-year soil (Table 3), resulting from American ginseng fertilization strategies. In our study and usually in China, farmers apply sufficient organic fertilizer before the sowing of American ginseng, and apply more chemical fertilizer mainly composed of available nutrients in the third and fourth year of growth of American ginseng. Soil chemical properties also affected the soil microbial community (Table 3), which is consistent with previous reports (Bell et al., 2013; Dong et al., 2016). Soil AK and AN were positively correlated with DI, while pH and organic matter were negatively correlated with DI (Figure 6), indicating that the alternation of soil nutrients aggravates the soil microbial community, which makes root rot disease occur more easily. As a result, we recommend using more organic fertilizer and less chemical fertilizer during the third and fourth years of the growth of American ginseng.
## 5. Conclusion
The root rot disease in American ginseng grows more severe with each year that it is cultivated. The second year is the vital period for shifting the American ginseng's rhizosphere microbial community. Disease aggravation after the third year is related to the deterioration of the rhizosphere microecosystem. Increases in soil available nutrients and decreases in organic matter may be related to changes in rhizosphere microbial community composition, diversity, and DI, implying that, in the third and fourth years of American ginseng growth, we should reduce chemical fertilizer input and increase organic fertilizer input.
## Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material.
## Author contributions
W-WG and Y-MB conceived and designed the research. Y-MB, X-MZ, X-LJ, YW, and J-FL performed the experiments. Y-MB and G-LT analyzed the data. Y-MB, W-WG, and NP wrote the manuscript. All authors read and approved the paper.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1097742/full#supplementary-material
## References
1. Bao S. D.. *Soil and Agricultural Chemistry Analysis* (2000)
2. Barbara R.-H., Bunger W., Burbano C. S., Sabale M., Hurek T.. **Roots shaping their microbiome: global hotspots for microbial activity**. *Annu. Rev. Phytopathol.* (2015) **53** 403-424. DOI: 10.1146/annurev-phyto-082712-102342
3. Bell T. H., Yergeau E., Maynard C., Juck D., Whyte L. G., Greer C. W.. **Predictable bacterial composition and hydrocarbon degradation in Arctic soils following diesel and nutrient disturbance**. *ISME. J.* (2013) **7** 1200-1210. DOI: 10.1038/ismej.2013.1
4. Berendsen R. L., Pieterse C. M., Bakker P. A.. **The rhizosphere microbiome and plant health**. *Trends Plant Sci.* (2012) **17** 478-486. DOI: 10.1016/j.tplants.2012.04.001
5. Berendsen R. L., Vismans G., Yu K., Song Y., De Jonge R., Burgman W. P.. **Disease-induced assemblage of a plant-beneficial bacterial consortium**. *ISME J.* (2018) **12** 1496-1507. DOI: 10.1038/s41396-018-0093-1
6. Berg G., Smalla K.. **Plant species and soil type cooperatively shape the structure and function of microbial communities in the rhizosphere**. *FEMS Microbiol. Ecol.* (2009) **68** 1-13. DOI: 10.1111/j.1574-6941.2009.00654.x
7. Bi Y.-M., Tian G.-L., Wang C., Zhang Y., Wang D.-N., Zhang F.-F.. **Differential effects of two earthworm species on Fusarium wilt of strawberry**. *Appl. Soil Ecol.* (2018) **126** 174-181. DOI: 10.1016/j.apsoil.2018.02.024
8. Caporaso J. G., Kuczynski J., Stombaugh J., Bittinger K., Bushman F. D., Costello E. K.. **QIIME allows analysis of high-throughput community sequencing data**. *Nat. Methods* (2010) **7** 335-336. DOI: 10.1038/nmeth.f.303
9. Cha J.-Y., Han S., Hong H.-J., Cho H., Kim D., Kwon Y.. **Microbial and biochemical basis of a Fusarium wilt-suppressive soil**. *ISME J.* (2016) **10** 119-129. DOI: 10.1038/ismej.2015.95
10. Chen L., Yang X., Raza W., Li J., Liu Y., Qiu M.. *Appl. Microbiol. Biotechnol.* (2011) **89** 1653-1663. DOI: 10.1007/s00253-010-2948-x
11. Chen P., Wang Y.-Z., Liu Q.-Z., Zhang Y.-T., Li X.-Y., Li H.-Q.. **Phase changes of continuous cropping obstacles in strawberry (Fragaria × ananassa Duch.) production**. *Appl. Soil Ecol.* (2020) **155** 103626. DOI: 10.1016/j.apsoil.2020.103626
12. Cregger M. A., Schadt C. W., Mcdowell N. G., Pockman W. T., Classen A. T.. **Response of the soil microbial community to changes in precipitation in a semiarid ecosystem**. *Appl. Environ. Microbiol.* (2012) **78** 8587-8594. DOI: 10.1128/AEM.02050-12
13. Dong L., Xu J., Feng G., Li X., Chen S.. **Soil bacterial and fungal community dynamics in relation to**. *Sci. Rep.* (2016) **6** 31802. DOI: 10.1038/srep31802
14. Dong L., Xu J., Zhang L., Yang J., Liao B., Li X.. **High-throughput sequencing technology reveals that continuous cropping of American ginseng results in changes in the microbial community in arable soil**. *Chin. Med.* (2017) **12** 18. DOI: 10.1186/s13020-017-0139-8
15. Edgar R. C.. **UPARSE: highly accurate OTU sequences from microbial amplicon reads**. *Nat. Methods* (2013) **10** 996-998. DOI: 10.1038/nmeth.2604
16. Eisenhauer N.. **Plant diversity effects on soil microorganisms: spatial and temporal heterogeneity of plant inputs increase soil biodiversity**. *Pedobiologia* (2016) **59** 175-177. DOI: 10.1016/j.pedobi.2016.04.004
17. Farh M. E.-A., Kim Y.-J., Kim Y.-J., Yang D.-C.. *J. Ginseng Res.* (2018) **42** 9-15. DOI: 10.1016/j.jgr.2017.01.004
18. Gao Z., Han M., Hu Y., Li Z., Liu C., Wang X.. **Effects of continuous cropping of sweet potato on the fungal community structure in rhizospheric soil**. *Front. Microbiol.* (2019) **10** 2269. DOI: 10.3389/fmicb.2019.02269
19. Garbeva P., Van Elsas J. D., Van Veen J. A.. **Rhizosphere microbial community and its response to plant species and soil history**. *Plant Soil* (2008) **302** 19-32. DOI: 10.1007/s11104-007-9432-0
20. Guo X., Yuan M., Lei J., Shi Z., Zhou X., Li J.. **Climate warming restructures seasonal dynamics of grassland soil microbial communities**. *mLife* (2022) **1** 245-256. DOI: 10.1002/mlf2.12035
21. Hu L., Robert C. a,.M, Cadot S., Zhang X., Ye M.. **Root exudate metabolites drive plant-soil feedbacks on growth and defense by shaping the rhizosphere microbiota**. *Nat. Commun.* (2018) **9** 2738. DOI: 10.1038/s41467-018-05122-7
22. Hu S., He R., Zeng J., Zhao D., Wang S., He F.. **Lower compositional variation and higher network complexity of rhizosphere bacterial community in constructed wetland compared to natural wetland**. *Microb. Ecol* (2022) **2022** 1-15. DOI: 10.1007/s00248-022-02040-6
23. Huang W., Sun D., Fu J., Zhao H., Wang R., An Y.. **Effects of continuous sugar beet cropping on rhizospheric microbial communities**. *Genes (Basel)* (2019) **11** 13. DOI: 10.3390/genes11010013
24. Hudec K., Muchova D.. **Influence of temperature and species origin on**. *Plant Protect. Sci.* (2010) **46** 59-65. DOI: 10.17221/12/2009-PPS
25. Janvier C., Villeneuve F., Alabouvette C., Edel-Hermann V., Mateille T., Steinberg C.. **Soil health through soil disease suppression: which strategy from descriptors to indicators?**. *Soil Biol. Biochem.* (2007) **39** 1-23. DOI: 10.1016/j.soilbio.2006.07.001
26. Jiang J., Yu M., Hou R., Li L., Ren X., Jiao C.. **Changes in the soil microbial community are associated with the occurrence of**. *Plant Soil* (2019) **438** 143-156. DOI: 10.1007/s11104-018-03928-4
27. Jiao S., Liu Z., Lin Y., Yang J., Chen W., Wei G.. **Bacterial communities in oil contaminated soils: biogeography and co-occurrence patterns**. *Soil Biol. Biochem.* (2016) **98** 64-73. DOI: 10.1016/j.soilbio.2016.04.005
28. Jiao X. L., Zhang X. S., Lu X. H., Qin R., Bi Y. M., Gao W. W.. **Effects of maize rotation on the physicochemical properties and microbial communities of American ginseng cultivated soil**. *Sci. Rep.* (2019) **9** 8615. DOI: 10.1038/s41598-019-44530-7
29. Jin Z. X., Wang C., Dong W., Li X.. **Isolation and some properties of newly isolated oxalate-degrading**. *J. Appl. Microbiol.* (2007) **103** 1066-1073. DOI: 10.1111/j.1365-2672.2007.03363.x
30. Kennedy A. C.. *Microbial Diversity in Agroecosystem Quality* (1999)
31. Kotan R., Okutucu A., Ala Görmez A., Karagoz K., Dadasoglu F., Karaman I.. **Parasitic bacteria and fungi on common mistletoe (**. *J. Phytopathol.* (2013) **161** 165-171. DOI: 10.1111/jph.12048
32. Larkin R. P.. **Soil health paradigms and implications for disease management**. *Annu. Rev. Phytopathol.* (2015) **53** 199-221. DOI: 10.1146/annurev-phyto-080614-120357
33. Li D., Yang Y., Zhao Y., Tian G., Li M., Qiu H.. **Differing roles of bacterial and fungal communities in cotton fields by growth stage**. *Agronomy* (2022) **12** 657. DOI: 10.3390/agronomy12030657
34. Li Q., Wu Y., Wang J., Yang B., Chen J., Wu H.. **Linking short-chain N-acyl homoserine lactone-mediated quorum sensing and replant disease: a case study of Rehmannia glutinosa**. *Front. Plant Sci.* (2020) **11** 787. DOI: 10.3389/fpls.2020.00787
35. Li Y., Qu Z., Xu W., Chen W., Hu Y., Wang Z.. **Maize (Zea mays L.) genotypes induce the changes of rhizosphere microbial communities**. *Microbiol* (2022) **204** 321. PMID: 35567648
36. Liu Y., Liu Y., Zeng C., Wang J., Nyimbo W. J., Jiao Y.. **Intercropping with Achyranthes bidentata alleviates Rehmannia glutinosa consecutive monoculture problem by reestablishing rhizosphere microenvironment**. *Front. Plant Sci.* (2022) **13** 1041561. DOI: 10.3389/fpls.2022.1041561
37. Magoc T., Salzberg S. L.. **FLASH: fast length adjustment of short reads to improve genome assemblies**. *Bioinformatics* (2011) **27** 2957-2963. DOI: 10.1093/bioinformatics/btr507
38. Mendes R., Kruijt M., De Bruijn I., Dekkers E., Van Der Voort M., Schneider J. H. M.. **Deciphering the rhizosphere microbiome for disease-suppressive bacteria**. *Science* (2011) **332** 1097-1100. DOI: 10.1126/science.1203980
39. Moore V. M., Schlautman B., Fei S. Z., Roberts L. M., Wolfe M., Ryan M. R.. **Plant breeding for intercropping in temperate field crop systems: a review**. *Front. Plant Sci.* (2022) **13** 843065. DOI: 10.3389/fpls.2022.843065
40. Mukherjee P. K., Chandra J., Retuerto M., Sikaroodi M., Brown R. E., Jurevic R.. **Oral mycobiome analysis of HIV-infected patients: identification of Pichia as an antagonist of opportunistic fungi**. *PLoS Pathog.* (2014) **10** e1003996. DOI: 10.1371/journal.ppat.1003996
41. Punja Z. K., Wan A., Goswami R. S., Verma W., Rahman M., Barasubiye T.. **Diversity of**. *Can. J. Plant Pathol.* (2007) **29** 340-353. DOI: 10.1080/07060660709507480
42. Punja Z. K., Wan A., Rahman M.. **Role of**. *Can. J. Plant Pathol.* (2005) **27** 474-474
43. Rosinger C., Keiblinger K. M., Rousk J., Sandén H.. **Shifts in microbial stoichiometry upon nutrient addition do not capture growth-limiting nutrients for soil microorganisms in two subtropical soils**. *Biogeochemistry* (2022) **159** 33-43. DOI: 10.1007/s10533-022-00911-1
44. Sasse J., Martinoia E., Northen T.. **Feed your friends: do plant exudates shape the root microbiome?**. *Trends Plant Sci.* (2018) **23** 25-41. DOI: 10.1016/j.tplants.2017.09.003
45. Sen S., Chen S., Feng B., Wu Y., Lui E., Chakrabarti S.. **Preventive effects of North American ginseng (**. *Phytomedicine* (2012) **19** 494-505. DOI: 10.1016/j.phymed.2012.01.001
46. Shen Z., Xue C., Penton C. R., Thomashow L. S., Zhang N., Wang B.. **Suppression of banana Panama disease induced by soil microbiome reconstruction through an integrated agricultural strategy**. *Soil Biol. Biochem.* (2019) **128** 164-174. DOI: 10.1016/j.soilbio.2018.10.016
47. Singh R. K., Lui E., Wright D., Taylor A., Bakovic M.. **Alcohol extract of North American ginseng (Panax quinquefolius) reduces fatty liver, dyslipidemia, and other complications of metabolic syndrome in a mouse model**. *Can. J. Physiol. Pharmacol.* (2017) **95** 1046-1057. DOI: 10.1139/cjpp-2016-0510
48. Tan Y., Cui Y., Li H., Kuang A., Li X., Wei Y.. **Rhizospheric soil and root endogenous fungal diversity and composition in response to continuous Panax notoginseng cropping practices**. *Microbiol. Res.* (2017) **194** 10-19. DOI: 10.1016/j.micres.2016.09.009
49. Trivedi P., Leach J. E., Tringe S. G., Sa T., Singh B. K.. **Plant-microbiome interactions: from community assembly to plant health**. *Nat. Rev. Microbiol.* (2020) **18** 607-621. DOI: 10.1038/s41579-020-0412-1
50. Vandenkoornhuyse P., Quaiser A., Duhamel M., Le Van A., Dufresne A.. **The importance of the microbiome of the plant holobiont**. *New Phytol.* (2015) **206** 1196-1206. DOI: 10.1111/nph.13312
51. Venter Z. S., Jacobs K., Hawkins H.-J.. **The impact of crop rotation on soil microbial diversity: a meta-analysis**. *Pedobiologia* (2016) **59** 215-223. DOI: 10.1016/j.pedobi.2016.04.001
52. Vinale F., Sivasithamparam K., Ghisalberti E. L., Marra R., Woo S. L., Lorito M.. **Trichoderma-plant-pathogen interactions**. *Soil Biol. Biochem.* (2008) **40** 1-10. DOI: 10.1016/j.soilbio.2007.07.002
53. Vukicevich E., Lowery T., Bowen P., Úrbez-Torres J. R., Hart M.. **Cover crops to increase soil microbial diversity and mitigate decline in perennial agriculture. A review**. *Agron. Sustain. Dev.* (2016) **36** 10. DOI: 10.1007/s13593-016-0385-7
54. Wang H., He Y., Li H., Wu F., Qiu Q., Niu W.. **Rumen fermentation, intramuscular fat fatty acid profiles and related rumen bacterial populations of Holstein bulls fed diets with different energy levels**. *Appl. Microbiol. Biotechnol.* (2019) **103** 4931-4942. DOI: 10.1007/s00253-019-09839-3
55. Wang H., Li X., Li X., Wang J., Li X., Guo Q.. **Long-term no-tillage and different residue amounts alter soil microbial community composition and increase the risk of maize root rot in northeast China**. *Soil Tillage Res.* (2020) **196** 104452. DOI: 10.1016/j.still.2019.104452
56. Wei X., Wang X., Cao P., Gao Z., Chen A. J., Han J.. **Microbial community changes in the rhizosphere soil of healthy and rusty**. *Biomed Res. Int.* (2020) **2020** 8018525. DOI: 10.1155/2020/8018525
57. Wong D. H., Barbetti M. J., Sivasithamparam K.. **Effects of soil-temperature and moisture on the pathogenicity of fungi associated with root-rot of subterranean clover**. *Aust. J. Agric. Res.* (1984) **35** 675-684. DOI: 10.1071/AR9840675
58. Xia Z., He Y., Korpelainen H., Niinemets Ü., Li C.. **Sex-specific interactions shape root phenolics and rhizosphere microbial communities in**. *For. Ecol. Manage.* (2022) **504** 119857. DOI: 10.1016/j.foreco.2021.119857
59. Xia Z., He Y., Yu L., Li Z., Korpelainen H., Li C.. **Revealing interactions between root phenolic metabolomes and rhizosphere bacterial communities in**. *Biol. Fertil. Soils* (2021) **57** 421-434. DOI: 10.1007/s00374-020-01527-z
60. Xie Y., Ouyang Y., Han S., Se J., Tang S., Yang Y.. **Crop rotation stage has a greater effect than fertilisation on soil microbiome assembly and enzymatic stoichiometry**. *Sci. Total Environ.* (2022) **815** 152956. DOI: 10.1016/j.scitotenv.2022.152956
61. Xiong W., Li R., Ren Y., Liu C., Zhao Q., Wu H.. **Distinct roles for soil fungal and bacterial communities associated with the suppression of vanilla**. *Soil Biol. Biochem.* (2017) **107** 198-207. DOI: 10.1016/j.soilbio.2017.01.010
62. Yang J., Jiao X., Gao W.. **Advances in the root diseases of**. *Plant Protect.* (2009) **35** 30-35
63. Zhang J., Zhang N., Liu Y.-X., Zhang X., Hu B., Qin Y.. **Root microbiota shift in rice correlates with resident time in the field and developmental stage**. *Science China Life Sciences* (2018) **61** 613-621. DOI: 10.1007/s11427-018-9284-4
64. Zhang X., Zheng S., Zhang H., Duan S.. **Autotrophic and heterotrophic nitrification-anoxic denitrification dominated the anoxic/oxic sewage treatment process during optimization for higher loading rate and energy savings**. *Bioresour. Technol.* (2018) **263** 84. DOI: 10.1016/j.biortech.2018.04.113
65. Zheng C. J., Liu R., Xue B., Luo J., Gao L., Wang Y.. **Impact and consequences of polyphenols and fructooligosaccharide interplay on gut microbiota in rats**. *Food Funct.* (2017) **8** 1925-1932. DOI: 10.1039/C6FO01783E
66. Zhou X., Zhang X., Ma C., Wu F., Jin X., Dini-Andreote F.. **Biochar amendment reduces cadmium uptake by stimulating cadmium-resistant PGPR in tomato rhizosphere**. *Chemosphere* (2022) **307** 136138. DOI: 10.1016/j.chemosphere.2022.136138
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.